Land Use Change

Land UseChange

Science, Policyand Management

42963_C000.indd 1 11/6/07 11:18:13 AM

42963_C000.indd 2 11/6/07 11:18:13 AM

Land UseChange

Science, Policyand Management

Edited byRichard J. Aspinall

Michael J. Hill

CRC Press is an imprint of theTaylor & Francis Group, an informa business

Boca Raton London New York

42963_C000.indd 3 11/6/07 11:18:13 AM

CRC PressTaylor & Francis Group6000 Broken Sound Parkway NW, Suite 300Boca Raton, FL 33487-2742

© 2008 by Taylor & Francis Group, LLC CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government worksPrinted in the United States of America on acid-free paper10 9 8 7 6 5 4 3 2 1

International Standard Book Number-13: 978-1-4200-4296-2 (Hardcover)

This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with permission, and sources are indicated. A wide variety of references are listed. Reasonable efforts have been made to publish reliable data and information, but the author and the publisher cannot assume responsibility for the validity of all materials or for the conse-quences of their use.

No part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC) 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe.

Library of Congress Cataloging-in-Publication Data

Land use change : science, policy, and management / Richard J. Aspinall and Michael J. Hill [editors].

p. cm.Includes bibliographical references and index.ISBN 978-1-4200-4296-2 (hardcover : alk. paper)1. Land use--Environmental aspects. 2. Land use--Management. 3. Land

use--Case studies. I. Aspinall, Richard J. II. Hill, Michael J. (Michael James), 1937- III. Title.

HD108.3.L362 2008333.73’13--dc22 2007027752

Visit the Taylor & Francis Web site athttp://www.taylorandfrancis.com

and the CRC Press Web site athttp://www.crcpress.com

42963_C000.indd 4 11/6/07 11:18:15 AM

RJA — to Chloe, for her company

MJH — to my mother, Mavis Hill, for her lifelong friendship

42963_C000.indd 5 11/6/07 11:18:16 AM

42963_C000.indd 6 11/6/07 11:18:16 AM

vii

ContentsFigures......................................................................................................................ixPermissions...............................................................................................................xiPreface.................................................................................................................... xiiiAcknowledgments...................................................................................................xvIntroduction...........................................................................................................xviiEditors.................................................................................................................. xxiiiContributors..........................................................................................................xxv

Part I theory and Methodology

Chapter 1 Basic.and.Applied.Land.Use.Science....................................................3

Richard J. Aspinall

Chapter 2 Developing.Spatially.Dependent.Procedures.and.Models.for.Multicriteria.Decision.Analysis:.Place,.Time,.and.Decision.Making.Related.to.Land.Use.Change................................................. 17

Michael J. Hill

Part II Comparative regional Case Studies

Chapter 3 Spatial.Methodologies.for.Integrating.Social.and.Biophysical.Data.at.a.Regional.or.Catchment.Scale........................... 43

Ian Byron and Robert Lesslie

Chapter 4 An.Integrated.Socioeconomic.Study.of.Deforestation.in.Western.Uganda,.1990–2000.............................................................. 63

Ronnie Babigumira, Daniel Müller and Arild Angelsen

Chapter 5 Modeling.Unplanned.Land.Cover.Change.across.Scales:.A.Colombian.Case.Study.................................................................... 81

Andres Etter and Clive McAlpine

Chapter 6 Landscape.Dynamism:.Disentangling.Thematic.versus.Structural.Change.in.Northeast.Thailand...........................................99

Kelley A. Crews

42963_C000.indd 7 11/6/07 11:18:16 AM

viii Contents

Chapter 7 Developing.a.Thick.Understanding.of.Forest.Fragmentation.in.Landscapes.of.Colonization.in.the.Amazon.Basin........................... 119

Andrew C. Millington and Andrew V. Bradley

Chapter 8 Urban.Land.Use.Change,.Models,.Uncertainty,.and.Policymaking.in.Rapidly.Growing.Developing.World.Cities:.Evidence.from.China......................................................................... 139

Michail Fragkias and Karen C. Seto

Part III Synthesis and Prospect

Chapter 9 Synthesis,.Comparative.Analysis,.and.Prospect............................... 163

Michael J. Hill and Richard J. Aspinall

Index....................................................................................................................... 179

42963_C000.indd 8 11/6/07 11:18:17 AM

ix

FiguresFigure.I. Conceptual.model.of.sustainability.choice.space.Figure.2.1. The.concept.of.grazing.piospheres.interacting.with.landscape.structure.Figure.2.2. Time.series.approaches.to.extracting.signal.summary.indicators.Figure.2.3. Nonbiophysical.time.series.also.provide.potential.indicators.Figure.2.4. The.spatial.pattern.of.temporal.signals.may.be.grouped.by.applying.prin-

cipal.components.analysis.Figure.2.5. Temporal.signals.are.usually.based.on.biophysical.or.human.phenomena.Figure.2.6. A.framework.for.application.of.multicriteria.analysis.of.ecosystem.services.

to.the.range.land.environment.Figure.2.7. Cognitive.mapping.interface.Figure.2.8. Schema.for.transformation.of.spatiotemporal.information.Figure.3.1. Land.use.change. in. the.Barmera,.Berri,.and.Renmark.areas.of.South.

Australia.Figure.3.2. Location.of.the.Glenelg.Hopkins.region.Figure.3.3. Land.managers’.perception.of.salinity.and.mapped.salinity.discharge.sites.Figure.3.4. Land.managers.who.manage.properties.near.areas.of.high.conservation.

value.Figure.4.1. Uganda.study.area.showing.the.distribution.of.reforestation.Figure.4.2. Conceptual.framework.for.analysis.Figure.4.3. Cumulative.deforestation.by.parishes.across.Uganda.Figure.4.4. Relationship.between.deforestation.and.distance.from.roads.Figure.5.1. Location.of.Colombia.Figure.5.2. Predicted.forest.presence.according.to.the.best.model.Figure.5.3. Predicted.deforestation.hot.spots.Figure.5.4. Forest.maps.of.the.colonization.front.Figure.5.5. Spatial.location.of.the.local.hot.spots.of.deforestation.and.regeneration.Figure.5.6. Logistic.pattern.of.forest.cover.decline.during.the.transformation.process.Figure.5.7. Forest.cover.change.at.the.local.scale.in.the.Colombian.Amazon.Figure.6.1. The.panel.process,.conducted.at.both.the.pixel.and.patch.levels.Figure.6.2. Typical.nuclear.village.settlement.as.seen.in.1:50,000.scale.Figure.6.3. LULC.in.the.greater.study.area.in.several.water.years.Figure.6.4. Stylized.LULC.trends.observed.and/or.reported.in.northeast.Thailand.Figure.6.5. Stylized.LULC.pattern.metric.change.for.the.interspersion/juxtaposition.

index.Figure.6.6. The.change.in.configuration.from.1972/1973.to.1975/1976.Figure.7.1. Thick.and. thin.understandings.of. forest. fragmentation.along. roads. in.

lowland.forests.of.the.Amazon.Basin.Figure.7.2. Sequence.of.images.showing.progressive.deforestation.Figure.7.3. Six-phase.conceptual.model.of.forest.fragmentation.Figure.7.4. Metrics.calculated.for.the.conceptual.model.Figure.7.5. Selected.metrics.

42963_C000.indd 9 11/6/07 11:18:17 AM

x Figures

Figure.7.6. Metrics.calculated.for.Communidad.Arequipa.Figure.8.1. Population.distribution.of.the.world’s.urban.agglomerations.Figure.8.2. The.Pearl.River.Delta.in.southeast.China.Figure.8.3. Predicted.probability.of.change.to.urban.areas.between.2004.and.2012.

for.Shenzhen.Figure.8.4. Predicted.probability.of.change.to.urban.areas.between.2004.and.2012.

for.Foshan.Figure.8.5. Predicted.probability.of.change.to.urban.areas.between.2004.and.2012.

for.Guangzhou.Figure.8.6. Shenzhen,.Guangzhou,.and.Foshan.urban/nonurban.prediction.for.2012.Figure.9.1. Integrating.frameworks.that.seek.interdisciplinary.definition.and.focus.

on.key.questions.of.management.of.land.use.change.in.coupled.human.environment.systems.

Figure.9.2. A.diagram.of. the.major. elements.of. the. coupled.human.environment.system.

Figure.9.3. Connecting.the.paradigms.Figure.9.4. The.context.for.spatial.and.temporal.analysis.in.determining.the.mix.of.

landscapes.of.fate.and.desire.

42963_C000.indd 10 11/6/07 11:18:18 AM

xi

PermissionsFigure.I. Potschin,.M..and.Haines-Young,.R.,.“Rio+10,”.sustainability.science.

and.landscape.ecology,.Landscape and Urban Planning,.75,.162–174,.2006..Reprinted.with.permission.from.Elsevier.

Figure.5.2. Etter,.A..et.al.,.Regional.patterns.of.agricultural.land.use.and.defor-estation.in.Colombia,.Agriculture, Ecosystems & Environment,.114,.369–386,.2006..Reprinted.with.permission.from.Elsevier.

Figure.5.3. Etter,.A..et.al.,.Regional.patterns.of.agricultural.land.use.and.defor-estation.in.Colombia,.Agriculture, Ecosystems & Environment,.114,..369–386,.2006..Reprinted.with.permission.from.Elsevier.

Figure.5.4. Etter,. A.. et. al.,. Characterizing. a. tropical. deforestation. wave:. the.Caquetá. colonization. front. in. the. Colombian. Amazon,. Global Change Biology,.12,.1409–1420,.2006..Reprinted.with.permission.from.Blackwell.Publishing.

Figure.5.5. Etter,. A.. et. al.,. Characterizing. a. tropical. deforestation. wave:. the.Caquetá. colonization. front. in. the. Colombian. Amazon,. Global Change Biology,.12,.1409–1420,.2006..Reprinted.with.permission.from.Blackwell.Publishing.

Figure.5.6. Etter,.A..et.al.,.Modeling.the.conversion.of.Colombian.lowland.eco-systems.since.1940:.drivers,.patterns.and.rates,.Journal of Environmental Management,.79,.74–87,.2006..Reprinted.with.permission.from.Elsevier.

Figure.5.7. Etter,. A.. et. al.,. Unplanned. land. clearing. of. Colombian. rainforests:.spreading.like.disease?,.Landscape and Urban Planning,.77,.240–254,.2006..Reprinted.with.permission.from.Elsevier.

Table.5.1. Etter,.A..et.al.,.Modeling.the.conversion.of.Colombian.lowland.eco-systems.since.1940:.drivers,.patterns.and.rates,.Journal of Environmental Management,.79,.74–87,.2006..Reprinted.with.permission.from.Elsevier.

Figure.9.1a. IGBP.Secretariat,.Global Land Project: Science Plan and Implementation Strategy,.GLP,.IGBP.Report.53/IHDP.Report.19,.64.pp,.2004..Reprinted.with.permission.from.the.International.Geosphere.Biosphere.Program,.Stockholm..

Figure.9.1d. Steinitz,.C.. et. al.,.Alternative Futures for Changing Landscapes: The Upper San Pedro River Basin in Arizona and Sonora,.2003..Reprinted. in. modified. form. with. permission. from. Island. Press,.Washington.D.C.

42963_C000.indd 11 11/6/07 11:18:18 AM

42963_C000.indd 12 11/6/07 11:18:18 AM

xiii

PrefaceThere.is.a.growing.international.community.of.scholars.who.work.on.themes.and.issues.that.are.central.to.understanding.land.use.change.as.a.fundamental.factor.in.the.operation.of.environmental.and.socioeconomic.systems.at.scales.from.local.to.global..This.book.presents.a.series.of.chapters.that.address.spatial.theories,.methodologies,.and.case.studies.that.support.an.integrated.approach.to.analysis.of.land.use.change..Case.studies.provide.a.series.of.regional.test.beds.for.theories.and.methodologies,.and.the.empirical.content.of.the.case.studies.allows.a.comparative.analysis.of.land.use.change.issues.from.diverse.places..Case.studies.thus.are.an.important.mecha-nism,.not.only. for.understanding. the.multiscale.nature.and.consequences.of. land.use.change.and.the.particular.history.of.changes.in.case.study.locations,.but.also.for..eliciting.general.principles. and. factors.of. importance. in.different. socioeconomic,.cultural,.and.environmental.contexts..This.generalization.from.case.studies.provides.input.to.decision.making.related.to.possible.future.trajectories.of.change.

The.chapters. in. this.book.were.written. to.present. this. interaction.of. theories,.methodologies,.and.case.studies..Additionally,.all. the.authors.are.concerned.with.links. between. science. and. decision. making,. especially. in. relation. to. policy. and.practice..Each.author.attempts.to.enable.effective.communication.between.the.aca-demic.content.of.his.or.her.work.with.decision.makers,.including.those.concerned.with.policy.and.those.concerned.with.land.management.

All.the.chapters.were.first.presented.in.a.paper.session.at.the.6th.Open.Meeting.of.the.International.Human.Dimensions.of.Global.Environmental.Change.Programme.in.Bonn,.October.2005—Spatial.Theory. and.Methodologies. for. Integrated.Socio-economic.and.Biophysical.Analysis.and.Modeling.of.Land.Use.Change:.An.Inter-national.Test.of.Theory.and.Method.and.a.Comparative.Synthesis.of.Change.at.Local.and.Regional.Scales.

42963_C000.indd 13 11/6/07 11:18:19 AM

42963_C000.indd 14 11/6/07 11:18:19 AM

xv

AcknowledgmentsWe.thank.the.authors.of.the.chapters.for.their.prompt.responses.to.all.our.requests..We.also.thank.Professor.Roy.Haines-Young.and.Dr..Marion.Potschin,.University.of. Nottingham,. for. their. participation. in. the. session. at. the. International. Human..Dimensions.of.Global.Environmental.Change.Programme.open.meeting.in.Bonn,.Germany,. in.October.2005.at.which. the.papers.and. ideas. in. this.book.were.first..presented.and.discussed..We.also.thank.Tai.Soda.and.Amber.Donley.from.Taylor.&. Francis/CRC. Press. for. their. patient. prompting. and. management. during. the..preparation.of.this.book.

42963_C000.indd 15 11/6/07 11:18:19 AM

42963_C000.indd 16 11/6/07 11:18:19 AM

xvii

IntroductionThis.book.addresses.spatial.theories.and.methodologies.that.support.an.integrated.approach.to.analysis.of.land.use.change..The.work.focuses.on.spatial.representation.and.modeling. for. scientific.study.and.development.of.management.understanding.of. complex,.dynamic. land.use. systems..Case. studies. are.used. to.develop. specific.examples,.not.only.of.change.in.the.study.areas.used.by.the.different.case.studies,.but.also.to.illustrate.the.variety.and.commonality.of.data.sources,.methods,.and.issues.faced.when.studying.and.developing.understanding.of.land.use.change.

Land.use.and.land.cover.change.reflect.a.variety.of.environmental.and.social..factors.1,2.This.necessitates.that.an.equally.varied.suite.of.data.be.used.for.effective.analysis..Remote. sensing,. from.both. satellites. and.air. photos,. provides. a. central.resource.for.study.of.land.use.and.land.cover.change..Socioeconomic.surveys.and.censuses. provide. an. equally. important. source. of. data. on. social. and. economic.systems..Atlases.and.other.map.sources.can.provide.data.on.specific.environmental.and. socioeconomic. characteristics. of. an. area.. Similarly,. household. and. other..surveys.can.give.information.on.motivations,.values,.behaviors,.and.actions.of.deci-sion.makers.and.land.managers,.and.there.are.many.other.specific.data.of.relevance.for.study.of.land.use.change..These.different.data.do.vary,.however,.in.their.avail-ability,.currency,.and.relevance.for.study.of.land.change,.and.this.is.reflected.in.the.variety.of.case.studies.in.the.published.literature.and.the.specific.issues.and.ques-tions.they.address.

Similarly,.methods.that.are.appropriate.for.analysis.of.the.different.data.sources.are.varied,.and.study.of. land.change. typically.may. involve.methods. from.remote.sensing,.GIS.(geographic.information.systems),.process.and.empirical.modeling,.and.spatial.and.statistical.analysis,.among.others,.as.well.as.methods.that.can.address.qualitative.and.quantitative.data..The.need.to.study.land.use.as.a.coupled.natural.and.human.system.adds.to.the.complexity.of.methods.needed.since.interaction.of.systems.and.integration.in.analysis.present.major.methodological.challenges.

Thus,.the.purpose.of.this.book.is.to.illustrate.issues.and.opportunities.for.study.of.land.change.by.discussing.relevant.science.and.methods.and.by.providing.a.series.of. exemplary. case. studies. that. present. approaches,. data,. and. methods. to. study.land.use.change..Various.themes.run.through.the.book,.including.the.wide.use.of.remote.sensing,.GIS.(often.implicit.and.supporting.analysis.rather.than.explicit.and.a.goal.of.the.work),.and.a.variety.of.statistical.and.other.modeling.methods.aimed.at.understanding.spatial.and.temporal.trends,.patterns,.and.dynamics..The.central.importance.of.understanding.the.social,.economic,.policy,.cultural,.and.institutional.contexts.and.influences.on.land.change.is.also.a.recurrent.theme..

A.further.set.of.issues,.and.a.major.context.for.interpretation.of.results.of.the.chapters,.is.associated.with.sustainable.development.and.the.emergence.of.sustain-ability.science.3,4,5.Sustainable.development.provides.a.context.and.focus.for.scientific.research4,6,7.as.well.as.a.guide.and.impetus.for.civic.debate.and.decision-making.5,8.Potschin.and.Haines-Young7.present.a.conceptual.model.(the.“tongue.model”).of.the.

42963_C000.indd 17 11/6/07 11:18:20 AM

xviii Introduction

state.of.a.landscape.and.its.development.trajectory.over.time.(Figure.I)..This.treats.landscapes.as.multifunctional.spaces.and.is.founded.on.a.paradigm.of.natural.capital.and.ecosystem.services.9,10.The. tongue.model.places.boundaries.on. sustainability.defined.by.a.combination.of.biophysical.limits.of.ecological.systems.in.a.landscape;.this.reflects.not.only.environmental.conditions.and.ecosystem.services.that.can.be.achieved,.but.also.social.and.cultural.values,.as.well.as.costs.and.uncertainties.that.can.be.accepted..The.tongue.model7.provides.a.useful.framework.for.development.and.evaluation.of. the.models.and.case.studies.of. land.use.and. land.cover.change.presented.in.Chapters.3.through.8..The.case.studies.also.provide.a.test.of.the.gen-eral.applicability.of.the.tongue.model.since.they.explore.a.diverse.range.of.socio-economic,.cultural,.and.biophysical.contexts.for.land.use.systems..Effective.linkage.of.science,.policy,.land.management.practices,.and.decision-making.related.to.land.and. ecological. systems. will. increasingly. require. frameworks. such. as. this. with. a.focus. on. integration. of. human. and. natural. systems. that. are. explicitly. directed. at.achieving.consensus.based.on.a.strong.scientific.foundation.

The.case.studies.also.focus.attention.on.terminology,.especially.related.to.land.use. and. land. cover..Land.use. refers. to. the. social,. economic,. cultural,. political,. or.other. value. and. function. of. land. resources.. This. contrasts. with. land. cover,. which.refers.to.the.biophysical.properties.of.the.land.surface.11.As.such,.land.use.and.land.cover,.although.related,.are.distinct.from.one.another..The.distinctions.are.important.for. understanding. causes. and. consequences.of. change. in. land. systems,. as.well. as.for.mapping.and.other.measurement.of. land.system.properties.. Indeed,.one.of. the.main.challenges.to.the.study.of.land.systems.may.be.the.tendency.for.conflation.and.confusion.of. the.terms. land use.and. land cover..This.partly.relates. to.wide.use.of.remotely.sensed.imagery.as.a.source.of.land.system.data.but.also.to.ambiguity.and.lack.of.distinction.in.definition.of.categories.of.interest..For.example,.the.Anderson.

Figure i Conceptual.model.of.sustainability.choice.space..(From.Potschin.and.Haines-Young.7.With.permission.)

Development options defined bydifferent interest groups

State indicator (e.g., area of a particular habitat)

Set of landscape configurations that more or less sustainrequired landscape outputs for society. Boundary ofenvelope set by minimum and maximum value of statevariable required to deliver outputs, given people’sattitudes to risks and costs, and long-term uncertainties.

+

–

Z1 – Z0

time

42963_C000.indd 18 11/6/07 11:18:22 AM

Introduction xix

classification,12.which.is.widely.used.as.a.taxonomy.for.land.surface.description.based.on.imagery.from.different.sources.(as.well.as.field.mapping),.and.many.other.clas-sifications.confound.cover.and.use.classes.within.the.same.formal.mapping.legend..This.may.be.based.on.apparent.similarity.in.cover.and.use..For.example,.forestry.(use).and.forest.(cover).may.refer.to.the.same.place.on.the.ground..It.does,.however,.have.important. consequences. for. process-level. understanding. and. modeling. of. change,.presenting. a. limitation. for. recording. change. and. consequently. for. understanding..Recently,.efforts.have.been.made.to.produce.separate.classifications.of.land.use13.and.land.cover,14.and.the.Global.Land.Project15.identifies.cover.and.use.along.a.continuum.spanning.natural.and.social.systems.

STruCTure OF THe BOOK

The.book.is.organized.in.nine.chapters..Chapters.1.and.2.address.theoretical.and.methodological. issues. and. mechanisms. for. study. of. land. use. systems. as. coupled.human.and.natural.systems..Chapters.3.through.8.provide.a.suite.of.regional.case.studies,.including.a.discussion.of.change.in.rapidly.urbanizing.areas..Collectively,.these.six.chapters.provide.insight.into.the.nature.of.both.land.use.change.and.the.diverse.range.of.socioeconomic,.cultural,.and.biophysical.contexts.of.land.use.sys-tems.across.the.planet..The.case.study.chapters.provide.a.series.of.illustrations.for.many.of.the.frameworks,.issues,.and.methodologies.described.in.the.first.two.chap-ters..The.empirical.content.of.the.case.studies.also.allows.a.comparative.analysis.of.land.use.change.issues.from.diverse.places..

Chapter. 1. considers. basic. science. questions. that. underpin. study. of. land. use.change,. as. well. as. a. suite. of. applied. science. issues. that. influence. the. utility. and.use. of. scientific. information. about. land. use. change. for. policy. and. land. manage-ment..The.chapter.focuses.on.issues.that.influence.the.analysis.of.dynamics.of.land.change;.the.need.for.understanding.the.integration.and.feedbacks.between.different.components.of.land,.climate,.socioeconomic,.and.environmental.systems;.resilience,..vulnerability,.and.adaptability.of.land.use.systems;.scale.issues;.and.accuracy..The.chapter.also.examines.some.issues.relevant.to.effective.communication.and.partner-ship. between. scientists. and. practitioners. responsible. for. decision-making,. either.policy.or.management,.about.land.use.

Full. understanding.of.both. land.use. change.and.decision-making. requires. an.appreciation.of.multiple.characteristics.of. land.use.systems.and.factors. that. influ-ence.change..Multicriteria.decision.analysis.can.provide.a.series.of.tools.to.help.with.both.scientific.study.and.decision-making..Chapter.2.examines.the.development.of.spatially.dependent.procedures.and.models.for.such.multicriteria.decision.analysis..The.approaches.focus.on.reducing. the.complex.responses.and.patterns.associated.with.land.use.change.to.simpler.meaningful.metrics..These.are.based.on.use.of.both.quantitative.and.qualitative.measures.of.patterns.and. trends.and. include.methods.from.signal.processing,.time.series.analysis,.and.analysis.of.spatial.patterns..

The.next.six.chapters.present.a.series.of.case.studies.that.are.exemplars.of.dif-ferent. aspects. of. land. use. science.. In. Chapter. 3. Byron. and. Lesslie. explore. inter-actions.between.people.and.the.environment.as.a.critical.component.in.regional-.and..catchment-scale.natural.resource.management..The.chapter.uses.a.case.study.from.

42963_C000.indd 19 11/6/07 11:18:22 AM

xx Introduction

Australia.to.bring.data.on.the.environment,.from.a.time.series.of.orthophotos,.and.social.systems,.from.a.property.survey,.together..This.integrated.analysis.is.complemented.by.use.of.spatially.referenced.survey.data.to.understand.how.land.managers’.values,.perceptions,.and.practices.relate.to.the.biophysical.environments.they.manage.

Chapter.4,.by.Babigumira,.Müller,.and.Angelsen,.links.deforestation.in.western.Uganda.in.the.1990s.to.the.socioeconomic,.spatial,.and.institutional.contexts.within.which. it.occurred..The.authors.develop.an.empirical.model. that. integrates. socio-economic.data.from.a.national.census.with.spatial.data.derived.from.remote.sensing..The.socioeconomic.survey.informs.on.poverty.and.economic.opportunities,.whereas.the.remotely.sensed.data.represent.the.costs.and.feasibility.of.forests.

Chapter.5,.by.Etter.and.McAlpine,.examines.the.patterns,.processes,.and.drivers.of.unplanned.land.cover.change.in.Colombia.as.representative.of.change.in.the.tropics..Statistical.modeling.is.used.to.predict.changes.in.forest.cover.at.local,.regional,.and.national.levels,.over.times.ranging.from.a.decade.to.a.century..Explanatory.variables.include. both. biophysical. and. socioeconomic. data,. and. these. are. obtained. from. a.range.of.sources,.including.remote.sensing,.maps,.and.surveys..

Chapter.6,.by.Crews-Meyer,.is.a.case.study.of.land.change.in.northeast.Thailand.that.uses.a.time.series.of.satellite-derived.data.within.a.longitudinal.approach,.panel.analysis,.for.modeling.temporal.dynamics..The.approach.also.draws.on.landscape.ecology. to. emphasize. the. importance. of. the. spatial. scale. of. observation. on. the.inference.of.process. and.attempts. to. examine. changes. in. landscape. composition.and.configuration..

Chapter.7,.by.Millington.and.Bradley,.uses.a.case.study.of.deforestation.associ-ated.with.planned.colonization.schemes.in.the.Amazon.Basin.to.develop.a.detailed—thick—understanding.of.deforestation..They.argue.that.the.impacts.of.roads.on.forest.fragmentation.are.agents.of.deforestation.at.one.scale.only.and.that.at.another.scale.the.pattern.of.property.ownership.represents.that.scale.at.which.land.owners.make.decisions.about.forest.clearance.and.regrowth.as.household.responses.to.economic.and.policy.signals.

Chapter. 8,. by. Fragkias. and. Seto,. discusses. issues. of. urban. land. use. change.modeling.and.explores.the.intersection.of.land.use.modeling.with.urban.policy-making.at.different.scales..The.work.also.concentrates.on.the.effects.of.uncertainties.in.data.sources.and.reviews.a.predictive.model.of.rapid.urban.transformation.that.extends.a.standard.modeling.approach.to.provide.a.policy-making.framework.that.explicitly.reduces.uncertainty..Chinese.cities.are.used.as.a.case.study.as.important.exemplars.of.developing.world.cities.

Chapter.9.reviews.findings.of.the.full.set.of.chapters.and.case.studies.and.points.to.the.potential.developments.in.spatial.methods.needed.to.advance.integrated.model-ing.of. land.use. change. and. improve. linkage.between. scientific. study.of. land.use.change. with. policy. and. management. decision-making. at. regional,. national,. and.international.scales.

Richard J. Aspinall Michal J. Hill

42963_C000.indd 20 11/6/07 11:18:23 AM

Introduction xxi

reFereNCeS

. 1.. Geist,.H.,.and.Lambin,.E..F..Proximate.causes.and.underlying.driving.forces.of.tropical.deforestation..Bioscience.52(2),.143–150,.2002.

. 2.. Geist,.H..J.,.and.Lambin,.E..F..Dynamic.causal.patterns.of.desertification..Bioscience.54(9),.817–829,.2004.

. 3.. Kates,.R..W.,.Sustainability Science. Research and Assessment Systems for Sustainability Program. Discussion.Paper.2000-33..Belfer.Center.for.Science.and.International.Affairs,.Kennedy.School.of.Government,.Harvard.University,.Cambridge.MA,.2000..Available.at.http://ksgnotes1.harvard.edu/BCSIA/sust.nsf/pubs/pub7/$File/2000-33.pdf.

. 4.. Kates,.R..W..et.al..Sustainability.science..Science.292,.641–642,.2001.

. 5.. Clark,.W..C.,.and.Dickson,.N..M..Sustainability.science:.The.emerging.research.program..Proceedings of the National Academy of Sciences.100(14),.8059–8061,.2003..

. 6.. Haines-Young,.R..Sustainable.development.and.sustainable.landscapes:.defining.a.new.paradigm.for.landscape.ecology..Fennia.178(1),.7–14,.2000.

. 7.. Potschin,. M.,. and. Haines-Young,. R.. “Rio+10,”. sustainability. science. and. landscape.Ecology,.Landscape and Urban Planning.75,.162–174,.2006.

. 8.. Cash,.D..W..et.al..Knowledge.systems.for.sustainable.development..Proceedings of the National Academy of Sciences.100(14),.8086–8091,.2003.

. 9.. Costanza,.R.,.and.Daly,.H..E..Natural.capital.and.sustainable.development..Conservation Biology.6(1),.37–46,.1992.

. 10.. Daily,.G..C..Nature’s Services: Societal Dependence on Natural Ecosystems..Washington,.DC:.Island.Press,.392.pp..1997.

. 11.. Comber,.A..J.,.Fisher,.P..F.,.and.Wadsworth,.R..A..What.is.land.cover?.Environment and Planning B: Planning and Design.32,.199–209,.2005.

. 12.. Anderson,.J..R..et.al..A LandUse and Land Cover Classification System for Use with Remote Sensor Data..Geological.Survey.Professional.Paper.No..964..Washington,.DC:.U.S..Government.Printing.Office,.1976.

. 13.. Jansen,.L..J..M..Harmonization.of. land.use.class.sets. to. facilitate.compatibility.and.comparability. of. data. across. space. and. time.. Journal of Land Use Science. 1(2–4),.127–156,.2006.

. 14.. Herold,.M..et.al..Evolving.standards.in.land.cover.characterization..Journal of Land Use Science.1(2–4),.157–168,.2006.

. 15.. GLP Science Plan and Implementation Strategy.. IGBP.Report.No.53,. IHDP.Report.No.19..Stockholm,.IGBP.Secretariat..64.pp..2005.

42963_C000.indd 21 11/6/07 11:18:24 AM

42963_C000.indd 22 11/6/07 11:18:24 AM

xxiii

EditorsRichard J. Aspinall. received. a. BSc. (Hons). from. the. University. of. Birmingham.and.a.PhD.from.the.University.of.Hull,.both. in.geography..He.has.worked.in. the.United.Kingdom.including.10.years.at.the.Macaulay.Land.Use.Research.Institute.in..Scotland,.and.in.the.United.States.in.the.Department.of.Earth.Sciences.at.Montana.State.University,.as.chair.of.the.School.of.Geographical.Sciences.at.Arizona.State.University,. and. as. program. director. for. geography. and. regional. science. at. the.U.S..National.Science.Foundation..He.is.now.professor.and.chief.executive.at. the.Macaulay. Institute. in. Scotland,. an. interdisciplinary. research. institute. addressing.sustainable.development.and.land.use..His.research.interests.are.in.the.areas.of.land.use.and.land.cover.change,.analysis,.and.modeling.of.coupled.natural.and.human.systems,.and.methods.and.applications.in.environmental.geography.including.GIS,.landscape.ecology,.biogeography,.geomorphology,.and.hydrology.

Michael J. Hill. received. the. BAgrSci. and. MAgrSci. from. LaTrobe. University,..Bundoora,.Australia,.and.a.PhD.from.the.University.of.Sydney,.Australia,.in.1985..He.spent.12.years.in.the.CSIRO.Division.of.Animal.Production.and.then.6.years.in. the.Bureau.of.Rural.Sciences. in. the.Department. of.Agriculture,.Fisheries. and.Forestry.of. the.Australian.government.where.he.carried.out. research. in.and.con-tributed. to. the.management.of. the.Co-operative.Research.Centre. for.Greenhouse.Accounting..In.2006,.he.became.professor.of.earth.systems.science.in.the.Depart-ment.of.Earth.Systems.Science.and.Policy.at.the.University.of.North.Dakota,.Grand.Forks..He.has.a.background.in.grassland.agronomy,.but.has.been.working.with.spa-tial. information. and. remote. sensing. of. land. systems. for. the. past. 12. to. 15. years..He.has.published.widely.on.agronomy,.ecology,.biogeography,.production.of.grass-lands,.and.radar,.multispectral,.and.hyperspectral.remote.sensing.of.grasslands,.and.more.recently.has.been.involved.in.the.development.of.scenario.analysis.models.for.assessment.of.carbon.dynamics.in.agricultural.and.rangeland.systems..His.current.interests. are. in. the. use. of. MODIS. land. product. data. in. model-data. assimilation,.application.of.quantitative.information.from.hyperspectral.and.multiangle.imaging.to.vegetation.description,.multicriteria.and.decision.frameworks.for.coupled.human.environment.systems,.and.methods.and.approaches.to.application.of.spatial.data.for.land.use.management.

42963_C000.indd 23 11/6/07 11:18:25 AM

42963_C000.indd 24 11/6/07 11:18:25 AM

xxv

Contributors

Arild AngelsenArild. Angelsen. is. an. associate. professor. of. economics. in. the. Department. of.Economics.and.Resource.Management.at.the.Norwegian.University.of.Life.Sciences..His. research. interests. are. in. economic. analysis. and. assessment. of. projects. and..policies. in. developing. countries,. particularly. within. agriculture. and. forestry,. and.related.to.poverty,.environmental.effects,.and.use.of.natural.resources.

Richard J. AspinallRichard. J..Aspinall. is.professor.and.chief. executive.at. the.Macaulay. Institute,. an.interdisciplinary.research.institute.addressing.sustainable.development.and.land.use..His.research.interests.are.in.the.areas.of.land.use.and.land.cover.change,.analysis,.and.modeling.of.coupled.natural.and.human.systems,.and.methods.and.applications.in.environmental.geography.including.GIS,.landscape.ecology,.biogeography,.geo-morphology,.and.hydrology.

Ronnie Babigumira.Ronnie.Babigumira.is.a.PhD.student.in.the.Department.of.Economics.and.Resource..Management.at.the.Norwegian.University.of.Life.Sciences..His.research.interest.is.in.land.use.change.in.Africa.

Andrew V. BradleyAndrew. V.. Bradley. is. a. research. scientist. at. the. Natural. Environmental. Research..Council.(NERC).Centre.for.Ecology.and.Hydrology.(CEH).at.Monks.Wood,.United..Kingdom..His.research.interests.focus.on.understanding.socioeconomic.drivers.of.forest.loss,.forest.fragmentation,.and.agricultural.land.use.and.land.cover.change.

Ian ByronIan.Byron’s.research.interests.include.analysis.of.sociological.survey.data.and.integra-tion.of.survey.results.with.biophysical.spatial.data..He.co-authored.the.chapter.while.a.social.scientist.with.the.Bureau.of.Rural.Sciences,.a.science-policy.agency.within.the.Australian.Department.of.Agriculture,.Fisheries.and.Forestry.

Kelley A. CrewsKelley. A.. Crews. is. associate. professor. of. geography. and. the. environment. at. the..University. of. Texas,. Austin,. where. she. also. directs. the. GIScience. Center.. Her..thematic.research. interests. include.remote.sensing,.population–environment. inter-actions,. landscape.ecology,.and.policy.analysis..Geographically.her.work. focuses.on.tropical.and.subtropical.forest/savanna/wetland.ecotones.in.the.western.Amazon,.the.Okavango.Delta.of.Botswana,.and.Northeast.Thailand.

42963_C000.indd 25 11/6/07 11:18:25 AM

xxvi Contributors

Andres EtterAndres.Etter.is.a.professor.in.the.Department.of.Ecology.and.Territory.of.the.School.of.Environmental.and.Rural.Studies.at.Javeriana.University.(Bogotá-Colombia)..He.has.20.years’.experience.in.landscape.ecological.mapping.and.research.in.the.Colombian.Amazon.forests,.Orinoco.savannas,.and.Andean.Montane.forest.regions.integrating.biophysical,.socioeconomic,.and.historic.data,.using.remote.sensing.and.GIS..His.cur-rent.research.deals.with.the.modeling.and.understanding.of.land.use.change.processes.aimed.at.more.informed.and.dynamic.conservation.planning.processes.

Michail FragkiasMichail. Fragkias. is. the. executive. officer. of. the. IHDP. Urbanization. and. Global..Environmental.Change.Core.Project.hosted.by.the.Global.Institute.of.Sustainability.at.Arizona.State.University..His.interests.focus.on.urban.land.use.change.and.the.interaction. of. urban. spatial. structure. with. the. environment.. He. employs. (spatial)..statistical. analysis,. simulations,. and. geographical. information. systems. (GIS). to.study.the.significance.of.social,.economic,.and.political.drivers.of.urban.land.use.change. in. China.. He. completed. his. undergraduate. studies. in. economics. at. the.National.University.of.Athens.in.Greece.and.his.MA.and.PhD.in.economics.at.Clark..University.in.Massachusetts..He.co-authored.the.present.chapter.while.being.a.post-doctoral.scholar.at.the.Center.for.Environmental.Science.and.Policy.(CESP).at.the.Freeman.Spogli.Institute.for.International.Studies.(FSI).at.Stanford.University.

Michael J. HillMichael.J..Hill.is.a.professor.of.earth.systems.science.in.the.Department.of.Earth.Systems.Science.and.Policy.at.the.University.of.North.Dakota..His.research.inter-ests.include.hyperspectral.remote.sensing,.biogeochemical.processes,.and.land.use.change. in. savanna. systems,. and. analysis. of. coupled.human.environment. systems.using.spatial.multicriteria.analysis.and.spatial.analysis.methods.

Robert LesslieRob. Lesslie. is. a. principal. research. scientist. in. the. Bureau. of. Rural. Sciences,. a..science.policy.bureau.with.the.Australian.government’s.Department.of.Agriculture,.Fisheries.and.Forestry..He.is.an.ecologist.by.training.and.retains.a.keen.interest.in.landscape.ecology..He.currently.manages.the.national-.and.catchment-scale.land.use.mapping.project.for.Australia,.which.includes.work.on.land.management.practices..He. has. been. developing. multicriteria. software. and. has. a. specific. interest. in. arid.lands.and.rangelands.

Clive McAlpineClive. McAlpine. is. a. senior. research. fellow. with. the. Centre. for. Remote. Sensing.and.Spatial.Analysis.with.the.School.of.Geography,.Planning.and.Architecture,.the..University.of.Queensland,.Brisbane..His.research.interests.are.in.landscape.ecology,.biodiversity.conservation,.land.cover.change.modeling,.and.the.climate.impacts.of.land.cover.change.

42963_C000.indd 26 11/6/07 11:18:26 AM

Contributors xxvii

Andrew C. MillingtonAndrew.C..Millington.is.a.professor.in.the.Department.of.Geography.at.Texas.A&M.University..He.has.previously.worked.in.three.other.geography.departments:.he.was.formerly.professor.and.departmental.chair.at.the.University.of.Leicester,.England;.reader.in.geography.at.the.University.of.Reading,.and.lecturer.at.Fourah.Bay.College.(part.of.the.University.of.Sierra.Leone)..His.research.focuses.on.the.impacts.of.human.and.institutional.agents.on.spatiotemporal.patterns.of.land.use.and.land.cover.change.and.the.impacts.of.land.use.and.land.cover.change.on.biological.phenomena..He.uses.hybrid.methodologies.from.GIScience,.environmental.science,.ecology,.and.social.science. to.research. these.phenomena..His.current.research. includes.analyzing.the.effects.of.policy.initiatives.on.land.use.and.land.cover.change.in.Argentina,.Bolivia,.and.Peru,.and.he.is.initiating.work.on.land.use.and.land.cover.change.in.forest.lands.in.Texas..He.received.his.BSc.in.geography.and.geology.from.Hull.University,.his.MA.in.geography.from.the.University.of.Colorado.at.Boulder,.and.his.DPhil.in.geog-raphy.from.Sussex.University.

Daniel MüllerDaniel. Müller. is. an. agricultural. economist. now. working. in. the. Department. of..Economic. and. Technological. Change,. Center. for. Development. Research. at. the..University.of.Bonn..He.contributed.to.this.book.during.a.postdoctoral.appointment.at.Humboldt.University..His.research.interests.are.in.development.and.natural.resource.economics,.patterns,.and.process. in. land.cover.and. land.use.change,.econometric.analysis,.and.spatial.data.analysis.

Karen C. SetoKaren.C..Seto.is.an.assistant.professor.in.the.Department.of.Geological.and.Environ-mental.Sciences.and.a.fellow.at.the.Woods.Institute.for.the.Environment.at.Stanford.University.. Her. research. focuses. on. monitoring. and. forecasting. land. use. change,.especially.urban.growth.in.Asia..Her.current.research.efforts.include.analyzing.the.effects. of. policy. reforms. on. urban. growth. in. China,. India,. and. Vietnam.. She. is.the.Remote.Sensing.Thematic.Leader.for.the.World.Conservation.Union’s.(IUCN)..Commission. on. Ecosystem. Management. and. is. a. recipient. of. the. NASA. New.Investigator. Program. in. Earth. Science. Award. and. an. NSF. Faculty. Early. Career.Development.(CAREER).Award..She.received.her.BA.in.political.science.from.the.University.of.California.at.Santa.Barbara,.her.MA.in.international.relations,.resource.and.environmental.management.from.Boston.University,.and.her.PhD.in.geography.from.Boston.University.

42963_C000.indd 27 11/6/07 11:18:27 AM

42963_C000.indd 28 11/6/07 11:18:27 AM

Part I

Theory and Methodology

42963_S001.indd 1 8/17/07 2:29:35 PM

42963_S001.indd 2 8/17/07 2:29:35 PM

�

1 Basic and Applied Land Use Science

Richard J. Aspinall

Contents

1.1 Introduction.....................................................................................................31.1.1 TheoreticalFoundations......................................................................4

1.2 BasicScience...................................................................................................51.2.1 DynamicsofChangeinSpaceandTime.............................................51.2.2 IntegrationandFeedbacksbetweenLandscape,Climate,

Socioeconomic,andEcologicalSystems.............................................71.2.3 Resilience,Vulnerability,andAdaptabilityofLandSystemsas

CoupledNaturalandHumanSystems.................................................81.2.4 ScaleIssues..........................................................................................81.2.5 Uncertainty..........................................................................................8

1.3 AppliedScience..............................................................................................91.3.1 AddressingEvolvingPublicandPrivateLandManagement

IssuesandDecisions.......................................................................... 101.3.2 InterpretationandCommunicationofScientificKnowledgefor

AdaptiveManagementofChangeinLandUseSystems................... 111.3.3 HumanandEnvironmentalResponsestoChange............................. 11

References................................................................................................................ 11

1.1 IntroduCtIon

Landusesciencecanbedefinedasaninclusive,interdisciplinarysubjectthatfocusesonmaterialrelatedtothenatureoflanduseandlandcover,theirchangesoverspaceand time, and the social, economic, cultural, political, decision-making, environ-mental,andecologicalprocessesthatproducethesepatternsandchanges.1Avarietyoftheories,methodologies,andtechnologiesunderpinresearchonlandusescience,and,consequently,anumberofbasicandappliedsciencethemesthatarecharac-teristic of land use research can be identified. These reflect the interdisciplinaryand integrated analysis required to comprehend landuse, aswell as the role andimportanceoflanduse,landusechange,andlandmanagementandpolicy,andtheimportanceoflanduseforsustainability.2LanduseisalsoconsideredacentralpartofthefunctioningoftheEarthsystem3aswellasreflectinghumaninteractionswiththeenvironmentatscalesfromlocaltoglobal.

42963_C001.indd 3 10/29/07 10:47:46 AM

� Land Use Change: Science, Policy and Management

Basicsciencequestionsinlandusescienceincludethosethatfocuson(a)dynamicsofchangeinspaceandtime;(b)integrationandfeedbacksbetweenlandscape,climate,socioeconomic,andecologicalsystems(c)resilience,vulnerability,andadaptabilityof coupled natural and human systems (d) scale issues and (e) accuracy. Appliedscienceaddressespolicyandmanagementquestionsinlandusescienceincluding(a)addressingevolvingpublicandprivate landmanagement issuesanddecisions;(b)interpretationandcommunicationofscientificknowledgeforadaptivemanage-mentofchangeinlandusesystems;and(c)humanandenvironmentalresponsestochange.Theappliedissuesalsoshouldbesetagainstaneedforexplicitmanagementofuncertainties.Thiswillincludedefinitionofthelimitsofapplicabilityofchangeprojectionsandotheranalyses,particularlyastranslatedintodecisionsupportandparticipatoryapproaches.Theneedandroleforspatiallyintegrateddynamicmodelsofcouplednaturalandhumansystemsinthecontextsofstudyandmanagementoflandusechangeunderpinthisdiscussion.

1.1.1 TheoreTical FoundaTions

Therehasbeensomediscussionofthepotentialandneedforanintegrated,orover-arching,theoryforlandchange.Lambinandcolleagues4notethreerequirementsforanoverarchingtheory:(a)toengagethebehaviorofpeopleandsocietyandreciprocalinteractionwith landuse, (b) tobemultilevelwith respect tobothpeopleand theenvironment,and(c)tobemultitemporalinordertoincludeboththecurrentandpastcontextsinwhichland,people,andenvironmentinteract.Integratedstudyoflanduseandlandcoverchangestypicallyisinterdisciplinaryormultidisciplinaryinapproachand thus involves theories from multiple participating disciplines.4 The practicalneedsofinterdisciplinaryresearchhaveledempiricalcasestudiestouseavarietyofmechanismsforencouragingdialoguebetweendisciplines,includingarangeofinte-gratingframeworks,mostbasedonsomeformofsystemsrepresentation.5,6Empiricalstudiesalsorecognizesomequalitiesoflandsystemsthatarecommonacrosscasestudies,andthesesuggestcharacteristicsthatatheoryoflandchangeneedstobeabletoincorporate:

(a) Complexcauses,processes,andimpactsofchange7

(b) Differencesandinter-relationshipsbetweenlanduseandlandcover8,9,10

(c) Interactionofsocioeconomicandbiophysicalprocesses11,12,13,14

(d) Multiplespatialandtemporalscalesatwhichprocessesoperate11,15,16

(e) Interactionacrossmultipleorganizationallevels17

(f) Feedbacksandconnectionsinbothsocialandgeographicalspaces (g) Multiplelinksbetweenpeopleandland7

(h) Influenceofsocial,historical,andgeographicalcontextonlandusechange18

(i) Importance of individual, social, demographic, economic, political, andculturalfactorsindecisionmaking19,20

(j) Combineduseofqualitativeandquantitativedataandmethods21,22

42963_C001.indd 4 10/29/07 10:47:47 AM

Basic and Applied Land Use Science �

Reviewofmultiplecasestudiesinmeta-analyseshasalsoprovidedinsightintolandchange,providinggeneralizationaboutfactorsthatleadtochange.Examining152publishedcasestudiesoftropicaldeforestation19and132casestudiesofdeserti-fication20 Geist and Lambin identified a relatively small set of underlying causescommon to the landusechangesobserved indifferent regionsandplaces.Theseunderlyingcausesaredescribedasproximate,havingapparent immediate impactonchange,andultimate,whichrepresentfundamentalcausesofchange.23Therearefivebroadgroupsofunderlyingfactorscommontobothsetsofcasestudies:demo-graphic,economic,technological,policyandinstitutional,andcultural;desertifica-tion also included climatic factors. Proximate causes of change common to bothtropicaldeforestationanddesertificationincludedinfrastructureextension,agricul-tural activities and expansion, and wood extraction; increased aridity also was aproximate cause for desertification. A meta-analysis of 91 published case studiesofagriculturallandintensificationinthetropics,24intendedasacompaniontothemeta-analysesoftropicaldeforestationanddesertification,usedthesamefactorsasGeistandLambin’sstudiesandrecordedaverydetailedandvariedlistofprocessesassociatedwithagriculturalintensification.Themainfactorsidentifiedweredemo-graphic,market,andinstitutional,particularlypropertyregimes.Themostcommonprocesses of agricultural intensification in the tropics included adoption of newcrops,plantingoftrees,anddevelopmentofhorticulture.24

Theseconcerns fordevelopment anduseof theory, and for improvingunder-standing of social and natural processes, as well as their interaction, in study oflanduse,provideaguideforcasestudiesandattemptsatintegrationandsynthesisacrosscasestudies.IntheremainderofthischapterIdiscusssomebasicandappliedscienceissuesthatmayhelpnotonlytheprocessofstudyinglanduseandchange,butalsothecommunicationandinvolvementofawidevarietyofinterestedparties,includingdecisionmakersandlandmanagers,inboththeconductofresearchanditsimplicationsforlandmanagementandpolicy.

1.2 BasIC sCIenCe

1.2.1 dynamics oF change in space and Time

Landuseandlandcoverchangesareinherentlyspatialanddynamic.Themagnitudeandimpactofchangesinlanduseandlandcoveraresuchthatlandusechangeisrecognizedasachangethatisglobalinextentandimpact.25,26,27Landusedynamicsisalsorecognizedasoneofthegrandchallengesinenvironmentalscience.28

Avarietyoflanduseandlandcoverchangeprojectshavemeasuredandmonitoredchangeinlanduseandlandcover,particularlyforperiodsoverthepast30years,usingsatelliteimagery,29,30,31butalsoforperiodsuptothepastseveralhundredyears.32,33Theseindividualprojectshavenotonlyidentifiedcommonlanduseandlandcovertransitions but also raised awareness and interest in the processes that producechange.Forexample,theInternationalHumanDimensionsProgrammeonGlobalEnvironmental Change/International Geosphere-Biosphere Programme Land-UseandLand-CoverChange(IHDP/IGBPLUCC)program,6aninternationalprogram

42963_C001.indd 5 10/29/07 10:47:47 AM


running from 1995 to 2005, coordinated a broad network of local, regional, andcontinentalscaleprojectsexamininglandcoverandlandusechange.14Thetaskofestablishingcauseandeffect inrelation to landusechangedynamicsfacesmanyofthechallengesofotherempiricalfield-basedsciencesthatrelyonobservationofEarthsurfacephenomena,suchasgeology.34

Studyoflandusedynamicsisfurthercomplicatedbyavarietyoftime-relatedfactors.Landusesystems,aswellas theunderlyingfactorsandprocesses, them-selvesmaychangethroughtime.Thisproducesavarietyofpathdependence35andlegacyeffects,36resultinginlandusepatternsandsystemsthatmayreflectavarietyofnot only contemporaryprocesses, but alsoprocesses and responses tohistoricdrivers of change. Additionally, land cover change involves both conversion andmodificationofcover37andmaybegradualorepisodic.7Typicallychangethroughtime,especiallyforspatialmodels,isstudiedquantitativelyforaplacewithaseriesofsnapshotsof landcover (sometimes treatedasequivalent toor interchangeablewithlanduse).Thismaynotonlyunderestimatetheextentofchangebutalsofailto capture whether changes are gradual or episodic. Measurement, analysis, andmodeling of land use systems need frameworks, tools, and methods that help toseparatemultipleinfluencesandasynchronouscausesofchangeinlandusesystemdynamics,aswellasprovideanimprovedabilitytodetectagreaterrangeoftypesandratesofchange.Crews’research38(Chapter6)usespanelanalysistofocusonthelongitudinal(time)sequenceofchange.

The use of snapshots of land cover to study change and develop quantitativemodels may also ignore the rich source of insight and methods from a study ofhistoricaldevelopmentinthatplace.7,36Forexample,environmentalhistory,narrative,andstorytellingallofferinsightintochangethroughtime.39Combiningquantitativeapproacheswithenvironmentalhistorywouldofferadvantagesforstudyofchangeinspatialpatternandtemporaldynamicsoflandsystems,andforunderstandingtherolesofmultiplecausesandprocessesofchange.

Arangeofotherissuesneedstobeaddressedtosupportresearchondynamicsof land change, especially if the goals are improved understanding of processesandchangesproducedandbettermodelingofplace-basedchangeinlandsystems.Specifically,thereisaneedfornewexperimentalandobservationaldesignsaswellas researchprotocols that supportquantitativeandqualitativeanalysisofchange.Thiswillenablecloseranddirectcomparisonofresultsfromdifferentcasestudies,leadingtoimprovedmeta-analysesandsynthesisbetweencasestudies.Thereisalsoaneedforresearchintohowspatialandtemporaldynamicsmightbestberepresentedtosupportquantitativeandqualitativeanalysis.Althoughgeographicalinformationsystems(GIS)andremotesensingprovidemechanismsforrecording,representation,measurement,andanalysisofthespatialstructureofpropertiesofthelandsurface,GISisstillpoorlydevelopedforrepresentationandanalysisofspatiotemporaldata.ImprovedGISdatastructuresforlandcoverandusedatathatexplicitlyincorporatetimeareneeded.Thiswillbeabenefitnotonlyfornewformsofanalysis,butitalsowillprovideanempiricalfoundationforthebroadrangeofnoveltools, includingagent-basedmodeling40andcellularautomata41,42thathaveaddedtoourabilitytorepresentandmodelsocialandspatialprocessesthatarecentraltoamorefullunder-standingoflandsystemdynamics.Regular,repeatedremotelysensedmeasurements

42963_C001.indd 6 10/29/07 10:47:48 AM


arealsoneededtohelpdevelopthespatialandtemporalhistoryoflanduseandlandcoverasaprecursortoimproveddynamicspatialmodelsofchange.43,44

1.2.2 inTegraTion and Feedbacks beTween landscape, climaTe, socioeconomic, and ecological sysTems

Landsystemsare increasinglybeing treatedasexemplarsof couplednatural andhuman systems (frequently socioecological systems,45 which are considered heretobeexactlysynonymous).Thissystems-basedviewof landrecognizes that landuseandlandcoversystemsaredefinedattheinterfaceofnatural/biophysicalandhumansystems.Indeed,theIHDP/IGBPGlobalLandProjectestablishesaframe-work for combined study of land change, ecosystem services, and sustainabilitywithinanEarthsystemsciencecontextthroughexploringlanduseandlandcoverasrepresentingacontinuumbetweensocioeconomic(use)andbiophysical(cover)systems.3 This attempt at a coupled and holistic view of land systems presents arichframeworkforunderstandingmultidirectionalimpactsandfeedbacksbetweenelementsoflandsystemsandthemanyunderlyingfactorsandprocessesthatoperatetoshapeandchangeland.Forexample,notonlydolandcoverandlandusereflectenvironmentalsystemsandtheopportunitiesandconstraintstheyprovide,butlandcover and land use also have a strong and direct influence on climate and otherenvironmentalchangeandmustbeincludedinnewmodelsoftheEarth’sclimatesystem.27,46Similarly,changesinlandsystemsreflectsocioeconomicprocessesthatoperateataverywiderangeofspatialandtemporalscales,includingglobalization,tradeandmarkets,policyandlandmanagementdecisionsatthenational,regional,local,orhousehold/individuallevel.

Therehavebeenseveralframeworksforstudyoflandsystemsasexemplarsofcouplednaturalandhumansystems.Forexample,Machlisandcolleagues5describeahumanecosystemmodelthatattemptstolinksocialsystemsandnaturalsystemsthroughafocusonsocialsystemslinkedtonatural,cultural,andsocioeconomicsascriticalresources.Thisexplicitlyincludessocialstructures,butdoesnotintegrateanecologicalsystemsview,anddoesnotaddressprocess-levelunderstanding.Despitethese attempts and the claims of certain disciplines to provide an overarchingintegrativecontent, there remainsaneedforbetterconceptualmodelsof integra-tion and feedbacks between landscapes, climate, ecological (environmental), andsocioeconomicsystems.Steinitzandcolleagues47providearesearchframeworkforlandscapedesignbasedondifferent typesofmodeland thequestions themodelsanswer.Again,thishasmanyelementsthatsuggestintegrationbutwithafocusonmodelsandcorequestionsthatmaynotbefullyinclusive.

Ingeneral,qualitiesofsuccessfulframeworksmightinclude(a)afocusoninter-disciplinaryortransdisciplinary48approaches,(b)atheoreticalgroundinginarangeof sciences to allow for participation from multiple disciplines, (c)being systemsbased,toaddressprocessesandstructures,(d)beingspatiallyandtemporallyexplicittoallowgeographicandhistoricalcontextsandcontingenciestobeincludedinunder-standingandanalysisofdynamics,and(e)beingexplicitinaddressingscale,includingspatial,temporal,andorganizationalscales,49includingmultiscaleeffects.50

42963_C001.indd 7 10/29/07 10:47:49 AM


1.2.3 resilience, VulnerabiliTy, and adapTabiliTy oF land sysTems as coupled naTural and human sysTems

Inconjunctionwithagrowingunderstandingoflandsystemsasexemplarsofcouplednatural and human systems, there is a move to interpret land systems within thecontextofsustainabilityscience.3,51Asanintegralpartofthis,researchaddressesthevulnerability(fromthehazardsliterature),resilience(fromtheecologicalliterature),andadaptabilityoflanduseandlandsystems.Sustainability,vulnerability,resilience,andadaptabilityallhavestrongtemporalelementsthatrefernotonlytotheabilityoflandsystemstorespondtochangebutalsototheabilityoflandsystemstoinflu-encetheresponsesofsocioeconomicandenvironmentalsystemstochange.Whatwouldbethecharacteristicsofaresilientlandsystemoravulnerablelandsystem?Canweassesswhetheralandsystemorlanduseisresilientorvulnerabletochangesintheexternaldrivingprocesses(suchasininternationalpricesforcrops)?Doesalandsystemconferresilienceorvulnerability tosocialsystems?Forexample,aretherepatternsof landuseand landcover,andasuiteofassociated landmanage-mentmechanisms, thatmakea community resilient/vulnerable todrought,flood,or other environmental or socioeconomic change? What are the time scales forresilience/vulnerability thataremostappropriateforsustainability?Landsystemsclearlyofferthepotentialtoexploreeconomic,social,institutional,environmental,andecosystemresilience,vulnerability,andsustainabilitywithbenefitstosociety,environment,andland.

1.2.4 scale issues

Scale issues are always central to discussions of land system change,52 but theirresolutionislinkedtothemanydifferentmeaningsof“scale”intheinterdisciplinaryandmultidisciplinarycommunitiesworkingonlandusescience.Scalereferstoallof(a)spatialextentandresolution,(b)time,includingthedurationofastudyandtheresolutionofdatasnapshots,(c)taxonomiclevel,forexample,thelevelofinterestinlanduseorlandcoverasinthedetailinalanduse/coverkeyorintheinstitutional/geopolitical hierarchy of global-international-national-regional-local, or (d) otheranalyticaldimensionsusedforstudy.Giventheinterdisciplinarynatureofmuchoftheresearchandliteratureonlanduse,someconsistencyandexplicitreportingoftheuseofterminologyrelatedtoscalewouldbevaluable.52Gibsonandcolleagues49provideadetaileddiscussionofscaleissues.Scaleissuesinlandusechangemayalso benefit from the space-time approaches developed in landscape ecology.53,54Asystematicanalysisandreviewofscalesatwhichprocessesandresponsesoperateinlandusesystemsmaybevaluableinguidinganalysisandmodelingofchange.

1.2.5 uncerTainTy

Uncertaintyunderliesmanyaspectsoflandusescienceandisassociatednotonlywithmeasurement,analysis,andmodeling,butalsowithdecisionmakingandtheconsequencesofmodelsforprojectingcurrentandfutureconditionsandstatesoflandusesystems.Theaccuracyofmodelsthatpredictchangeisaconstantconcernof land use scientists,55 but there are also concerns related to the accuracy and

42963_C001.indd 8 10/29/07 10:47:50 AM


uncertaintiesofsystemsmodelsasawholecomparedwiththeaccuracyanduncer-taintiesofcomponent(sub-)modelsoflandusesystems.Errorpropagationthroughcoupledmodels of landuse systems as coupledhumanandnatural systems is ofparticularconcernandinterest.

Although statistical methods for management of uncertainty associated withmeasurement,models,andpredictionsarerelativelywelldeveloped,othertypesofuncertaintyrequireattention.Conroy56describesfourtypesofuncertainty:

1.Statisticaluncertainty:reflectinginabilitytomeasure 2.Structural uncertainty: reflecting inability to describe and model system

dynamics 3.Partialcontrollability:reflectinginabilitytocontroldecisions 4. Inherentuncertainty:presentinallsystems(stochasticprocesses)

All of these aspects of uncertainty need research for effective use of models topredict changeand tousemodels fordecision support related topolicy and landmanagement.Betterintegrationandunderstandingofscientificuncertaintyrelatedtodecisionmakingunderconditionsofuncertaintyshouldhelptoenhancecommu-nicationbetweenscientists,policymakers,andlandmanagers.

Policy-relevantmodelsraisesomespecialconcernsbeyondthoseofuncertainty.Lee57andKingandKraemer58havediscussedrequirementsformodels thatare tobeused inpolicycontexts.Lee57 identifiesfivequalitiesofmodels themselves: (1)transparency,(2)robustness,(3)reasonabledataneeds,(4)appropriatespatiotemporalresolution,and(5)inclusionofenoughkeypolicyvariablestoallowforlikelyandsig-nificantpolicyquestionstobeexplored,whileKingandKraemer58concentrateontheroleservedbymodels:(1)toclarifytheissuesinthedebate,(2)toenforceadisciplineofanalysisanddiscourseamongstakeholders,and(3)toprovideaninterestingformofadvice,mainlyintheformofwhatnottodo.Agarwalandcolleagues59usethesecriteria, criteria from Veldkamp and Fresco60 describing space, time, biophysicalfactors,andhumanfactors,aswellasscaleandcomplexityofmodelsascross-cuttingcriteriatoreviewandcomparelandusechangemodels.Theytakeapragmaticviewthatlandusemodels,althoughnotideal,willbegoodenoughtobetakenseriouslyinthepolicyprocess.Theyconcludebyidentifyinganeedforgreatercollaborationbetweenlandusemodelersandpolicymakers.Suchcollaborationwillneedtoiden-tifythekeyvariablesandsectorsofinterest,scalesofanalysis,andchangescenariosanticipated.Thiswillalsorequiretranslationbetweenthedemographic,economic,technology,institutionalandpolicy,andculturalfactors19,20,24usedforanalysis,under-standing,andmodelingoflandusechangeandthespecificneedsofpolicymakers.59Collaboration in this form requires closer focuson applied science from landusescientistsandconsiderationofmechanismsusedforcollaborationandparticipationbetweenscientistswithpolicymakersandlandmanagers.

1.� applIed sCIenCe

Landusescienceisalsoanappliedsciencewithclearlinkstopolicyandpracticethroughdecisionmakingandotherhumaninterventionandactiononlanduseand

42963_C001.indd 9 10/29/07 10:47:50 AM

10 Land Use Change: Science, Policy and Management

landcover.Threeissuesseemparticularlyimportantinrelationtoemergingtrendsandneedsforlandusescienceasanappliedscience:

1.Addressingevolvingpublicandprivatelandmanagementissuesanddecisions 2. Interpretation and communication of scientific knowledge for adaptive

managementofchangeinlandusesystems 3.Understandinghumanandenvironmentalresponsestochange

1.3.1 addressing eVolVing public and priVaTe land managemenT issues and decisions

The multiscale nature and consequences of land use change present a compel-lingcase for translating science intopolicyandpractice.Since landuse isat theinterfaceofhumanandnaturalsystems,improvedunderstandingofthesocialandbiophysicalprocessesthatproducechangeinlandusesystemscanplayausefulroleinbothpolicyandpracticeforbothprivateandpubliclandmanagementissuesanddecisions.Improvedunderstandingoflandusechangecanprovideinputtoevidence-basedpolicyandbeusedtodevelopalternativescenariosforlanduseresponsetodifferent policies.61,62,63 This may help to inform wide participation of interestedgroupsandindividualsindebateanddiscussionandalsohelptoleadtoimproveddecisionmaking.Thismaybethroughimprovedandinformedconsensusbyhelp-ingtoexploreimpactsandconsequencesofparticularpolicies(andalternatives)orotheractions.Landusechangealsoreflectslinksthatoccuracrossorganizationalscalesand thus is important in linkingbroadernational, international,andglobaltrendsandconditionswithconsequentunderstandingandconcern for livelihoodsandsustainabilityofcommunitiesatmorehumanlocalscales.

In this context, case studies of land use change can be valuable in two ways.First,theycanprovidegenericunderstandingthatcanbeusedasevidenceinsupportofdiscussion toproduceevidence-basedpolicy.Second, theycanprovide specificscenarios and impacts for an area of interest to help focus decision making in adeliberateandlocallyrelevantmanner.Thusstudiesoflandusechangecanbeusednotonlytoelucidategeneralissuesandprinciplesconcerninglandusechange,butalso to provide insight into applied issues associated with land use and potentialmeaningofchangeforcommunitiesandplaces.Thelatterwouldbeofvalueinestab-lishingimprovedunderstandingofsocialfeedbacksandindevelopingparticipatoryprocessesandconsultationindecisionmaking.Theseopportunitiesandusessuggestthatthereareanumberofimportantconcernsthatshouldbemadeexplicitincasestudiesoflandusechange.Forexample,arecasestudiesconcernedwithmanagementand/orpolicy?Arestakeholders involved in the researchand inwhatways?Whatarethereactionstoresultsandprocessoftheresearch,especiallyfromavarietyofdifferentgroups?Whatlessonsarelearnedfromacasestudyaboutthepoliciesandmanagementpracticesthatinfluencelandusechangeinthestudyareaandsystems?Betterunderstandingof theseaspectsof landusechangecasestudieswillhelp toimprovelinksbetweenscienceandpractice37andalsoplacelandusechangeevidenceinbothascientificanddecisionandmanagementcontext,gettinglanduse/systemssciencetranslatedfromsciencetomanagementandpolicy,andviceversa.

42963_C001.indd 10 10/29/07 10:47:51 AM

Basic and Applied Land Use Science 11

1.3.2 inTerpreTaTion and communicaTion oF scienTiFic knowledge For adapTiVe managemenT oF change in land use sysTems

Interpretationandcommunicationofscientificknowledgeareofincreasingimpor-tanceacrossallofscience.Landuse–relatedresearch,notablyforagricultureandforestry, has a longhistoryofdirect and strong application to landmanagement;examples may be found in agricultural extension services or forestry worldwide.Therearealsoexamplesoflanduseinformationdirectlyguidingpolicy(e.g.,LandUtilisation Survey in Great Britain64). Communication of case study content andinformation on land use change is no less important today, not only for input topolicyandlandmanagementdecisions,butalsoforunderstandinglocalhumanpro-cessesthatproducemanyaspectsofcontemporarychangeinlanduseandlandcover(asreflectedinthegenericunderlyingcausespreviouslydiscussed).Thisraisesthequestionof theextent towhichcase studies involve landmanagersandpolicyorotherdecisionmakersaspartoftheresearchteam(transdisciplinaryinthesenseofTressandcolleagues48).Suchatwo-waycollaborationwouldbebeneficialbothtoscientistsandpractitioners.65,66

1.3.3 human and enVironmenTal responses To change

Applicationoflandusesciencetounderstandingpracticalconsequencesofchangeforhumanandenvironmentalsystemsandtheircomponentsubsystemspotentiallyprovideslinkstoawidevarietyofareasofconcerninenvironmentalmanagementandprovisionofecosystemservices.67Improvedunderstandingandpredictiveandscenario-basedtoolsalsohaveapplicationinlanduseplanningandenvironmentalmanagementandcare.68Keyappliedissuesincludethemannerinwhichdifferentinstitutions,society,andindividualsrespondtochange;howenvironmentalsystemsrespond; space and time scales of response and consequences; development ofstrategiestoadapttoandmanagechangeanditsimpacts;andlong-andshort-termconsequencesofchangeanddecisions.

Theseapplied issues shouldbesetagainstaneed forexplicitmanagementofuncertainties, which is a recurrent theme in scientific communication. Increasedawarenessofuncertaintyshouldincludedefinitionofthelimitsofapplicabilityofprojectionsoflandusechangeandotheranalyses,includingscenarios,particularlyastranslatedintodecisionsupportandparticipatoryapproaches.Boundaryorgani-zations65 and transdisciplinary48 approachescanhelp this explicitmanagementofuncertaintysinceawidercommunityisengagedinmodelingandanalysis,whichboth helps the scientists and practitioners through improved communication andgreaterunderstandingofbothdecision-makingneedsandscientificprocesses.

referenCes

1. Aspinall,R.J.Editorial.Journal of Land Use Science1(1),1–4,2006. 2. Raquez,P.,andLambin,E.F.Conditionsforasustainablelanduse:Casestudyevidence.

Journal of Land Use Science1(2–4),109–125,2006. 3. GLP.Science Plan and Implementation Strategy. IGBPreportNo.53,IHDPReport

No.19.Stockholm,IGBPSecretariat.64pp,2005.

42963_C001.indd 11 10/29/07 10:47:51 AM


4. Lambin,E.F.,Geist,H. J., andRindfuss,R.R. Introduction: Local processeswithglobal impacts. In:Lambin,E.F. andGeist,H. J.,Eds.Land-Use and Land-Cover Change: Local Processes and Global Impacts.IGBPSeries.Springer-Verlag,Berlin,1–8,2006.

5. Machlis,G.E.,Force,J.E.,andBurch,W.R.Thehumanecosystem,PartI:Thehumanecosystemasanorganizingconceptinecosystemmanagement.Society and Natural Resources10,347–367,1997.

6. Turner,B.L.etal.,eds.Land Use and Land Cover Change: Science/Research Plan.IGBPReportNo.35,HDPReportNo.7.Stockholm,Sweden,InternationalGeosphere-BiosphereProgramme,132pp,1995.

7. Lambin,E.F.,Geist,H.J.,andLepers,E.Dynamicsofland-useandland-coverchangeintropicalregions.Annual Review of Environment and Resources28,205–241,2003.

8. Brown,D.G.,Pijanowski,B.C.,andDuh,J.D.Modelingtherelationshipsbetweenland use and land cover on private lands in the upper midwest, USA. Journal of Environmental Management59(4),247–263,2000.

9. Brown,D.G.,andDuh,J.D.Spatialsimulationfortranslatingfromlandusetolandcover.International Journal of Geographical Information Science 18(1),35–60,2004.

10. Xu,J.C.etal.Land-useandland-coverchangeandfarmervulnerabilityinXishuang-bannaprefectureinsouthwesternChina.Environmental Management36(3),404–413,2005.

11. Walsh,S.J.etal.Scale-dependentrelationshipsbetweenpopulationandenvironmentinnortheasternThailand.Photogrammetric Engineering and Remote Sensing 65(1),97–105,1999.

12. Veldkamp, A., and Verburg, P. H. Modelling land use change and environmentalimpact.Journal of Environmental Management 72(1–2),1–3,2004.

13. Urama, K. C. Land-use intensification and environmental degradation: Empiricalevidencefromirrigatedandrain-fedfarmsinsoutheasternNigeria. Journal of Environ-mental Management75(3),199–217,2005.

14. Lambin, E. F., and Geist, H. J., eds. Land-Use and Land-Cover Change: Local Processes and Global Impacts.IGBPSeries.Springer-Verlag,Berlin,222pp,2006.

15. Riebsame,W.E.,Gosnell,H.,andTheobald,D.M.Landuseandlandscapechangein the Colorado Mountains. 1. Theory, scale, and pattern. Mountain Research and Development16(4),395–405,1996.

16. Veldkamp, A., and Fresco, L. O. CLUE-CR: An integrated multi-scale model tosimulate land use change scenarios in Costa Rica. Ecological Modelling 91(1–3),231–248,1996.

17. Bousquet,F., andLePage,C.Multi-agent simulations and ecosystemmanagement:Areview.Ecological Modelling176(3–4),313–332,2004.

18. Adger,W.N.Evolutionofeconomyandenvironment:Anapplication to landuse inlowlandVietnam.Ecological Economics 31(3),365–379,1999.

19. Geist,H.,andLambin,E.F.Proximatecausesandunderlyingdrivingforcesoftropicaldeforestation.Bioscience52(2),143–150,2002.

20. Geist,H.J.,andLambin,E.F.Dynamiccausalpatternsofdesertification.Bioscience54(9),817–829,2004.

21. Hill, M. J. et al. Multi-criteria decision analysis in spatial decision support: TheASSESSanalytichierarchyprocessandtheroleofquantitativemethodsandspatiallyexplicitanalysis.Environmental Modelling and Software20(7),955–976,2005.

22. Madsen,L.M.,andAdriansen,H.K.Understandingtheuseofruralspace:Theneedformulti-methods.Journal of Rural Studies20(4),485–497,2004.

23. Geist,H.J.,andLambin,E.F.What Drives Tropical Deforestation? A Meta-Analysis of Proximate and Underlying Causes of Deforestation Based on Subnational Case Study Evidence. LUCC Report Series No. 4., Land-Use and Land-Cover Change

42963_C001.indd 12 10/29/07 10:47:52 AM

Basic and Applied Land Use Science 1�

(LUCC)ProjectIV.InternationalHumanDimensionsProgrammeonGlobalEnviron-mentalChange(IHDP)V.InternationalGeosphere-BiosphereProgramme(IGBP),116pp,2001.

24. Keys,E.,andMcConnell,W.J.Globalchangeandtheintensificationofagricultureinthetropics.Global Environmental Change15,320–337,2005.

25. Vitousek, P. M. Beyond global warming: Ecology and global change. Ecology 75,1861–1876,1994.

26. Vitousek,P.M.etal.HumandominationofEarth’secosystems.Science277(15July),494–499,1997.

27. Foley,J.A.etal.Globalconsequencesoflanduse.Science 309,570–574,2005. 28. National Research Council (NRC). Grand Challenges in Environmental Sciences.

ReportfromtheCommitteeonGrandChallengesinEnvironmentalSciences.NationalAcademyPress,Washington,DC.96pp,2001.

29. Skole, D., and Tucker, C. Tropical deforestation and habitat fragmentation in theAmazon:Satellitedatafrom1978to1988.Science260(5116),1905–1910,1993.

30. Brown,D.G.LanduseandforestcoveronprivateparcelsintheuppermidwestUSA,1970to1990.Landscape Ecology18(8),777–790,2003.

31. Ferraz,S.F.D.etal.LandscapedynamicsofAmazoniandeforestationbetween1984and2002incentralRondonia,Brazil:Assessmentandfuturescenarios.Forest Ecology and Management 204(1),67–83,2005.

32. Etter,A.,andvanWyngaarden,W.PatternsoflandscapetransformationinColombia,withemphasisintheAndeanregion.Ambio29(7),432–439,2000.

33. Goldewijk,K.K.Estimatinggloballandusechangeoverthepast300years:TheHYDEDatabase.Global Biogeochemical Cycles15(2),417–433,2001.

34. Schumm, S. D. To Interpret the Earth: 10 Ways to Be Wrong. 2nd ed. CambridgeUniversityPress,Cambridge.143pp,2006.

35. Brown,D.G.etal.Pathdependenceandthevalidationofagent-basedspatialmodelsoflanduse.International Journal of Geographical Information Science19(2),153–174,2005.

36. Aspinall,R.J.Modellinglandusechangewithgeneralizedlinearmodels—amulti-modelanalysisofchangebetween1860and2000inGallatinValley,Montana.Journal of Environmental Management72(1–2),91–103,2004.

37. Lesslie, R., Barson, M., and Smith, J. Land use information for integrated naturalresources management—a coordinated national mapping program for Australia.Journal of Land Use Science1(1),45–62,2006.

38. Crews-Meyer, K. A. Characterizing landscape dynamism using paneled-patternmetrics.Photogrammetric Engineering and Remote Sensing 68(10),1031–1040,2002.

39. Wyckoff,W.,andHansen,K.Settlement,livestockgrazingandenvironmentalchangeinsouthwestMontana,1860–1990.Environmental History Review15,45–71,1991.

40. Parker,D.C.etal.Multi-agentsystemsforthesimulationofland-useandland-coverchange:Areview.Annals of the Association of American Geographers93(2),314–337,2003.

41. Batty,M.Agents,cells,andcities:newrepresentationalmodelsforsimulatingmulti-scaleurbandynamics.Environment and Planning A 37(8),1373–1394,2005.

42. Lau,K.H.,andKam,B.H.Acellularautomatamodelforurbanland-usesimulation.Environment and Planning B-Planning and Design32(2),247–263,2005.

43. Turner, D. P., Ollinger, S. V., and Kimball, J. S. Integrating remote sensing andecosystemprocessmodelsforlandscape-toregional-scaleanalysisofthecarboncycle.Bioscience54(6),573–584,2004.

44. Tralli,D.M.etal.Satelliteremotesensingofearthquake,volcano,flood,landslideandcoastal inundationhazards.ISPRS Journal of Photogrammetry and Remote Sensing59(4),185–198,2005.

42963_C001.indd 13 10/29/07 10:47:53 AM

1� Land Use Change: Science, Policy and Management

45. Carpenter, S. et al. From metaphor to measurement: Resilience of what to what?Ecosystems4(8),765–781,2001.

46. Foley,J.A.etal.Incorporatingdynamicvegetationcoverwithinglobalclimatemodels.Ecological Applications 10(6),1620–1632,2000.

47. Steinitz,C.etal.Alternative Futures for Changing Landscapes: The Upper San Pedro River Basin in Arizona and Sonora.Washington,D.C.:IslandPress.202pp,2003.

48. Tress, B. et al., eds. Interdisciplinary and Transdisciplinary Landscape Studies: Potential and Limitations,DELTASeries2.Wageningen,192pp,2003.

49. GibsonC.C.,OstromE.,andAnhT.K.Theconceptofscaleandthehumandimen-sionsofglobalchange:Asurvey.Ecological Economics32,217–239,2000.

50. Redman,C.L.,Grove,J.M.,andKuby,L.H.Integratingsocialscienceintothelong-termecologicalresearch(LTER)network:Socialdimensionsofecologicalchangeandecologicaldimensionsofsocialchange.Ecosystems7(2),161–171,2004.

51. Kates, R. W., and Parris, T. M. Long-term trends and a sustainability transition.Proceedings of the National Academy of Sciences100,8062–8067,2003.

52. Verburg,P.H.etal.Landusechangemodelling:Currentpracticeandresearchpriori-ties.GeoJournal61,309–324,2004.

53. Delcourt, H.R.,Delcourt, P. A., andWebb T. A., III. Dynamic plant ecology: Thespectrum of vegetation change in space and time. Quaternary Science Reviews 1,153–175,1983.

54.Delcourt,H.R.,andDelcourt,P.A.Quaternarylandscapeecology:Relevantscalesinspaceandtime.Landscape Ecology2,23–44,1988.

55. Pontius,R.G.,andSpencer,J.Uncertaintyinextrapolationsofpredictiveland-changemodels.Environment and Planning B-Planning and Design 32(2),211–230,2005.

56. Conroy,M.J.Conservationandlandusedecisionsunderuncertainty:Models,data,andadaptation.In:Hill,M.J.,andAspinall,R.J.,Eds.,Spatial Information for Land Use Management.OPAOverseasPublishingAssociates,Ltd.,Reading,U.K.,145–158,2000.

57. Lee,D.B.,Jr.Requiemforlarge-scalemodels.AIP Journal (May),163–177,1973. 58. King,J.L.,andKraemer,K.L.Models, facts,and thepolicyprocess:Thepolitical

ecology of estimated truth. In: Goodchild, M. F., Parks, B. O., and Steyaert, L. T.,eds.Environmental Modeling with GIS.OxfordUniversityPress,NewYork,353–360,1993.

59. Agarwal,C.etal.A Review and Assessment of Land-Use Change Models: Dynamics of Space, Time and Human Choice.CIPECCollaborativeReportSeriesNo.1.CIPEC,IndianaUniversity.90pp,2002.

60. Veldkamp,A.,andFresco,L.O.CLUE:Aconceptualmodeltostudytheconversionoflanduseanditseffects.Ecological Modelling85,253–270,1996.

61. Steinitz, C. et al. A delicate balance: Conservation and development scenarios forPanama’sCoibaNationalPark.Environment47(5),24–39,2005.

62. Berger,P.A.,andBolte,J.P.Evaluatingtheimpactofpolicyoptionsonagriculturallandscapes:Analternative-futuresapproach.Ecological Applications14(2),342–354,2004.

63. vanDijk,T.ScenariosofCentralEuropeanlandfragmentation.Land Use Policy20(2),149–158,2003.

64. Sheail,J.Scott revisited:Post-waragriculture,planningandtheBritishcountryside.Journal of Rural Studies13(4),387–398,1997.

65. Guston,D.H.Boundaryorganisationsinenvironmentalpolicyandscience:Anintro-duction.Science, Technology and Human Values26(4),399–408,2001.

66. Lemos, M. C., and Morehouse, B. J. The co-production of science and policy inintegratedclimateassessments.Global Environmental Change15,57–68,2005.

42963_C001.indd 14 10/29/07 10:47:53 AM

Basic and Applied Land Use Science 1�

67. Rounsevell, M. D. A. et al. Future scenarios of European agricultural land use II.Projectingchangesincroplandandgrassland.Agriculture Ecosystems and Environ-ment107(2–3),117–135,2005.

68. Haberl,H.,Wackernagel,M.,andWrbka,T.Landuseandsustainabilityindicators.Anintroduction.Land Use Policy 21(3),193–198,2004.

42963_C001.indd 15 10/29/07 10:47:54 AM

42963_C001.indd 16 10/29/07 10:47:54 AM

17

2 Developing Spatially Dependent Procedures and Models for Multicriteria Decision AnalysisPlace, Time, and Decision Making Related to Land Use Change

Michael J. Hill

Contents

2.1 Introduction................................................................................................... 172.2 Concept......................................................................................................... 192.3 TransformationIssues................................................................................... 192.4 TransformationDomainsandMethods........................................................ 212.5 ExampleLandscapeContext—AustralianRangelands................................23

2.5.1 SpatialPatternsandRelationships.....................................................252.5.2 TemporalPatternsandInfluences......................................................262.5.3 DataandInformation:ScaleofRepresentation.................................292.5.4 SomeSpatiotemporalInputstoaRangelandMCA...........................30

2.6 AFrameworkforaMulticomponentAnalysiswithMCA........................... 322.7 Conclusions................................................................................................... 332.8 Acknowledgments......................................................................................... 37References................................................................................................................ 37

2.1 IntRoDUCtIon

Land use change occurs within a space-time domain. Frameworks for assessingappropriatelanduseandprioritiesforchangemustcapturethecomplexity,reduce

42963_C002.indd 17 11/6/07 7:17:13 AM

18 Land Use Change

dimensionality,summarizeahierarchyofmaineffects,transfersignalsandpatterns,andtransforminformationintothelanguageofthepoliticalandeconomicdomains,1yetretainthekeydynamics,interactions,andsubtleties.Spatialinteraction,temporalcycles,responsesandtrends,andchangesinspatialpatternsthroughtimeareimpor-tant sources of information for condition, planning, and predictive assessments.Spatiallyappliedmulticriteriaanalysis2enablesdiversebiophysical,economic,andsocialvariablestobemappedintoastandardizedrankingarray;usedasindividualindicators;combinedtodevelopcompositeindexesbasedonobjectiveandsubjec-tivereasoning;andusedtocontrastandcomparehazards,risks,suitability,andnewlandscapecompositions.3,4,5,6Themulticriteriaframeworkallowsthecombinationofmulti-andinterdisciplinarity.7Thesystemdefinitiondependsuponthepurposeoftheconstruct,scaleofanalysis,andsetofdimensions,objectives,andcriteria.7

Whenmappingbothquantitativeandordinaldataintofactorlayers,retentionof,andaccessto,rationaleandreasoningforinclusionandweightingorcontributiontocompositesisimportantformaintenanceofthelinkbetweentheoutcomeoftheanalysisandtherealorapproximatedatausedasinput.Thisparticularlyappliestospatialandtemporalinformation.Hereitisimportanttoknowwhatthemeaningofaspatialortemporalmetricmightbewhenitisincludedamongotherdataindevel-opmentofanassessmenttoaiddecisionmakers.Themeaninghastwocomponents:(1)thefirstrelatestothedirectdescriptionofthemetricsuchastheaveragepatchsizeofremnantvegetationwithinaparticularanalyticalunit,or theamplitudeoftheseasonaloscillationingreennessfromanormalizeddifferencevegetationindex(NDVI)profile;(2)thesecondrelatestowhatthemetricmeasuresintermsofinflu-enceon the target issue; forexample,patchsizesgreater thanx indicateahigherwaterextractiontowaterrechargeratio,resultinginaloweringofthewatertable,oranamplitudeequaltoyindicatesa75%probabilitythattheareaisusedforcerealcroppingandhencehasnowaterextractioncapacityinsummer.Inthecontextofmulticriteriaanalysis(MCA),assignmentofmeaningtospatialandtemporalmetricsdepends on project-based research, wherein a relationship is established betweensomeaspectoflandusechangeorcondition,orsomederivedpropertyofaninputvariablelayer,andametricthatisrobustandtranslatablefromstudytostudy.Intrin-sically,somemetricshavemoreeasilyascribedmeaningthanothers—themeaning-fulnessbeinginverselyproportiontothedegreeofabstractionandextentofremovalfrom biophysical, economic, or social measures that are a directly related to themanifestationoflandusechange.

ThereisaverywidearrayofpotentialanalyticaladjunctstoMCA.8Thesecanbesummarizedintoseveralgroupsofmethods:thosefordealingwithinputuncer-tainty;thoseappliedtoweightingandranking;modelsanddecisionsupportsystems(DSS)deliveringhighlyprocessedandsummarizedderivedlayersintotheanalysis;various cognitive and soft systems methods requiring transformation for use, orperhapssittingoutsideofthestandardMCA;optimizationapproaches;andintegratedspatialDSS,participatorygeographical information(GIS)andmultiagentsystems.However, the quantification, metrication, and summary of spatial and temporalsignals and temporal change in spatial patterns represent a level of sophisticationandderivationthathasyet tobefullyexplored.Recentexperiencewith thedevel-opmentof simple scenario tools forassessingcarbonoutcomes frommanagement

42963_C002.indd 18 11/6/07 7:17:14 AM

Developing Spatially Dependent Procedures and Models 19

change inrangelands9,10hasemphasized the importanceofspatialgradients, inter-actionsandpatterns,and temporal trendsand transitions in response toanthropo-genicandenvironmentalforcing.Inthischapter,theAustralianrangelandsareusedasanexamplecoupledhuman-environmentsystemtoexaminetherolethatspatialandtemporalinformationcanplayinamulticriteriaframeworkaimedatinformingpolicyandbydefinitionrequiringasubstantialelementofsocialcontext.

Thereisalargeandlong-standingliteraturebasedealingwithsignalprocessing11andtimeseriesanalysis12,13,14andmergingmethodsacrossthesetwoareas.15Thisliteratureindicateshowthepropertiesofdemographic,economic,social,andbio-physicalpoint-basedtimeseriesdatacanbecaptured.Withspatiallyexplicittimeseriesweareinterestedinhowthesepropertiescanbemeaningfullymappedintoamulticriteriaanalysisframework.

2.2 ConCePt

Thepremisebehindthischapteristhatsomeformofmulticriteriaframeworkisuse-fulforexplorationofcomplexcoupledhumanenvironmentsystemsandforinformingpolicydecisionmaking.Integrationofnonscientificknowledgeisofkeyimportance,andtheuserperspectivemaybetheultimatecriterionforevaluation.16Arequirementofthisanalysisisthatitissimpleandtransparenttotheclient,stakeholder,partici-pant,anddecisionmaker,butthatithasthecapabilitytocapturecomplexspatialandtemporalinteractionsandtrendsthatinfluencethenatureofbothsystembehaviorand evolution and the consequencesof decisions. Inprinciple, it is necessary formulticriteriaframeworkstoincludemeasuresofsystemdynamics—bothspatialandtemporal.Therefore,theunderlyingthemeinthischapteristheefficacy,efficiency,andinformationcontentoftransformationsofspatialandtemporaltrends,patterns,and dynamics into standardized, indexed layers for use in spatial multicriteriaanalysis.Theensuingdiscussiondoesnot imply thatmulticriteriaapproachesareeithertheonlywayorthebestwaytoapproachanalysisforpolicydecisionmakingincoupledhumanenvironmentsystems.Itissimplyoneapproachthathasproventobeuseful,4,6,17,18anditprovidesacontextfordiscussionoftheissueoftransformationofspatialandtemporalsignalsoutofacomplexmultidimensionalresponsespaceintostandardized,unitless,ordinalscalarstoassistinhumanproblemexplorationanddecisionmaking.

2.3 tRAnsFoRMAtIon IssUes

Intermsofdefinition,transformationistakentomeanamethodbywhichamorecomplexspatialpatternorrelationship,ortemporalpatternortrend,ismappedintoonetomanyquantitativemetrics thathavesomefunctionalrelationshiporunder-standabledescriptivecontribution thatcanbe ranked in termsof theobjectiveofamulticriteriaapproach.Thistransformationcanthereforebeasimpleregressionfunctionwhereintheslopeisusedasthemetric,oritcanbeasetofpartialmetricsthat together provide a composite indicator capable of being ranked.Examples ofthelattermightincludeseveralspatialpatchmetricssuchasnumber,size,andedgelengthorseveralcurvemetricssuchastimings,amplitude,andareaunderthecurve.

42963_C002.indd 19 11/6/07 7:17:14 AM

20 Land Use Change

Sextonetal.19definefourdimensionsofscale:

(a) Biological—fromcell,organism,population,community,ecosystem,land-scape,biometobiosphere;withfourusefullevels:(1)genetic,(2)species,(3)ecosystem,and(4)landscape.

(b) Temporal—differentspansoftimefordifferenteventsandprocesses. (c) Social—examplescheme:(1)primaryinteraction—physicalhumancontact

withecosystem,(2)secondaryinteraction—emotional(laws,policies,regu-lation,votes,plans,assessments,andso forth), (3) tertiary—indirectandqualitative(values,interests,cultures,heritage,andsoforth).

(d) Spatial—manyhierarchiesbasedonnumerousattributes.

Possiblythegreatestissueintransformationrelatestoscale-dependenteffects.Thisisparticularlysoinhumanenvironmentinteractionswheregeographicalvaria-tioninhumanbehaviorandbiophysicalfactorsatdifferentscalesinteract.20Thisisalsoparticularlysowhencombiningbiophysicaldatawitheconomicandsocialdatawherepixelsandpolygonswithdiscretespatialpropertiesmustbecombinedwithindividualbehaviors and institutional arrangements thatoperate in amultivariatepseudospatialsphereofinfluence21andhavenonequivalentdescriptions.7Forexample,aregionmaybeboundbycertainrulesthatgovernthedegreeofeconomicsupportforcertainactivities.Thepotentialspatialdimensionsaretheregionboundary,buttheeffectivespatialpatterninsidetheregionisgovernedbyarangeofexistingconditions,humancharacteristicsandbehaviors,economicconditions,andbiophysicallimita-tions, someofwhichcanbedirectly suppliedas spatialdata layers, andsomeofwhichrequireamodelofpotentialinfluenceoreffecttocreateanindexoflikelihoodofadoptionorcompliance.Itispossibletoestablishequivalencerulesbetweenbio-physicalandsocial landscapeelementsusingstructural (e.g., speciescompositionandhydrologicalsystemversuspopulationcompositionandtransportationandcom-municationinfrastructure),functional(e.g.,patchconnectivityversuscommuting),andchange-based(e.g.,desertificationversusurbanization)22approaches.Itisalsopossibletoestablishdemographicscaleequivalencebetweenbiophysicalandsocialdomains using a spatial hierarchy based on individuals (e.g., plants and people),landscapes (e.g., watersheds and counties), physiographic regions (e.g., ecoregionandcensusregion),andextendedregions(e.g.,biomeandcontinent).22

Relationships of information derived within one scale category are relianton assumptions from others.19 In a more general sense, the modifiable areal unitproblem(MAUP),wherecorrelationsbetweenlayersvarywithdifferentreportingboundaries, requires excellent transformation methods, using finer scale data toinformthebroaderscaleanalysis,23andconstantawarenessofthepotentialproblemofunderstandingandmanagingpatterns,processes,relationships,andhumanactionsatseveralscales.19Multiagentsimulationapproaches24haveconsiderablebenefitsindealingwithindividualbehaviorsinurbananddenselypopulatedsystemproblems25aswellaslandcoverchangeproblems26andtechnologydiffusionandresourceusechange.27Theymayalsobeappliedtoexamineemergentpropertiesatthemacro-scalefromdifferentmicroscaleoutcomesandincorporatespatialmetrics.28

Thesecondmajorissueintransformationrelatestoameaningorquantifiablerela-tionship with an attribute that affects or contributes to assessment of the objective

42963_C002.indd 20 11/6/07 7:17:15 AM


oftheanalysis.Akeyelementhereis thefieldworkandanalyticalworktodevelopspecific and general quantitative, probabilistic, or qualitative relationships betweenpatternsandprocesses29,30,31thatcanbeusedeitherlocallyorgloballytoassignarankin termsof somemulticriteriaobjective.Laney32describes twoapproaches: studiesidentifying the land cover and change pattern, then seeking to develop amodel toexplainthesepatterns(pattern-ledanalysis)andstudiesthatdevelopatheorytoguidepatterncharacterization(process-ledanalysis).Bothapproachesmayhaveflaws,withpattern-ledanalysisbeinghighlydatadependentandabletoidentifyonlyprocessesassociatedwiththatdata,andprocess-ledanalysisdependentonthepriortheoreticalmodel,adherencetowhichmayprecludetreatmentofotherequallyvalidprocessesandpaths.Theultimateintegrationoftransformationandmeaningmightberepresentedbythe“syndrome”approach,33whereinalternativearchetypal,dynamic,coevolutionpatternsofcivilization-natureinteractionaredefined(e.g.,desertificationsyndrome).Thesesyndromesmightbecharacterizedbyhighlydevelopedcomposite indicatorsthatincorporatecomplexderivedspatiotemporalrelationshipsandpatterns.

2.4 tRAnsFoRMAtIon DoMAIns AnD MetHoDs

Theeffectivenessofamulticriteriaframeworkisprobablyproportionaltotheextenttowhichsystemelementsandinteractionsarecaptured.Representationoftimeintradi-tionalGISplatformsisverypoor,34whileimage-processingsystemsthathandletimeseriesofspatialdatalackthetoolsforextractionandsummaryofinformationfromthetimedomain.Moreaccessiblespace-timeanalyticalfunctionalityisneededtomakeawidevarietyoftransformationapproachesavailabletothoseotherthanexpertspatialanalystsandsignalprocessors.Thechallengeliesinacquiringdatainallofthepoten-tialresponsedomainsatasuitablescaleandwithacceptablequality.AlistofpossibleinformationdomainsisgiveninTable2.1alongwiththekindoftransformationissueinvolvedandsomepossiblemethods.Where individualsare involved,demographicinformationcoupledwithsurveysandunitsofcommunityaggregationformthebasisfortransformation—spatiallyintermsofthelocationofbehaviorsandrecordedpref-erencesinrelationtolandusepatternsandchanges,andtemporallyinthesensethattrajectoriesinopinionandbehaviorleadtolandusechange.Socialsystemsarereflex-ivelycomplex(i.e.,havingawarenessandpurpose).Therefore,withinasocialmulti-criteriaanalysiswithnonequivalentobserversandnonequivalentobservations,thereisaneedtodefineimportanceforactorsandrelevanceforthesystem.7Theactorsinsocialnetworksthatinfluencethelanduseoutcomemustbespatiallyrepresented,35butthereisachallengeincapturingthelinkbetweeninfluenceandbiophysicaloutcome.36

Atthelevelofsocialandeconomicstatistics,collectionunitsoftendeterminethenatureoftheanalysis.Socialindicatordatamaybeidiosyncraticatthelocalscale,haveincompletetimeseries,havedefinitionalchangesovertime,andhavemisalignedreportingboundaries.37ThisresultsinMAUP,ecologicalfallacy,expedientchoiceofstatistics,arbitrarychoiceofmeasures,anddifficultyinestablishinganycausalrela-tionships.37,38Transformationsarerequiredtosummarizetemporaltrendsandcyclesand todefinespatialpatternsandrelationshipsatafinerscale,whichmayhelp todistributetheinformationdownwardfromthecollectionunitinscaleinaspatiallyexplicitway.Dasymetricmappingcanbeusedonlytoassignpopulationstoremotely

42963_C002.indd 21 11/6/07 7:17:16 AM

22 Land Use Change

tAble 2.1transformation Domains for spatiotemporal Multicriteria Frameworks

Information Domain transformation Issue Methods

Individualbehaviorsandpreferences

Representationofindividualatresolutionofanalysis

Transformsurveyinformationintostatisticsandmetricsthatsummarizethetendenciesinthepopulationforthatspatialunit

Individualperceptions Representationofabstractconceptssuchasbeauty,degreeofspacecontamination,etc.

Uselandscapeimagemetrics,spatialdistancesandlandscapecontents

Institutionalarrangements,governmentregulations,andincentives

Representationoftheinfluenceorlikelihoodofadoptionorcompliance

Developprobabilitymodelsbasedonpriorsurveysofimpactandcreateprobabilitylayers

Economicvariables Relatingcollectionunittoanalysisunit

Self-organizationofspatialunits;temporaltrends,metrics,andtimeperiodsummaries

Socialstatistics,societalsystems,transportandsurveys

Conversiontoafactorlayer—attachingameaningandarank

Developprobabilitymodelsandpartialregressionmodelstoascribesomeofvariationintargetissuetothesocialfactors.Createfactorlayersbasedonthepercentagevariationdescribed,direction(+or–)andstrength(slope)oftrend

Climate Impact/responseanoutcomeofcomplextemporalsequencesandspatialpatterns

Developimpactthresholdandseveritylayersbasedonmultiplescenarioruns

Disturbances—fire,grazing,clearing,flooding,desertification,urbanization,abandonment

Representationofspatialextent,spatialgradient,timing,duration,impact,agents(i.e.,activeunitssuchasanimals)

Derivemetricsdescribingspatialandtemporalpatterns,harmonics,limits,responses,demographicsthatcanberankedintermsofthetargetissue

Landuse Representationofpersistenceandchangeatlevelofcovertype,species,managementpractice,seasonalmagnitude

Derivemetricsthatcapturepattern,change,persistence,sequence,andallquantitativepropertiesofthechangeinahierarchicalstructure

Bio/geochemicalprocess—hydrology,sedimentation,nutrients,gasexchange,emissions,consumption

Representationofprocessintermsofoutcomeaffectingorinfluencingtargetissue

Aggregated,averaged,summarizedandprobabilityconvertedoutputsfromprocessmodeling

42963_C002.indd 22 11/6/07 7:17:16 AM


sensedurbanclasses,andpopulationsurfacescanbecreatedbyassociatingthecountwithacentroidanddistributingitaccordingtoaweighteddistancefunction.38,39Therelationshipbetweenpeopleandtheirenvironmentiscapturedbycognitiveappraisalfromperceivedenvironmentalquality indicators.40 Indicatorsof residentialqualityand neighborhood attachment40 might be transformed into spatial properties byassigningproximityfunctionstoservices,assigningdistancemetricstoroadaccessand access to green space, ranking buildings for aesthetics and quality of humanenvironment,andmappingthesewithspatiallyexplicitviewabilityconstraints.

Climateprovidesanoverarchinginfluencethatisbothspatiallygeneralizedandlocally spatiallydependent, and it is a fundamentally time-dependentandcyclicalfactor.Here the transformations includespatialpatternsofmicroclimaticvariationandtemporaltrendsinclimatechange,metricsofseasonalcyclesandtrends,orvari-anceinextremes.Theremaininginformationdomainsarethemostspatiallyandtem-porallyinteractive,withbiogeochemicalprocessesinteractingwithlandusetypeandchangehighlyinfluencedbyhumanandotherdisturbances.Thesedomainsrequiremanyspatialandtemporalmetricsaswellashigherlevelmeasuresofsystemresponseintheformofoutputsofspatiallyandtemporallyexplicitmodels(e.g.,hydrology).

Somemethodsfortransformingcomplexspatialandtemporalpatterns,relation-ships,andsignalsaregiveninTable2.2.Theseareconsideredintermsofthegeneralspatialcontext,thespecificsocialnetworkdatawherespatialandnonspatialcogni-tivedomainsmix,40 thevisual contextwhereviewsandbeautyperceptions inter-mingle with functional and locational considerations,41 and the temporal contextwheremethodsfromnonspatialtimeseriesanalysiscomplementmethodsspecificallydeveloped for time series of satellite data. The spatial and temporal contexts arediscussedinmoredetailinthefollowingsections;however,theexamplelandscapecontextusedinthediscussionmustfirstbedescribed.

2.5 eXAMPle lAnDsCAPe ConteXt—AUstRAlIAn RAnGelAnDs

TheAustralianrangelandsprovideasuitablecombinationofspatialandtemporaldynamics and dependencies for illustration of issues surrounding transformationofspatialandtemporalsystempropertiesintoanMCAframework.Thissystemischaracterizedbyahierarchyofscaleswithinandacrosswhichinfluences,effects,relationships,andfunctionsoperate.Allofthescaledomainsofbiological,temporal,social,andspatialarerelevant.Thesystemisaffectedbyverylarge-scaleclimateandeconomicfactorsandverysmall-scalespatialdependenciesinhabitatsandland-scapefunction.Therangelandshavethefollowingcharacteristics:

1.Diversityinclimate,soils,andvegetationtypes(Figure2.1). 2.Heavilyutilizedbydomesticlivestock. 3.Substantiallyinfestedwithferalanimals. 4.Asignificantbiomassandsoilcarbonreserveandasourceofgreenhouse

gasemissionsthroughannualwildfire. 5.Systemprincipallylimitedbywateravailability. 6.Spatialinteractions,patterns,andgradientssubstantiallyrelatedtolandscape

scaleterrain–waterdynamicsandanthropogenicwatersupply(bores).

42963_C002.indd 23 11/6/07 7:17:17 AM

24 Land Use Change

7.Temporal dynamics heavily influenced by interaction between climate(watersupply),grazingandfire.

8.Meso-scalelandscapepropertiesstronglylinkedtooveralllandscapefunction,particularlyinrelationtowaterharvestingandconsequenthabitatdevelopment.

9.Significantsocialissuesthroughindigenousrightsandsacredsitesandsite-basedtourism.

10.Managementofthelandscapeisinfluencedbyexogenoustemporalvaria-tionincostoffinanceandinputs,tradebarriersandrestrictions,priceofcommodities,specificallybeefcattle,andchangesinfamilystructuresandruralemployment.

tAble 2.2Methods for exploration and transformation of Complex spatial and temporal Patterns, Relationships, and signals

spatial23

Convolutionfiltering(movingwindoworkernel)containingfunctionsfromsimplestatisticstotexturalindexestocomplexregressiontospatialautocorrelation

Distancemeasuresinspatialneighborhoods—associationofpatch,gapandshapewithsocioeconomicchangefactors29

Cost–distancegenerationofuserandpurposedefinedanalysisunitsGeographicallyweightedregression49toovercomenonstationarity,spatialdependencies,andnonlinearspatialdistributions,allowingclassificationofsystemparametersbyalearningalgorithm—self-organization

spatial/social networks35–50

Resilience,fastandslowadjustment,perturbation,catastrophe,turbulence,andchaosmodels51

Bioecologicalmodels—analysisofdynamicphenomenaofcompetition-complementarity-substitution(networkasaniche);sociallandscapeanalysisinlandscapeecology22

Neuralnetworks—noteasilyinterpretablefromeconomicviewEvolutionaryalgorithms—geneticalgorithmswithbinarystrings;evolutionaryalgorithmswithcontinuoussettingandfloatingpointvalues

Visual41

Characteristicfeatures—lower-upperfeaturerelationships;contourblockdrawings;imagetextures;contoursandhorizon;spatialrelationsofspacesandelements;proportionsoflandscapezonesinview;hierarchicalproperties;typologyoffringes

Spatialdistancemeasures—viewtexture;intrusionintoskylineandlandscapeline;relativestructuralcomplexity;relativeproportions;distance–sizerelationships

Sensitivity—functionaldistanceinlandscapes;structuraldistancestobekeptfree

temporal

Traditionaltimeseriesanalysis12,14,52,53—trends,cycles,seasonality,lags,phase,irregularity,smoothing,differencing,autocorrelations,spectralanalysis

Curvemetrics43,54—limits,amplitude,periodicity,timings,areas,slopes,trajectories55;phenologySignalprocessing—Fouriertransforms56;Wavelettransforms15,44

Principalcomponentanalysisoftimeseries44

Complexbio-socioeconomiccycles(e.g.,Kondratieffwaves57);syndromesofchange33

42963_C002.indd 24 11/6/07 7:17:17 AM


Thesystemrepresentsatypeofexamplewherehumandemographicsarenotamajor factorsince largepastoral leasesareessentiallyunpopulatedexcept for thestationhomesteadandassociatedbuildings.Humaninfluenceinthisenvironmentisprovidedthroughmanagement,whichreachesoutfromthehomesteadtoinfluenceverylargetractsofland.Hence,superficiallyitmightbedifficulttodrawmethod-ological parallelswith themany coupledhuman environment systems worldwideandhighhumanpopulationdensities.However, in this system,demographicsarestill important since themajor influential population is thatofdomesticatedbeefcattle,withancillaryinfluencefromferalanimalpopulations.Theyareindividualeconomicunitswithcostsassociatedwithparasiteanddiseasecontrolandhumanhandlingandvalueintermsoffoodandbreedingpotential.Thedecision-makingframework forcattle ismuch lesscomplex than forhumans;cattle requirewater,feed,shade,andsocializationandwilloptimizetheirbehaviorwithinthisresponsespace. Nevertheless, they influence and respond to spatial and temporal patterns,and,therefore,thissystemcanstillprovideusefulmethodologicalinsights.

2.5.1 SPATIAL PATTERNS AND RELATIONSHIPS

Thespatialinterrelationshipsinthisrangelandsystemcanbeillustratedbyastylizedlandscapecontainingartificialwaterpointssurroundedbypiospheresofinfluencebygrazinganimalsuponthevegetationuptoadistancelimit(Figure2.1).Thesewaterpointsoccurwithinfencedpaddocks,partsofwhichareinaccessibletostocksincethey are outside the water access limit. The paddocks also contain different landcovertypeswithdifferenthabitatsuitability,firesusceptibility,andlivestockcarryingcapacities.Thelandscapehasrockyareas,areaswiththickshrublandinaccessibletostock,swampyandsalineareaswithlowproductivity,andanaboriginalsacredsite.Theareaalsohasanaestheticcomponentwithaviewpointandrestarealocatedonamajorroad,withbasicpicnicfacilitiesoutsidethemappedextent.Themajorspatial

Water point piosphere of grazing intensity

Fenced paddocks

Heavily thickened woodland with shrubs

Poor, light soil

Inaccessible rocky outcrop

Swampy area with unpalatable plants

Saline scald area

Elevation contours

Sacred aboriginal site

FIGURe 2.1 TheconceptofgrazingpiospheresinteractingwithlandscapestructuretocreatespatiallyandtemporallydependentresponsezonesinAustralianrangelands.Thesearemoreprevalentwhererainfallislessreliable,paddocksaresmaller,andstockingpressureishigher.

42963_C002.indd 25 11/6/07 7:17:20 AM

26 Land Use Change

gradients in this landscape are createdby the effect of grazingonvegetation andhabitat,theconnectivitybetweenhabitats,thestructureinrelationtoshelter,waterharvestandstockaccess,andtheappearanceofthelandscapefromaspecificdirec-tionandangleofview.

Inordertocapturespatialattributes,alevelofspatialpatternreportingmustbedefined,andthislevelofaggregationmustbecompatiblewiththeresolutionofotherdataintheanalysis.Thescaleofaggregationmightrelatetosomefunctionaldistanceandsphereofinfluenceinthelandscape,andpatternextractionmightbeundertakenforanumberofdifferentaggregationunits,42anestedsetofpatchscales,30inordertospecificallycapturetheinfluenceoflandscapestructurefromdifferentelementsofthesystemsuchasbirdhabitat,cattlegrazingbehavior,scaleofmicrotopography,andsoforth.

2.5.2 TEMPORAL PATTERNS AND INFLUENCES

The temporal behaviors of, and influences on, this rangeland system could bedescribedbya timeseriesofweatherandsatellitedata,whichrecordssequencesofdetectablelandcoverchangeandvegetationstate,aswellasderivedmeasuresofsystemfunctionintegratedthroughmodels.Amonthlytimeseriesofnetecosystemcarbonexchange(Barrett,personalcommunication)providesanexampledatasetfor illustrationofapproaches todisaggregationanddecompositionof signals intomeaningfulindexes(Figure2.2).Aseriesofseasonallybasedsystemresponsespro-videthebasisforextractionof:

1.Curvemetricsthatdescribethetiming,duration,magnitudeandperiodicityoftheresponse43

2.Acumulativeaggregateofthenetsystembehaviorthroughtime 3.Trendinsignalfromwaveletorothertransforms15,44

4.Temporalautocorrelationtoseehowstrongthe“memory”isinthesystem—astrongmemoryindicatesmoreregularcyclicalbehavior

5.PowerspectrumandFourier transformsonoriginaldataandfirstdiffer-ences or first derivatives to detect major cyclical patterns—in this caseoccurringatabout22,44,and66months

6.Cumulative probability curves to identify the relative behavior for someproportionofcases(Figure2.2)

Thesemetricsandmeasuresoftimeseriesattributescanbederivedspatiallyandconvertedtosingleorpartialcomponentindicatorsofsystemproperties.

Thetemporalinfluencesarealsorepresentedbynonbiophysicaltimeseriessuchas livestock numbers, climate cycle indexes, prices and costs, and human activitymeasures(Figure2.3).Thesedatamayonlybeavailableatacoarselevelofspatialresolution,suchascattlenumbersfromtheagriculturalcensus,orindividualbehaviorsfromsocialsurveyswithlimitedsamples.Alternatively,theymaybeglobalvariablessuchascattleprices,interestrates,andclimateindexessuchasthesouthernoscilla-tionindex(SOI).Inthesecases,ameansmustbefoundtoapplythesespatiallyviasomefilteringlayerthatassignstheattributesonlytothosepixelswheretheinfluenceoccurs,ortothosepixelsnotconstrainedbyotherfactors.

42963_C002.indd 26 11/6/07 7:17:21 AM


–2

0

–4

0

–0.0 2

0.00

0.02

–0.4

0.0

0.4

0.8

0.0

0.2

–0.2

0.0

0

40

80

NEP 1981 – 2000

NEP

Cum

ulat

ive N

EPRe

spon

seSi

gnal

Aut

ocor

rela

tion

Pow

erSp

ectr

um

Pow

erSp

ectr

um

Firs

tD

iffer

ence

Cum

ulat

ive

perc

enta

ge

Normal score

Wavelet transformTrend line

Fourier transform

Running cumulation

Time in months Time intervals

Metrics

Amp Integral

IntervalPeriod

0 12T1 T24 8

Max

Min

0.5 Amp

0.2

0.0

0 24 48 72 96 120 144 168 192 216 240

0.00000

0.00008

0.00016

–2 –1 0 1 2 3 4

0.00000

0.00010

FIGURe 2.2 Timeseriesapproaches toextractingsignalsummary indicators.Atimeseriesofnetecosystemproductivityindicatesthebasepotentialforcarbonfixation.Thismaybetrans-formedinto indicatorsbycalculatingmetrics, includingarunningintegral,extractingthefre-quencyofcyclicpatternsusingfastFouriertransform(FFT)orpowerspectrumsonoriginaldataorfirstdifferencesandderivatives,definingdirectionoftemporalchangethroughtrendanalysisorwavelettransforms,andestimatinglikelihoodofvariouslevelsthoughcumulativeprobability.

42963_C002.indd 27 11/6/07 7:17:22 AM

28 Land Use Change

–20

0

20

SOI

0

2

0

4

8

12

Pow

er

0

40

80

120

0

200

400

600

0

0

20

40

60

1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000

2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32

1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000

1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000

1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000

Cattl

e No.

(mill

ion)

1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 20008

9

10

11

12

Interest paymentsHandling/marketingWages – hired hands

Age – ownerAge – spouseHours worked – ownerHours worked – spouse

Cattl

e ($/

head

)

Cost

s ($)

40000

80000

120000

Age

/Hou

rs

IPO

FIGURe 2.3 Nonbiophysicaltimeseriesalsoprovidepotentialindicatorsbutmaynotbespatially explicit at the required scale.Theseneed tobe transformed into indicators suchastrendsindemographics(e.g.,cattle),patternsofclimateandfrequencyofoccurrenceofcertainclimate types, trends inpricesforcommodities, trends incostsofproduction,andtrendsinhumanactivitiespotentiallyaffectingmanagementandeconomicoutcomes.

42963_C002.indd 28 11/6/07 7:17:24 AM


Thespatialextentofdiscreteandconsistenttemporalpatternsmaybedefinedby,forexample,aprinciplecomponentsanalysis(PCA)onthetimeseriesandsubsequentclassification (Figure 2.4a). This reveals distinct temporal patterns that representa regional summary (Figure2.4b; the temporal net ecosystemproductivity (NEP)signalforoneoftheseclassesisusedinFigure2.2).Bycontrast,timeseriesattributesmaybecalculatedonapixel-by-pixelbasis,andthedata,suchasstandarddeviationinNEP,aremappedtoprovideacontinuousfactorlayer.Theanalysisofthetrendsinthetimeseries(Figure2.2)mayprovideinformationaboutmajorperiodsofdifferingbehaviors,suchasthenetlosscarbonthroughoutnorthernAustraliabetween1985and1993,versusthenetgainincarbonbetween1994and2000(Figure4.2c).

2.5.3 DATA AND INFORMATION: SCALE OF REPRESENTATION

Movingdownscaletotheregionofinterestforanalysis,theVictoriaRiverDistrict(VRD)inthenorthernAustralianrangelandsprovidesagoodbasisforassessmentof

(a) (b)

(c) (d)

PCA classes12345678910111213141516171819202122

St. Dev. NEP0.004 – 0.0270.027 – 0.050.05 – 0.0730.073 – 0.0960.096 – 0.1180.118 – 0.1410.141 – 0.1640.164 – 0.1870.187 – 0.210.21 – 0.233

NEP 8593< –2–2 – –1–1 ––0.5–0.5 – –0.2–0.2 – 00 – 0.20.2 – 0.50.5 – 1> 1

NEP 9400< –2–2 – –1–1 ––0.5–0.5 – –0.2–0.2 – 00 – 0.20.2 – 0.50.5 – 1> 1

FIGURe 2.4 (See color insert following p. 132.) (a) The spatial pattern of temporalsignalsmaybegroupedbyapplyingprincipalcomponentsanalysis to the timeseriesandcreatingaclassificationbasedonthemajorprincipalcomponents.(b)Thetemporalmetricsmaybecalculatedonthespatialtimesseriestocreatemapsof,forexample,standarddevia-tionofnetecosystemproductivity(NEP).(c)Arunningintegration,trend,orwaveletanalysismaydefineperiodsofdistinctbehaviorinthetimeseriesthatcanthenbesummarizedbymetricssuchasanintegralofNEPforperiodsofdeclineandincrease.Shownfor1985–1993and1994–2000here.

42963_C002.indd 29 11/6/07 7:17:27 AM

30 Land Use Change

datascalesandtransformationthroughspatialfiltering(Figure2.5).TheregionhasthreeofthePCAclassesgiveninFigure2.4a.ThetemporalsignalandassociatedmetricsandtimeseriesattributescouldbebroadlyassignedtoallareaswithintheclasszoneintheVRD.However,theNEPdatarepresenttheresponseofthesystemundisturbedbylivestock.Therefore,intheabsenceofspecificestimatesthattakethespatiallyvariableimpactofgrazingintensityaroundwaterpointsintoaccount,onecouldapplyanarbitraryscalingofeffectonNEPthatvariesfromusingthesuppliedsignalsforungrazedareas,tocompletelysuppressingtheaccumulationsignalatthewaterpoint,wherestockpressureishighest.Verybroadscaleestimatesofprofitperhectareatfullequityprovideanindicationofprofitability.Minepresenceisindicatedbyacountforeach1kmpixel—adecisionmayneedtobemadeaboutabufferingruleforradiusofdisturbance.Thenumberofthreatenedbirdsisderivedfromaverycoarseresolutiondataset.However, ifany information isavailableon thehabitatforthesebirds, then,forexample,anindexofpotential threattogrounddwellingbirdscouldbecreatedwithveryfinespatialresolutionusingarulegoverningdegreeofdisturbancewithdistancefromcattlewateringpoints.Completelyaspatial,butimportant,systemattributesandmetricsmaybedownscaledtoappropriateresolu-tionifrelationshipstospatialdataatanappropriateresolutionareknownorcanbederived.Themajorchallengearises inascribingcausal relationshipsandareasofinteresttohumanpopulationcentersbasedonsocialstatisticsaboutactivitiesandpreferencesofhumans.However,certainkeyvariablessuchasbusinessenterprisedebttoequityratiosmaybeimportant.Ifspatialdataareofsufficientqualityandeffectsofdisturbanceonkeyenvironmentalmeasuresareknown,thenaspatialeco-nomicandsocialdatamaybeusedalongwithbiophysicaldatatoascribeintegrativeindexesof economicbenefit, costofdegradation, and socialbenefit to individuallandscapepixelshavingparticularsuitesofbiophysicalattributes.

2.5.4 SOME SPATIOTEMPORAL INPUTS TO A RANgELAND MCA

Therangelandexampleprovidesaveryspecificopportunitytoelucidatetheetiologyofspatialandtemporaltransformationstosummarizecomplexsystemresponsesandbehaviorsinmulticriteriaanalysis.Ifwetaketheeffectofgrazingonvegetationcondi-tionasanexampleofacomplexbiophysicalprocessinfluencedbyhumanmanagement,inordertocapturetheelementsoftheconditionofaparticularpixelonemightneed:

1.Distancetowaterpoint(seeFigure2.1)—simpledistancemetric. 2.Valueofaveragegrazingpressure—complexfunctionalcalculationbasedon

distancefunctionsdescribingthecost–distancerelationshipbetweendistancefromwaterpointandlivestocktendencytotraveldistancefromwater.Thenthevalueofaveragegrazingpressureiscalculatedfromtheratioofaveragedensityoflivestockunitstothenominalsafecarryingcapacityofthevegeta-tiontype.Thisisrelatedtotime-useanalysis,45wheretheperiodoftimethatanindividualunitisinaparticularspatiallocationisimportant.

3.Aproximityfunctiontodenseshrubbyvegetation—modifies(2)above,sincegrazingpatternmaybeperturbedbyanimalbehaviorwithrespecttoshelter.

4.Averagemaximumseasonalbiomassforthatpixel—simpletemporalcurvemetric.

42963_C002.indd 30 11/6/07 7:17:27 AM


(a) (b)

(c) (d)

FIGURe 2.5 (See color insert following p. 132.) Temporalsignalsareusuallybasedonbiophysicalorhumanphenomenathatoperateatalargescale(e.g.,climate,interestrates).Demographicchangesatafinescalemayhavescalelimitationsduetolevelofaggregationinreporting.Temporalsignalsandindicatorsarefilteredbyspatialvariation.TheVictoriaRiverDistrictintheNorthernTerritoryofAustraliaishighlyproductive.(a)Cattlearedis-tributedoffreehold-leaseholdlandbutconfinedbywaterpoints.(b)Bothproductivityandecologicalimpactvarywithvegetationtype,whichisassociatedwithsoils,topography,andrainfallgradient.(c)Costsarelowandenterprisesareprofitablebuttheincrementissmallonaperhectarebasis.(d)Miningwithmajorphysicaldisturbanceoccurssporadicallyacrossthearea.Therearethreatenedbirdspeciesintheregionandthesemaybegroundnestingandimpactedbygrazing.

42963_C002.indd 31 11/6/07 7:17:29 AM

32 Land Use Change

5.Average amplitude for annual biomass for that pixel (max minus min)temporalcurvemetric.

6.Timeofhalfmaximumseasonalbiomassforthatpixel—simpletemporalmetricforstartofperiodofgreenfeedavailability.

Thesedatamightbeusedtodefineanindexofanimalimpactforeachpixelthatintegrates spatial and temporal relationships, trends, and influences.Thiskindofcombinationofspatialandtemporalrelationshipsmaybeusedtoconstructspatiallyexplicit indexes forotherelementsof thesystemsuchas landscapefunction,bio-diversityimpact,andsocioeconomicbenefit(Table2.3).Thetemporalcomponentisfilteredonthebasisofspatialrelevance(i.e.,grazedareasarerelevantbutungrazedareasarenot),exceptwherethetemporaltrendofresponsehasasphereofinfluencebeyondthelocalpixelandisthengovernedbyaspatialfunctionforrelativeeffectandadjustedbyanyspatialconstraints.

2.6 A FRAMewoRk FoR A MUltICoMPonent AnAlysIs wItH MCA

Finally, a broad framework for application of MCA to assessment of ecosystemservicefromarangelandenvironmentisdescribed(Figure2.6).InthisframeworkMCAisusedtoprovideassessmentofcurrentecosystemservicelevels,andthentoassessstrategiesforimprovingecosystemservicesthroughmanagementchange.ThespatiallyexplicitecosystemserviceratingforanareacouldbeconstructedusingthesuiteofindividualandcompositeindicatorsinaMCAenvironmentsuchasMCAS-S(multicriteriaanalysisshell–spatial46;Figure2.7).Thisframeworkcouldmakeuseofspatialandtemporalmeasuresandmetricsaspartofthesuiteofindicatorsusedtodefinetheconditionofthelandscapeintermsofarangeofusesandfunctions.TheapproachdescribedinFigure2.6usesstateandtransitionmodelsofvegetationcondition.10,11Therangelandlandscapeisclassifiedintostatesbasedondisturbanceof original natural vegetation.The states are described in termsof structure andfoliagecoverandtypeofvegetation.47

Theecosystemservicefromthevegetationstatesisdescribedintermsofanumberofthemes:biodiversity,landscapefunction,waterharvesting,carbonstock,grazingpotential, indigenous utility, economic return, and aesthetic value. Each of thesethemesismadeupofasetofindicatorlayersdescribingattributes.Theseattributescouldbeindividualmeasuressuchasnumberofthreatenedbirds,carbonbiomass,orproximitytoaboriginalsacredsites.Alternativelytheycouldbecompositeindicatorsbasedonaggregationofindividualmeasuresincomplexspatiotemporalrelationshipstogiveanimalimpactperpixel.Theseindividualandcompositeindicatorsmaybecombinedbyvariousmethodsintoasingleindexofecosystemserviceforthattheme.Overallecosystemservicefromthatlandscapepixelisrepresentedbythecombina-tionoftheindividualthemeindicatorsintooneoverallindex.Theoverallecosystemserviceslevelcanbeimprovedbymovingthethemeindicatorstohigherlevels.Eachtheme can be improved by applying a number of strategies—individual strategiesmayimprovemorethanonetheme,butmayhaveanegativeeffectinotherthemes.Forexample,economicpotentialmaybeincreasedbyincreasingstockingrates,butthismayaffectvariouselementsoflandscapefunction.

42963_C002.indd 32 11/6/07 7:17:29 AM


tAb

le 2

.3D

eriv

ed F

acto

r la

yers

tha

t In

tegr

ate

spat

ial a

nd t

empo

ral t

rend

s w

ith

Func

tion

and

eco

syst

em o

utco

me

Fact

orIn

put

data

spat

ial

Inpu

t da

tate

mpo

ral

Inte

grat

ion

Ani

mal

impa

ct

per

pixe

lB

uffe

red

wat

erp

oint

sw

ith

anim

ald

istr

ibut

ion

assi

gned

by

expo

nent

ial

dist

ance

fun

ctio

n;m

apo

fin

acce

ssib

leo

run

attr

activ

eve

geta

tion

Prob

abili

tyo

fan

imal

ut

iliza

tion

ina

nyti

me

peri

od—

from

dis

tanc

efu

nctio

nof

fre

quen

cyo

fvi

sita

tion

mod

ified

by

spat

iala

cces

sibi

lity

and

divi

ded

byf

eed

requ

irem

ent

for

spec

ified

wei

ghtg

ain

Mod

eled

tim

ese

ries

of

gras

slan

dgr

owth

;es

timat

eof

fee

dre

quir

emen

tper

uni

tof

time;

est

imat

eof

saf

eut

iliza

tion

rate

for

ve

geta

tion

type

s

Lik

elih

ood

that

ani

mal

de

man

dw

illn

ote

xcee

dsa

feu

tiliz

atio

nra

tein

an

ype

riod

—fr

om

hist

oric

alti

me

seri

eso

ffe

eda

ccum

ulat

ion

in

grow

ing

seas

ons

Com

bine

d,s

cale

din

dex

ofr

elat

ive

anim

alim

pact

on

any

pios

pher

epi

xeld

ueto

spa

tialc

ontr

olb

yw

ater

acc

ess

and

tem

pora

lpr

obab

ility

of

over

graz

ing

Lan

dsca

pe

func

tion

Ele

vatio

n,s

lope

,hig

hre

solu

tion

imag

ery

defin

ing

tree

san

dbu

shes

;de

rive

dla

ndsc

ape

drai

nage

pat

tern

and

wat

er

accu

mul

atio

n

Text

ural

ana

lysi

sof

pat

tern

of

wat

erh

arve

stc

onve

rted

to

mag

nitu

dea

ndd

ensi

ty

mea

sure

s

Tim

ese

ries

of

rain

fall

and

evap

orat

ion;

mod

eled

de

man

dfo

rw

ater

by

vege

tatio

nty

pes

Tim

ing,

fre

quen

cy,a

nd

peri

odic

ityo

fw

ater

flo

ws

and

prob

abili

tyo

fw

ater

har

vest

suf

ficie

nt

tos

usta

ins

tabl

esy

stem

Com

bine

d,s

cale

din

dex

ofw

ater

ha

rves

tfun

ctio

nfo

ran

ypi

xeld

ue

tos

patia

lpat

tern

of

land

scap

eel

emen

tsa

ndte

mpo

ral

prob

abili

tyo

fsu

cces

sful

cap

ture

if

har

vest

able

rai

nfal

leve

nts

occu

r

Soci

oeco

nom

ic

bene

fitp

er

graz

edu

nit

Liv

ew

eigh

tgai

npe

ran

imal

;ani

mal

di

stri

butio

n;v

alue

of

live

wei

ght(

pric

epe

ran

imal

);

cost

of

prod

uctio

n(c

ost

per

anim

al)

Perc

enta

geo

fto

tall

ive

wei

ghtg

ain

attr

ibut

able

to

each

gra

zed

pixe

lthe

nco

st,

pric

ean

dpr

ofitp

erg

raze

dpi

xel;

spat

ialc

orre

latio

nw

ithv

eget

atio

nty

pe.

Tim

ese

ries

of

anim

al

pric

es,c

ostc

hang

es,

anim

aln

umbe

rs(

stoc

king

ra

tes)

,fee

dav

aila

bilit

y,

live

wei

ghtg

ain

per

unit

time.

Len

gth

ofp

erio

dso

fin

crea

seo

rde

clin

ein

te

rms

oftr

ade

and

in

profi

tper

uni

tof

graz

eda

rea

Com

bine

d,s

cale

din

dex

of

econ

omic

ret

urn

per

unit

area

due

to

spa

tiald

istr

ibut

ion

ofa

nim

als

and

feed

,and

tem

pora

lcha

nges

in

pri

ces

and

cost

spe

run

itar

ea

Bio

dive

rsity

im

pact

per

pix

elD

istr

ibut

ion

ofe

ndan

gere

dsp

ecie

s;a

ssoc

iatio

nof

sp

ecie

sw

ithv

eget

atio

n,

land

scap

efe

atur

es

Con

nect

ivity

,pat

tern

of

habi

tat;

focu

san

dex

tent

of

dist

urba

nce

from

ani

mal

im

pact

per

pix

el

Tim

ese

ries

of

spec

ies

num

bers

;clim

ate,

fee

dsu

pplie

s,s

tock

num

bers

,pr

edat

orn

umbe

rs

Peri

odic

ityo

fpo

pula

tion

fluct

uatio

ns;t

imin

gof

lo

ws

inp

opul

atio

n;

tren

dsin

pop

ulat

ion

Com

bine

d,s

cale

din

dex

of

prob

abili

tyo

fne

gativ

ebi

odiv

ersi

tyim

pact

fro

mr

educ

ed

conn

ectiv

itya

ndd

istu

rban

ce

duri

ngp

opul

atio

nlo

ws

42963_C002.indd 33 11/6/07 7:17:30 AM

34 Land Use Change

Strategies for improving ecosystem service within a land cover state may beassessed by a mixture of spatial and nonspatial MCA where drivers of changeswithineachstatearedefined,andfeasibilityoreffectivenessofadjustmenttothesedriversisassessedusingarangeofspatialattributes.AnexampleofthecombinationofavarietyofdatalayersbyarelativelysimplemethodtoformasingleindexofpotentialproductivityfromgrazinginAustralianrangelandsisgiveninFigure2.7.Thiscouldrepresentasinglethemewithinafullecosystemservices’assessment.Thesethemesmaybecomparedbytwo-wayanalysis18,46ormaybecombinedandanoverallconditionassessedusingspiderdiagrams/radarplots.Somedetailedcontex-tualanalysisofthiskindofecosystemserviceassessmentincorporatingtheuseofspatialdataandspiderplotsisprovidedbyDefriesetal.48

2.7 ConClUsIons

Transformation of complex spatial and temporal patterns and behaviors intosimpledata forms isan importantenabling technicalcapability required inordertomaximize the informationcontentandeffectivenessofdigitaldecisionsupportsystems.Anessentialframeworkofanalyticalcapabilityinvolvesseamlessaccesstotoolsandmethodsforspatialanalysisandextractionofspatialpatternsandinter-relationships,toolsfortimeseriesanalysisonspatialdata,toolsfordevelopingsimplemodelsanddefiningrelationshipsandcausalityacrossspatialscales,andgenerationof uncertainty and error measures and incorporation in analysis along with basedata.AschemafortransformationofspatiotemporalcomplexityintoindicatorsforMCA is given in Figure2.8. The schema uses rangeland and urban examples toillustratethekindsofspecificmeasuresandderivedinformationrequired.Process

FIGURe 2.6 Aframeworkforapplicationofmulticriteriaanalysisofecosystemservicestotherangelandenvironment.

VegetationStructure, Growthforms, Foliage cover

Vegetation stateTypes I to VII

Transitions betweenvegetation states–also changes ES

Vegetationzones – atany scale

Land useBiodiversity

Water quality

Water yieldAgriculturalproduction

Timber yield

Amenity uses

Carbonsequestration

Landmanagementpractices

Ecosystem Service (ES)states with categoryattributesBiodiversityWater qualityWater quantityFood and FiberAmenityWood productsCarbon stock

Transitions betweenES states withinvegetation states

Land usechange

Land managementchange

Climatechange

Drivers of change in ESS and vegetation states

Attributes of drivers of change, e.g., profitability, cost, social impact, feasibility,likelihood, effectiveness, permanence, negative effects, etc.

Water usechange

Climatevariability

State andTransitionModel

State andTransitionModel

MCA to define ecosystemservice states and indexes

MCA to assessmerits of changestrategies andattributes ofoutcomes

42963_C002.indd 34 11/6/07 7:17:32 AM


Temporal variation, e.g., climate; economic conditions

Space with gradients andobjects

e.g., landscape, city

Spatial relationshipsand patterns

e.g., terrain/soil;buildings/open space

Mobile agent makingdecisions

e.g., cattle/humans

Spatial relationshipsand patterns

e.g., feed and waterrequirement; play areas

Temporal dependenciesvariation in feed supplyprobability of demand > supplyfrequency of droughttravel timerecreation time

Spatial impactsgrazing/utilization pressuretrampling/trackingpre–condition for woodyingress ease of businessinteraction attractivenessfor tourists

Conceptual Basis for Transformation

Interrelation andcombination

Measures of system properties and processes – Indicators

Spatial constraints, e.g., distance from water or business clients and servicesTemporal constraints, e.g., time spent in location or decision/consulting time

Variation inavailable resources

function,appearance density,

range

texture,association

quantity,type

FIGURe 2.8 Schemafor transformationofspatiotemporal information into indicatorsofsystempropertiesformulticriteriaanalysisevaluations.

Potential productivity for grazingRainfall reliabilityRainfall rel. (w/sp)

Rainfall rel. (ann)

Forage potential

Growth Foliage proj cover Soil nutrient

Accessibility (ARIA)

Soil PSoil NSoil carbonNDVI meanNPP mean soil_fert

FIGURe 2.7 (See color insert following p. 132.) Cognitivemappinginterfacesuitableforcombinationofdiversespatiotemporalmetrics,indices,anddatalayersdescribingmean-ingfulpropertiesofasystemunderanalysis.Thisexampleshowstheconstructionofacom-positeindextorepresentpotentialgrazingproductivityfromrangelands.46,58

42963_C002.indd 35 11/6/07 7:17:34 AM

36 Land Use Change

andothercomplexmodelsoperateexternallytothisframeworkandprovidespatiallayersandtemporalsignalsrepresentinganintegrationofcomplexprocessesforusewithintheframework.

Theexampleusedinthischapterlargelyaddressesthebiophysicaldomainsincerangelandshavelowhumanpopulationsandindividualshavecontroloverlargelandareas.Thismaymean that sociological factors play a smaller relative role in themanagementofthesystemexceptwhenconcernedwithindigenousrightsandissues.However,theanalyticalproblemsareuniversal—forexample,thereisstillaneedtodeterminethelevelandimportanceofchangesinspousalworkcontributionstotheoperationofcattlestationsandtothefunctioningofisolatedfamilieseveniftheseeffectsareverysmall.However,throughtheprincipleofequivalenceofbiophysicalandsocial landscapeelements, cattledemographics, likehumandemographics incities,operateasamajordriverinthesystem,andassigningrulestobehavioranddefining spatial relationships and temporal trends and influences assume criticalimportance.Thischapterhasprovidedsomeexamplesofanalyticalmethodsand

(a)

(b)

FIGURe 2.9 VegetationtypesinAustralianrangelands.(a)Northernsavannawoodlandsmayhaveexcellentherbaceouscoverduetolargepaddocks,lowerstockdensityandreliableseasonalrainfall(PhotoM.J.Hill).(b)Saltbushplainsmaybeingoodcondition(asshownhere)butcanbesusceptibletodegradationwithdroughtsandovergrazing(PhotocourtesyofR.Lesslie).

42963_C002.indd 36 11/6/07 7:17:36 AM


approaches needed. The case studies to follow will examine spatial methods ingreaterdetailforarangeofgeographicallydistinctsystemsandproblems.

2.8 ACknowleDGMents

TheMCASsoftwareinterfaceillustratedinthischapter(Figure2.7)hasbeendevel-opedbyRobLesslieasteamleader,AndrewBarryasprogrammer,andtheauthorastechnicaladvisor.Iamgratefultomycolleaguesforpermissiontoreproduceworkfromourjointeffortsinthissoleauthorchapter.IthankDamianBarrettforaccesstocontinentalcarboncyclemodeloutputsusedinFigures2.2and2.4.

ReFeRenCes

1. Grant,W.E.,Peterson,T.R.,andPeterson,M.J.Quantitativemodellingofcouplednatural/human systems: simulation of societal constraints on environmental actiondrawingonLuhmann’ssocialtheory.Ecological Modelling158,143–165,2002.

2. Jankowski,P.Integratinggeographicalinformationsystemsandmulti-criteriadecisionmaking methods. International Journal of Geographical Information Systems 9,251–273,1995.

3. Varma,V.K.,Ferguson,I.,andWild,I.Decisionsupportsystemforthesustainableforestmanagement.Forest Ecology and Management128,49–55,2000.

4. Store, R., and Jokimaki, J. A GIS-based multi-scale approach to habitat suitabilitymodelling.Ecological Modelling169,1–15,2003.

5. Mazzetto,F.,andBonera,R.MEACROS:Atoolformulti-criteriaevaluationofalter-nativecroppingsystems.European Journal of Agronomy18,379–387,2003.

6. Giupponi, C. et al. MULINO-DSS: A computer tool for sustainable use of waterresourcesatthecatchmentscale.Mathematics and Computers in Simulation64,13–24,2004.

7. Munda,G.Socialmulti-criteriaevaluationforurbansustainabilitypolicies.Land Use Policy23,86–94,2006.

8. Hill, M. J. et al. Multi-criteria decision analysis in spatial decision support: TheASSESSanalytichierarchyprocessandtheroleofquantitativemethodsandspatiallyexplicitanalysis.Environmental Modelling and Software20,955–976,2005.

9. Hill,M.J.etal.VegetationstatechangeandconsequentcarbondynamicsinsavannawoodlandsofAustraliainresponsetograzing,droughtandfire:Ascenarioapproachusing113yearsofsyntheticannualfireandgrasslandgrowth.Australian Journal of Botany53,715–739,2005.

10. Hill,M.J.etal.AnalysisofsoilcarbonoutcomesfrominteractionbetweenclimateandgrazingpressureinAustralianrangelandsusingRange-ASSESS.Environmental Modelling and Software21,779–801,2006.

11. Smith,S.W.The Scientist and Engineer’s Guide to Digital Signal Processing,2nded.CaliforniaTechnicalPublishing,SanDiego,Calif.643pp,1999.

12. Hamilton,J.D.Time Series Analysis. PrincetonUniversityPress,N.J.820pp,1994. 13. Gourieroux, C., and Monford, A. Time Series and Dynamic Models. Cambridge

UniversityPress,UK.668pp,2000. 14. Brockwell,P., andDavis,R. Introduction to Time Series and Forecasting.Springer

Verlag,Netherlands.434pp,2002. 15. Percival, D.B.,andWalden,A.T.Wavelet Methods for Time Series Analysis (WMTSA).

CambridgeUniversityPress,Cambridge.594pp,2000.

42963_C002.indd 37 11/6/07 7:17:37 AM

38 Land Use Change

16. Siebenhuner,B.,andBarth,V.Theroleofcomputermodellinginparticipatoryinte-gratedassessments.Environmental Impact Assessment Review25,391–409,2004.

17. Walker,J.etal.Assessment of Catchment Condition.CSIROLandandWater,Canberra.36pp,2002.

18. Hill,M.J.etal.Multi-criteriaassessmentof tensions in resourceuseatcontinentalscale:aproofofconceptwithAustralianrangelands.Environmental Management 37,712–731,2006.

19. Sexton,W.T.,Dull,C.W.,andSzaro,R.C.Implementingecosystemmanagement:Aframeworkforremotelysensedinformationatmultiplescales.Landscape and Urban Planning40,173–184,1998.

20. Osborne,P.E.,andSuarez-Seoane,S.Shoulddatabepartitionedspatiallybeforebuild-inglarge-scaledistributionmodels?Ecological Modelling157,249–259,2002.

21. Geoghegan,J.etal.ModellingtropicaldeforestationinthesouthernYucatanpeninsularregion:comparingsurveyandsatellitedata.Agriculture, Ecosystems and Environment85,25–46,2001.

22. Field, D. R. et al. Reaffirming social landscape analysis in landscape ecology:aconceptualframework.Society and Natural Resources16,349–361,2003.

23. Nelson,A.AnalysingdataacrossgeographicscalesinHonduras:Detectinglevelsoforganisationwithinsystems.Agriculture, Ecosystems and Environment85,107–131,2001.

24. Bousquet,F., andLePage,C.Multi-agent simulations and ecosystemmanagement:Areview.Ecological Modelling176,313–332,2004.

25. Loibl, W., and Toetzer, T. Modeling growth and densification processes in suburbanregions—simulationoflandscapetransitionwithspatialagents.Environmental Modelling and Software18,553–563,2003.

26. Evans,T.P.,andKelley,H.Multi-scaleanalysisofahouseholdlevelagent-basedmodeloflandcoverchange.Journal of Environmental Management72,57–72,2004.

27. Berger,T.Agent-basedspatialmodelsappliedtoagriculture:Asimulationtoolfortech-nologydiffusion, resourceusechangesandpolicyanalysis.Agricultural Economics25,245–260,2001.

28. Parker,D.C.,andMeretsky,V.Measuringpatternoutcomesinanagent-basedmodelofedge-effectexternalitiesusingspatialmetrics.Agriculture, Ecosystems and Environment101,233–250,2004.

29. Peralta, P., and Mather, P. An analysis of deforestation patterns in the extractivereservesofAcre,Amazoniafromsatelliteimagery:alandscapeecologicalapproach.International Journal of Remote Sensing21,2555–2570,2000.

30. Crews-Meyer,K.A.AgriculturallandscapechangeandstabilityinnortheastThailand:Historicalpatch-levelanalysis.Agriculture, Ecosystems and Environment101,155–169,2004.

31. Laney,R.Aprocess-ledapproachtomodellinglandusechangeinagriculturalland-scapes:AcasestudyfromMadagascar.Agriculture, Ecosystems and Environment 101,135–153,2004.

32. McConnell,W.J.,Sweeney,S.P.,andMulley,B.Physicalandsocialaccesstoland:Spatio-temporalpatternsofagriculturalexpansioninMadagascar.Agriculture, Ecosystems and Environment101,171–184,2004.

33. Cassel-Gintz,M.,andPetschel-Held,G.GIS-basedassessmentof the threat toworldforestsbypatternsonnon-sustainablecivilisation-natureinteraction.Journal of Environmental Management 59,279–298,2000.

34. Sui,D.Z.GIS-basedurbanmodelling:Practices,problemsandprospects.International Journal of Geographical Information Science12,651–671,1998.

35. Faust,K.etal.SpatialarrangementsofsocialandeconomicnetworksamongvillagesinNangRongDistrict,Thailand.Social Networks21,311–337,1999.

42963_C002.indd 38 11/6/07 7:17:38 AM


36. Leenders, R. Th. A. J. Modeling social influence through network autocorrelation:Constructingtheweightmatrix.Social Networks24,21–47,2002.

37. Jackson, J. E., Lee, R. G., and Sommers, P. Monitoring the community impacts ofthe Northwest Forest Plan: An alternative to social indicators. Society and Natural Resources17,223–233,2004.

38. Martin,D.,andBracken,I.Theintegrationofsocio-economicandphysicalresourcedataforappliedlandmanagementinformationsystems.Applied Geography13,45–53,1993.

39. Martin, D. Automatic neighbourhood identification from population surfaces.Computers, Environment and Urban Systems22,107–120,1998.

40. Bonaiuto,M.,Fornara,F.,andBonnes,M.Indexesofperceivedresidentialenviron-mentqualityandneighbourhoodattachment inurbanenvironments:AconfirmationstudyonthecityofRome.Landscape and Urban Planning65,41–52,2003.

41. Krause,C.L.Ourvisuallandscape:managingthelandscapeunderspecialconsider-ationofvisualaspects. Landscape and Urban Planning54,239–254,2001.

42. Croissant,C.Landscapepatternsandparcelboundaries:Ananalysisofcompositionand configuration of land use and land cover in south-central Indiana. Agriculture, Ecosystems and Environment101,219–232,2004.

43. Reed,B.C.etal.Measuringphenologicalvariabilityfromsatelliteimagery.Journal of Vegetation Science 5,703–714,1994.

44. Li,Z.,andKafatos,M.InterannualvariabilityofvegetationintheUnitedStatesanditsrelationtoElNino/SouthernOscillation.Remote Sensing of Environment71,239–247,2000.

45. Zhang,M.Exploringtherelationshipbetweenurbanformandnonworktravelthroughtimeuseanalysis.Landscape and Urban Planning73,244–261,2005.

46. Hill,M.J.etal.Asimple,portable,multicriteriaanalysis tool fornatural resourcemanagement assessments. In: Zerger, A, and Argent, R. M., eds., MODSIM 2005, Proceedings of the International Symposium on Modelling and Simulation.UniversityofMelbourne,12–15December,2005,CDROM(availableat:http://www.mssanz.org.au/modsim05/authorsH-K.htm#h).

47. Thackway,R.,andLesslie,R.J.ReportingvegetationconditionusingtheVegetationAssets, States and Transitions (VAST) framework. Ecological Management and Restoration7,s53–s61,doi:10.1111/j1442-8903.2006.00292.x.,2006.

48. Defries,R.S.,Asner,G.P.,andHoughton,R.Trade-offsinlandusedecisions:Towardsa framework for assessing multiple ecosystem responses to land-use change. In:Defries,R.S.,Houghton,R.,andAsner,G.P.,eds.Ecosystems and Land Use Change.GeophysicalMonographSeries153,AmericanGeophysicalUnion,2004.

49. Platt, R. V. Global and local analysis of fragmentation in a mountain region ofColorado.Agriculture, Ecosystems and Environment101,207–218,2004.

50. Reggiani,A.,Nijkamp,P.,andSabella,E.Newadvancesinspatialnetworkmodelling:Towards evolutionary algorithms. European Journal of Operational Research 128,385–401,2001.

51. Nijkamp,P.,andReggiani,A.Non-linearevolutionofdynamicspatialsystems.Therelevance of chaos and ecologically-based models. Regional Science and Urban Economics 25,183–210,1995.

52. Roderick,M.L.,Noble,I.R.,andCridland,S.W.Estimatingwoodyandherbaceousvegetation cover from time series satellite observations. Global Ecology and Biogeography 8, 501–508,1999.

53. Lu,H.etal.DecompositionofvegetationcoverintowoodyandherbaceouscomponentsusingAVHRRNDVItimeseries.Remote Sensing of Environment86,1–18,2003.

54. Malingreau,J.-P.Globalvegetationdynamics:satelliteobservationsoverAsia.International Journal of Remote Sensing9,1121–1146,1986.

42963_C002.indd 39 11/6/07 7:17:38 AM

40 Land Use Change

55. Zhang,X.etal.MonitoringvegetationphenologyusingMODIS.Remote Sensing of Environment84,471–475,2003.

56. Moody, A., and Johnson, D. M. Land-surface phenologies from AVHRR using thediscretefouriertransform.Remote Sensing of Environment75,305–323,2001.

57. Devezas,T.,andCorredine,J.T.Thebiologicaldeterminantsoflongwavebehaviorin socio-economic growth and development. Technological Forecasting and Social Change68,1–57,2001.

58. Lesslie, R. et al. Towards Sustainability for Australia’s Rangelands. Analysing the Options.DepartmentofAgriculture,FisheriesandForestry,Australia,16pp,2006.

42963_C002.indd 40 11/6/07 7:17:39 AM

Part II

Comparative Regional Case Studies

42963_S002.indd 41 8/17/07 2:29:21 PM

42963_S002.indd 42 8/17/07 2:29:21 PM

43

3 Spatial Methodologies for Integrating Social and Biophysical Data at a Regional or Catchment Scale

Ian Byron and Robert Lesslie

Contents

3.1 Introduction...................................................................................................443.2 MappingChangeinLandUseataRegionalScale....................................... 45

3.2.1 WhyUnderstandingLandUseChangeIsImportant........................463.2.2 TypesofLandUseChange................................................................463.2.3 ChangesinLandUseintheLowerMurrayRegion..........................46

3.3 MappingCorrespondencebetweenBiophysicalDataandLandManagerPerceptions,ValuesandPractices..................................................483.3.1 IntegratingDataSources...................................................................48

3.3.1.1 TheGlenelgHopkinsLandholderSurvey............................483.3.1.2 SalinityDischargeSitesBasedonGroundwaterFlows

Systems.................................................................................483.3.1.3 LandUseCategorizedasConservationofNatural

Environment......................................................................... 493.3.2 SpatialMethodsforAssessingCorrespondenceinAssessments

andResponsestoSalinity..................................................................503.3.2.1 Context.................................................................................503.3.2.2 Approach..............................................................................503.3.2.3 Analysis................................................................................50

3.3.3 MappingtheRelationshipsbetweenAreasofHighConservationValueandLandManagers’ValuesandPractices....... 533.3.3.1 Context................................................................................. 533.3.3.2 Approach.............................................................................. 533.3.3.3 Analysis................................................................................54

3.4 InsightsandImplicationsfromIntegratingSocialandBiophysicalDataataRegionalScale...............................................................................54

42963_C003.indd 43 11/6/07 7:10:14 AM

44 Land Use Change

3.5 FutureDirections.......................................................................................... 593.6 Conclusions................................................................................................... 59References................................................................................................................60

3.1 IntroduCtIon

Catchment or watershed-based approaches to natural resource management andplanninghavebeenwidelyadopted inmanycountriesacross theglobe includingAustralia.1,2,3 These approaches seek to meld the benefits of building local com-munity engagement with the need to develop better integrated and coordinatedapproaches for addressing landscape-scale changes in the condition of land andwaterresources.4

In Australia, regional catchment management organizations now manage alargeproportionofnationalinvestmentinnaturalresourcemanagementthroughtheNaturalHeritageTrustandtheNationalActionPlanforSalinityandWaterQuality.TheNaturalHeritageTrustrepresentsAustralia’slargestinvestmentinenvironmentalmanagement.5Aspartof thedeliveryof funds,catchmentgroupsare required todevelopregionalplansthatsetouthowthenaturalresourcesoftheregionaretobemanaged.EachregionalplanistobeendorsedbystateandAustraliangovernmentagenciespriortotheirimplementation.Althoughtherearestateandregionaldiffer-ences,thesecatchmentgroupsaretypicallyaskedto:

Describetheircatchmentconditionintermsofenvironmental,economic,andsocialassetsIdentifythedesiredfutureconditionofthoseassetsIdentify the key processes that might mitigate the achievement of thedesiredconditionsIdentify management actions and targets that will help achieve desiredconditionsMonitorandevaluateprogress

Clearlytheserolesrequirecatchmentgroupstobeabletounderstandthedriversandbarriersaffectinglandmanagersandunderstandtheimpactsoflandmanagementpracticesonkeyregionalassets.Unfortunately,thereareverylimiteddataavailablethathavebeendesignedwiththesepurposesinmind.AlthoughmostregionalgroupsinAustraliahaveaccesstoarangeofbiophysicaldatasources,veryfewhaveaccesstodetailedsocialdataspecifictotheirregion.Endter-Wadaetal.6assertedthatnatu-ralscientistshavebeenreluctanttoincludesocialsciencedimensionsinecosystemassessments.Atthesametime,Brown7suggestedthatinpartthelackofsocialdataincorporatedinlandscapeplanningreflectsanabsenceofsystematicapproachesforcollectingandanalyzingthisinformationwithbiophysicaldata.

Nevertheless,thereisincreasingrecognitionoftheneedtointegratesocialandbiophysical data to achieve improved natural resource management outcomes. AreviewconductedaspartofAustralia’sNationalLandandWaterResourcesAuditconcludedthat thereisastrongneedforapproachesthat integratesocioeconomicandbiophysicaldataataregionalscale.8Similarly,Endter-Wadaetal.6concluded

•

••

•

•

42963_C003.indd 44 11/6/07 7:10:15 AM

Spatial Methodologies for Integrating Social and Biophysical Data 45

thateffectiveecosystemmanagement ispredicatedonbringing togetherscientificanalysisofsocialfactorsandbiophysicalfactors.

Therelativelyrecentemergenceofgeographicalinformationsystems(GIS)hasprovidedanimportantsetoftoolstofacilitateinterdisciplinaryresearchthatinte-gratessocialandbiophysicaldataatalandscapescale.9Inrecentyearstherehavebeenanumberofimportantstudiesusingcensusdatatolinkchangesinsocialstructurewithecologicalfactorsandvisaversa.9,10,11Althoughthesestudieshaveclearlyhigh-lightedhowintegratingsocialandbiophysicaldataiscriticaltoimprovingnaturalresource management outcomes, they also acknowledge significant limitationswithnationallycollectedcensusdata.Inparticular,censusdataareonlyavailableataggregatelevelsthatrequireresearcherstoassumethatthesocialvariablesarehomogenousacrossthesmallestcensusunit(typically200households).9

Inadditiontoconcernsaboutspatialresolution,Endter-Wadaetal.6suggestthatwhileimportant,understandingdemographictrendsaloneisinsufficientforunder-standingcomplexsocialsystemsandtheirrelationshiptoresourceconditionsanddynamics.Insummarizingthepotentialcontributionsofsocialsciencetoecosystemmanagement,Endter-Wadaet al.6 concluded thatunderstanding spatialvariabilityin resourceneeds,values,anduseswascriticalbuthighlighteda lackofsystem-aticdataanalysisrequiredtomovebeyondtherhetorictotherealityofintegratinghuman values in ecosystem management. According to Grove et al.,11 exploringquestionsabouthowmotivationsandcapacitiesinfluenceandareinfluencedbythebiophysicalenvironmentwillbebestexploredbyadaptingtraditionalsocialsciencefieldmethodsthathavebeenappliedtonaturalresourcemanagement.

Althoughnumerousresearchershaveintegratednationallycollectedcensusdatainto landscapeanalyses, therearevery fewexamplesofattempts topurposefullycollectsocialdatathatcanbeintegratedwithspecificbiophysicaldatalayers.Brown7providessomeinsightsintotheapplicationoftheseapproachesasdoesearlierworkbyCurtis,Byron,andMcDonald12andCurtis,Byron,andMacKay,13uponwhichthischapteraimstoextend.

Thischapterdrawsonfindingsfromspatiallyreferencedsurveysoflandmanagerstohighlightmethodologiesforintegratingsocialandbiophysicaldataataregionalorcatchmentscale.Specific issuesandapproachescovered includemapping landusechangeandexploringtheextentandnatureoflinksbetweenmappedbiophysicalresourceconditionsandlandmanagerperceptions,values,andpractices.

3.2 MAPPInG CHAnGe In LAnd use At A reGIonAL sCALe

3.2.1 Why Understanding Land Use Change is important

Acapacityfordetectingandreportinglandusechangeiscriticaltoevaluatingandmonitoring trends in natural resource conditions and the effectiveness of publicinvestment innatural resourcemanagement.Australia’sNationalLandandWaterResourcesAudit,forexample,hasidentifiedthereportingofchangeovertimeandtheintegrationoflanduseinformationwithothernaturalresourceinformationasakeytoeffectivelyaddressingmajorsustainabilityproblemssuchassalinity,waterquality,andsoilloss.

42963_C003.indd 45 11/6/07 7:10:15 AM

46 Land Use Change

3.2.2 types of Land Use Change

Aparticulardifficultywithlandusechangereportingisdiscriminatingthediffer-entdimensionsofchange.Protocolsforreportinglandusechangeinanagriculturalcontext,forexample,shouldbecapableofdistinguishingthetemporalcharacteris-ticsoffarmingsystems(e.g.,rotations),seasonalvariability,andlonger-termindus-tryandregionaltrends.Lesslie,Barson,andSmith14identifyfourbroadapproachestomeasuringandreportinglandusechange:

1.Arealchange:lossorgaininthearealextent.Thisprovidesanindicationofwhethertargetlandusesareincreasingordecreasinginareaovertime.Changescanbepresentedstatistically,graphically,orspatiallyandidenti-fiedchangescomparedandtrendsobserved.

2.Transformation:thepatternoftransitionfromonelandusetoanother.Forexample,anareamaybecroppedoneyear,grazedthenextyear,andthencroppedagaintheyearafter.Alternatively,landunderimprovedpasturefordairymaybepermanentlyconvertedtovineyards.15Landusetransforma-tionsbetweentimeperiodsmaybeexpressedusingachangematrix.

3.Dynamics:ratesofchangeandperiodicityinarealextentortransforma-tions.Thetemporalnatureofchangemaybefurtherexploredbyanalyzingwhetherratesofchangeareincreasingordecreasing,arelong-orshort-termtrends,orcyclic(forexample,changesasaresultofdifferencesingrow-ingseasons,structuraladjustment,farmingsystems,orrotationregimes).Thismayrevealkeytrendsinlanduseandlandmanagementnotevidentinexpressionsofsimplearealchangeortransformations.Successfulanaly-sisoflandusedynamicsrequiresconsistent,high-qualitytime-seriesdata.Oftenitisnotpossibletoobtainsufficientlyconsistentdataoverconsecu-tiveyearsorseasons.

4.Prediction: modeling spatial or temporal patterns of change. The useofmodels topredictpast,present,andfuture landusesbasedoncertainrules,relationships,andinputdatamayhelpidentifykeydriversoflandusechange,implementscenarioplanning,andfillgapsindataavailability.

3.2.3 Changes in Land Use in the LoWer mUrray region

Thecapacitytoreportchangealsodependsontheavailabilityofconsistenttime-seriesdatacapableofprovidinginsightsintorelevantaspectsofchange.Wherefine-scaledtime-seriesdataareavailable,spatialanalysiscanprovideimportantinsightintothenatureoflandusechange.

Using time-series data from fine-scaled mapping based on orthophoto inter-pretationanddetailedpropertysurveys,itispossibletohighlightspatialtrendsinlandusepatterns.Forexample,Figure3.1showstrendsintheexpansionofirrigatedhorticulturearoundthetownsofRenmark,Berri,andLoxtonintheLowerMurrayregionofsoutheasternAustraliaproducedbytheAustralianCollaborativeLandUseMappingProgram (ACLUMP), a partnershipofAustralian and state governmentagenciesproducingcoordinatedlandusemappingforAustralia.14Thistime-series,1990to2003,isdrawnfrom1:25,000catchment-scalelandusemappingcompleted

42963_C003.indd 46 11/6/07 7:10:16 AM


usingorthophotointerpretationanddetailedpropertysurveys.16Themappingrevealsapatternof landusetransformationandintensificationfromdrylandcerealcrop-pingandgrazingtoirrigatedhorticulture,andatrendfromsmall-scaleenterprisesclusteringaroundtownareastodispersed,large-scaleestablishmentsatincreasingdistancesfromirrigationwatersupply(rivers).

Time-series land use mapping at catchment-scale in Australia is producedby ACLUMP to agreed to national standards, facilitating its use in national andregionalnaturalresourceassessments.Themappingprocessinvolvesstagesofdatacollation, interpretation, verification, independent validation, quality assurance,andtheproductionoflandusedataandmetadata.Thisincludescollectingexistinglanduse informationandcompiling it intoadigitaldata setusingaGIS. Impor-tantinformationsourcesincluderemotelysensedinformation,landparcelboundaryinformation,forestandreserveestatemapping,landcover,localgovernmentzoning

Statistical Local AreasIrrigated horticulture first mapped prior to 1990 (mapped in 1988 for SA)Irrigated horticulture first mapped in 1995 and between 1990–1995Irrigated horticulture first mapped between 1995–1999Irrigated horticulture first mapped in 2001Irrigated horticulture first mapped in 2003

Irrigated horticulture data provided bySA Department of Environment and Heritage

Loxton

BerriBarmera

Renmark

FIGure 3.1 (See color insert following p. 132.) LandusechangeintheBarmera,Berri,andRenmarkareasofSouthAustralia.

42963_C003.indd 47 11/6/07 7:10:18 AM

48 Land Use Change

information, other land management data, and information collected in the field.Agreed-tostandards includeattribution toanationalclassification, theAustralianLandUseandManagement(ALUM)Classification.14Fine-scaleddataofthetypeillustratedinFigure3.1are,however,expensivetoproduceandarepresentlyoflim-ited availability. More cost-effective methods for wider application are presentlyunderdevelopment.

3.3 MAPPInG CorresPondenCe BetWeen BIoPHYsICAL dAtA And LAnd MAnAGer PerCePtIons, VALues And PrACtICes

3.3.1 integrating data soUrCes

The approachoutlined in this chapter used aGIS to integrate social survey datacollectedintheGlenelgHopkinsregionwithbiophysicaldata.TheGlenelgHopkinsregionislocatedintheStateofVictoriainthesoutheastofAustralia.Theregioncovers an area of approximately 26,000 square kilometres or approximately 11%ofthestate17(Figure3.2).Agriculturerepresentsamajorcontributortotheregionaleconomy,andin1999to2000itwasworthapproximatelyAU$650millionorapprox-imately10%ofthegrossvalueofagriculturalproductionintheStateofVictoria.

Thethreemajordatalayersusedinthischapterare:

1.Aspatiallyreferencedsurveyofrurallandholders18

2.Amapofsalinitydischargebasedonthegroundwaterflowsystems19

3.Land use categorized as Conservation of Natural Environment (Class 1)undertheALUMclassificationsystem20

3.3.1.1 the Glenelg Hopkins Landholder survey

In2003theBureauofRuralSciencesandGlenelgHopkinsCatchmentManagementAuthorityconductedasurveyofapproximately1,900rurallandholdersfromacrossthe region.18The survey focusedongatheringbaseline information regarding thekey social and economic factors affecting landholder decision making about theadoptionofpracticesexpectedtoimprovethemanagementofnaturalresourcesinthe Glenelg Hopkins region. The survey was sent to a random selection of ruralpropertyowners,withpropertiesover10hectaresinsize,identifiedthroughlocalrate payer databases. A final response rate of 64% was achieved for this survey.Allsurveydata(some250variables)wereenteredintoageographicalinformationsystem(ArcViewGIS)andassignedtoapropertycentroidusingxandycoordinatesincludedintheratepayerdatabases.

3.3.1.2 salinity discharge sites Based on Groundwater Flows systems

ThemapofsalinitydischargesitesintheGlenelgHopkinsregionwasundertakenaspartofthegroundwaterflowsystemsprojectconductedbyDahlhaus,Heislers,andDyson.19ThegroundwaterflowsystemsweredevelopedbytheNationalLandand Water Resources Audit as a framework for dryland salinity management in

42963_C003.indd 48 11/6/07 7:10:18 AM


Australia.21Thisworkcategorizeslandscapesbasedonsimilaritiesingroundwaterprocesses,salinityissues,andmanagementoptions.Dahlhaus,Heislers,andDyson19stated thatwhilegroundwaterflowsystemsareauseful tool inhelping tounder-standsalinity,therehasbeenlittlescientificvalidationoftheflowsystemsorsalinityprocessesintheGlenelgHopkinsregion.

3.3.1.3 Land use Categorized as Conservation of natural environment

LandusemappingfortheGlenelgHopkinsregionintheStateofVictoriawasunder-takenusingathree-stageprocess.20Thefirststageofmappinginvolvedthecolla-tionofexistinglanduseinformation,remotelysensedinformation(satelliteimageryandaerialphotography), andcadastre.Other important information sourceswerereserveestatedata,landcover,localgovernmentzoninginformation,andotherlandmanagementdata.ThesecondstageinthemappingprocessinvolvedinterpretationandassignmentoflanduseclassesaccordingtotheALUMclassificationtocreateaninitialdraftlandusemap.Thefinalstagesofmappingincludedfieldverification,theeditingofdraftlandusemaps,andvalidation.Themappingisdatedat2001andisproducedatscalesof1:25,000and1:100,000.

FIGure 3.2 LocationoftheGlenelgHopkinsregion.

V I C T O R I A

TownsSurvey AreaSLA BoundaryMain Road

Legend

Survey Area

N

BALMORAL

COLERAINECASTERTON

MERINO

DARTMOOR

HEYWOOD MACARTHUR

PORTLAND

Km150100500

PORT FAIRYWARRNAMBOOL

KOROITBUSHFIELD–WOODFORD

ALLANSFORD

TERANGNOORAT

MORTLAKE

PENSHURST

DUNKELD

WILLAURA

ARARAT

BEAUFORT

SKIPTONSNAKE VALLEY

LEARMONTHMINERS REST

HAMILTON

42963_C003.indd 49 11/6/07 7:10:20 AM

50 Land Use Change

3.3.2 spatiaL methods for assessing CorrespondenCe in assessments and responses to saLinity

3.3.2.1 Context

TheGlenelgHopkinsregionisoneof21priorityregionsidentifiedundertheNationalActionPlanforSalinityandWaterQualityasbeingaffectedbysalinityandwaterqualityproblems.TheGlenelgHopkinsSalinityPlan22identifiedheavyimpactsofsalinityonagriculture,theenvironment,andinfrastructurewithanestimatedcosttotheregionofoverAU$44millionannually.

3.3.2.2 Approach

The land manager survey included a question that asked respondents to indicateif they had any areas of salinity on their property. By assigning land managers’responsestothepointdatacontainingpropertycentroidsforeachpropertysurveyed,itispossibletoexploretheextentthatlandmanagerperceptionsarespatiallylinkedtomappedsalinitydischargesitesusingthegroundwaterflowsystems(representedaspolygons).

Asdatafromthelandmanagersurveycouldonlybejoinedtoapointfilebasedonapropertycentroid,anymeasureofdirectcorrespondencewouldfailtoallowfordifferencesinpropertysizeandshape.Althoughthereisawiderangeoftechniquesavailablefor interpolatingcontinuoussurfacesfrompointdata, theextent thatanychangeinsocialvariablescanbepredictedbyalgorithmsbasedonaspatialrelation-shipisquestionable,particularlywherethepointsaredispersedacrossalargearea.23

Forthesereasons,nearestneighboranalysis24wasusedtoidentifythedistanceto the closest edge of the nearest mapped salinity discharge site for each surveyrespondent.Thesedistancescanthenbecomparedforrespondentswhosaidtheyhadsalinityontheirpropertyandthosewhodidnotoracrossarangeofothervariables.

Althoughinterpolatingsurfacesfromthelandmanagerpointdataisproblem-atic, creating a raster-based surface of distance from any grid cell to the nearestsalinitydischargesiteprovidesaquickvisualdisplaythatcanbeoverlayedwiththelandmanagerperceptionsandsalinitydischargelayers(Figure3.3).

3.3.2.3 Analysis

Theresultsof thenearestneighboranalysisclearlyshow that landholders incloseproximitytomappedsalinitydischargesitesweresignificantlymorelikelytoidentifyareasofsalinityontheirproperty(Table3.1).Withoverhalfofallrespondentswithin0.5kmofadischargesiteidentifyingsalinityontheirproperty,applyingthismethod-ologyalsosuggeststhatmostlandholdersareawareofsalinityontheirproperty.

By adopting the nearest neighbor technique it is also possible to explore theextentthatlandmanagersclosertomappedsalinitydischargesitesaremorelikelytobeconcernedabouttheimpactsofsalinityandundertakingpracticesexpectedtohelpmitigatesalinity(Table3.2).

Althoughmostrespondentsclosetomappedsalinitydischargesitesappeartobeawareoftheissue,therewerestillalargenumberofrespondentsnearmapped

42963_C003.indd 50 11/6/07 7:10:20 AM


FIGure 3.3 (See color insert following p. 132.) Landmanagers’perceptionofsalinityandmappedsalinitydischargesites.

42963_C003.indd 51 11/6/07 7:10:21 AM

52 Land Use Change

tABLe 3.1Land Managers’ Perception of salinity and distance to Mapped salinity discharge

distance to nearest mapped salinity discharge site (m)

Land manager identified salinity (%)

Yes no

0–499 61 39

500–999 47 53

1,000–1,999 35 65

2,000–2,999 27 73

3,000–3,999 31 69

4,000–4,999 17 83

Over5,000 11 89

tABLe 3.2distance to Mapped salinity discharge and Land Manager Attitudes and Practices

Land manager attitudes and practicesMedian distance to nearest

mapped salinity discharge site (m)

Concern about salinity reducing productive capacity of their propertya

High 1,639

Moderate 2,028

Low 3,092

Concern about salinity reducing productive capacity of the local areaa

High 1,951

Moderate 2,386

Low 3,202

Concern about salinity reducing water quality in the local areaa

High 1,824

Moderate 2,262

Low 3,216

Planted trees or shrubsa

Yes 2,493

No 3,498

Established deep rooted perennial pastureb

Yes 2,764

No 2,648a Differencewasstatisticallysignificant(p<0.05)usingtheKruskal-Wallisnon-

parametricchi-squaretest.b Differencewasnotstatisticallysignificant(p>0.05)usingtheKruskal-Wallis

nonparametricchi-squaretest.

42963_C003.indd 52 11/6/07 7:10:22 AM


salinitythatappeartobeunawareoftheproblem.Forthepurposesofthisexamplewehaveassumedthatlandholderswithin0.5kilometerofadischargesitethathavenotidentifiedsalinityontheirpropertyareunawareoftheproblem.

Thespatialidentificationofrespondentswhoappeartobeunawareofsalinityontheirpropertyprovidesanimportantopportunitytoidentifykeycharacteristicsofthisgroupofrespondentsandthusdevelopbettertargetedcommunityengagementstrategies.Forexample,Table3.3clearlyhighlightsadistinctivesetofcharacteris-ticsoflandholdersthatappeartobeunawareofsalinityontheirproperty.

3.3.3 mapping the reLationships betWeen areas of high Conservation vaLUe and Land managers’ vaLUes and praCtiCes

3.3.3.1 Context

Akey aim fornatural resourcemanagement in theGlenelgHopkins region is tomaintainandenhanceremnantnativevegetation.TheGlenelgHopkinsregionhasanextensivehistoryoflandclearing,andnativevegetationnowcoverslessthan13%oftheregion,with8%inparksandreservesfragmentedacrosstheregion.

3.3.3.2 Approach

ThecombinationofdatacollectedthroughtheregionallandholdersurveywithlandusemappingdatafortheGlenelgHopkinsregionprovidesanopportunitytoidentifythoselandmanagerswhoaremostlikelytohaveanimpactonareasofhighconser-vationvalue.

Nearestneighboranalysisforeachsurveyrespondent(representedasapropertycentroid)tothenearestedgeofanareaclassifiedashavinghighconservationvalue(polygon)canbeusedtohelpidentifykeygroupsoflandmanagers.Oncethedis-tancetothenearestareaofhighconservationvaluehasbeencomputed,standard

tABLe 3.3differences between Land Managers Who Were Aware of salinity and those unaware

Characteristics of land managersa

Land managers within 500 m of salinity

Aware unaware

Primaryoccupationfarming 85% 56%

MemberofaLandcaregroup 70% 37%

Completedatrainingcourserelatedtopropertymanagement

69% 25%

Hadworkundertakenontheirpropertythatwasatleastpartiallyfundedbygovernment

53% 19%

Plantedtreesandshrubs 85% 67%

Establishedperennialpasture 73% 44%

Medianpropertysize 465ha 136haa Alldifferenceswerestatisticallysignificant(p<0.05)usingthePearsonchi-squaretest.

42963_C003.indd 53 11/6/07 7:10:22 AM

54 Land Use Change

statisticalanalysescanbeusedtohelpdiscernpatternsbetweenspatialproximityandkeycharacteristicsoflandmanagersandtheirproperty.Asoutlinedpreviously,araster-basedsurfaceofdistancefromanygridcelltothenearestareaofhighcon-servationvalueprovidesausefuloverlaytohelpgraphicallyrepresentrelationships(Figure3.4).

3.3.3.3 Analysis

WhenappliedtosurveyandlandusedatafromtheGlenelgHopkinsregiontheseanalyses show some very clear differences in the characteristics of land manag-ersandtheirpropertybasedontheirproximitytoareasofhighconservationvalue(Table3.4).

Applyingthesametechniqueitisalsopossibletoexploreifthevalueslandholdersattachtotheirpropertyintermsofthesocial,economic,andenvironmentalbenefitsarelinkedspatiallytoareasofhighconservationvalue.Theseanalysesshowthatrespondentswhosaidbeingclosetonaturewasanimportantvalueoftheirpropertywereinfactsignificantlyclosertoareasofhighconservationvalue,whileinturnthosewhosaidprovidinghouseholdincomewasimportantweresignificantlyfurtherfromtheseareas.However,itisinterestingtonotethatthepropertyprovidingthesortoflifestyledesiredwasnotlinkedtotheproximitytoareasofhighconservationvalue(Table3.4).

Finally,surveydataandlandusemappingdatacanbecomparedtoseeiflandmanagersnearareasofhighconservationaremorelikelytohaveadoptedmanage-mentpracticesaimedatimprovingbiodiversity.Theseanalysesshowmixedresults.Landmanagersclosertoareasofhighconservationvalueweresignificantlymorelikelytohaveencouragedregrowthofnativevegetationandfencedoffareasofnativevegetationontheirproperty.However,theserespondentswerealsosignificantlylesslikelytohaveplantednativetreesandshrubsontheirproperty(Table3.4).

3.4 InsIGHts And IMPLICAtIons FroM InteGrAtInG soCIAL And BIoPHYsICAL dAtA At A reGIonAL sCALe

Theuseofspatialmethodologiesforintegratingsocialandbiophysicaldata,asdem-onstratedinthischapter,hassomeimportantinsightsandimplicationsforeffortstoimprovenaturalresourcemanagementoutcomes.

In thefirst instance, these approacheshighlighted anumberof importantdis-crepanciesbetweenrespondents’assessmentsofsalinityandthosemadeusingthegroundwaterflowsystem.Overa thirdofall respondentswithin500metersofamappedsalinitydischargesitedidnotidentifyareasofsalinityontheirproperty,andoverhalfofallthelandmanagerswhoidentifiedareasofsalinitydidnotcorrespondwithmappedsalinitydischargesites.ManypartsoftheGlenelgHopkinsregionarecharacterizedbyrelativelyhighrainfall,anditispossiblethatsomelandholdershavemistakenwaterloggingasasignofsalinity. It isalsopossible thatsmall localizedareasofsalinityidentifiedbylandholderscouldbemissedthroughinterpretationoflarge-scale aerial photographs. Similarly, the identification of salt indictor speciesthrough the interpretationofaerialphotographsaspartof themappingofsalinity

42963_C003.indd 54 11/6/07 7:10:22 AM


N

S

W E

Kilometers

0 – 0.0124435330.012443533 – 0.0248870660.024887066 – 0.0373305980.037330598 – 0.0497741310.049774131 – 0.0622176640.062217664 – 0.0746611970.074661197 – 0.0871047300.087104730 – 0.0995482620.099548262 – 0.1119917950.111991795 – 0.124435328

Glenelg Hopkins regionFarmerNon–farmerConservation of natural environments

Std. Dev = 1595.02Mean = 1951N = 640.00

Std. Dev = 1249.48Mean = 1276N = 355.00

8000 – 8500

7000 – 7500

6000 – 6500

5000 – 5500

4000 – 4500

3000 – 3500

2000 – 2500

1000 – 1500

0 – 500

8000 – 8500

7000 – 7500

6000 – 6500

5000 – 5500

4000 – 4500

3000 – 3500

2000 – 2500

1000 – 1500

0 – 500

Distance to nearest area of high conservation value (meters)


Non–farmer

Farmer

Freq

uenc

y

0 30 60 120

Freq

uenc

y

140

120

100

80

60

40

20

0

140

120

100

80

60

40

20

0

Distance to nearest area of high conser vation value<VALUE>

Legend

FIGure 3.4 (See color insert following p. 132.) Landmanagerswhomanagepropertiesnearareasofhighconservationvalue.

42963_C003.indd 55 11/6/07 7:10:42 AM

56 Land Use Change

tABLe 3.4distance to nearest Area of High Conservation Value and Land Managers’ Characteristics, Values, and Practices

Land managers’ characteristics, values, and practicesMedian distance to nearest area of

high conservation value (m)

Primary occupationa

Farmer 1,614

Non-farmer 794

Property sizea

Small(<100ha) 829

Medium(100–499ha) 1,425

Large(500haandover) 1,642

Landcare membershipa

Yes 1,063

No 1,600

Completed a short course related to property managementa

Yes 1,429

No 1,129

Had work undertaken on their property that was at least partially funded by governmenta

Yes 1,512

No 1,152

Value attached to property in providing an attractive place to liveb

High 1,276

Moderate 1,470

Low 969

Value attached to property in providing habitat for native animalsa

High 1,074

Moderate 1,357

Low 1,479

Value attached to property in providing majority of household incomea

High 1,616

Moderate 1,012

Low 794

Planted tree and shrubsa

Yes 1,470

No 881

42963_C003.indd 56 11/6/07 7:10:43 AM


dischargesitescouldalsobeconfoundedbyhighrainfallandatendencyforwater-logging.Incorporatingdataonrainfallandrainfallreliabilityacrosstheregionisoneoptionthatmayhelpfurtheridentifyreasonsforconflictingperceptionsofsalinity.

Totheextentthatweassumethatthesalinitydischargemapshavecorrectlyiden-tifiedareasofsalinity,theintegrationofspatiallyreferencedsurveydataprovidesimportantinsightsforthedevelopmentoftargetedmanagementstrategies.Throughthis process it is possible to spatially locate not only areas affected by salinitybut areas within those, where the land managers are unaware of, or not activelymanaging,theproblem.Takeninisolationeitherthemapsofsalinitydischargeorlandholderidentifiedsalinityareinadequatefortargetedstrategiestomitigatesalin-ity.Forexample, ifmost landholders inaregionarenotaffectedordirectlycon-tributingtosalinity,thecapacityoflandmanagersatanaggregatescaletomodifyland management practices is largely irrelevant. That is, rather than undertakingbroadoutreachactivitiesaboutsalinity,themethodspresentedinthischapterallowregionalcatchmentmanagerstodevelopmorestrategicinvestments.

Similarly, linking remnant areasofvegetation through targeted investment isanintegralpartofplanstoimprovebiodiversityoutcomesintheGlenelgHopkinsregion(Figure3.5).Theintegrationoflandusemappingofareasofhighconserva-tionvaluewithspatiallyreferencedlandholderinformationclearlyhighlightedthatlandmanagersnearareasofhighconservationareaswerequitedifferentandappeartopreferentiallypurchasepropertieswithhighamenityvalue.

Thefactthatlandmanagerswhowereunawareofsalinityandthosenearareasofhighconservationvaluetendedtobenonfarmers,ownsmallerproperties,werelesslikelytobeinvolvedintheLandcareprogramorhavecompletedacourserelatedtopropertymanagementhaveimportantimplications.Thesmallnumbersoflargefamily farming operations that manage much of the land area of Australia havebeen a key target of policies and programs aimed at improving natural resourcemanagement.Althoughtheseprogramsandpolicieshavebeenlargelysuccessful,as

tABLe 3.4 (continued)distance to nearest Area of High Conservation Value and Land Managers’ Characteristics, Values, and Practices

Land managers’ characteristics, values, and practicesMedian distance to nearest area of

high conservation value (m)

Encouraged regrowth of native vegetationa

Yes 1,122

No 1,352

Fenced areas of native busha

Yes 856

No 1,357a Difference was statistically significant (p < 0.05) using the Kruskal–Wallis non-parametric

chi-squaretest.b Differencewasnotstatisticallysignificant(p>0.05)usingtheKruskal–Wallisnon-parametric

chi-squaretest.

42963_C003.indd 57 11/6/07 7:10:43 AM

58 Land Use Change

(a)

(b)

(c)

FIGure 3.5 (a) Pastures, (b) rolling country-side and (c) canola cropping in theGlenelgHopkinsregion.

42963_C003.indd 58 11/6/07 7:10:45 AM


inthecaseofflagshipprogramssuchasLandcare,thereisincreasingevidenceofagrowingnumberofrurallifestylemotivated“tree-change”orsubcommerciallandholdings.25There isaclearneedforregionalcatchmentmanagers tobetter targetandengagetheselandmanagerswhoarelikelytooperateoutsidetheirnormalinfor-mationchannelsandnetworks,particularlytotheextentthattheymanagecriticalnaturalresources,asisthecaseintheGlenelgHopkinsregion.

3.5 Future dIreCtIons

Theexamplespresented in thischapter linking spatially referenced social surveydatawithbiophysicaldataata regional scaleprovidesonlyafirstglimpseat thepotentialoftheseapproaches.Otherapplicationsmayincludeexploringon-propertyprofitabilityagainstrainfallreliability,croppingtechniquesagainstsoilerosionandturbiditylevels,ormanagementofriparianzonesandwaterquality.

Moresophisticated techniquesformappingsalt,suchasairborneelectromag-netics(AEM),alsoprovideanopportunity tomoreprecisely identify locations inacatchmentwherevariousmanagementactions (suchasplanting trees)willhelpmitigate salinity.AEMprovides a three-dimensionalmodel of salt bymeasuringtheelectricalconductivityofthegroundatdifferentdepths.26Furthermore,arecentstudybyBakerandEvans27usingAEMfoundthattreeplantinginareaspreviouslythoughttobebeneficialmayactuallycontributetosalinitybyreducingfreshwaterflows.ThecombinationofAEMdatawithspatiallyreferencedsurveydataappearslikely to hold much promise in allowing better targeted and more site-specificapproachesformanagingdrylandsalinity.

Generally,thecostoflandusechangedetectionandreportingusingfine-scaledtime-seriesdatabasedonorthophotointerpretationanddetailedpropertysurveys,asoutlinedinthischapter,isprohibitiveandmorecost-effectivemethodsareneeded.The capacity to adequately characterize change is highly dependent on matchingthespatiotemporalaccuracyandprecisionofavailabledatatotherelevantlandusedynamic.Thecouplingof time-seriessatellite imagery toregularlycollectedagri-culturalstatisticspresentsonepracticalandwidelyapplicableapproachtomappingagriculturallandusechange.Aprocedureadoptedforregional-scalelandusemappinginAustraliabyACLUMPusingagriculturalcensusandsurveydata,AdvancedVeryHighResolutionRadiometer(AVHRR)imagery,andastatisticalspatialallocationprocedureispromisingforthedevelopmentofannualtime-seriesanalyses,providingthatlimitationsinspatialaccuracycanbeaccommodated.28,29Thereisscopetooforimprovementifhigherresolutionsatelliteimagery(e.g.,ModerateResolutionImagingSpectroradiometer [MODIS]) can be applied successfully. ACLUMP partners arecurrentlypursinginvestigationsinthisarea.

3.6 ConCLusIons

Thischapterhaspresentedarangeofsimpleandpracticalapproachesforintegratingspatiallyreferencedsocialdatawithbiophysicaldatalayers.Furthermore,bypre-sentingresultsfromcasestudiesdemonstratingtheapplicationofthesetechniques,

42963_C003.indd 59 11/6/07 7:10:46 AM

60 Land Use Change

hopefullywehavehelpedshowhowtheseapproachescancontributetoimprovedmanagementoutcomesatalandscapescale.

Thespatialmethodologiesoutlined in thischapterarederivedfromourearlyattemptsatcreatingmeaningfulintegrationbetweensocialandbiophysicalsciences,andbarelyscratchthesurfaceofthemanypotentialapplicationsofthistechnique.Inparticular,theuseoflongitudinalsurveydatacombinedwithlongitudinallandusechangedataislikelytohelpmoredefinitivelyunderstandtheinteractionsbetweenlinkedsocialandbiophysicalsystems.

reFerenCes

1. Kenney,D.S.Arecommunity-basedwatershedgroupsreallyeffective?:Confrontingthethornyissueofmeasuringsuccess.ChronicleofCommunity3(2),33–38,1999.

2. McGinnis,M.V.,Woolley,J.,andGamman,J.Bioregionalconflictresolution:Rebuild-ing community in watershed planning and organizing. Environmental Management24(1),1–12,1999.

3. Curtis,A.,Shindler,B.,andWright,A.Sustaininglocalwatershedinitiatives:Lessonsfrom Landcare and watershed councils. Journal of the American Water ResourcesAssociation38(5),1–9,2002.

4. Dale, A., and Bellamy, J. Regional Resource Use Planning in Rangelands: AnAustralianReview.LandandWaterResourcesResearchandDevelopmentCorporation,Canberra,1998.

5. CommonwealthofAustralia.NHTAnnualReport2003–2004.NaturalHeritageTrust,Canberra,2004.

6. Endter-Wada, J. et al. framework for understanding social science contributions toecosystemmanagement.EcologicalApplications8(3),891–904,1998.

7. Brown,G.Mappingspatialattributesinsurveyresearchfornaturalresourcemanage-ment:Methodsandapplications.SocietyandNaturalResource18,17–39,2004.

8. CommonwealthofAustralia.Australia’sNaturalResources1997–2002andBeyond.NationalLandandWaterAudit,Canberra,2002.

9. Radeloff,V.C.etal.Exploringthespatialrelationshipbetweencensusandland-coverdata.SocietyandNaturalResources13,599–609,2000.

10. Field, D. R. et al. Reaffirming social landscape analysis in landscape ecology: Aconceptualframework.SocietyandNaturalResources16,349–361,2003.

11. Grove,J.M.etal.DataandmethodscomparingsocialstructureandvegetationstructureofurbanneighbourhoodsinBaltimore,Maryland.SocietyandNaturalResources19,117–136,2006.

12. Curtis,A.,Byron,I.,andMcDonald,S.Integratingspatiallyreferencedsocialandbio-physicaldatatoexplorelandholderresponsestodrylandsalinityinAustralia.JournalofEnvironmentalManagement68,397–407,2003.

13. Curtis,A.,Byron,I.,andMacKay,J.Integratingsocio-economicandbiophysicaldatatounderpincollaborativewatershedmanagement.AmericanJournalofWaterResourcesAssociation41(3),549–563,2005.

14. Lesslie, R. Barson, M., and Smith, J. Land use information for integrated naturalresources management—a coordinated national mapping program for Australia.JournalofLandUseScience1(1),1–18,2006.

15. Flavel,R.,andRatcliff,C.MountLoftyRangesImplementationProject:EvaluatingRegionalChange.PrimaryIndustriesandResourcesSouthAustralia,Adelaide,2000.

16. Smith,J.,andLesslie,R.LandUseDataIntegrationCaseStudy:TheLowerMurrayNAP Region. Unpublished report to the National Land and Water Resources Audit.BureauofRuralSciences,Canberra,2005.

42963_C003.indd 60 11/6/07 7:10:46 AM


17. GlenelgHopkinsCMA.GlenelgHopkinsRegionalCatchmentManagementStrategy.GHCMA,Hamilton,2003.

18. Byron,I.,Curtis,A.,andMacKay,J.ProvidingSocialDatatoUnderpinCatchmentPlanningintheGlenelgHopkinsRegion.BureauofRuralSciences,Canberra,2004.

19. DahlhausP.,Heislers,D.,andDysonP.GHCMAGroundwaterFlowSystems.DahlhausEnvironmentalConsultingGeology,GHCMA,Hamilton,2002.

20. BureauofRuralSciences.GuidelinesforLandUseMappinginAustralia:Principles,ProceduresandDefinitions,3rded.AustralianGovernmentDepartmentofAgriculture,FisheriesandForestry,Canberra,2006.

21. NationalLandandWaterResourcesAudit.AustralianDrylandSalinityAssessment2000.Extent, Impacts,Processes,MonitoringandManagementOptions.NLWRA,Canberra,2001.

22. GlenelgHopkinsCMA.SalinityPlan.GHCMA,Hamilton,2002. 23. Sinischalchi,J.M.etal.Mappingsocialchange:Avisualisationmethodusedinthe

Monongahelanationalforest.SocietyandNaturalResources.19,71–78,2006. 24. Jenness,J.Nearestfeatures(nearfeat.avx)extensionforArcView3.x,v.3.8a.Jenness

Enterprises. (Available at http://www.jennessent.com/arcview/nearest_features.htm).2004.

25. Aslin,H.et al.Peri-urbanLandholdersandBio-security Issues—AScopingStudy.BureauofRuralSciences,Canberra,2004.

26. Spies,B.,andWoodgate,P.SalinityMappingMethodsintheAustralianContext.LandandWaterAustralia,Canberra,2004.

27. Baker, P., and Evans W. R. Mid Macquarie Community Salinity Prioritisation andStrategicDirectionProject.BureauofRuralSciences,Canberra,2002.

28. BureauofRuralSciences.LandUseMapping for theMurray-DarlingBasin:1993,1996,1998,2000Maps.AustralianGovernmentDepartmentofAgricultureFisheriesandForestry,Canberra,2004.

29. Walker, P., and Mallawaarachchi, T. Disaggregating agricultural statistics usingNOAA-AVHRRNDVI.RemoteSensingofEnvironment63,112–125,1998.

42963_C003.indd 61 11/6/07 7:10:47 AM

42963_C003.indd 62 11/6/07 7:10:47 AM

63

4 An Integrated Socioeconomic Study of Deforestation in Western Uganda, 1990–2000

Ronnie Babigumira, Daniel Müller and Arild Angelsen

Contents

4.1 Introduction................................................................................................... 634.2 Background...................................................................................................64

4.2.1 Uganda...............................................................................................644.2.2 TheForestrySector...........................................................................65

4.3 Deforestation.................................................................................................664.3.1 DefinitionsofDeforestation...............................................................664.3.2 GoodorBadDeforestation................................................................ 674.3.3 AConceptualFramework..................................................................68

4.4 DataandMethods.........................................................................................694.4.1 DataSources......................................................................................694.4.2 EconometricModel............................................................................ 704.4.3 MethodologicalIssues....................................................................... 71

4.5 ResultsandDiscussion.................................................................................. 724.5.1 DescriptiveStatistics.......................................................................... 724.5.2 EconometricResultsandDiscussion................................................. 74

4.5.2.1 SocioeconomicContext........................................................ 754.5.2.2 SpatialContext..................................................................... 764.5.2.3 InstitutionalContext............................................................. 76

4.6 ConcludingRemarks.....................................................................................77References................................................................................................................ 78

4.1 IntRoDUCtIon

Thepast20yearshasbeenaperiodofintensivestatisticalinvestigationintothecausesoftropicaldeforestation,withtheworkofAllenandBarnes1commonlyreferredtoasthearticlethatkicked-offthiseffort.Yetthereissurprisinglylimitedconvergenceonthebasicquestion:“whatdrivesdeforestation?”Thereareanumberofreasonsfor

42963_C004.indd 63 11/6/07 7:16:10 AM

64 Land Use Change

this.First,thesimplefactisthattheanswertothisquestioniscontextspecific—itisnotthesameconstellationoffactorsthatcanexplaindeforestationacrossthetropics.Second,onecanexpectsomeresearcherbias,inthesensethattheanswersprovidedreflect the researchers’background:geographical focus,discipline,politicalview,andsoforth.Third,thevariablesincludedhavedifferedgreatly—oftendeterminedbywhateverdataareeasilyavailable.Thesefactorshaveleadtodifferentandevencontradictorydeforestationstoriesbeingtold.Onewaytowardaconsensuswouldbebetterandmoreintegratedandholisticmethodologies.Thisbookmakesthecasefortheneedandroleforspatiallyintegratedmodelsofcouplednaturalandhumansystemsinthecontextsofstudyandmanagementoflanduse.

ThischapterisanempiricalapplicationofanintegratedapproachusingdatafromWesternUganda.Ourobjectiveistoanalyzetherolethatthecontextwithinwhichlanduseagentsoperateplaysintheirlandusedecisions.Todothisweintegratespatiallyexplicit socioeconomicandbiophysicaldata aswell asdataon landcover changesderivedfromremotesensingtoestimateaneconometricmodelofdeforestation.

Wearguelikeothersthatdeforestationismainlyaresultofactionsofagentsrespondingtoincentives.Indeed,overthepast20yearsmostanalystshavearguedthattropicaldeforestationoccursprimarilyforeconomicreasons,thatis,landusersconvertforesttononforestusesifthenewlandrenttheycangetishigherthanforforestuses.Thisapproachisbasedonthefactthatpeopleandsocialorganizationsare themost substantial agentsof change in forested ecosystems throughout theworld.2 Although this perspective is important, it is not the complete story oftropical deforestation. The incentives (land rent) are determined by the contextwithinwhichagentsoperate,andamorecomprehensiveanalysisneedstoincorpo-ratetheseaswell.

Followingabroadreviewofeconomicmodelsofdeforestation,AngelsenandKaimowitz3 recommended incorporation of agricultural census and survey datainto a geographic information systems framework. They argued that models thatcombineremoteobservationswithgroundbasedsocialdatawouldallowmodelerstotakeintoaccounttheroleofsocioeconomicfactorsandhavepotentialtoimproveourunderstandingofthedeterminantsoflandcoverchanges.3,4

Thischapterintroducesthreekeyaspectsofcontext,namelythesocioeconomic,spatial,andinstitutionalaspects.AfterabriefbackgroundonUgandaandthedefores-tationdebate,wepresentaframeworkofanalysisandthendataandmethods.Thekeyresultsarethenpresentedanddiscussed.

4.2 BACKGRoUnD

4.2.1 Uganda

Ugandaisalandlockedcountrycoveringabout236,000km2,81%ofwhichissuit-ableforagricultureowingtoarichendowmentofsoilsandaclimatethatisgenerallyfavorableforfarmingthroughouttheyear.5,6,7UgandaistoalargeextentdependentonnaturalresourcesbecausethemajorityofUgandansliveintheruralareaswithlow-inputlow-outputagricultureasthemainsourceoflivelihood.7,8

42963_C004.indd 64 11/6/07 7:16:11 AM

An Integrated Socioeconomic Study of Deforestation in Western Uganda 65

The country has enjoyed an impressive economic growth rate since the early1990s,amongthehighestinSub-SaharanAfrica.Thisisinsharpcontrasttoitsrecentpast.Thelate1970sandtheearly1980swerecharacterizedbyeconomicchaosthatresultedfromthecivilunrestoftheperiod.Macro-andmicroindicatorsofeconomichealthwerepoor,with low savings rates, high inflation rates, and ahigh externaldebtburden.Atippingpointinthistrend,however,wasthechangeingovernmentin1986.Thenewgovernmentthenembarkedonanumberofinitiativestorehabilitate,stabilize,andexpand theeconomy.Theresultof these initiativeswas theonsetofUganda’sownroaringnineties.Theexceptiontothispictureisthenorthernpartofthecountry,wherepoliticalinstabilityandviolencehaveemptiedthecountrysideinmanydistricts.Itisforthisreasonthatwedonotfocusonthewholecountry.

Additionally,populationhasbeengrowingatanaverageof2.5%peryear,9almostdoublinginjust22yearsfrom12.6millionin1980toabout24.7millionin2002.Duringthelatterpartofthisperiodgrowthwasevenhigher,withanaveragegrowthrateof3.4%between1991and2002.10Thepopulation isprojected to increase to32.5millionby2015.7

Giventhehighdependenceonnaturalresources,thecombinationofeconomicandpopulationgrowthwillundoubtedlyexertalotofpressureontheseresources.Ugandathereforeprovidesaninterestingstudyintohowthesesocioeconomicdimen-sionscouldhaveimpacteddeforestation(Figure4.1).

4.2.2 The ForesTry secTor

Priortothelate1990s,theextentofUganda’sforestestatewasbasedoneducatedguesses.Lackofcomprehensivedata limitedthedeterminationofforestareaandratesofdeforestation. Initialestimatesby theFoodandAgricultureOrganization(FAO)puttheforestandwoodlandcoverat45%ofthetotallandcoverin1890.Morerecentfigureshavebeeninthe20%to25%range.Forestandwoodlandareimportantbecauseonly3%ofUgandanhouseholdsinruralareasand8%inurbanareashaveaccesstogridelectricity;therestrelyonbiomassforenergysources.11Itisestimatedthatforestsprovideanannualeconomicvalueof$360million(6%ofGDP).Treesthroughfuelwoodandcharcoalprovide90%oftheenergydemandswithaprojec-tionof75%in2015.

ThefirstefforttomapUganda’soriginalvegetationwasdonebyLangdale-Brown12in1960whoestimatedtheextentofforestcoverfor1900,1926,and1925.13Thesedatashowanincreasingtrendintheannualrateofchangeofforestcover(Table4.1).

Thenext effort tomapUganda’s forest estatewasundertakenbyHamilton.13Usingsatelliteimagery,Hamiltontriedtomapoutclearstandingforest.Ourunder-standingofthismapisthatitfocusesonwhatissubsequentlyreferredtoastropicalhighforestbytheNationalBiomassStudy(NBS).ThemaprevealsthatforestisnotaparticularlycommontypeofvegetationinUganda.ThisledHamiltontoconcludethat visions of vast sweeps of mahogany-rich jungles, such as are entertained bysomeplanners,werequiteillusory.

AmorerecentandcomprehensiveattemptwasundertakenbyNBSinaprojectstartedin1989withtheobjectiveofprovidinguniqueinformationonthedistribu-tionandindirectlyconsumptionofwoodybiomassinthecountry.

42963_C004.indd 65 11/6/07 7:16:11 AM

66 Land Use Change

4.3 DeFoRestAtIon

4.3.1 deFiniTions oF deForesTaTion

Deforestationhasbeenusedtodescribechangesinmanydifferentecosystems.Itisgenerallydefinedaslossofforestcoverorforestedland,1,14whileVanKootenandBulte15defineitastheremovaloftreesfromaforestedsiteandtheconversionoflandtoanotheruse,mostoftenagriculture.FAOappliesasimilardefinition—aperma-nentchangefromforesttononforestlandcover,withforestbeingdefinedasanareaofminimum0.5hawithtreesofminimum5mheightinsitu,minimum10%canopycover,andthemainusenotbeingagriculture.

N

S

W E

Kilometers

Major roadAll–year roadWater body

YesNo

0 5025 100 150

Legend

TanzaniaRwanda

Kenya

Sudan

Congo, DRC

Kampala

North

Central East

West

Deforestation

FIGURe 4.1 (See color insert following p. 132.) Ugandastudyareashowingthedistri-butionofdeforestationwithinthewesternregionofthecountry.

42963_C004.indd 66 11/6/07 7:16:13 AM


Moredetaileddefinitionstakeintoaccountwhathappenstothedeforestedland,transitions among classes, the magnitude of change, the threshold in area abovewhichdeforestationissaidtohaveoccurred,aswellthetemporaldimensionsofthechange.16,17Astheprecisionindefinitionincreases,sodoesthelevelofcomplexityandthechallengesofempiricalwork.However,evenrecognizingtheimportanceofexactdefinitions,thecaseforprecisionshouldnotbeexaggerated.Causesofmajorundesirableforestinterventionscanbeanalyzedandpracticalimplicationsforpolicymakingderived,eveninaworldwitharelativelackofpureconceptualdefinitions.18

4.3.2 good or Bad deForesTaTion

Thedebateondeforestationcentersonwhethertropicaldeforestationisanimpendingenvironmentaldisaster,onewhichifnotaddressedwouldhavedireenvironmentalconsequences,orisjustanotheroverhypedagendabyenvironmentalistsandsomealarmistresearchers.

Fortheever-worseningschoolofthought,tropicaldeforestationisconsideredtobeamajorenvironmentalcrisis,becauseofitsinternationalimpactsonbiodiversitylossandclimateandbecauseofitslocalimpactssuchasanincreaseinfloodoccurrence,thedepletionofforestresources,andsoilerosion.19Suchfearsabouttheimminentextinctionoflargenumbersofplantsandanimalshavepromptedanoutpouringofconcernandanalysisabouttropicaldeforestationinthepasttwodecades.20

However, there is an it’s-not-that-bad school that is a less pessimistic schoolarguingthattherearenogroundsforthealarmistclaims.21Proponentsofthisschoolwouldgoontoarguethatdeforestationisanatural,beneficialcomponentofeconomicdevelopmentespeciallyindevelopingcountriesandisthereforenothingmorethanagradualhumanalterationofanabundantnaturalresource(land)inordertoincreaseproductivityandwelfare.

Theformerschool isgenerallymoreprominent,owingto thevisibilityof theimpactsofchangesinlocalandinternationalclimate,andhasresultedintheemer-genceofthesocialmovementdevotedtoreducingdeforestation.Importantquestionstherefore remainaboutwhy,despite the emergenceof this and thepublicationofhundredsofstudiesthatanalyzeditscauses,thedestructionoftropicalrainforestsdidnotappeartoslowdownmuch,ifatall,duringthe1990s.20

tABle 4.1early estimates of Forest Cover and Deforestation Rates

YearForest and moist thicket

(Ha) total area (%)Annual forest lossa

(HaY–1)

1900 3.1×106 12.7

1926 2.6×106 10.8 1.8×104

1958 1.1×106 4.8 4.7×104

a Owncalculations.Source: Langdale-Brown(1960).12

42963_C004.indd 67 11/6/07 7:16:14 AM

68 Land Use Change

4.3.3 a concepTUal Framework

Deforestationistheresultoftwobroadsetsofprocesses:naturalandhumaninducedprocesses. In the former, forest reduction is inducedbybioticandabioticgrowthreducingfactorswithintheforestecosystemorasaresultofbroadclimaticchangesorcatastrophessuchasfiresandlandslides.1Thesenaturalprocesses,however,areoftensoslowandsubtleastobeimperceptible.

Ontheotherhand,thechangesinitiatedbyhumanactivitytendtoberapidinprogression,drasticineffects,widespreadinscale,andthusmorerelevanttousonaday-to-daybasis.Understandingtherelationshipbetweenhumanbehaviorandforestchangethereforeposesamajorchallengefordevelopmentprojects,policymakers,andenvironmentalorganizationsthataimtoimproveforestmanagement.22

Toshedsomelightonthisrelationship,wetakeasourstartingpoint,ashaveothermodelsofdeforestationinthevonThünen(1826)tradition,thatanypieceoflandisputintotheusethathasthehighestnetbenefitsorlandrent.Thecenterofthediscussionisthenhowvariousfactorsdetermineandinfluencetherentaccruedfromforestversusnonforestuses,andtherebytherateofdeforestation.ArecentextensivereviewofthisapproachisgivenbyAngelsen.23

This approach is operationalized by modeling an agent (land use decisionmaker)livingatorwithaccesstotheforestmargin,whoseaimistomaximizethelandrent.(Wearemindfulofthepitfallsofapplyingaprofitmaximizingapproachtoruralhouseholds;however,westillbelievethisapproachisinformative.)Agentsareindividuals,groupsofindividuals,orinstitutionsthatdirectlyconvertforestedlandstootherusesorthatinterveneinforestswithoutnecessarilycausingdeforest-ationbutsubstantiallyreducetheirproductivecapacity.Theyincludeshiftingculti-vators, private and government logging companies, mining and oil and farmingcorporations, forestconcessionaires, and ranchers.18Themainculpritoragent isgenerally thought tobe theagriculturalhouseholddwellingat the forest frontier(this setting is plausible in Uganda given the dependence on forests for energyhighlightedabove).

Theagent’sdecisionsare influencedbyanumberof factorssuchaspricesofagricultural outputs and inputs, available technologies, wage rates, credit accessandconditions,householdendowments, forestaccess (bothphysicalandpropertyrights),andbiophysicalvariableslikerainfall,slope,andsoilsuitability.Location,thecenterofattentioninvonThünen’soriginalwork,doesinfluenceanumberofthesevariables(e.g.,pricesandwagerates).Thesefactorsaffecttheagent’sdecisionsdirectlyandare,therefore,referredtoasdecisionparametersorimmediatecausesofdeforestation(cf.theterminologyusedbyAngelsenandKaimowitz3).

Atthenextlevelisthecontextwithinwhichtheagentsoperate.Thesecontextualforcesdeterminedeforestationviatheir impactonthedecisionparameters.Thesecausesaremorefundamentalandoftendistancedinthesensethatitisdifficulttoestablishclearlinksbetweenthissetoffactorsanddeforestation.Theyareacomplexdynamicmixofthesocioeconomic,spatial,andinstitutionalsystemsofcommuni-tiesrepresentingthefundamentalorganizationofsocietiesandinteractinginwaysthataredifficulttopredict.TheabovediscussioncanbesummarizedinFigure4.2.

42963_C004.indd 68 11/6/07 7:16:14 AM


4.4 DAtA AnD MetHoDs

4.4.1 daTa soUrces

Landuseandlandcoverdataforthisstudycomefromlanduse/covermapsfromtheUgandaNBSandFAOAfricover.Althoughwerefertothemasthe1990and2000maps,thesatelliteimagesusedintheirproductionarefrom1989to1992and2000to2001,respectively,owingtotheneedtousecloud-freeimages.

The1990mapwasproducedbyvisualinterpretationofSpotXSsatelliteimageryfromFebruary1989 toDecember1992.Followingpreliminary interpretation, themapwasverifiedthroughsystematicandextensivegroundtruthing.The2000mapistheFAOAfricoverlandcovermapproducedfromvisualinterpretationofdigitallyenhancedLandsatThematicMapper(TM)images(Bands4,3,2)acquiredmainlyin theyear2000/2001.ThelandcoverclassesweredevelopedusingtheFoodandAgriculture Organization/United Nations Environmental Program (FAO/UNEP)internationalstandard(LCCS)landcoverclassificationsystem.The2000mapwasreclassifiedbystaffatNBStoenablecomparisonbetweenthetwomaps.

Administrative boundaries, infrastructure, and river maps come from theDepartment of Surveys and Mapping, Ministry of Lands, Housing and UrbanSettlementsandtheDepartmentofSurveysandMapping.Socioeconomicdataare

Natural Causes

Agents

Context

Subsistenceorientedfarmers LoggersCommercial

farmers

Spatial InstitutionalSocioeconomic

Rent (Agricultural)

Deforestation

FIGURe 4.2 Conceptual framework for analysis. Deforestation is influenced by naturalcausesandhumanactivities.Thehumanactivitiesaredrivenbytherentalcostoflandwithinsocioeconomic,spatial,andinstitutionalcontexts.

42963_C004.indd 69 11/6/07 7:16:16 AM

70 Land Use Change

fromtheNationalPopulationandHousingCensus1991,bytheStatisticsDepartment,MinistryofFinanceandEconomicPlanning.

The slopeandelevationwere calculated from thedigital elevationdataof theShuttleRadarTopographicMission(SRTM)(CGIAR-CSISRTM).Void-filledseam-lessSRTMdataV1,accessedJanuary2005,availablefromtheCGIAR-CSISRTM90m Database: http://srtm.csi.cgiar.org. Soil data are from Uganda’s agroecologi-calzones(AEZ)database24andfromtheresultsofasoilreconnaissancesurvey.25Followingconsultationswithoneof the authorsof thismap,weuse soil organicmatterandsoiltextureasthevariablestocapturesoilsuitability.Wethencalculateaweightedindexfrombothrastermaps.Thisindexactsasaproxyforagriculturalpotentialinherentinaparcel.

ThedifferentmapswereprojectedintoUniversalTransverseMercator(UTM)Zone36,southoftheequatorandthenassembledinarastergeographicalinforma-tionsystem(GIS)whereweresampledthedatatoacommonspatialresolutionof250m.Thechoiceofresolutionwasprimarilyguidedbytheneedforamanageabledatasize.

AGISwasused togenerateadditionalspatialvariables,specifically thecost-adjusteddistancetoroads,theeuclideandistancetowater,andtheeuclideandistancetoprotectedareas.WethenexportallthegridsasASCIIfilesandimportthemintoStata9,26whichweusetocarryoutthedescriptiveandeconometricanalysis.

4.4.2 economeTric model

Toanalyzetherolethatcontextplaysinlandusechange,weestimateaneconomet-ricmodelfortheprobabilitydeforestation.Ourunitofanalysisisa6.25hapixel.Underlyingthiseconometricmodelisalatentthresholdmodelbasedontheideathatthelandusedecisionregardingtheparcelismadebyanoperatorwhocanbeasingleperson,household,orgroupofpeopleinthecaseofcommonpropertyownership.27Thisoperatormayormaynotowntheparcel(ourdatadoesnotallowustomakethatdistinction).However,weassumethatforanygivenparcel,thereisanoperatorwhoisabletomakealandusedecisionpertainingtothisparcel.Aparcelwillbeclearedifitiseconomicallyprofitable.Thatis:

R R

nft f ft f+ +≥1 1| |

where Rnft f+1| representsthepresentvalueoftheinfinitestreamofnetreturnsfrom

convertingaparcelthatwasoriginallyunderforest( f)inperiodttononforest(nf)landuse inperiod t +1, whichwewill refer toasagricultural rent.This typeofmodelisfurtherdiscussedelsewhere.27,28Inlinewiththisintegratedapproach,theeconomicprofitabilityofaparcel isa functionof threesetsof factors: thesocio-economic,spatial,andinstitutionalcontexts.

1.The socioeconomic context within which the parcel is embedded has abearingonoutputpricesandinputcosts.Higheroutputpriceswillincreaseagricultural rent, while higher wages translate into higher input costs,whichreducetherentandmaythusreducetheprobabilityofdeforestation.

42963_C004.indd 70 11/6/07 7:16:18 AM


Wearguethatbecausetheopportunitycostoflaborinpoorcommunitiesistypicallyverylow,theprobabilityofdeforestationwillbehigherinpoorercommunities.Moreover,inequalitymayhaveabearingwithinthisframe-work. For any given average income, higher inequality implies a largerproportion of the population has an opportunity cost of labor below thelevelthatmakesforestclearingprofitable.Thuswehypothesizethathighinequalitywillbecorrelatedwithhigherprobabilitiesofdeforestation.

2.Thespatial contexthasaninfluenceontheagriculturallandrent.Includedinthisistheinsituresourcequality,thatis,theresponseofthelandtotheusewithout regard to its locationdetermines thequantityofagriculturalharvestpossiblefromagivenparcel,whichinturnaffectstheprobabilityofclearance.Alsoincludedistheaccessibilityand,byextension,allcostsandbenefitsassociatedwithaspecificlocationasopposedtoresourcequalityaswellasidiosyncraticlocation-specificcharacteristicsoftheparcel.Moreaccessibleparcelsaremorelikelytobecleared,andthisdoesnotnecessarilymeanthatagriculturewillbethesubsequentlanduse.Theseparcelswillbeclearedmainlyforthesaleoftimber.

3.Finally,theinstitutional contextwithinwhichtheagentsoperatealsohasaninfluenceonagriculturallandrent.Thisprimarilyreferstothepropertyrightsregimesinthecommunitiesthatdetermineaccessanduserights.Totheextentthattheyareenforceable,restrictionsonclearancetranslateintoacostandtherebyloweragriculturalrent.

We therefore select a number of explanatory variables that best capture the con-textsurroundingthemanagementoftheparcel.ThevariablesandtheiroriginsaredescribedinTable4.2togetherwithouraprioriexpectationsontheirrelationshipwiththelikelihoodofdeforestation.

Ourfocusisonagriculturalrentonly,whileforestrentisignored.Thissimpli-ficationcanbejustifiedontwogrounds:First,muchoftheforestisofdefactoopenaccessandtheforestrentthereforeisnotcapturedbytheindividuallanduser(unlikeagriculturalrent).Second,duringearlystagesintheforesttransition(characterizedbyhighlevelsofdeforestation,suchasinWesternUganda),changesinagriculturalrentratherthanforestrentarethekeydriver(cf.Angelsen23).

4.4.3 meThodological issUes

Conventional statistical analysis frequently imposes a number of conditions orassumptions on the data it uses. Foremost among these is the requirement thatsamplesbe random.Spatialdata almost alwaysviolate this fundamental require-ment,andthetechnicaltermdescribingthisproblemisspatial autocorrelation.29

Spatial autocorrelation (dependence) occurs when values or observations inspacearefunctionallyrelated.Spatialautocorrelationmayarisefromanumberofsourcessuchasmeasurementerrorsinspatialdatathatarepropagatedintheerrortermsorfrominteractionbetweenspatialunits.Itmayalsoarisefromcontiguity,clustering,spillovers,externalities,orinterdependenciesacrossspace.

42963_C004.indd 71 11/6/07 7:16:18 AM

72 Land Use Change

Threeapproachesforcorrectingforspatialeffectsareoftenmentionedin theliterature:regularsamplingfromagrid,purespatiallagvariablesusinglatitudeandlongitude indexvalues, and spatial lagvariables involving ageophysical variablesuchasaslopeorrainfall.30

Beforecarryingouttheeconometricestimation,wetestforspatialdependenceusing the SPDEP package31 in R language.32 We find evidence of spatial auto-correlationatboththepixelandparishlevels.Weminimizetheeffectsofspatialautocorrelationbyincludinglatitudeandlongitudeindexvariables,andbydrawingasamplefromagridwithadistanceof500mbetweencells.

4.5 ResUlts AnD DIsCUssIon

4.5.1 descripTive sTaTisTics

Mostdeforestationwasconcentratedinafewareas.Aplotofcumulativedistribu-tionofdeforestationshowsthat15%oftheparishesaccountedfor70%ofthetotaldeforestation(Figure4.3).Furthermore,mostofthedeforestation(60%)waswithin10kmfrommainroads(Figure4.4).

tABle 4.2Description of Variables

Variable Description source expected signa

socioeconomic Context

head_emp Employedhouseholdheads Census91 –

educ_Gini EducationGinicoefficientb Census91 +

popdens Populationdensity Census91 +

mig_share Shareofmigrantsinparish Census91 +

spatial Context

cdcity_allrds Costadjusteddistancetoroads Infrastructuremap –

dwater Distancetowater Infrastructuremap –

slp Slope DEM –

elev Elevation DEM –

soil+2cl Proportionofsuitablesoils CIAT +

rain Rainfall CIAT ?

x Latitudeindexvalue LUC&infrastructuremaps ?

y Longitudeindexvalue LUC&infrastructuremaps ?

Institutional Context

dprotect Distancefromprotectedareas LUC&infrastructuremaps ?

prtct Protectedareadummy LUC&infrastructuremaps –a A priori expectationson theeffectofvariablesondeforestation (–) less; (+)moredeforestation;

(?)ambiguous.CIAT,InternationalCenterforTropicalAgriculture;DEM,digitalelevationmodel;LUC,landusecover.

b TheEducationGinicoefficientisameasureofinequalityrangingfromzero(perfectequality)toone(perfectinequality).

42963_C004.indd 72 11/6/07 7:16:19 AM


Somedescriptive statistics arepresented inTable4.3.Compared to all otherparcels,forestparcelshadmorerainfall,wereatalowerelevation,andwereonlesssteepslopes.Not surprisingly, forestparcelsweregenerally located farther fromurban centers and farther from themain roads.This is typical of avonThünendevelopment process, with areas close to urban centers being cleared first and

1009080706050403020100Parishes – cumulative %

Obs

erve

d D

efor

esta

tion

– cum

ulat

ive %

Deforestation is accumulated starting with highest valueSource: NBS, FAO Africover

100

90

80

70

60

50

40

30

20

10

0

FIGURe 4.3 CumulativedeforestationbyparishesacrossUgandastartingwiththelargestpercentagearea.

Parcels within 10 KM

60.75

28.32

10.93

Parcels between 10 KM and 20 KM Parcels 30 KM or more

Perc

enta

ge lo

ss

60

40

20

0

FIGURe 4.4 Relationshipbetweendeforestationanddistancefromroads.

42963_C004.indd 73 11/6/07 7:16:23 AM

74 Land Use Change

expandingaspopulationand theeconomygrows.Most forestparcelswere inorclosetoprotectedareas.

Withintheforestedparcels,theonesthatweredeforestedwereatalowereleva-tionwithlesssteepslopes.Thissuggeststhataccessibilitywasakeyfactorinthedecisiontoclearaforestparcel,consistentwithourexplanationabove.

4.5.2 economeTric resUlTs and discUssion

Giventhebinarynatureofthedependentvariable,thatis,thelandiseitherclearedor it isnot,weestimateabinary logitmodel.Wecorrect forpossiblecorrelationintheerrortermsofpixelswithinaparishandusetheHuberandWhitesandwichestimatortoobtainrobustvarianceestimates.

The econometric results are presented in Table4.4. The dependent variableisacategoricalvariable indicatingwhetherornot theparcelwasdeforested.The

tABle 4.3Descriptive statistics

All Parcels

(n = 697,060)

Mean Minimum Maximum

Proportionofsuitablesoils 64.89 32 91

Rainfall(mm) 1,091.05 701 1,949

Elevation(meters) 1,296.52 601 4,391

Slope 6.01 0.00 63.66

Cost-adjusteddistancetoroads 0.51 0.00 1.67

Distancetourbancenters(km) 64.68 0.00 167.38

Distancetoroad(km) 8.74 0.00 38.57

Distancefromprotectedareas(km) 6.83 0.00 38.21

Distancefromwater(km) 16.29 0.25 57.34

Forest Parcels

All Deforested nondeforested

(n = 194,601) (n = 46,420) (n = 148,181)

Mean Mean Mean

Proportionofsuitablesoils 62.51 63.78 62.11

Rainfall(mm) 1,177.46 1,148.02 1,186.69

Elevation(m) 1,230.12 1,052.06 1,285.91

Slope 5.2 3.77 5.65

Cost-adjusteddistancetoroads(km) 0.65 0.62 0.66

Distancetourbancenters(km) 82.91 85.97 81.95

Distancetomainroad(km) 9.98 8.97 10.29

Distancefromprotectedareas(km) 3.23 4.78 2.75

Distancefromwater(km) 19.3 18.28 19.62

42963_C004.indd 74 11/6/07 7:16:23 AM


regressionmodelwas significantat thep= .001 level.Mostcoefficientshave theexpectedsigns.Belowwediscussthestatisticallysignificantresults.

4.5.2.1 socioeconomic Context

Theresultsshowthatdeforestationismorelikelytooccurinbetter-offcommunities.Ourproxy forwealth in thecommunity—proportionofhouseholdheads that areemployedoutsidethefarm—isstatisticallysignificantatthe5%level,withanoddsratio(foronestandarddeviationincrease)of1.2.Therearetwocontradictoryeffectsof the better-off farm employment opportunities reflected by this variable. First,goodemploymentopportunitiesincreasetheopportunitycostsoflaborandtherebyloweragriculturalrentandreducethepressureonforestconversion.But,ahigher

tABle 4.4logit Model Results of Deforestation in Western Uganda, 1990–2000

Variable Coef. p-value e^(b*sdx)a

socioeconomic Context

Proportionofheadsthatareemployed 0.828* 0.034 1.202

EducationGinicoefficient 0.329 0.675 1.028

Populationdensity(pp/ha) 0.291** 0.008 1.166

Shareofmigrantsinparish 1.363** 0.008 1.302

spatial Context

Costadjusteddistancetoroads –1.318** 0.004 0.709

Distancetowater –0.004 0.503 0.948

Slope 0.051*** 0.000 1.366

Elevation –0.002*** 0.000 0.339

Proportionofsuitablesoils –0.004 0.579 0.948

Rainfall –0.002 0.054 0.784

X –0.003*** 0.000 0.450

Y –0.003*** 0.000 0.409

Institutional Context

Protecteddummy –1.912*** 0.000 0.388

Distancefromprotectedarea –0.003 0.843 0.985

Constant 8.250*** 0.000

No.ofparcels 43,760

Modelp-value 0.000

PseudoR2 0.191a ChangeinoddsforoneSDincreaseinx.Dependentvariable=1ifparcelchangedfromforesttononforestclass,0otherwise.Percentageofcorrectpredictions=76.3.*,p<.05;**,p<.01;***,p<.001.

42963_C004.indd 75 11/6/07 7:16:24 AM

76 Land Use Change

shareofoff-farmemploymentisalsocorrelatedwitheconomicdevelopment,creat-inghigherdemandforagriculturalproducts.OurresultssuggestthatintheWesternUgandancontextthelattereffectisdominating.

Consistentwith thisexplanation,wealsofindthathigherpopulationdensitieshaveapositiveimpactonforestconversion.Ahighshareofmigrantsintheparishalsopullsinthesamedirection.Migrantshaveinitiallynoorverysmallparcelsofagriculturallandandcanthereforebeexpectedtobecomemajoragentsofdeforesta-tion.Thustheempiricalmodelsuggeststhatmigrationtotakeadvantageofforestlandsuitableforagriculturalconversionplaysamajorrole.Inequality,asmeasuredbytheeducationalGini,doesnotappeartohaveanysignificanteffectonthelikeli-hoodofdeforestation.

4.5.2.2 spatial Context

Parcelsclosertoroadsweremorelikelytobedeforested.Thedescriptivestatisticshave already shown this, and the econometric results confirm the importance ofdistanceasafactor.Thisresultsuggests,inlinewithnumerousotherstudies,thatinterventionsthatreducethecostofaccesstoforestedlandwillincreasethelikeli-hoodofdeforestation.

Notsurprisinglywefindthatparcelsatlowerelevationweremorelikelytobedeforested, suggestingagain that the spatial contexthadabearingon thecostofaccess.Aonestandarddeviationincreaseintheelevationreducedtheoddsofdefor-estationby0.34.Asurprisingresultis,however,theeffectofslopeontheprobabilityofdeforestation.Weanticipatedthattheprobabilityofdeforestationwouldbenega-tivelycorrelatedwiththeslope(mainlyowingto thehighercostsassociatedwithworkingathigherelevationandsteeperslopes).However,wefindinsteadthatsteeperslopesweremorelikelytobedeforested,andwedonothaveasatisfactoryexpla-nationof this result.Aplausibleexplanationcouldbe that lower landshavebeenconvertedandthepressuremayhaveshiftedtothemarginallands.Supportforthiscanbefoundintheargumentthatfragilelandsinsub-SaharanAfricaarefacingaworseningsocialandenvironmentalcrisis.33

Distance towater, rainfall, andproportionof suitable soilswerenot statisti-cally significant. One reason for the insignificance of the soil variable might bethatthemaponwhichthisvariableisbasedisrathercoarse,andthereforedoesnotcapturetherelevantlocalspecificvariationthatmayexist.Thesamemaybetrueforrainfall.

4.5.2.3 Institutional Context

Aninterestingresultthatemergesfromthisworkisthefactthatinstitutionalinterven-tionsseemtohavemitigateddeforestation.Parcelsinprotectedareaswerelesslikelytobedeforested.Thisisalsowhatonecanobserveonthemapsandontheground:theprotectedareasareindeed“greener.”Theconservationareashavebeenbackedbyrelativelystrongenforcementatthelocallevelswithpunishmentofviolators.

In addition, we tested if the conservation led to more land being deforestedoutsidetheconservationareas,akindofnegativespillovereffect.Ourhypothesiswould thenbe that,aftercontrolling forother factors, landclose to theprotected

42963_C004.indd 76 11/6/07 7:16:24 AM


areasshouldexperiencehigherforestconversion.Butthisvariableisnotstatisticallysignificantthuswecouldnotrejectthenullhypothesis.

4.6 ConClUDInG ReMARKs

The process of land use change is driven by a complex web of factors that cutsacrossdisciplines.Thismeansthateffortstoaddressthelandusechangeprocessshouldsimilarlybeholisticandcutacrossdisciplines.Thischapterisanexampleof how such a study couldbe empirically carriedout.Weargued that additionalinsightscouldbegainedfromintegratingspatiallyexplicitsocioeconomic,institu-tional,biophysical, and landcoverdata. Increasingavailabilityofhigh resolutionspatialdatameansthatsuchanapproachispossibleinmostplaces.Thefact thatthevariablesusedtocapturethesocioeconomiccontextaresignificantshowsthatsuchaframeworkcanbeofpolicyrelevance,forexample,byincludingtheeffectonnaturalresourcesinthedesignandimplementationofhumanresettlementprogramsorinfrastructuredevelopmentprojects.

Four main stories emerge from our econometric analysis: the poverty cumaffluence,thepopulation,theprotection,andthespatialstory.Althoughwesetoutthinkingdeforestationwasdrivenbypoorhouseholds,wedonotfindanyevidenceinsupportofthis.Ratheritappearsthatdeforestationismorelikelyinthebetter-offcommunities. Second, high population densities, together with a high proportionofmigrantswhomaybe ingreaterneedforagricultural land,hasalsoplayedanimportantrolefordeforestationinWesternUgandaduringthe1990s(Figure4.5).

(a) (b)

(c)

FIGURe 4.5 Landusechange inUganda. (a)Forestclearing. (b)Bananaplantationsonclearedland.(c)Pastorallanduseonclearedland.

42963_C004.indd 77 11/6/07 7:16:26 AM

78 Land Use Change

Third,thereisalsoastrongspatialstorytothisintermsoffactorssuchasclosenesstoroadsandlowelevation,leadingtomoredeforestation.Finally,ourstudyshowsthatprotectionhadbeeneffectiveinreducingthelikelihoodofdeforestation.

ReFeRenCes

1. Allen, J.C., andBarnes,D.F.The causesof deforestation indeveloping countries.Annals of the Association of American Geographers75(2),163–184,1985.

2. Wear,D.N.,Abt,R.,andMangold,R.People,spaceandtime:Factorsthatwillgovernforestsustainability.In:Transactions of the 63rd North American Wildlife and Natural Resources conference.WildlifeManagementInstitute,Washington,D.C.,1998.

3. Angelsen,A.,andKaimowitz,D.Rethinkingthecausesofdeforestation:lessonsfromeconomicmodels.World Bank Research Observer14(1),73–98,1999.

4. Rindfuss,R.R.,andStern,P.C.Linkingremotesensingandsocialscience:Theneedand challenges. In: Liverman, D., et al., eds., People and Pixels: Linking Remote Sensing and Social Science.NationalAcademyPress,Washington,D.C.,1–27,1998.

5. Mukiibi,J.K.AgricultureinUganda.In:Mukiibi,J.K.,ed.,Agriculture in Uganda.FountainPublishers,Kampala,2001.

6. Esegu,J.F.O.ForesttreegeneticresourcesinUganda.In:Mukiibi,J.K.,ed.,Agricul-ture in Uganda.FountainPublishers,Kampala,2001.

7. NEMA.State of the Environment Report for Uganda 2000/2001.MinistryofWater,LandsandtheEnvironment,Kampala,2000.

8. MAAIFandMFPED.Plan for Modernisation of Agriculture: Eradicating Poverty in Uganda.Kampala,2000.

9. NEMA.State of the Environment Report for Uganda 1998.MinistryofWater,LandsandtheEnvironment,Kampala,1998.

10. MFPED.Background to the Budget for Financial Year 2004/05,Kampala,2004. 11. MFPED.Poverty Eradication Action Plan (2004/5–2007/8),Kampala,2004. 12. Langdale-Brown, I., The VegetationofUganda.UgandaDepartment ofAgriculture

2(6),1960. 13. Hamilton,A.C.Deforestation in Uganda.OxfordUniversityPress,Nairobi,1984. 14. Geist,H.J.,andLambin,E.F.What Drives Tropical Deforestation?Louvain-la-Neuve,

2001. 15. Van Kooten, G. C., and Bulte, E. The Economics of Nature: Managing Biological

Assets.Malden,MA,Blackwell,2000. 16. FAO. State of the Worlds Forests, 1997. Food and Agriculture Organization of the

UnitedNations(FAO),Rome,1997. 17. Myers,N.Conversion of Tropical Moist Forests: A Report Prepared by Norman Myers

for the Committee on Research Priorities in Tropical Biology of the National Research Council.NationalAcademyofSciences,Washington,D.C.,1980.

18. Contreras-Hermosilla,A.The Underlying Causes of Forest Decline.CenterforInter-nationalForestryResearch(CIFOR),Bogor,Indonesia,2000.

19. Caviglia,J.L.,andKahn,J.R.Diffusionofsustainableagriculture in theBraziliantropical rain forest: A discrete choice analysis. In: Orlove, B. S., ed., Current Anthropology.UniversityofChicagoPress,Chicago,2001.

20. Rudel,T.K.Tropical Forests: Regional Paths of Destruction and Regeneration in the Late Twentieth Century.Chichester,ColumbiaUniversity,NewYork,2005.

21. Lomborg,B.The Skeptical Environmentalist: Measuring the Real State of the World.CambridgeUniversityPress,Cambridge,2001.

42963_C004.indd 78 11/6/07 7:16:27 AM


22. Southworth, J., and Tucker, C. The influence of accessibility, local institutions, andsocioeconomicfactorsonforestcoverchangeinthemountainsofwesternHonduras.Mountain Research and Development21(3),276–283,2001.

23. Angelsen, A. Forest cover change in space and time: combining von Thünen andthe forest transition. In: World Bank Policy Research Working Paper, World Bank,Washington,D.C.,2006.

24. CIAT.Bean Database for Africa and Uganda’s Agroecological zones (AEZ) Database. CD-ROM,Ver.1.0,Kampala,1999.

25. MemoirsoftheResearchDivision,KawandResearchStation.DepartmentofAgricul-ture,Uganda,1960.

26. StataCorp.Stata Statistical Software: Release 9,CollegeStation,TX,2005. 27. Nelson, G. C., and Geoghegan, J. Deforestation and land use change: Sparse data

environments.Agricultural Economics27(3),201–216,2002. 28. Munroe,D.,Southworth,J.,andTucker,C.M.Thedynamicsofland-coverchangein

western Honduras: Spatial autocorrelation and temporal variation. Prepared for the2001AAEAAnnualMeetings,2001.

29. O’Sullivan,D.,andUnwin,D.Geographic Information Analysis.JohnWiley&Sons,Hoboken,NJ,2003.

30. de Pinto, A., and Nelson, G. C. Correcting for spatial effects in limited dependentvariableregression:Assessingthevalueof“ad-hoc”techniques.PresentedatAmericanAgriculturalEconomicsAssociationAnnualMeeting,LongBeach,CA,2002.

31. Bivand, R. Implementing spatial data analysis software tools in R. Geographical Analysis38(1),23–40,2006.

32. RDevelopmentCoreTeam.R: A language and environment for statistical computing.RFoundationforStatisticalComputing,Vienna,Austria,2005.

33. Babu,S.,andHazell,P.Growth,povertyalleviation,andtheenvironmentinthefragilelandsof sub-SaharanAfrica. In:Strategies for Poverty Alleviation and Sustainable Resource Management in the Fragile Lands of Sub-Saharan Africa.DeutscheStiftungfürInternationaleEntwicklung,Kampala,1998.

42963_C004.indd 79 11/6/07 7:16:27 AM

42963_C004.indd 80 11/6/07 7:16:28 AM

81

5 Modeling Unplanned Land Cover Change across Scales

A Colombian Case Study

Andres Etter and Clive McAlpine

Contents

5.1 Introduction................................................................................................... 815.2 StudyRegion................................................................................................. 83

5.2.1 RegionalLevel...................................................................................845.2.2 LocalLevel........................................................................................85

5.3 Methods.........................................................................................................855.3.1 NationalLevel....................................................................................865.3.2 RegionalLevel...................................................................................865.3.3 LocalLevel........................................................................................88

5.4 Results...........................................................................................................885.4.1 NationalDeforestationThreatHotSpotsandMainDrivers.............885.4.2 AmazonColonizationFronts.............................................................905.4.3 DeforestationPatternsattheLocalLevel.......................................... 91

5.5 DiscussionandConclusions..........................................................................92Acknowledgments....................................................................................................96References................................................................................................................96

5.1 IntRoDUCtIon

Landuseistheinteractionbetweenhumansandthebiophysicalenvironmentwithcumulativeimpactsonthestructure,function,anddynamicsofecosystemsatthelocal, regional, and global levels of ecological organization.1 Human impacts ontheEarth’senvironmentareleavinganincreasingecologicalfootprint,whichnowthreatensmanyglobalecosystems.2,3Humanactivitiesthatproducechangesinlandcover, such as agriculture, mining, and urban development, are a major cause ofthe ecological footprint. Over the past century, the global human population hasincreased3.5 times,while the areaof agricultural landhasdoubled.4From1980thetropicshaveexperiencedhigherdeforestationthantemperateregions,withthe

42963_C005.indd 81 11/6/07 7:14:27 AM


largestconcentrationintheAmazonBasin.5Changesinlandcoversuchasdefor-estationaffect the functioningofecological systemsatmultiple scales,withcon-sequencesrangingfromglobalclimatechange,soilandhydrologicaldegradation,toincreasedbiologicalextinctions.6,7,8,9,10Humanpressuresontheenvironmentareexpectedtoincrease,atleastinthenearfuture,asaresultoftheexpandingglobalpopulationandincreasinglevelsofconsumptionandwasteaccumulation.3Althoughthestakesarehigh,knowledgeoftheprocessesthatunderliehuman-inducedlandcoverchangeisstilllimited.7Thisisanemergingissueinspatiallyexplicitenviron-mentaldisciplinessuchaslandscapeecology.11,12

Theproximatecausesofhuman-inducedchangesinlandcoverarisefrombothbroad-scaleclearingofnaturalvegetationforagriculture,miningandurbandevel-opment,andfromhabitatmodificationsresultingfromforestryandalteredgrazingand fire regimes. However, the ultimate causes are the biophysical and culturalfactors that influence where and at what pace habitat clearing and modificationoccurs,fueledbytheglobalpopulationgrowthandtheconsumptionofresources.7Toreducethehumanecologicalfootprintandencouragesustainablelanduse,weneedtounderstandtheseultimatecauses.

Most human-induced land cover change is directly related to land use. Landuse involves the exploitation of land and its resources and largely represents theinteractionbetweenhumansandthebiophysicalenvironment.1Landcoverchangeoccursatdifferentspatialscales,includingthelocal(102to103km2),regional(104to106km2), and national (106 to 108 km2), or continental and global geographicalscales.Landuse isaprocessdrivenbysocial,economic,political, technological,and cultural factors, but is spatially and temporally constrained by biophysicalconditions such as climate, soils, water availability, accessibility, and biologicalresources. These drivers and constraints, and their interactions, need to be con-ceptualizedandanalyzed inan integratedmultidisciplinary framework ifwearetodevelopanadequateunderstandingofthewaysinwhichlanduseprocessesleadto land cover change. Bürgi et al.12 point out that convincing integrative studiesarestilllargelylacking,withseveralkeyareasrequiringinvestigation:(i)movingfrompattern-orientedtoincreasinglyprocess-orientedstudies,whichincludebroadtime frames and historical approaches; (ii) devising methods for extrapolationandgeneralitybuilding; (iii)findingways to linkdataof different qualities; and(iv)effectivelyincorporatingcultureasadrivingforceofchange.Globalpressuresofhuman-landscapeinteractionsandtheirfeedbackshaveincreasedasaresultoftheworld’ssociopolitical,economic,andculturalglobalizationprocessesduringthelate20thcentury.Becauseofthis,opportunitiesandconstraintsforlandusechange,inparticularnewlanduses,increasinglydependontheglobalfactorsinfluencingmarketsandpoliciesatdifferentorganizationallevels.

Historically, the major changes in global land cover occurred in temperateregions, but land cover change, especially deforestation, is now concentrated intropicalregions,leadingtolargereductionsinnetprimaryproduction.8ItiswidelyrecognizedthatdeforestationratesinLatinAmerica,especiallyinthetropicalrainforestsoftheAmazonBasin,arecurrentlythehighestintheworld.6,13,14AbouthalfofthelandclearinginLatinAmericahastakenplacesince1960,15withtheAmazonBasinaccounting fora largeproportionof recentglobaldeforestation.5Themost

42963_C005.indd 82 11/6/07 7:14:27 AM

Modeling Unplanned Land Cover Change across Scales 83

studied region of the Amazon is the Brazilian Amazon,16,17,18,19,20,21,22,23,24 whichshowsthatmostofthelandclearinghasresultedfromgovernment-orientedcoloni-zationschemesandinfrastructurebuilding.

TheColombianAmazonregion is lesswellstudied,although it iscurrentlyamajor global deforestation hot spot.5,25 Unlike the Brazilian Amazon, clearing isoccurring in an unplanned way, with peasants migrating freely into the regionattractedbyanillegaldrugeconomyandthepossibilitytoownland.OtherregionsofColombiasuchastheAndesandtheCaribbeanarealreadyheavilyimpactedbyhuman land use.26 In Colombia, a growing population and increasing integrationintotheglobaleconomyarebringingincrementalpressuresonnaturalecosystems.Colombiaisinternationallyrecognizedforitshighbiologicaldiversity,bothintermsof species richness and endemism.27,28,29,30 However, knowledge of the processesanddriversofdeforestationandconsequentthreatsarelimited.28Thereneedstobeamuchbroaderanddeeperunderstandingof thespatialand temporalpatternsofchangeandthefactorsdrivingthesechanges.Thisnewknowledgewillallowmoremeaningfulassessmentsoftheimpactsofdeforestationonecosystemsandbiodiver-sityandprovideusefulinputsforlanduseandconservationplanning.

Inthischapterwepresentamultiscalesynthesistocontributetotheunderstandingofthepatterns,processes,anddriversofunplannedlandcoverchangeinthetropics,basedonColombiaasastudycase.31Weaddresslandcoverchangeatthenational,regional,andlocallevels,andanalyzethevariabilityofdrivers,spatialpatterns,andratesofchange,focusingonthespatiallydominantforestecosystems.Atthenationallevel,attentionisdirectedtowardinvestigatingthebroadpresentandfuturelandcovertrendsacrossthemainbiogeographicregions.Attheregionallevel,wefocusonthecolonizationoftheAmazonregion,whiledetailedanalysisatthelocalleveliscarriedoutinsixselectedsampleareasoflowlandhumidtropicalforests.Finally,wediscusstheimplicationsoftheresearchoutcomesforconservationplanning.

5.2 stUDY ReGIon

Colombia covers a land areaof 1.1millionkm2,with largevariations in altitude(0to5,800m),meanannualrainfall(300to10,000mm),lengthofgrowingperiod(60to360daysperyear1),andahighdiversityofgeologicalsubstrates,givingrisetoanunusualhighenvironmentalvariabilityrelativetoitsgeographicsize.Colombianecosystemsrangefromdesertandtropicalsavannastoveryhumidrainforestsandtropical snow-coveredmountains.Asa consequenceof thisvariability,Colombiahashighlevelsofendemismandspeciesrichnessandhasbeenclassedasamega-diversecountry.28,32Thecountrycanbedividedintofivebiogeographicregionswithspecificbiophysicalandlandusecharacteristics(Figure5.1a):Andes(278,000km2),Caribbean (115,400 km2), Pacific Coast (74,600 km2), the Colombian Amazon(455,000 km2), and Orinoco plains (169,200 km2); and two smaller regions, theMagdalena(37,100km2)andCatatumbo(7,000km2),whicharegenerallyincludedintheAndeanregion.

ThetotalpopulationofColombiaiscurrentlyaround40million,25%ofwhichstill live in rural areas. The population has been historically concentrated in theAndean andCaribbean regions.33,34These regions currentlyhave an average rural

42963_C005.indd 83 11/6/07 7:14:28 AM


populationdensityofapproximately33personsperkm2,whilethePacific,Orinoco,andAmazonregionshavelowerdensitiesrangingfrom5to17personsperkm2.Duringthe20thcentury,thepopulationgrew10-fold,havingastrongimpactonthelandscape.Howeversincethe1970s,Colombiahasbecomeanincreasinglyurbanandindustrial-izedcountry,with thenationalpopulationgrowthratefallingbelow2%in the late1990s,andtheruralpopulationstabilizing.AlthoughthecountryunderwentastrongracialmixingresultinginadominantMestizopopulation,theculturaldiversityisstillhigh,withregionallycontrastingruralculturesof99ethnicgroupswith101languages,varyingfromAmerindiantoAfro-AmericanandEuropean-American.35

The economy is based on mining (oil, coal, and nickel), agriculture (coffee,flowers),andindustrialexports.Inrecentdecades,Colombiahasexperiencedcon-siderablesocialandpoliticalunrest,drivenbyextremistleft-andright-wingarmedforces, triggering large internalpopulationmovements andeconomicdestabiliza-tion.Paralleltothisunrest,apervasiveeconomyofillegalcrops(cocaandopium)forexporthasdevelopedinmanyremotefrontierareas,causingfurthersocialandpolitical instability. These internal social and political pressures have importantconsequencesforregionalpatternsoflandcoverchange,includingaredirectionofcolonizationpatternsandlandabandonmentincertainareas.

5.2.1 Regional level

Theregionalstudyarea(23,000km2)islocatedintheCaquetáDepartmentintheAndeanfoothillsoftheeasternAmazonregionofColombia(Figure5.1b).Itincludes17municipalities(wholeorinpart)withatotalruralpopulationofabout180,000in1993.Populationdensitiesrangefrom7to16inhabitantsperkm2ofclearedland.The

Regional case study

N

a)

Caribb

eanM

agdala

na

Paci

ficAndes

b)Cleared forestsExtant forests

Local‒level case studies

750 Km5002500

Catatumbo

Orinoco

Amazon

FIGURe 5.1 LocationofColombia showing: (a) the seven regions; and (b) extentof theoriginal(gray+black)andremnantforestedecosystems(black),andnonforestedecosystems(white);andlocalandregionalstudyareas(red).

42963_C005.indd 84 11/6/07 7:14:30 AM


mostextensivelanduseiscattleranching,withillegalcropsofcoca(Erythroxylumcoca)coveringasmallerareabutincreasingineconomicimportancesincetheearly1980s.Thelatestfiguresaredebated,butindicateadropofalmost50%inillegalplantationsareainColombiafromahighin2001.36

5.2.2 local level

The local-levelanalysis focusedonsix landscapes located in thehumid lowlands(>2,000mmmeanannualrainfall)ofthecentral(Magdalena)andeastern(Orinoco,Amazon, and Catatumbo) regions of Colombia. The landscapes comprise areasrangingfrom68to128km2,whichhavebeensubjecttosubstantiallandclearingduringthepast30to60years(Figure5.1b,Table5.1).Lessthantwomillionpeople(5%)currentlyliveintherurallowlands.Intheseareas,theruralpopulationdensityisverylow,varyingfromless than5to15inhabitants/km2,with lowestdensitiesoccurringintheextensivecattlegrazinglandscapesintheCaribbean,Orinoco,andAmazonregions.

5.3 MetHoDs

Thedatasetsandmethodsdescribedhereinwereusedtoanalyzethespatialpatternsof forest conversion, determine underlying drivers of this process, and predict

table 5.1General biophysical Characteristics and Data sets of the local study sites

study area

area (km2) Region

Dates of air-photo coverage

Mean altitude

(m)

Mean annual rainfall (mm)/

no. dry months (<100 mm)

original vegetation in

landscape

LaBalsa 128 Orinoco 1938,1961,1979,1987,1992,2001

250 2500/4 Forest–Savannamosaic

Guamal 140 Orinoco 1938,1961,1979,1987,1992,1997,2001

300 3000/3 Forest–Savannamosaic

Caquetá 100 Amazon 1946,1975,1985,1992,2000

200 3100/2 Forest

Opón 68 Magdalena 1971,1985,1996,2002

200 3200/1 Forest

Berrío 82 Magdalena 1950,1961,1977,1985,1996,2002

120 3000/3 Forest

Tibú 74 Catatumbo 1961,1975,1985,2000

150 2700/3 Forest

Source: Etter,A.etal.40Withpermission.

42963_C005.indd 85 11/6/07 7:14:31 AM


forestedareaswithahighprobabilityofconversioninthefuture.Thestudyapplieda multiple-scale modeling approach using logistic and ordinal regression, regres-siontreeanalysis,andgeographicalinformationsystems(GIS).Landcoverchangesweremodeledwithbiophysicalandsocioeconomicvariablestoanalyzeandpredictdeforestationandregenerationprocesses.Foralllevelsofanalysis,theresponsevari-ablewasabinary(1,0)mapofforestcover,whiletheexplanatoryvariablesincludedsoilfertility,rainfall,moistureavailabilityindex,slope,accessibility,andprotectedareasforthenationallevel,andsoilfertility,accessibility,andlandcoverneighbor-hoodofforestandsecondaryvegetationfortheregionalandlocalanalyses.Weusedabroadrangeof informationsources, includingremotelysenseddata fromaerialphotographsandsatelliteimages,andsecondarysourcesofbiophysicalandsocio-economicdata,aswellashistoricaldata(Table5.2).

5.3.1 national level

Atthenationallevel,thestatisticalmodelingcomprisedtwosteps:(i)predictingthespatial location of forest conversion using logistic regression and regression trees;(ii)refiningthepredictionswiththepopulationgrowthratedata.Weemployedbothmodeling techniquesas they treat thespatialvariation in theeffectofexplanatoryvariablesdifferently.37Logisticregressionmodelstheeffectofvariablesinaspatiallyhomogeneousmanner,whereasclassificationtreescantreattheeffectinaspatiallyheterogeneousmanner.Usingtheresultsofthebest-performingmodel,weidentifiedareas with a high probability of conversion as those with a predicted probabilitygreaterthan70%.Weoverlaidtheseareaswitharecentruralpopulationgrowthratemaptoidentifyareaswithahighprobabilityofconversionandapopulationgrowthrategreaterthan2%.Wedefinedtheseareasas“deforestationhotspots.”AllanalyseswereperformedusingS-PLUS.38AqualitativeassessmentoftherelativeriskofforestconversionforthedifferentremnantnaturalforestecosystemswasdonebyoverlayingthepredicteddeforestationhotspotswiththeecosystemmapbyEtter.26

5.3.2 Regional level

Using thebinary forest/nonforestmaps, forestproportionzoningmapswerepro-ducedbysmoothingthedataovera50by50gridcellwindowandthenslicedinto10%intervals.39Toanalyzethespatialdynamicsofdeforestationandregeneration,wecomparedhowdeforestationandregenerationchangedineach10%forestcoverzoneforeachtimeperiod.Inasecondstep,weperformedmapcross-tabulationsofthe10%zonemapsforthethreeperiods(1989to1996,1996to1999,1999to2002)(e.g.,cellsmakingtransitionfromthe70%to80%forestcoverzonetothe60%to70%forestcoverzoneinagivenperiodoftime).Thisallowedustoclassifyzonetransitionsinaspatiallyexplicitwayandtoidentifytheintensityofchange(negativefordeforestationandpositive for regeneration)dependingon thenumberof zonejumpsperperiod.Wethendefinedhotspotsofchangeonlyasthoseareaswith“highspeed”transitions,definedasareasshowingtwoormorezonejumps(e.g.,fromthe30%to40%zonetothe10%to20%zone).

42963_C005.indd 86 11/6/07 7:14:31 AM


tab

le 5

.2D

ata

and

Dat

a so

urce

s fo

r th

e th

ree

leve

ls o

f ana

lysi

s

leve

l of

anal

ysis

Dep

ende

nt v

aria

ble

Inde

pend

ent

vari

able

type

Dat

a so

urce

type

Dat

a so

urce

Nat

iona

lB

inar

yfo

rest

/non

fore

stE

cosy

stem

map

(E

tter,

1998

)C

limat

e,s

oil,

slop

e,d

ista

nce

tor

oads

,di

stan

ceto

riv

ers,

rur

alp

opul

atio

ngr

owth

(19

85–1

993)

Inst

ituto

de

Est

udio

sA

mbi

enta

les

-ID

EA

M,2

004;

In

stitu

toG

eogr

áfico

Agu

stín

Cod

azzi

,198

3,1

985,

20

00;I

nter

natio

nalW

ater

Man

agem

entI

nstit

ute

(IW

MI)

,200

4)

Reg

iona

lB

inar

yfo

rest

/non

fore

stL

ands

atE

TM

(19

89,1

996,

19

99,2

002)

Soil,

dis

tanc

eto

roa

ds,d

ista

nce

tor

iver

s,

rura

lpop

ulat

ion

grow

th(

1985

–199

3),

neig

hbor

ing

land

cov

er

Inst

ituto

de

Est

udio

sA

mbi

enta

les

-ID

EA

M,2

004;

In

stitu

toG

eogr

áfico

Agu

stín

Cod

azzi

,200

0;

Mal

agón

eta

l.,1

993)

Loc

alB

inar

yfo

rest

/non

fore

stA

ir-p

hoto

s(1

938–

2000

);

Lan

dsat

ET

M(

2000

,200

2)So

il,d

ista

nce

tor

oads

,dis

tanc

eto

riv

ers,

ne

ighb

orin

gla

ndc

over

Inst

ituto

Geo

gráfi

coA

gust

inC

odaz

zi,1

983,

200

0;

Air

-pho

tos

(193

8–20

00)

42963_C005.indd 87 11/6/07 7:14:32 AM


5.3.3 local level

Forestdynamics,asquantifiedthroughforestclearingandregrowth,weremodeledandanalyzedusingfivedifferentlogisticregressionmodels,includinganincreasingnumberofindependentvariables,inordertoexplainthespatiotemporalpatternsandsomeofthedriversofchange.40Thepurposeofmodelingwasthreefold:(i)toanalyzethegeneraltemporaltrendsofthedeforestationprocessofhumidlowlandforestsindifferent regions; (ii) assess the relative importanceof thepredictorvariables foreachapriorimodel;and(iii)selectabestapproximatingmodelfromtheapriorisetofcandidatemodels.

5.4 ResUlts

5.4.1 national DefoRestation thReat hot spots anD Main DRiveRs

The best performing model to predict deforestation was the regression tree thatincludedtheregionsasavariable(Figure5.2).Atthenationallevel,themostimpor-tantvariablesexplainingthepresenceandabsenceofforestcoverweredistancetoroadsanddistancetotownsforboththelogisticregressionandclassificationtreemodels.However,theeffectofthesevariablesonthepresenceofforestcovervariedfromregiontoregion.Theeffectofothervariablessuchassoilfertility(Andean,Pacific,Orinoco)andnumberof raindays (Caribbean,Magdalena)are importantexplanatoryvariablesinonlyafewregions.Ingeneral,thefollowingrelationships

N

Low

High

0 . 0.1110.111 – 0.2220.222 – 0.3330.333 – 0.4440.444 – 0.5560.556 – 0.6670.667 – 0.7780.778 – 0.8890.889 – 1No Data

Km5002500

FIGURe 5.2 Predicted forest presence according to the best model (classification tree).(FromEtteretal.50Withpermission.)

42963_C005.indd 88 11/6/07 7:14:33 AM


between deforestation and the explanatory variables were observed across allmodels: deforestation was predicted to be greater in unprotected areas that havefertilesoils,gentleslopes,andareneartosettlements,roads,andrivers.Therela-tionshipbetweenthenumberofraindaysandforestconversionwaspositiveintheCaribbeanandAmazon,whilenegativeintheAndeanregion.Theregressiontreemodeladjustedwiththepopulationgrowthdatashowsasetofdistinctgeographical“hotspots”ofpredictedfuturedeforestationareas(Figure5.3).

Thefiveforestedecosystemspredictedtobemostvulnerabletoforestconver-sionaccordingtothepredictedfuturedeforestationhotspotsrankedaccordingtopredicteddeforestationareawere:thehumidtropicalforestsoftheundulatingplains

N

Km5002500

1

2

3

45 6

7

8

FIGURe 5.3 (See color insert following p. 132.) Predicted deforestation hot spotsobtainedbycombiningareaspredictedtohavethehighestprobabilityofforestconversion(>70%)fromthebestmodel(theregion-specificclassificationtree)withtheareaswithgreaterthan2%ruralpopulationgrowthrate(1985–1993).Reddepictsthedeforestationhotspots(areaswith>70%probabilityofforestconversionand>2%ruralpopulationgrowth).Orangeandreddepictareaswith>70%probabilityofforestconversion.Greendepictsforestedareas,grayrepresentsclearedforestedareas,andwhiterepresentsnonforestedareas.Whitecircledareasindicatecurrenthotspotsofdeforestation,whicharealsoareasofhigh-valuebiodiversityvalue:(1)Quibdó-Tribugá,(2)Farallones-Micay,(3)Patía-Mira,(4)Fragua-Patascoy,(5)AltoDuda-Guayabero,(6)Macarena,(7)Guaviare,and(8)Perijá.BlacklineistheAndeanregion,andlightgreenlinesarenationalparks.(FromEtteretal.50Withpermission.)

42963_C005.indd 89 11/6/07 7:14:34 AM


of thenorthernAmazon, thehumid sub-Andean forests, thehumidhigh-Andeanforests,thehumidmid-altitudeAndeanforests,andthehumidtropicalforestsoftheundulatingplainsintheMagdalena.However,whenriskwasmeasuredintermsoftheproportionoftheremnantecosystemareapredictedtobetransformed,therank-ingof riskchanged to:verydry tropical forests in theCaribbean,humid tropicalforestsof the rolling landscapes in theMagdalena, the tropicaldry forestsof thehills,andthelowlandswampforestsoftheCaribbean.

5.4.2 aMazon colonization fRonts

TheanalysesshowthecolonizationfrontintheAmazonmovinglikeawaveorigi-nating in themainurbancentersof the regionand followingprimarily the rivers(Figure5.4). Between 1989 and 2002, the 90% forest cover boundary, used as aproxy for the “colonization frontline,” moved to the east an average distance of11.2km,representingarateof0.84kmperyear.However,theratevariedbetween0.5to3.2kmperyeardependingonthelocationalongthefront.Thetotalannualnetdeforestationratesshowedapeakof40,400hain1996to1999,increasingfrom18,600hain1989to1996,anddecliningto23,830hain1999to2002.Anaverageof25,400haofforestswasclearedeachyearduring1989to2002,ofwhich3,400haperyearwasclearedintheperforationprocessbeyondthe90%frontline.

The net forest change rate within the colonization front was the result ofthe combined effect of deforestation and forest regeneration and varied with theforestcoverzones.Thehighest rateofdeforestationoccurred in the50% to80%forest cover zones, with rates exceeding 4% per year during 1996 to 1999. Theforestregenerationratepeakedinthe20%to50%forestcoverzones,withannualratesofupto0.9%,suggestingtheprogressiveincreaseofsecondaryforestsagainst

Forest cover zone (%)

Cleared

1989 1996 1999 2002

Forest

0–10 10–20 20–30 30–40 40–50 50–60 60–70 70–80 80–90 90–100

(a)

(b)

FIGURe 5.4 (See color insert following p. 132.) Forestmapsofthecolonizationfrontforeachstudydate:(a)extentofforestcover(black=forest);and(b)percentageforestcoverat10%incrementzones.(FromEtteretal.39Withpermission.)

42963_C005.indd 90 11/6/07 7:14:36 AM


“matureforests”astheleveloflandscapetransformationincreases.Thesamepatternwasmaintainedforalltimeperiods.

Highratesofforestcoverchangewereobservedinlocalizedareas,or“localhotspots”offorestlossandforestregeneration(Figure5.5).Someofthesedeforestationhotspotswerespatiallystableduringtheentirestudyperiod,suchastheoneinthenortheastofthestudyarea;othersadvanced;whilesomelocationsshowedintensedeforestationinoneperiodfollowedbyaperiodofintenseregeneration,suchasthesouthwesternpartofthestudyarea.Themoredynamiclandscapeswereinvariablysituatedin theforestcoverzoneswithanintermediateamountofremnantforests(30%to70%).Thenortheastof thestudyarea,whichwasdeclared inDecember1999asamilitaryexclusionzonebythegovernmentforthepeaceprocesswiththeFARCguerrillas,showedthemostrapiddeforestationinthe1999to2002period,whiletherestoftheregionshowedtheoppositetrend.

5.4.3 DefoRestation patteRns at the local level

Theanalysisofthedeforestationtrendinallofthesixlowlandforestareasduringthepast60yearsshowedarapiddeclineoftheforestcover,andwastypicallyfollowedby an increase of pasturelands, initially supporting extensive cattle grazing, butintensifying where infrastructure development permits. However, a major findingwasthat inallcasestheforest lossfollowsanasymmetrical logisticpattern,withfourrecognizablephasesofforestloss(Figure5.6):

Phasei:initialphasewithlowrateofchangePhaseii:middlephasewithhighestrateofchangePhaseiii:midtolatephasewithrateofchangeslowingdownPhaseiv:finalphasewithanapparentnewdynamicequilibriumofre-growth

balancingforestloss

LEGENDRapid deforestationRapid forest regenerationMinor changes

TownsRiversRoads

1999–2002 militaryexclusion zone

1989–1996 1996–1999 1999–2002

FIGURe 5.5 (See color insert following p. 132.) Spatiallocationofthelocalhotspotsofdeforestation(red)andregeneration(green)forthethreetimeperiodsofstudy.(FromEtteretal.39Withpermission.)

42963_C005.indd 91 11/6/07 7:14:37 AM


Anadditionalresultisthatasaconsequenceofthelogisticdeclinepatternoftheforestcoverinthelandscape,theratesofdeforestationshowaquadraticrelationshipwiththeproportionofforestinthelandscape(Figure5.7a)andadirectlinearrela-tionshipwiththeamountofexposedforestedge(Figure5.7b).Themaximumratesofclearingoccurredatintermediatevaluesofforestproportioninthelandscapeandatthehighestvaluesofexposedforestedgedensities,ofapproximately40mha–1.

Also,when comparing thevarious regions, differences in colonizationwavesand rates of deforestation were evident for this period in the Amazon, Orinoco,andMagdalenaCaribbean.The logistic curvedepicting thedeforestationprocessshiftsintimedependingonwhenthelandclearingprocessbegan.Onaverage,thetimespanbetweenbeginning(phasei)andend(phaseiv)oftheprocessisattainedafter30to40years,atatimewhentheremnantsofforestecosystemsreachvaluesbetween2%and10%oftheiroriginalcover.Inallstudiedlandscapes,thedominantandmorepersistentreplacementcoverwasintroducedpastures,withcropsrepre-sentingaminorproportionoflandusefollowingclearing.

5.5 DIsCUssIon anD ConClUsIons

Our study contributes to broadening the understanding of patterns and drivers oftropicaldeforestationprocessesinunplannedcolonizationfronts.Therearefourmajorcontributionsaboutpatternsandprocessesof landcoverchangeinColombia thatcanbedrawnfromthisstudywithpotentialapplicationinothertropicalregions.

First, lossof forests followsa logisticpattern.Thecasestudies from lowlandforestsinColombiashowthatthedeclineofforestcoverduringtheprocessofdefor-estationconformstoalogisticpattern,reachingasemistabletransformedlandscapestateafterundergoingfourtransformationphaseswithvaryingratesofdeforestation.

FIGURe 5.6 Logisticpatternofforestcoverdeclineduringthe transformationprocessinColombia,showingphasesitoiv.LargeplotcorrespondstoLaBalsacasestudy,withequation: Y = 0.033 + [0.967 / (1 + exp(0.219 * (X – 1947)))]. (From Etter, A. etal.40Withpermission.)

1925 1945 1965 1985 2005

Ori

gina

l for

est c

over

1

0.8

0.6

0.4

0.2

0

i ii iii iv

42963_C005.indd 92 11/6/07 7:14:39 AM


Thismeansthattherateofforestlossassumesaquadraticshapewithapeakrateoflosswhenapproximatelyhalfthelandscapeistransformed.

Second, deforestation fronts move as waves. In the unplanned colonizationfronts inColombia, deforestationprogresses in a relativelyuniformpattern at anapproximatespeedof1kmperyear,mimickingawave.Forestcoverdeclineswithintheadvancingwave,followingasimilarlogisticpatternofforestdeclinementionedabove,therebylinkingthelocalandregionallevelsofanalysis.

Third,changesinlandscapestructure,suchasanincreaseinedgedensityperunitarea,facilitatetheclearingprocess.Thereisadirectrelationshipbetweentheexposed forest edge density in the landscape and the rate of deforestation, withmaximumratesofclearingoccurringatthehighestforestedgedensitiesofapproxi-mately40mha–1.Therelationshipbetweenproportionofforestinthelandscapeandforestedgeassumesaquadraticform,withmaximumedgedensityatintermediateforestproportionswhentheclearingrateisalsohighest.Thissuggestsastronglink

Forest proportion in landscape

Forest edge (m.ha–1)

R2 = 0.8819

R2 = 0.7643

Ann

ual d

efor

esta

tion

rate

Ann

ual d

efor

esta

tion

rate

0

0 10 20 30 40

0.2 0.4 0.6 0.8 1

5.0%4.5%4.0%3.5%3.0%2.5%2.0%1.5%1.0%0.5%0.0%

5.0%4.5%4.0%3.5%3.0%2.5%2.0%1.5%1.0%0.5%0.0%

iii

iii

ii

ii

i

i

a)

b)

FIGURe 5.7 ForestcoverchangeatthelocalscaleintheColombianAmazon:(a)Relationbetweendeforestationratesandforestproportioninlandscape;(b)Relationbetweenratesofdeforestationandforestedgedensity.(FromEtteretal.51Withpermission.)

42963_C005.indd 93 11/6/07 7:14:40 AM


between spatial pattern (forest proportion and edge density) and process (rate offorest clearing) in unplanned tropical colonization fronts. Furthermore, this rela-tionshipalsoconfirmsthebroaderapplicabilityofthelogisticpatternofforestlossacrossanentirecolonizationfront.

Fourth, secondary forests replaceclearedmature forestsandbecomeadomi-nantcomponentofthetropicalforestmosaicinthehighlytransformedlandscapes.Thehistoricalanalysisofdeforestationacrossallcasestudiesdemonstratesthatinhighlytransformedtropicalforestlandscapestheforestcomponentismadeoftwoparts: astable componentofmature remnant forest andadynamiccomponentofsecondaryforestsofdifferentages.Thesecomponentsarespatiallymixed,andformasuccessionalforestmosaic.

Landscapemetrics1,41aremeanttoprovidethemeanstointegrateissueslinkingpatternandprocess.However, landscape indicesarestillmostlyused todescribethespatialpatternsoflandscapechangewithoutmakingalinktoprocessesinthelandscape.42Ourstudyprovidesanexamplewherealinkbetweenlandscapemetrics,suchasforestproportionandamountofexposedforestedges,andthespeedoflandconversionismade.Thelogisticmodelofforestdeclineatthebaseofthesefindingsprovidesan importantgeneralprinciple forexplainingdeforestation inunplannedcolonization fronts (Figure5.8).However, itneeds tobe tested in tropical forestsoutsideColombia.

Studiesandassessmentsonmultiscaleanalysesoflandusemodelshavebeendone for Ecuador43 and Central America.44,45 We provide, for the first time inColombia,amultiple-leveloverviewofdeforestation,whichshouldgivebothgeneral(i.e.,whereareforestslikelytobelostinthefuture)andspecific(i.e.,howandwhyisdeforestationadvancing)guidelinesfordetectingspatiallyexplicitlandcoverchangedynamicsboth for further studiesand todirect theattentionofplanners.Weusespecificdatasetsforeachlevelofanalysis.

LanduseplanninginColombiaislimitedwithregardtotheuseofup-to-dateandrelevantspatialdata,whichiscriticalconsideringthatColombiaisacountrywherethebiologicalnaturalresourcesareanimportantassetandthegainsandlossesfrominappropriate landcoverchangeareveryhigh.Understandingthe levelof threat tonaturalecosystemsisfundamentaltohelpmakemoreinformeddecisionsinconser-vationplanning.46Thegeneralpatternsofdeforestationdescribedmakeasignificantcontributiontothisaimandcanpotentiallybeappliedinothertropicalandsubtropicalcountriesoftheworldwhereunplannedlandclearingoccurs.Thestudyhasspecificrelevancetolanduseandconservationplanninginhighlydynamiclandscapesbecausethemodelshelppredictandanticipatethespatialprogressionofunplanneddeforesta-tionfronts.Theresultsshould,inprinciple,betransferableandapplicabletootherunplannedcolonization fronts in the tropicsandsubtropicsandhelp improve landuseandconservationplanningby: (i)providingmethods tocalculate the threat toforestecosystemsandcharacterizethedynamicsofclearingassociatedwithcoloni-zationfronts;(ii)assistingtheanalysisofcarbonsequestrationandgreenhousegasemissionsbudgets,47,48 through the spatial and temporaldynamicsofdeforestationandforestregeneration;(iii)guidingbiodiversityconservationplanningandrestora-tionecologyinhighlytransformedanddynamiclowlandforestlandscapes,therebyhelpingdiscriminatebetweenrapidlychanging(hotspots)andmorestableareas.49

42963_C005.indd 94 11/6/07 7:14:41 AM


(a)

(b)

(c)

FIGURe 5.8 (a)Andeanlandscapeshowinghighlevelsoftransformationwithdominatingcattle grazing land uses. (b) High diversity forests in the Amazon region of Colombia.(c)ColonistforestclearingsinanAmazoncolonizationfrontinColombia.

42963_C005.indd 95 11/6/07 7:14:44 AM


Weshowedthatanunplannedtropicalforest landscape transformationresultsinseverelyclearedand impacted landscapes.The logisticmodelof forestdeclineprovidesan importantgeneralprinciple forexplainingdeforestation inunplannedcolonizationfronts.However,itneedstobetestedintropicalforestssuchasotherregionsofLatinAmerica,Africa,andAsia.However,animportantquestionremains:does government controlled land-use planning make a difference? An intriguingresearchagendawouldbetoinvestigatetheextenttowhichtheselandcoverchangepatternsandprocessesoccur incountrieswherelandclearingisplanned,suchasBrazil,orpartiallyregulated,suchasAustralia.

aCknowleDGMents

We especially thank Hugh Possingham, Kerrie Wilson, Stuart Phinn, and DavidPullarfortheirinputatseveralstagesofthisresearch.UniversidadJaverianaandtheUniversityofQueenslandprovidedfinancialsupporttoAE.

ReFeRenCes

1. Forman,R.T.T.LandMosaics:TheEcologyofLandscapesandRegions.CambridgeUniversityPress,NewYork,1995.

2. Wackernagel,M.etal.Trackingtheecologicalovershootofthehumaneconomy.PNAS99(14),9266,2002.

3. Vitousek,P.M.etal.HumandominationofEarth’secosystems.Science277(5325),494,1997.

4. Houghton,R.A.Theworldwideextentofland-usechange:Inthelastfewcenturies,andparticularly in the lastseveraldecades,effectsof land-usechangehavebecomeglobal.Bioscience44(5),305,1994.

5. Lepers,E.etal.Asynthesisofinformationonrapidland-coverchangefortheperiod1981–2000.Bioscience55(2),115,2005.

6. Laurance,W.F.Reflectionsonthetropicaldeforestationcrisis.BiologicalConservation91,109,1999.

7. Lambin,E.F.etal.Thecausesoflanduseandland-coverchange:Movingbeyondthemyths.GlobalEnvironmentalChange11,261,2001.

8. DeFries,R.,andBounouaL.Consequencesoflandusechangeforecosystemservices:Afutureunlikethepast.GeoJournal61,345,2004.

9. Fearnside, P. M. Global warming and tropical land-use change: Greenhouse gasemissionsfrombiomassburning,decompositionandsoilsinforestconversion,shiftingcultivationandsecondaryvegetation.ClimateChange46(1–2),115,2000.

10. Myers, N. et al. Biodiversity hotspots for conservation priorities. Nature 403, 853,2000.

11. Wu, J., and Hobbs R. Key issues and research priorities in landscape ecology: Anidiosyncraticsynthesis.LandscapeEcology17,355,2002.

12. Bürgi, M., Hersperger, A. M., and Schneeberger, N. Driving forces of landscapechange—Currentandnewdirections.LandscapeEcology19(8),857,2004.

13. Bilsborrow,R.E.,andOgendo,O.H.W.O.Population-drivenchangesinlanduseindevelopingcountries.Ambio21(1),37,1992.

14. Achard, F. et al. Determination of deforestation rates of the world’s humid tropicalforests.Science297(5583),999,2002.

42963_C005.indd 96 11/6/07 7:14:45 AM


15. Houghton,R.A.,Skole,D.L.,andLefkowitz,D.S.ChangesinthelandscapeofLatinAmericabetween1850and1985II.Net releaseofCO2to theatmosphere.EcologyManagement38(3–4),173,1991.

16. Frohn,R.C.etal.Usingsatelliteremotesensinganalysistoevaluateasocio-economicandecologicalmodelofdeforestation inRondonia,Brazil. InternationalJournalofRemoteSensing17(16),3233,1996.

17. Fearnside,P.M.SoybeancultivationasathreattotheenvironmentinBrazil.EnvironmentalConservation28(1),23,2001.

18. Moran,E.F.etal.Effectsofsoilfertilityandland-useonforestsuccessioninAmazo-nia.ForestEcologyandManagement139(1–3),93,2000.

19. Portela, R., and Rademacher, I. A dynamic model of patterns of deforestation andtheireffecton theabilityof theBrazilianAmazonia toprovideecosystemservices.EcologicalModelling143(1–2),115,2001.

20. Skole, D. L. et al. Physical and human dimensions of deforestation in Amazonia.Bioscience44(5),314,1994.

21. Alves, D. S. Space-time dynamics of deforestation in Brazilian Amazônia. InternationalJournalofRemoteSensing23(14),2903,2002.

22. Laurance,W.F.etal.PredictorsofdeforestationintheBrazilianAmazon.JournalofBiogeography29,737,2002.

23. Laurance,W.F.,andFerreira,L.V.EffectsofforestfragmentationonmortalityanddamageofselectedtreesinCentralAmazonia.ConservationBiology11(3),797,1997.

24. Laurance,W.F.etal.The futureof theBrazilianAmazon.Science291(5503),438,2001.

25. Viña,A.,Echavarria,F.,andRundquist,D.C.SatellitechangedetectionanalysisofdeforestationratesandpatternsalongtheColombia-Ecuadorborder.Ambio33(3),118,2004.

26. Etter,A.MapaGeneraldeEcosistemasdeColombia(1:2000000).In:Chaves,M.E.and Arango, N., eds., Informe Nacional sobre el Estado de la Biodiversidad enColombia—1997.InstitutoAlexandervonHumboldt,Bogotá,1998.

27. Myers,N.Biodiversityhotspotsrevisited.Bioscience53(10),916,2003. 28. Chaves,M.E.,andArango,N.,eds.InformeNacionalsobreelestadodelaBiodiversidad

enColombia—1997.I.A.vonHumboldt,Bogotá,689,1998. 29. ConventionofBiologicalDiversity(CBD).GlobalBiodiversityOutlook,2005.Available

from:http://www.biodiv.org/gbo/. 30. Orme, C. D. L. et al. Global hotspots of species richness are not congruent with

endemismorthreat.Nature436(7053),1016,2005. 31. Etter,A.Modelingunplannedlandscapechange:AColombiancasestudy.Dissertation,

UniversityofQueensland,Brisbane,2006. 32. Hernández, J. et al. Estado de la biodiversidad en Colombia. In: Halffter, G., ed.,

LaDiversidadBiológicadeIberoamérica,Vol.I.CYTED,Mexico,1992. 33. Herrera,M.OrdenamientoTerritorialyControlPoliticoenlasLlanurasdelCaribeyen

losAndesCentralesNeogranadinos,SigloXVIII.Dissertation,UniversityofSyracuse,NewYork,2000.

34. Colmenares,G.HistoriaEconomicaySocialdeColombia.5thed.Vol.1,TMEditores,Bogotá,357pp,1999.

35. Loh,J.,andHarmon,D.Aglobalindexofbioculturaldiversity.EcologicalIndicators5(3),231,2005.

36. UnitedNationsOfficeforDrugControlUNDOC.Colombia:Cocacultivationsurvey.UnitedNationsOfficeforDrugControlandGovernmentofColombia,Bogotá,72pp,2004.

37. McDonald,R.I.,andUrban,D.L.Spatiallyvaryingrulesoflandscapechange:lessonsfromacasestudy.LandscapeandUrbanPlanning74(1),7,2004.

42963_C005.indd 97 11/6/07 7:14:46 AM


38. Insightful-Corporation.S-Plus6.1forWindows.Seattle,Washington,2002. 39. Etter,A.etal.Characterizingatropicaldeforestationfront:Adynamicspatialanalysis

of a deforestation hotspot in the Colombian Amazon. Global Change Biology 12,1409–1420,2006.

40. Etter,A.etal.ModelingtheconversionofColombianlowlandecosystemssince1940:drivers,patternsandrates.JournalofEnvironmentalManagement79(1),74–87,2006.

41. McGarigal,K. et al.FRAGSTATS:Spatialpatternanalysisprogram for categoricalmaps. Computer software program produced by the authors at the University ofMassachusetts, Amherst. Available at: www.umass.edu/landeco/research/fragstats/fragstats.html,2002.

42. Li,H.,andWu,J.Useandmisuseoflandscapeindices.LandscapeEcology19(4),389,2004.

43. deKoning,G.H.J.etal.Multi-scalemodellingoflandusechangedynamicsinEcuador.AgriculturalSystems61,77,1999.

44. Kok, K., and Veldkamp, A. Evaluating impact of spatial scales on landuse patternanalysisinCentralAmerica.Agriculture,EcosystemsandEnvironment85(1–3),205,2001.

45. Kok,K.etal.Amethodandapplicationofmulti-scalevalidationinspatiallandusemodels.Agriculture,EcosystemsandEnvironment85(1–3),223,2001.

46. Wilson,K.etal.Measuringandincorporatingvulnerabilityintoconservationplanning.EnvironmentalManagement35(5),527,2005.

47. Houghton,R.A.Abovegroundforestbiomassandtheglobalcarbonbalance.GlobalChangeBiology11(6),945,2005.

48. Foley,J.A.etal.Globalconsequencesoflanduse.Science309(5734),570,2005. 49. Meir,E.,Andelman,S.,andPossingham,H.Doesconservationplanningmatterina

dynamicanduncertainworld?EcologyLetters7(8),615,2004. 50. Etter,A.etal.RegionalpatternsofagriculturallanduseanddeforestationinColumbia.

AgriculturalEcosystemsandEnvironment114,369–386,2006. 51. Etter,A.etal.UnplannedlandclearingofColumbianrainforests:Spreadinglikedis-

ease?LandscapeandUrbanPlanning77,240–254,2006.

42963_C005.indd 98 11/6/07 7:14:47 AM

99

6 Landscape DynamismDisentangling Thematic versus Structural Change in Northeast Thailand

Kelley A. Crews

Contents

6.1 Introduction...................................................................................................996.1.1 TemporalFrequency:TensionsandLimits...................................... 1006.1.2 FromPatternandStructuretoProcessandFunction...................... 101

6.2 APanelApproach....................................................................................... 1026.2.1 ExtensiontoPatternMetrics............................................................ 103

6.3 ThaiTestingGrounds.................................................................................. 1046.3.1 LocalLessonsLearnedThusFar.................................................... 107

6.4 PaneledPatternMetrics:MeansorEnd?.................................................... 112Acknowledgments.................................................................................................. 115References.............................................................................................................. 115

6.1 IntRoDUCtIon

Landchangeresearchnecessarilydrawsuponaninterdisciplinarymilieuoftheoriesand practices ranging from ecology to geography to policy and beyond; a domi-nantapproachsuccessfullyusedin thisarenaover thepast fewdecadeshasbeenthatofscale-pattern-process.1Choiceofscaleinfluenceswhichlandscapepatternscanbediscerned,inturnusedtoinferprocess.Thenumberofresultinglandscapestudieshaveincreasedsubstantiallyoverthepastdecade.2,3,4,5Assessingsensitivitiesofpatterndetectionandsubsequentinferableprocessestochangesinscale(typicallyspatial resolutionorpixelsize)of remotelysenseddatahasbecomean importantresearchagendaforremotesensingspecialists.6,7Thisworkdrawsinparticularonprinciplesoflandscapeecologythatpositthepossibleimpactsthatscalecanhaveonlandscapecharacterization.8Scaleiscomprisedoftwoprimaryfacets:grain,thesizeofanobservationalunit(e.g.,thedimensionsofasinglepixel),andextent,typicallyrepresentedasthesizeoftheoverallstudyarea.Althoughthesecomponentpartsaretypicallyappliedtospatialscale,theyaseasilymaybeappliedtotemporalscale(e.g.,scaleasthefrequencyofobservationandextentasthetotallengthofstudy).

42963_C006.indd 99 11/6/07 7:13:44 AM


Implicitintheseargumentsistheseparationoflandscapeconfigurationfromland-scapecomposition.Inotherwords,thespatialassociationofdifferentelementsisasimportantastheoverallproportionofthelandscapeoccupiedbytheelements.Manylanduse/landcoverchange(LULCC)applications,rangingfrombiologyconserva-tion tohydrological assessments to landuseplanning,nowroutinelyprovide thisdecoupledinformation.4Thisworkreviewsthesuccesses,limitations,andpossibili-tiesofenrichingLULCCresearchwithincreasedtemporalgrainorobservationalfrequencyforextricatingcompositionalor thematicchangefromconfigurationorstructuralchange.AcasestudyfromNortheastThailand isused to illustrate thispairedapproach,underscoringtheneedtofurtherdevelopandrefinethismethodinecosystemsfromelsewhereonthenaturallyoranthropogenicallydrivenspectrumandwithvaryingdegreesofspatiotemporalheterogeneity.

6.1.1 Temporal Frequency: Tensions and limiTs

Although most scale-pattern-process work has focused on spatial scale, temporalscalehasnonethelessbeenexplicitlyincludedintheoreticaldiscussionsevenifseldomanalyzed.4,9 Environmental remote sensing defines scale more broadly to includespatial,temporal,spectral,radiometric,anddirectionalscales.10Spatialandtemporalscaleareparticularly importantwhenextracting thematicdata for satellite image-basedchangedetection.11Spatialscale,bothgrainandextent,isregardedasamajorinfluenceondetectionanddefinitionoflandscapepatterns.12,13,14Temporalfrequency,orthetimescalebetweenavailabledataacquisitions,islessstudiedinLULCCwork,inlargepartduetothelimitedavailabilityofhighqualityandhighresolutionmulti-temporalimagesequences.6,15,16Thetemporalgrainofimagery,thoughtypicallynotreferredtoassuch,hasbeenexaminedinenvironmentswhereseasonality(whetherduetophenological,climatic,oranthropogenicchanges)caninterferewithassessmentof longer-term(read: interannual)LULCC.17,18,19,20The temporalextentofLULCCprojectstypicallydefaultstoeithertheearly1970s(concomitantwiththe1972launchofEarthResourceTechnologySatelliteorERTS1,laterrenamedLandsat1)or,inafewcases,toafewdecadesearlierwhenmilitaryreconnaissanceaerialphotographywas available.Thenecessarily truncated temporal extentof these studiespresentsproblemsinestablishingbaselines,acriticalissuegiventhenecessityofdeterminingwhatchangehasoccurredandplacingitintheappropriatehistoricalcontext.21,22Thetermspatiotemporal scaleordomain iswidelyusedwithin theLULCCmodelingcommunity,butthisdescriptionmaycausesomeconfusion.Thecoinageofthetermpresentsanunderstandablecommitmenttoconsiderhowlandscapeschangeacrosstimeandspace,thoughcurrentlythestateofthesciencetendstomodelspatialinter-actionsovertimeratherthanofferingapathfordigitallyrepresentingaspatiotemporalscaleasinteractiveratherthanonlycombinatory.

Compounding the confusion is the dialogue concerning the use of the termslandscape scale and landscape level. For geographers, the term landscape scaleconnotesacertainsize(spatialextent)ofastudyarea—largerthanaplot,smallerthan a continent,4 and the term level is often used to denote study across spatialscales,23whetherornotsuchworkisspatiallyexplicit(e.g.,multilevelmodeling24).For many ecologists who tend to focus first on biotic components of landscapes

42963_C006.indd 100 11/6/07 7:13:45 AM

Landscape Dynamism 101

(e.g.,populationsandcommunities),thetermlandscapescaleisnonsensical,sincedepending upon organism size and range, a landscape can be incredibly small(considermicrobeslivinginasmallpuddle)oraslargeasaplanet.25Thistensionmanifests itself in incongruities and inconsistencies in theuseof these terms (aswellaswhether theyareseenas interchangeableornot),a troublesomeglitchforLULCCscholarsdrawinguponecologythroughthelensof landscapeecology.1,26ForthepurposesofdiscussingLULCCinthiswork,thetermlandscape scalewillbeusedtoconnotespatialandtemporalgrainandextentcommonlyusedinLULCCwork.Thetermlandscape levelwillbeusedtorefertoorganizationalortheoreticalconstructswherethelandscapeliesonaspectrumoffunctionalunits,rangingfrompatchestolandscapestometalandscapes.27

Notetheaboveissuesrevolveprimarilyaroundspatialscale;rarelydoLULCCpractitionersmentionalandscapescalewhenreferringtoacertaintime,asopposedto those studying longer-term landscapes (e.g., ingeomorphology, sedimentology,or palynology). Temporal matters are receiving more scholarly attention of late,particularly in both empirical and process-based modeling efforts.28 LandscapesintemporallyshallowLULCCstudiesarebeingincreasinglyconsideredasactingupon their previous incarnations22 and seen, therefore, as temporally contingentuponthosepastdrivers.Pathdependenciescanandarebeingtrainedintomodel-ing scenarios, and presumably other temporal analogues of spatial concepts willbeoperationalized(e.g.,spatialneighborhoodeffectscouldbeusedasamodelformoresophisticatedrepresentationsofpathdependencyviatemporalneighborhoodeffects). In spatialneighborhoodeffects, it isunderstood that theprecise locationoftheneighborrelativetotheareaofinterestisoftenunimportantaslongasthatneighboriswithinacertainthresholdeddistance.Sowhileanalysismaytakeplaceinaspatiallyexplicitenvironment,conditionsorrulescanbewrittentoloosenthatexplicitnessandquery,forexample,forneighborswithinacertainspatialdistance(withoutregardtothatdistance),withoutregardtodirection.Theparallelintemporalstudieswouldbetorelaxtheassertionoftemporalexplicitnesssuchthatitmaynotmatterwhenaparticularprecedingeventhappened,onlythatitdidhappenorhowoften ithappened.Although temporalmodeling is fairly straightforward in termsofassessingcausality(giventhepresumptionoflineartimemovingonlyforward),thechallengeremainstosortoutfrommyriadperiodicitiesoflandscapedriversandchange,whichareimportantenoughinanygivenlandscape,andhowthentobestdefineatemporallandscapescale.

6.1.2 From paTTern and sTrucTure To process and FuncTion

Several decades of LULCC research have shown that understanding landscapechangerequiresdetectingchangesinbothcompositionandconfiguration.6,8Typi-callythesecomponentsareassessedsequentially:first,alandscapeisclassifiedintoathematiclanduse/landcover(LULC)schemeforatleasttwotimes;second,theconfigurationofeachofthoseclassifiedlandscapesarequantifiedthroughsometypeofpatternassessment(oftenpatternmetrics29);andthird,postclassificationchangedetectionisperformedonthosethematicclassificationstoproducethematicchangemap(s).30Althoughtheprocessusuallystopsthere,someresearchershavealsothen

42963_C006.indd 101 11/6/07 7:13:45 AM


quantified the configuration of the change map(s) with pattern metrics as well,11thoughconcernsoferrorpropagationhavelimitedthisapproach.31Theimportanceofascertainingspatialstructure(andchangesinsaidspatialstructure)stemsfromlandscapeecology,wherespatialconfigurationfacilitatesandmitigatestheflowofenergyandmaterialsacrossthelandscape.8Thatis,thelandscapeinteractionsthatbothcauseandaremanifestedaslandscapechangenecessarilyoccurinspace,andlocationmatters.Thatisnottosayallchangesoccurasdiffusivemovementssince,dependinguponthevectorofmovement,energyormaterialsmaybeimpartedbyjumpingorpercolatingacrossthelandscape.4Definingthetemporalnatureofspatialstructurewillassistintakingthesemeasurementsandconvertingthescale-patternintoprocess.

Process can be defined in two primary ways. The first will be referred to asdynamics, and it is a mechanistic concept rooted in patterns of change, growth,and activity; this definition is embraced by the Geographic Information Sciencecommunity(GISc)studyinglandscapedynamics,andfitsthenecessarilypiecemealfashion by which LULC is extracted, studied, and modeled. The second type ofprocess will be referred to as dynamism, which is a more gestaltic concept thatinvolves continuous change, growth, or activity; this definition comes from theecology community (particularly landscape ecology) and fits the more continualnatureoftheprocessesstudiedbyecologists,whetherparticulartolandscapestudiesornot.32Thenuanceofthedifferenceinthesetwoapproachesisslight,buttheimpli-cationsareeasilyobservableinthevaryingoperationalizationofbothepistemologyandmethodologynowevidencedinlandscapechangestudiesfromthesetwocom-munities.Here,panelanalysisofLULCandpaneledpatternmetricsareofferedasonemethodofbridgingthisgap,suggestingthatLULCCscholarsshifttowardanapproachofunderstandinglandscapedynamismviaimprovedassessmentofLULCCdynamics.11,27Thatis,improveddescriptionofmechanicsshouldleadtoimprovedexplanationandpredictionofprocessand,perhaps,function.

6.2 A PAneL APPRoACH

Panelanalysissimplyreferstoalongitudinalmethodwherebyunitsofanalysisareheldconstant.Longusedinpsychologyandsociology,panelorlongitudinalanalysisfollowed the same subjects over time (as opposed to a census or cross-sectionalapproach,wheredifferentsubjectsareevaluatedineachobservationperiod).Tech-nically,allfrom-toremotesensing-basedchangedetectionispanelanalysis,sinceeachpixelor instantaneousfieldofview(IFOV)isfollowedindividually throughtime,presumingaccurategeometricrectification.30However,from-tochangedetec-tionusuallyisperformedonpairsofimages,whereaspanelanalysis(inLULCC)isnowusedtorefertoatimeseriesofthreeormoreclassifications.33,34,35Inpanelanalysis,pixelhistoriesortrajectoriesareconstructedthatmaintaintheentiretem-poralpatternofLULCinordertorevealgreaterinformationaboutprocess(es)behindobservablepatterns.Forexample,considerahumidtropicalareaclassifiedonlyintoforest(F)andnonforest(N)andobservedovertwodecadeseveryotheryear.Withpanel analysis, trajectories that might be calculated would include those suggest-ingsemipermanentdeforestation(e.g.,F-F-F-F-N-N-N-N-N-N-N),deforestationand

42963_C006.indd 102 11/6/07 7:13:46 AM


successionalregrowth(e.g.,F-F-F-F-N-N-N-N-N-F-F),afforestationorreforestation(e.g., N-N-N-N-N-N-N-F-F-F-F), silviculture of fast growing tree species (e.g.,F-N-N-N-F-N-N-N-F-N-N-N),orfallowcycling(e.g.,N-F-F-F-F-F-N-F-F-F-F-F-N).Withtraditionalfrom-tochangedetectionofthefirstandlastyears,thosetrajectorieswouldhavehadtheirchangecharacterizedasfollows:semipermanentdeforestationwithF-Nwouldstillbecalleddeforestation(correct);deforestationandsuccessionalregrowthwithF-Fwouldbecalledstableorpermanentforest(incorrect);afforesta-tionorreforestationwithN-Fwouldstillbecalledassuch(correct);silviculturewithF-Nwouldbecalleddeforestation(incorrect),andfallowcyclingwithN-Nwouldbecalledpermanentnonforest(incorrect).UltimatelythepanelapproachtoLULCCdoesnothingtoimproveattributionofclassesthatarestableovertime,andlittletoimproveattributionofclasseswhosechangeisunidirectional.Butlandscapecompo-nentsthatundergoveryquickchange,cyclethroughmultiplestages,switchbetweentwoormoreclassesfrequently,orareinfluencedbyrelativelyshort-termphenomena(e.g.,seasonality)areopentobettermultitemporalcharacterization.Thatis,panelanalysisimprovesourabilitytodetectthekindsofchangethatLULCCresearchislargelydesigned tocapture,model, andmanage;bycorollary, traditional from-tochangedetectionisbiasedtowarddetectingstable,slow-changing,orunidirection-allychangingclasses.Asthenumberofclassificationsinthetimeseriesincreases,quite obviously the ability to detect greater nuanced or more quickly switchingchangeincreases.ThequestionforLULCCprojectsthenishowmanyimagesareenough?Thetextbookansweristhatitdependsuponthetimefootprintoflandscapeprocessesonthelandscape(e.g.,humidtropicalforestsreachsuccessionalcanopyclosuremorequicklythantheaveragetemperateforest);thepracticalansweristhatitdependsonhowmanyqualityimagesareavailableinanareagivenatmosphericinterference,sensorproblems,costofacquisition,andaccesstoarchives,tonameonlyafewoftheproblemsfacingtheLULCCcommunity.

6.2.1 exTension To paTTern meTrics

Thoughpatternmetricanalysisistypicallyoutputasstatisticsatthepatch,class,andlandscapelevels,somepackagessuchasFragstats29allowforoutputtingpatch-basedimageswherebyeachpatch(fromwhichallpatch,class,andlandscapestatisticsaregenerated)ismappedwithauniqueidentifierorobject(whethercomputedinrasterorvector,bitdepthlimitationsnotwithstanding).Thegoalofpaneledpatternmetricanalysisistoassessthechangingstructureoflandscapepatcheswithoutregardtothematic class. That is, in building pattern metric panels we explicitly choose toexaminethenatureof,forexample,fragmentationwithoutregardtowhether it isanurbanarea,forestedexpanse,oragriculturalfieldthatisbeingfragmented.Bydoingso,theexplicitcontributionofconfigurationasopposedtocompositioncanbetested,assessed,andmodeled.Currentresearchatthispointhasfocusedontheformulation and sensitivity analyses of paneled pattern metrics, and this methodrequiresfurthertestinginotherecosystemsandlandscapeswithdifferinglevelsofspatial,temporal,andspatiotemporalheterogeneity.Incaseswheretherobustnessandsensitivityofthepaneledpatternmetricmethodisvalidated,thenextstepisto

42963_C006.indd 103 11/6/07 7:13:47 AM


notonlytesttheseparateimpactsofcompositionandconfiguration,butalsotheirinteractionandconfoundingaswell.

Theconstructionofpaneledpatternmetricsfollowslogicallyfrompanelanalysis,andtheentirepanelmethodispresentedinFigure6.1.First,atimeseriesofimageryiscategorized into thematicclassifications;aminimumoffour temporalobserva-tionsissuggested,thoughifpatchboundariescanbederived,generated,orfoundelsewhere,threeclassificationsmaysuffice(inabsenceofpreexistingpatchdelinea-tions,abaselineyearofthetimeseriesisused,requiringthreefurtherclassificationsformovingbeyondtraditionaltwo-imagechangedetection).Fromtheseclassifica-tions,apanelLULC iscreatedasdepictedandasdescribedabove.Additionally,patternmetricanalysis isrunoneachclassification,outputtingbothstatisticsandpatchimagesforallobservationsforeachmetricofinterest.Forpurposesofthisdis-cussion,presumethemetricofinterestistheinterspersion/juxtapositionindex(IJI).Changeimagesbetweenconsecutivepairsofpatchimagesarecalculatedandmayinitiallybeleftasfloatoutputbutmusteventuallybebinnedintocategoriesofchange(e.g.,increaseby>20%,increaseby10%to20%,increaseby5%to10%,changeby±5%,decreaseby5%to10%,decreaseby10%to20%,decreaseby>20%).Oncebinnedappropriately,thechangebetweeneachsetofIJImetricimagesisstackedtobuildatrajectoryofchangeatthepatchlevelandthenexportedtoindividualpixelsandbuiltback toafinalmappedproductofpaneledpatternmetricsoutputat thepatchlevel.11Theprocessisrepeatedforeachmetricofinterest,witheachmetricbinnedaccordingtoappropriatehypothesizedorobservedthresholdsorflippoints.

Currentlyboundedorconstrainedmetricshavebeentestedinordertolimitthesubjectivityinvolvedincategorizationofthemetricoutput.Thatis,metricssuchasIJI,doublelogfractaldimension,andpercentagelandscapeall—asoperationalizedinFragstatsandotherpatternmetricprograms—havetheoreticalboundswhereboththeupperandlower limitsareknown.Unboundedorunconstrainedmetrics(e.g.,meanpatchsize,showninFigure6.1forcontrast)presentgreatersubjectivityincat-egorizationsincethereisnotheoreticallimitforthesemetrics(thoughinanygivenlandscape and with any given classification scheme an empirical limit obviouslyexists).Ascurrentlywritten,thepaneledpatternmetricalgorithmpresumesequalintervals between time steps since the original time series used for testing metthoseconditions;modificationtoaccountfordifferingtimelagsiseasilydoneviaaweightingmechanismoncecategorizationthresholds(numberandplacement)havebeendetermined.As such, themethod is suitable forboth interannual and intra-annualanalyses.

6.3 tHAI testInG GRoUnDs

The concern over interannual and intraannual LULCC stems from building thisapproach in an environment with strong phenological, climatic, and anthropogenicseasonalpulses,renderingassessinglonger-termLULCCproblematicwhenanythingbut anniversary date imagery was used for deriving LULC information. NortheastThailandishometoaregionknownasIsaan,wheretheformerNangRongdistrictresides(duetogrowthandredistrictingthisareanowincludesnotonlytheNangRongdistrictbutalsoNonSuwan—denotedonsomemapsasNonguWuan,Chamni,and

42963_C006.indd 104 11/6/07 7:13:47 AM


IJI = 14.2MPS = 8.1

IJI = 17.9MPS = 4.7

IJI = 23.7MPS = 2.2

IJI = 19.2MPS = 3.1

(2)

(2a)

(2b)(3a)

(3)

(1)

FIGURe 6.1 (See color insert following p. 132.) The panel process, conducted atboth the pixel and patch levels: (1) four multispectral satellite images are each catego-rized intoa thematicLULCclassification; (2)patternmetricsare runoneachof the fourLULCclassifications,eachproducingasetofpatch,class,andlandscapestatistics(heretheinterspersion/juxtapositionindex[IJI]andmeanpatchsize[MPS]areshown)aswellasanoutputimageofthedelineatedpatches;(2a)patternmetricoutputforeachofthefourtimesisusedtocalculatethreepiecemealchangemapsforeachpatternmetricandeachconsecu-tivepairofimages(e.g.,showingfluctuationsinIJIorMPSbetweentwotimeperiods)asperCrews-Meyer11,27;(2b)threepatternchangemapsarestackedintoonepanelofallstruc-turalchangeforeachgivenmetric(e.g.,showingfluctuationinIJIorMPSthroughalltimeperiods)asperCrews-Meyer11,27;(3)threethematicchangemapsarecreatedforeachofthetimeperiodsrepresentedbythefourclassifications;(3a)thethreethematicchangemapsarestackedtorepresentthefullrecordofallthematicchangeacrossthefourclassificationsasperCrews-Meyer.3

42963_C006.indd 105 11/6/07 7:13:49 AM


ChalermPrakeat;thisworkwastestedprimarilyincurrentdayNangRongandNonSuwan).SituatedinbothBuriramProvinceandthenorth-flowingMekongRiverDeltasystem,theareaisthepoorestareaofapoorcountry36,37anddominatedculturally,ecologically,andfinanciallybyastrongmonsoonalpulse,poorsoils,38andconcomi-tant lowland wet rice production.39 Villagers typically live in a nuclear settlementpattern(seeFigure6.2),withresidenceslocatedinlowlandwoodedremnantsandricefieldsradiatingoutinmostdirectionsforthetypical2to5kmdailywalktofields.40,41ThoughthisareawasnotinfluencedbytheGreenRevolution,agriculturehasdriventheconversionofthelandscapeinitiallyopenedbymilitaryroadbuildingeffortsandfacilitated by the gradual building toward a market economy.37 Wet rice replacedsavannainthelowlands,whiledrought-deciduouscropssuchascassavaandsugar-canefollowedthe1970sfactorpriceincreaseintotheuplanddrydipterocarpforests.Followingacurrencycollapse in the late1990s,manyyoungadultswhotypicallymigratedtoBangkokortheeasternseaboardforlaborreturnedtothedistrictatthesametimethegovernmentunderwentanewwaveofdecentralizationacrossfederaltolocallevels.40Anincreasinglydensenetworkofroadbuildingandwaterimpound-ments,33combinedwithpoorenvironmentalmanagement(e.g.,lackofdrainingriceirrigation waters increases soil salinity), has compounded the intensification cycleseeninpartsofSoutheastAsiaandelsewhere.Althoughtheselonger-termdynamicshavebeendocumentedthroughanextensivehouseholdandcommunitysurveyseries

FIGURe 6.2 Typical nuclear village settlement as seen in 1:50,000 scale panchromaticaerial photo from 1994, with approximate settlement boundary indicated. Note remnantforestpatchesusedforshaderelief,andricepaddysurroundingvillageradially.

42963_C006.indd 106 11/6/07 7:13:50 AM


aswellasremotesensingandgeographicalinformationsystems(GIS)analyses,theseasonalpulsesalsodetected(whenimagery,fieldwork,andweatherpermit)cancausedetectablelandscapechangeaslargeinmagnitude(ifnotecologicalimportance)astwodecadesofinterannualchange.18,19Thepresenceofamonsoonallydrivenclimateaddstothelogisticalproblemsofobtainingcloud-freeimageryforderivingLULCinformation.However,adeep timeserieshasbeenestablishedaspartofa largerprojectandhasprovenmorethanadequatefortestingthepanelLULCandpaneledpatternmetricmethods.41,42Figure6.3 illustrates interannual trends inLULCCinthelargerstudyareaovera25-yearperiod;easilydiscerniblearetherapiddeclineinmorehighlyvegetatedLULC(particularlyintheuplandsouthwesternsection)andtheexpansionofriceintothelowlandsavannas.

6.3.1 local lessons learned Thus Far

Figure6.4illustratesastylizedrepresentationoffourLULCclassesandtheircompo-sitionalchangeovertimeasobservedand/orreportedelsewhere.Figure6.4ashowsthe interannualor longer-termchange in forest (primarilyuplanddrydipterocarpandgallery remnant forestsalong ripariancorridors), savanna (primarily lowlandgraminoids with some standing trees), wet rice agriculture, and other agriculture(uplandordroughtdeciduouscropsandcashcrops,includingcassava,kenaf,jute,andsugarcane).33These“real”changescanbecontrastedwiththestylizedrepresen-tationofintraannualchangeinagivenyearduetopreviouslymentionedseasonalityshowninFigure6.4b.ThisgraphisorderedbytheThaiwateryearthatrunsApril1throughMarch31,withearlymonsoonalshowers(knownasmangorains)commenc-inginMayandfollowedbyseveralmonthsofheavyprecipitationthatisextremely

(a) (b) (c)

FIGURe 6.3 (See color insert following p. 132.) (a)LULCinthegreaterstudyareainthe1972/1973wateryear;(b)1985;and(c)1997.

BackgroundHigher Density ForestLower Density ForestSavannaBare SoilRice AgricultureMixed AgricultureCash Crop AgricultureWater

42963_C006.indd 107 11/6/07 7:13:53 AM


variable in both time and space; rice is typically harvested in late November orDecember,withfieldsburnedusuallyinJanuaryandthedriestmonthsendingthewateryear.These“changes”arepartreal(e.g.,phenologicalchangewithagriculturalcropsordeciduouscycles)andinpartartifact(e.g.,green-upfromshowerswithoutactualcanopyorbiomasschange).

Typical forest changes in this part of northeastThailand represent a familiarstory:fromthe1970sthroughthe1990s,forestsgenerallydeclined(asdidsavannas)duetoagriculturalextensification.Anearlyriseinotheragricultureintheuplandsattheexpenseofforests(nowrelegatedtoextremelythinripariancorridorsandsmallremnantsatopthemostuplandsitesonvolcanicsoils)wasfollowedbyasharpriseinwetriceagricultureinthelowlandareas.Villagesettlementandexpansionoccurintheselowlandareasaswell,althoughtheseareasaccountforlittlechangeintermsof

FIGURe 6.4 (a) Stylized LULC trends observed and/or reported in Northeast Thailandfrom the1970s to late1990s (annualchange,holdingseasonalityconstant). (b)Thesametrendswithinagiventypicalyear(intraannual).

Time (Interannual)

Perc

ent L

ands

cape

504540353025201510

50

ForestSavannaRice Ag.Other Ag.


0

10

20

30

40

50

60

Time (Intraannual)

Perc

ent L

ands

cape

(a)

(b)

42963_C006.indd 108 11/6/07 7:13:54 AM


spatialfootprint.Sometradeoffissuggestedbetweenwetriceagricultureandotheragricultureindryversuswetyears,likelyexplainedbyterraceposition(notanthro-pogenic rice terraces,butfluvial terracesormiddleelevationgrounds thatdonotfloodeachyear)sincericemaybeplantedatslightlyhigherelevationsinwetyearswithcashcropsoccupyingthoseareasindryeryears.Overalltheinterannualcom-positionalchangesexhibitsomesharpincreasesordecreasesovertimebutrepresentfairlystabletrajectories(notethisgraphisoverallcomposition;aperpixelcompari-sonwouldrepresentmuchmoreswitchingamongclasses,andagreaterclassificationschemedepthormoveawayfromananniversarydate/stylizedrepresentationwouldshowmuchgreaterfluctuationaswell).

Theintraannualrepresentationdepictsmuchmoremarkedfluctuationsincom-positionof these fourLULCclasses.Forests appear tohavegreater coverdue tophenologicalchanges(green-upfromrains)andlessercoverduringdeciduousevents.Rice agriculture changes dramatically with crop calendar and related monsoonaltiming,asricepaddymovefromfloodedtoplantedtofloweringtoharvesttoburn-ingtobarren;sotoodolandsofotheragriculturechange,althoughtoalesserextentthan rice owing to the drought resistant nature of some of these crops. Savannacoverappears tochangedramaticallyaswellanddoessoinresponsenotonlytogreen-upofgrassesbutmoreover toallagricultural landsspectrallymixingwithgrassysavannasduringdryerperiods.Themagnitudeofintraannualchangecom-paredtointerannualchangeunderscorestheneedforanniversarydateimagery,butalsoforunderstandingtheprocessimplicationsoftheperiodicitiesofdifferenttypesofchange.Withoutaphysicalprocessguideintermsofthecriticalityofacertainlossorgaininonecovertypeoveranother,perhapsLULCCscholarscanatleastbracketinterannualchangeinthecontextofseasonaldynamicstypicallyobservedtounderstandwhenthelandscapehaschangedbeyondits“natural”resilience.9,43

Figure6.5movestotheconsiderationofstructuralchangebyshowingtwotypi-calmetrictrajectoriesforinterannualchanges,againfromthe1970stothe1990s.Figure6.5agraphstheIJIforthesamefourclasses,andfromthisillustrationthelandscapenarrativequicklybecomesapparent.AmappedviewofthisforasubsetofthestudyareaisalsopresentedinFigure6.6,astypicallyininterpretationbothpatch-derivedstatistics(e.g.,Figure6.5)andmapsofchangeinconfiguration(Figure6.6)areused.Here,bothforestcoverandwetriceagricultureexperiencedasharpincreaseinIJIfollowedbyasharpdecrease,indicatingincreasesininterspersionfollowedby decreases. Taken together, these two trends exhibit evidence of an importantecologicalchange in the landscape,andone thatwhenmappedshowselevationaldifferences. Forested areas, notably in the southwest in Non Suwan district, hadbeenthematrixordominantclassinuplandareasintheearly1970s,butbecameincreasinglyfragmentedand interspersedasotheragriculturewas introduced.Bytheearly1990s, the foresthadbeendesiccated to littlebut remnantpatches, andalthoughstillinterspersedwithotherLULCtypes,theseforestedpatchesweresosmallthatthemetricplummetsaslessandlessforestedgeremainstoneighborotherLULCtypes.Asimilartrendofriceagricultureoccursinthelowlandareasbutfortheoppositereason:riceexperiencesasharpincreaseininterspersionasitbecamemoreandmorewidespread,untilbythelate1980sitbecamesoubiquitousthatitsspatialcohesionresultsinlowerIJIscores.SmallerchangesinotheragricultureIJI

42963_C006.indd 109 11/6/07 7:13:55 AM


supportthehypothesisthatnotonlythecompositionbutalsotheconfigurationofthisclasschangesinthemiddleelevationareas,whileexperiencingsomechangesinuplandareasattheexpensesofforestedlands.Savanna,particularlyintheeasternportion of the area, becomes increasingly interspersed with wet rice agriculture,particularlyinareasmoreproximatetotheprimaryriverchannels(i.e.,thatfloodthemostfrequently).Thislandscapenarrativeisbolsteredbyinterestinglyparallelchangesinmeanpatchsize(MPS)(Figure6.5b),wheretypicalagriculturepatches(regardlessoftypeortopographicposition)increaseasagricultureoverwhelmsthelandscapeatthespatialexpenseofbothforestandsavanna.

Movingbeyondthecombinedchangesincompositionandconfigurationtocon-figuration only11,27 illustrates trends in structure related to topography and acces-sibilitynotdetectableintheaboveefforts.First,wavesoffragmentationareeasilydiscernibleinthesouthwesternoruplandareasaspeopleliterallymovedupthehillsfromsurroundinglow-lyingareas.Typicallyuplandcropsarelesslaborintensive,

Inte

rspe

rsio

n / J

uxta

posit

ion

Inde

x 30

25

20

15

10

9876543210

(a)

(b)

Mea

n Pa

tch

Size

(Hec

tare

s)

Time (Interannual)

Time (Interannual)



FIGURe 6.5 (a)StylizedLULCpatternmetricchangefortheinterspersion/juxtapositionindex (IJI) observed and/or reported in northeast Thailand from the 1970s to late 1990s(annualchange,holdingseasonalityconstant).(b)Metricoutputformeanpatchsize(MPS)forthesametimeandlocation.

42963_C006.indd 110 11/6/07 7:13:57 AM


(a)

(b)

(c)

FIGURe 6.6 (See color insert following p. 132.) (a) The change in configurationfrom1972/1973to1975/1976,revealingthattheentiresubsetareahasexperiencedagreaterthan10%increaseininterspersion/juxtapositionindex(IJI)scores(duetoincreasedfragmentationandconcomitantinterdigitation).Notethatmostoftheareaexperiencedthesametypeofchange.(b)Illustrationofadifferent trendbetween1975/1976and1979,wherebyincreases,decreases,andrelativestabilityinIJIvaryspatially.Moreuplandareas(mostcentral in thesubset)experiencedaconsolida-tionon the landscape,whileperipheral areas remain relatively stable in termsofconfigurationwithnotableexceptionsonthesoutheasternperimeter.(c)IllustrationofthecontinuedspatialheterogeneityinIJI,withlowland/peripheralareasundergo-ingcontinuedfragmentation,whilethelessaccessible,uplandareasappeartohaveleveledoff intermsoflarger-scalefragmentationorconsolidationbutcontinuetoexperiencesmallpocketsoffragmentationthroughout.

42963_C006.indd 111 11/6/07 7:13:58 AM


anddonotrequireproximitytothenuclearvillagesettlementsthatricepaddydo.Thus,whenfactorpricesforcassavaincreasedinresponsetoEuropeandemandforcattlefeed,itwasrelativelyeasyforlowlandvillagerstoplantuplandareasnotpre-viouslyclaimedgiventhattheyhadbeenseenasundesirableforriceorresidence,giventheir(uphill)distancetorivers.Interestingly, thesouthernedgeof thisareadidnotseeanadvancementofagricultureasdidtheotherareas,possiblyduetothatside’sproximitytotheKhmer(Cambodian)borderandrelatedhistoryofmilitaryviolenceandminefields.40Inthelowlandareas,thetrendsovertimerespondmoretoaccessibility toripariancorridorsandthusassuranceofyearlyfloodingfor thebestricepaddylands.27Structurally,though,theuplandandlowlandareaswereverysimilar,despitethedifferentdominantLULCtypesandvariousdriversofLULCC.First,intermsofIJI,thecoreareas(eithermostcentralintheuplandareasorclosesttoriparianchannels)exhibitedthegreatestpossiblefluctuationinIJIacrosstime.RegardlessofLULCtype,theseareasexperiencedincreases,decreases,andperiodicepisodesofstabilityinIJIoverthe1970sand1980s.Sowhilethecovertypesmayhavebeenrelativelyslowtochangeovertime,bycomparisonthestructure,specifi-callytheinterspersionofLULCclasses,wasconstantlychangingandchangingindifferentways.Thatis,notonlywastherefirstorderchangebutalsosecondorderchange.TheareasproximatetothesecoreregionsalsoexperiencefairlydramaticstructuralchangeasreflectedbyIJI,althoughslightlymoreconsistentinthesecondorder(roughlytwoperiodsofincreasingIJIwithoneperiodofdecrease).Intermsofpercentagelandscape(PCT),thesimilaritiesinuplandversuslowlandareaswerelesspronounced.Thecoreuplandareas (the leastaccessible) sawearlydecreasesinPCT, followedby increases and subsequentdecreases, similar to the trends inLULCCandLULCpatternchangediscussedearlier.Theregionsproximatetotheuplandcorebutmoreaccessible(lowerelevations)showedtwoperiodsofdecreaseas well, although the increase in PCT came at slightly different times (a notableexceptionwas the southernmost regionmentionedearlier,whichwas structurallystable in thedynamic sense in that it experiencedadecrease inPCT throughallobservations).Lowlandareas,incontrast,nearlyuniformlydisplayedthedecrease-increase-decreasepatternforPCT,despiteobservedstabilityinLULCcomposition,suggestingthateveninthefaceofrelativelystableLULCCthestructuraldynamismoftheareawasin“constantfluctuation.”

6.4 PAneLeD PAtteRn MetRICs: MeAns oR enD?

Thesubjectivityofthepanelapproach,andparticularlythepaneledpatternmetricmethod,presentssignificantchallengesandmerits testinginecosystemswithdif-ferentlevelsofhumanimpactandlandscapeheterogeneity(acrosstimeandspace).Aparticularconcernisthedelineationofpatchboundaries;whentestingexistingpatches(betheyforestrefugia,controlplots,orcadastre-definedparcelsofland),theboundariestouseinanalysisareclear.Butotherwise,boundariesareconstructedfromayearof the imagery,and thedeterminationof thebaseyear impactswhattrends arepossible todetect and inwhatdirection.By testing the sensitivity androbustnessofresultstochangesinthebaseyear,researchershaveinpaneledpatternmetricsapotentialtoolforcreatingecologicallymeaningfulunitsofanalysisthat

42963_C006.indd 112 11/6/07 7:13:58 AM


fitwithinahierarchical(nestedornon-nested)frameworksuitablefordrawingupontheoriesoflandscapeecologyandhierarchicalpatchdynamics.4,8,9,44Withincreas-inggraininthetimeseries(greaternumberofobservationsorimages),thepossiblebiasinbaselinedeterminationshoulddrop.However,traditionalnotionsofaccuracyassessmentcalculationpositthatasthenumberoftemporalobservationsincreases,theamountofaccuracyassessmentdataneededincreasesatanincreasingrate31andthelikelihoodofanacceptablecumulativeproductdecreasessharply.If10imageswith95%accuracywereused,thetraditionalmannerofrepresentingchangeaccuracywouldmeanthat,atbest,the10-imagechangeproductwouldhaveanaccuracyofjustunder60%(0.9510):anaccuracylevelhardlyworthpursuing.Andyetitappearsthatthesemulti-inputproductsofferthemostpromiseforextractingprocess.Perhapsanewframeworkforconsideringhowtothinkaboutchangeandaccuracyisneeded.Fromphysics,Griffiths’s45consistenthistoryapproachholdspromiseasametaphorforaccuracy,wherethegreatertheobservations,thelessimportantanygivenerrorinanygivenobservationis,sincethepatternisindicativeoflargertemporalrelation-shipsatworkandnotaproductofanyonesingulareventormistakenpixelclassifica-tion.Rectifyingthislineofreasoningwithdisturbanceand(dis)equilibriumtheorypresents a challenging but fruitful avenue of epistemological and methodologicalreasoning, especially given continued sensor systems failures requiring analyststobuild timeseries frommultiplesensorplatformsand thuspossibly introducingartifactsintoclassificationswhencomparedacrosstime.30,31

IntegratingnotonlyclassificationsderivedfrommorethanonesensorsystembutotherkindsofproductspresentsanotherfrontierofLULCCwork.Duetofactorsvaryingfromcost toclimaticconditions,establishingsufficient temporalgrainorfrequencychallengesmanyresearchteams;moreover,thetemporalextentofmostLULCC work lacks the historical depth necessary to test, assess, or account forlandscape impactsoccurringover longer timeframesor in themoredistantpast:geomorphologicalchange,speciesevolution,pastcivilizationlanduses,andclimaticchangeresearchallofferevidencethattheforcesatworkonlandscapesthousandstohundredsofthousandsofyearsagostillimpactlandscapesandlandscapecompo-nentstoday(Figure6.7).22LinkingofmappedproductstorecentLULCclassifica-tionshasbeendoneinfrequentlytoextendatimeseries,buttherealchallengeliesinlinkingnon–wall-to-wallornonspatiallyexplicitproductstotoday’sdigitalproducts.Traveljournals,agriculturaltaxationrecords,andartworkofferrareglimpsesintouncharteredtemporalextentsoflandscapesoflongago,22eveninthoselackinginspatialextentandexplicitness.

The panel method generally is, perhaps obviously, of most use in data setswith a rich time series. As digital archives of LULC and LULCC become morewidelyavailable,thisapproachcanbefurthertestedindifferentlyimpactedsocialandenvironmentallandscapes.Panelanalysishasmostcommonlybeenappliedtoinformationextractedfromopticalsensorsystems,butitcouldbeextendedtootherimagerysources.Panelanalysisoffersparticularappealforlandscapeswithtemporalheterogeneity.Paneledpatternmetrics,morespecifically,offerthegreatestpotentialinsightwhenlandscapecoverchangeappearstobeseparablefromchangesinland-scapeconfiguration.Forexample,inareasofshiftingorswidden(anareaclearedfortemporarycultivationbycuttingandburningthevegetation)cultivation,theamount

42963_C006.indd 113 11/6/07 7:13:59 AM


(b) (c)

(a)

(d)

FIGURe 6.7 (a)Seasonalityimpactsthislandscapeinseveralways.Here,ricestubbleisatypical“winter”ordryseasonlandscapecomponent.Hazeinbackgroundissmokefromtraditionalburningofthericefieldstoboostnutrientsinthesedegradedsoils.(b)Cashcropagriculture tends tooccur in topographic regionsnot suitable forannual riceproduction.Shownaresugarcane(formarket)andeucalyptus(typicallyforlocalbuildingsupplies,fieldborders,orsoilstability).(c)Typicalrecentlyimproved(elevated)roadwithriceagricultureinbackground.“Borrowpits”provideroadelevationmaterialsandareoftensubsequentlyusedassmallwaterimpoundmentsforirrigation.(d)Whilemostwaterimpoundmentsinthe area are small in spatial extent, the impact of the network of such impoundments iscriticalinpopulation-environmentimpacts.Waterimpoundments,whileusedforirrigationandfloodcontrolthroughoutmanyfields,aretypicallyringedbyasmallfringeofwater-demandingcropsforhouseholdconsumption,suchaswatermelonandbananas.

42963_C006.indd 114 11/6/07 7:14:01 AM


oflandsunderanyparticularusemaychangeverylittlebetweenobservations;buttheconfigurationcouldstillbefluctuating,particularlyunderlocalizedmanagementstrategies.LandscapesundergoingevendrasticLULCCmayinfactbestructurallystablewhenwholesalereplacementofclassesoccurs.Landscapesundergoingbothcompositional and configurational change present the most complex situation forlandscapeassessment,anddisentanglingthesetypesoflandscapechangeiscriticalforextractingabetterunderstandingofprocessandfunctionfrompattern.Tempo-rally,landscapeswithheterogeneouschange(differentratesofchangeoverdifferentperiods) would benefit more from the general pattern approach than those withconsistenttemporalprocesses,butproperextractionofthesepatternspresumesanadequatelyrichtimeseries.

The panel method generally, as used here and elsewhere,34 offers LULCCresearchersonewayofslidingcloser toprocesson thepattern-processspectrum.Assessing dynamism (continuous change) from dynamics (changes and driversassessed via snapshots in time) remains a critical area of concentration for theLULCC community. The paneled pattern metric approach provides a means forexploring stronger linkages to process and function from patterns of LULC andLULCC.Moreover,paneledpatternmetricsconstituteanexplicittestofthevalueof landscapeecologytoLULCCworkbyexploring therelativecontributionsandinteractionsoflandscapecompositionandconfiguration.Theimplementationofthistool is,currently,pronetosubjectivity.Althoughpatternmetricshavebeenfoundtorevealcriticaldifferencesinlandscapes,rarelyaretheyexplicitlyandquantita-tively linked to human or biophysical processes. As such, determining appropri-ate thresholds forcategorizationofconstrainedorunconstrainedmetrics remainscriticaltodobutdifficulttojustify.Aswiththeintroductionofvegetationindicesseveraldecadesago,astatisticalcorrelationmayconvincesomeoftheutilityofanapproach,butultimatelyitistheempiricalandquantitativetietoprocessthatcon-vincespractitionersoftheapproach’sworth.ThepursuitofthatlinkageisaripeareaforresearchofecologicallyandbiophysicallygroundedLULCCteams.Itmaybethattheprocesslinkagebetweenpaneledpatternmetricsandlandscapeprocessesisneverdiscoveredorverified,oreventhatitisrejectedanddisproven.Butuntilsucha time,requiringaprocess linkagemaybepremature,whenthegreatestpromiseofpaneledpatternmetricsmaylieinapplicationofadataminingapproachtofirstuncovercriticalthresholdsorflippointsofLULCC,andthenbringtobeartheoriesandmethodsforfleshingouttheprocessesatwork.

ACKnoWLeDGMents

Iamgratefultothefollowingforsupportofthiswork:CarolinaPopulationCenterandDepartmentofGeography,UniversityofNorthCarolina;SigmaXiScientificResearchSociety;andMahidolUniversity,Thailand.

ReFeRenCes

1. Walsh,S.J.,andCrews-Meyer,K.A.Linking People, Place, and Policy: A GIScience Approach.KluwerAcademicPublishers,Boston,2002.

42963_C006.indd 115 11/6/07 7:14:02 AM


2. Meyer, W. B., and Turner, B. L., II, eds. Changes in Land Use and Land Cover: A Global Perspective.CambridgeUniversityPress,Cambridge,1994.

3. Liverman,D.etal.People and Pixels: Linking Remote Sensing and Social Science.NationalAcademyPress,Washington,D.C.,1998.

4. Turner,M.G.,Gardner,R.H.,andO’Neill,R.V.Landscape Ecology in Theory and Practice: Pattern and Process.Springer,NewYork,2001.

5. NationalResearchCouncil.Population, Land Use, and Environment: Research Directions.PanelonNewResearchonPopulationandtheEnvironment.Entwisle,B.,andStern,P.C.,eds.,CommitteeontheHumanDimensionsofGlobalChange,DivisionoftheBehavioralandSocialSciencesandEducation.NationalAcademiesPress,Washington,D.C.,2005.

6. Frohn, R. C. Remote Sensing for Landscape Ecology: New Metric Indicators for Monitoring, Modeling, and Assessment of Ecosystems.LewisPublishers,BocaRaton,Fla.,1998.

7. Rindfuss,R.R.etal.Developingascienceoflandchange:Challengesandmethodolog-ical issues. Proceedings of the National Academy of Science 101(39), 13976–13981,2004.

8. Forman,R.T.T.Land Mosaics: The Ecology of Landscapes and Regions.CambridgeUniversityPress,Cambridge,1995.

9. Wu, J., and Loucks, O. L. From balance of nature to hierarchical patch dynamics:Aparadigmshiftinecology.Quarterly Review of Biology 70(4),439–466,2005.

10. Jensen, J. R. Remote Sensing of the Environment: An Earth Resource Perspective.PrenticeHall,UpperSaddleRiver,N.J.,2000.

11. Crews-Meyer,K.A.Characterizinglandscapedynamismviapaneled-patternmetrics.Photogrammetric Engineering and Remote Sensing 68,1031–1040,2002.

12. Ahl,V.,andAllen,T.F.H.Hierarchy Theory: A Vision, Vocabulary, and Epistemology.ColumbiaUniversityPress,NewYork,1996.

13. Gustafson, E. J. Quantifying landscape spatial pattern: What is state of the art?Ecosystems1,143–156,1998.

14. O’Neill,R.V.etal.Landscapepatternmetricsandecologicalhealth.Ecosystem Health5,225–233,1999.

15. Skole, D., and Tucker, C. Tropical deforestation and habitat fragmentation in theAmazon:Satellitedatafrom1978to1988.Science260,1905–1910,1993.

16. Plummer, S. E. Perspectives on combining ecological process models and remotelysenseddata.Ecological Modelling 129,169–186,2000.

17. Oetter,D.R.etal.LandcovermappinginanagriculturalsettingusingmultiseasonalThematicMapperdata.Remote Sensing of Environment76,139–155,2001.

18. Walsh,S.J.etal.Amulti-scaleanalysisofLULCandNDVIvariationinNangRongDistrict,NortheastThailand.Agriculture, Ecosystems, and Environment 85, 47–64,2001.

19. Walsh, S. J. et al. Patterns of change in land use, land cover, and plant biomass:separating inter- and intra-annual signals in monsoon-driven northeast Thailand. In:Millington,A.,Walsh,S.J.,andOsburn,P.,eds.,GIS and Remote Sensing Applications in Biogeography and Ecology.KluwerAcademicPublishers,TheNetherlands,2001.

20. Norman,A.L.IsolatingSeasonalVariationinLanduse/LandcoverChangeUsingMulti-temporal Classification of Landsat ETM Data in the Peruvian Amazon. MAthesis,UniversityofTexas,Austin,Tex.,2005.

21. Hall,F.G.,Strebel,D.E.,andSellers,P.J.Linkingknowledgeamongspatialandtem-poralscales:Vegetation,atmosphere,climateandremotesensing.Landscape Ecology2(1),3–22,1988.

22. Butzer, K., and Helgren, D. Livestock, land cover and environmental history: TheTablelandsofNewSouthWales,Australia,1820–1920.Annals of the Association of American Geographers95,80–111,2005.

42963_C006.indd 116 11/6/07 7:14:02 AM


23. Ramos, F. A multi-level approach for 3D modelling in geographical informationsystems.International Archives of Photogrammetry and Remote Sensing34(4),43–47,2002.

24. Axinn,W.G.,Barber, J.S.,andGhimire,D.J.Theneighborhoodhistorycalendar:Adata collection method designed for dynamic multilevel modeling. Sociological Methodology27,355–392,1997.

25. Allen,T.F.H.Thelandscape“level”isdead:Persuadingthefamilytotakeitofftherespirator.In:Peterson,D.L.,andParker,V.T.,eds.,Ecological Scale: Theory and Applications.ColumbiaUniversityPress,NewYork,1998.

26. Millington,A.C.,Walsh,S.J.,andOsborne,P.E.GIS and Remote Sensing Applica-tions in Biogeography and Ecology.KluwerAcademicPublishers,Boston,2001.

27. Crews-Meyer,K.A.AgriculturallandscapechangeandstabilityinnortheastThailand:Historicalpatch-levelanalysis.Agriculture, Ecosystems and Environment101,155–169,2004.

28. Brown,D.G.,Aspinall,R.J.,andBennett,D.A.Landscapemodelsandexplanationin landscape ecology—A space for generative landscape science? Professional Geographer,Focussection58(4),2006.

29. McGarigal,K.,andMarks,B.J.FRAGSTATS: Spatial Pattern Analysis Program for Quantifying Landscape Structure.ForestScienceDepartment,OregonStateUniversity,Corvallis,Oregon,1993.

30. Jensen,J.R.Introductory Digital Image Processing: A Remote Sensing Perspective,3rded.PrenticeHall,UpperSaddleRiver,N.J.,2005.

31. Congalton,R.G., andGreen,K.Assessing the Accuracy of Remotely Sensed Data: Principles and Practices.LewisPublications,BocaRaton,Fla.,1999.

32. Crews-Meyer,K.A.Temporalextensionsof landscapeecology theoryandpractice:LULCCexamplesfromthePeruvianAmazon.Professional Geographer,Focussection58(4),421–435,2006.

33. Crews-Meyer,K.A.IntegratedLandscapeCharacterizationviaLandscapeEcologyandGIScience:APolicyEcologyofNortheastThailand.Doctoraldissertation,UniversityofNorthCarolina,ChapelHill,2000.

34. Mertens,B.,andLambin,E.F.Land-cover-changetrajectoriesinSouthernCameroon.Annals of the Association of American Geographers90,467–494,2000.

35. Crews-Meyer,K.A.Assessing landscape change andpopulation-environment inter-actionsviapanelanalysis.Geocarto International16,69–80,2001.

36. Fukui, H. Food and Population in a Northeast Thai Village. University of HawaiiPress,Honolulu,1993.

37. Donner,W.The Five Faces of Thailand: An Economic Geography.St.Martin’sPress,NewYork,1978.

38. Arbhabirama,A.etal.Thailand Natural Resources Profile.OxfordUniversityPress,Singapore,1988.

39. Craig, I. A., and Pisone, U. Overview of rainfed agriculture in northeast Thailand.In:Proceedings of a Workshop on Soil, Water and Crop Management Systems for Rainfed Agriculture in Northeast Thailand. Khon Khaen University, Khon Khaen,Thailand,1985.

40. Crews-Meyer,K.A.Personalcommunicationwithvillagersduringfieldwork,JanuarytoFebruary,unpublished,1999.

41. Walsh,S.J.etal.Scale-dependentrelationshipsbetweenpopulationandenvironmentin Northeastern Thailand. Photogrammetric Engineering and Remote Sensing 65,97–105,1999.

42. Fox, J. et al. People and the Environment: Approaches for Linking Household and Community Surveys to Remote Sensing and GIS.KluwerAcademicPublishers,Boston,2003.

42963_C006.indd 117 11/6/07 7:14:03 AM


43. Holling,C.S.Resilienceandstabilityofecologicalsystems.Annual Review of Ecology and Systematics4,1–23,1973.

44. Peterson,D.L.,andParker,V.T.Ecological Scale: Theory and Applications.ColumbiaUniversityPress,NewYork,1998.

45. Griffiths, R. B. Consistent histories and the interpretation of quantum mechanics. Journal of Statistical Physics36,219–272,1984.

42963_C006.indd 118 11/6/07 7:14:03 AM

119

7 Developing a Thick Understanding of Forest Fragmentation in Landscapes of Colonization in the Amazon BasinAndrew C. Millington and Andrew V. Bradley

Contents

7.1 Introduction................................................................................................. 1197.2 CaseStudyandMethods............................................................................. 122

7.2.1 Chapare............................................................................................ 1227.2.2 Methods........................................................................................... 123

7.3 AConceptualModelofFragmentation......................................................1247.4 BehaviorofLandscapeMetrics.................................................................. 1287.5 Discussion................................................................................................... 133Acknowledgments.................................................................................................. 135References.............................................................................................................. 135

7.1 IntRoDUCtIon

Landscapefragmentationisakeyconcernoflandscapeecologistsandconservationbiologists.Landscapesprovidethehabitatsanddeterminetheresourcesnecessaryforplantandanimalspeciestosurvive.Aslandscapesfragment,theproportionsofdifferentelementsinanylandscapechange,asdotheirspatialproperties.1Effortsto understand spatial patterns of, and relationships between, these elements havebeenakeyobjectiveoflandscapeecology,whileconservationbiologistshavetriedto relate the responsesof species to these spatialpatterns,2,3,4 sometimes throughdirectmanipulationoflandscapes(cf. reviewofsuchexperimentsbyDebinskiandHolt5)butmoreoftenbyobservation.Manycommentatorsonenvironmentalissuesin the humid tropics cite landscape fragmentation, along with the more general“forestloss,”asmajordeterminantsofbiodiversitylossandecologicaldeterioration.

42963_C007.indd 119 10/29/07 10:48:54 AM

120 Land Use Change

Researchtounderstandecologicalresponsestoforestlossovertimeisfragmentaryin itself, but research carried out under the auspices of the Biological DynamicsofForestFragments Project in northern Amazonia6,7,8 has provided a plethora ofnotableexceptions,theresultsofwhichwererecentlyreviewedbyLaurenceetal.9

Researchundertakenbybiologistson theecologicalandconservationaspectsof fragmentationof tropical forests isvoluminouswhencompared to thenumberofresearchpapersthatlinksocioeconomicprocessestothespatialphenomenonofforestfragmentationinthehumidtropics.Therelativepaucityofresearchonhowparticularpatternsoffragmentationaregeneratedshouldbeofconcernbecauseonlywhenwe fullyunderstandhowandwhy fragmentationoccursatawide rangeofgeographicalscaleswillwebeable toplanfor,manage,andaccrueconservationbenefits.Researchundertakenonthistopicgenerallyfocusesonthespatialpatternsthatresultfrom,oraretheend pointsof,ageneralizedprocess(orsetofprocesses)oftropicalforestconversion.10,11,12,13Thesespatialpatternshavebeentermedclearance typologiesbyHussonetal.10orclearance morphologiesbyLambin.11Examplesofthelinkagesbetweenspatialtypologiesormorphologiesandgeneralizedeconomicactivities are: planned settlement creates the so-called fishbone pattern of forestloss;spontaneoussettlercolonizationalongroadnetworkscreateslinearcorridorsofclearance;large-scalecommercialranchingandothertypesofcommercialagri-culturecreate largeblocksofpasture, cultivation, and forest; subsistenceagricul-turecreatesadiffusemosaicofsmallclearings;veryhighruralpopulationdensitiesleaveanagriculturallandscapewithforestislands;andislandsofforestsurroundingurbanareasoccurwhenperi-urbanplantationspredominate.Itisthefirstofthesetypologies—plannedsettlementleadingtofishbonepatternsoflandclearance—thatweaddressindetailinthischapterbydrawingonevidencefromChapare,Bolivia.

There are limitations to these generalized process-pattern relationships.ImbernonandBranthomme13notedvariations in theprocess-pattern relationshipswithin relatively small studyareasacross the tropics,possibly indicating that thespatialscaleatwhichmuchofthisresearchhasbeencarriedoutonlyallowsverygeneralizedobservationsofthelinkagesbetweendriversoflandusechangeandtheresultingpatternsoffragmentationtobemade.Lambin11andHargisetal.14 describehow spatial patterns can morph from one to another over time as fragmentationevolves,indicatingthatsometemporaldependencymightexistandthatpatternsmaynot,inthemselves,beend points.Moregermanetoourworkisthattheseprocess-patterngeneralizationstypifya“thin”understandingofhowparticularpatternsofforestfragmentationarecreatedbythedrivers(agents)oflandcoverchange,despitethefactthattheactionsofsuchagentsinthehumidtropicsarewellunderstoodfromthesynthesesthathavebeenundertaken.15,16,17

Acase inpoint is the relatively“thin”understandingbetween roadconstruc-tionandforestfragmentationthathasdevelopedfortheAmazonBasin.18Figure7.1illustrateswhat ismeantbya“thin”understandingandindicateswherea“thick”understandingneedstobedevelopedtomeet theneedsofconservationplanning.We acknowledge that research in a limited number of colonization zones in theAmazonBasinhasmodeledlandcolonization.19,20,21,22Thefociofthesestudieshasgenerallybeenonsocietalprocessesandimpactsandontheresultingforestcoverinageneralsense.Althoughthishasdeepenedourunderstandingoflandusedynamics

42963_C007.indd 120 10/29/07 10:48:55 AM

Developing a Thick Understanding of Forest Fragmentation in Landscapes 121

incolonizationzones,knowledgeofthecausallinkagesbetweenprocessesthatleadindividuallandownerstofragmenttheforestsontheirpropertiesinparticularways,andhowthefragmentationpatternsonindividualpropertiesmeshtogetheracrossa community or a number of communities, has rarely been investigated, thoughits importance is recognized.23,24Anotableexception is the researchbyPerzandWalker25whoappliedaneo-Chayanoviananalysistosecondaryforestregrowthonsmallcolonistfarms,arguingthatmoreattentionneedstobepaidtohouseholdsasthemostproximatecontextinlandusedecisionmaking.

Forest

City A

City A

City B

City B

Conservationunit

Major road

Flow of species betweenconservation units

1

2

3

FIgURe 7.1 ThickandthinunderstandingsofforestfragmentationalongroadsinlowlandforestsoftheAmazonBasin.Theprogressionfromstage1to2showsaforestblockdissectedbyaroadconnectingcitiesAandB.Stage3illustrateslarge-scalefragmentationbetweentwoconservationunitsandtheconnectivitybetweenthemthatisrequiredisillustratedbytheblackarrow.A“thin”understandingoffragmentationisrepresentedinStages2and3;thedevelopmentof the“thick”understandingthatwearguefor in thischapter isrequiredforthewhiteareaalongthemainroad.

42963_C007.indd 121 10/29/07 10:49:01 AM

122 Land Use Change

Ifplanningandmanagementinterventionsaretobemadeduringforestconver-sioninareasundergoingcolonization,thenadetailedunderstandingofhowagentsareoperatinginthelandscapeatdifferentgeographicalscalesisessential.Forexample,theinfluenceofroadsandotherlinesofaccessoccuratonescale.Generalizationsabouttheenvironmentalimpactsofroadsatthisscalehavebeenrecognized26,27,28andused,somewhatcontentiously,tomodeltheimpactofdevelopmentpoliciesinBrazil.29,30,31,32,33Butnestedbelowtheroadnetworkingeographicalspaceisalmostalways a cadastre or land property grid. Although the roads can be constructedbothbefore andafter a cadastrehasbeen surveyed,weargue that the roadsand the cadastre provide two spatial imprints connected through a scale hierarchy inforestedlandscapesthataredestinedtofragment.Moreover,attemptstomodelfrag-mentationspatiallybasedonroadbuildinghavemissedafundamentalpoint.Thatis,itisthecolonisthouseholdswithinthelimitsoftheirpropertiesthatcreatethepatternsofforestfragmentationbyrespondingtoeconomicandpolicysignalswithmachetes and chain saws, rather than planners and road builders with maps andbulldozers.Itcouldbearguedthattheplannersandroadbuildersspatiallyconstrainwhat farmers canclear, aswell asprovide thewherewithal to extract timber andproduce fromtheir farms.Weacknowledge that theargument thatwemakeheremayonlyapplytofarmersinplannedcolonizationschemes:itmaynotapplytoothertypesofhumidtropicalforestcolonizationintheAmazonBasinorelsewhere.

Developinga“thick”understandingoffragmentationatcontemporarydeforesta-tionfrontsthereforerequiresintegratingtheactionsoflandmanagersonindividualpropertiesovertimeandmeshingthemtogetherwithintheroadnetworksandlandpropertygrids.Inanappliedvein,whatisrequiredspecificallytoplanandmanagelandscapesofcolonizationisresearchintothespatialandtemporaldynamicsofforestfragmentation,whichcovermultiplescales,consideringallagentsofchange,andthelinksbetweenagentsandscales.Theresultsofsuchresearchwillallowanimportantquestiontobeanswered.Thatis,howdothecollectiveactionsoflandmanagersinaparticulararealeadtoparticularpatternsofforestfragmentationovertrajectoriesoftime?Ifthisquestioncanbeansweredrobustly,thentwofurtherquestionsofconcerntolandscapeecologistsandconservationplannerscanalsobetackled:

1.Canzonesofcolonizationbeplannedsothattheycandevelopintomulti-purpose landscapes that allow rural production systems to co-exist withbiologicalconservation?

2.Howcanexisting,partiallyfragmentedlandscapesbeplannedfor?

Inthischapterweexplainhowwehaveattemptedtodevelopa“thick”under-standingofthedynamicsoflandscapefragmentationinacolonizationzoneinthelowlandhumidtropicsofBolivia,andthenreflectonfurtherresearchneeds.

7.2 CAse stUDY AnD MetHoDs

7.2.1 Chapare

WeusedobservationsfromtheChapareregionofBoliviainthisresearch.ChapareisacolonizationzoneinthehumidtropicallowlandsofBoliviadatingbacktothe

42963_C007.indd 122 10/29/07 10:49:02 AM


1930s, though most colonization and forest conversion has taken place since the1960s.34,35,36,37,38Theareaisboundedtothenorthbyrelativelyundisturbedlowlandtropicalforestsandtothesouthbymontaneforests.Theseforestsarelikelytoremainrelativelyundisturbedintheforeseeablefuturebecausetothenorththeyareeitherpermanentlyorseasonallyinundated,andtothesouththeyareprotectedbyParqueNacional(PN)Carrasco.Thezoneofcolonizationcreatesawedgeoflivelihoodsanddisturbancebetweenthesetwoforestblocks,therebycompromisingtheexchangeofanimalsand,lessobviously,plantmaterialbetweenthetwo.Giventhestrongaffini-tiesbetweentheanimalsandplantsinthelowlandmontaneforestsinPNCarrascoandthelowlandforests,thisisacauseofconcernforconservationists.

Chaparebenefits fromadensenetworkofprimaryandsecondary roadsaug-mentedbyfoottracks.34,37,39Thisnetworkhasdevelopedprogressivelysincethe1960sintwoways.First,byitsphysicalextension;second,byupgradingtheroadsurfacesfromdirttotarmacorcobble.Alandpropertygridhasdevelopedinparallelwiththeroadnetwork,andthetwoareintegraltocolonizationofthearea.ThelandnowoccupiedbyeachofthecommunitiesinChaparewassurveyedandmarkedoutdownto the limitsofeach landparcelby theInstitutoNacionaldeColonización(INC)before itwassettled.Titlesweregiven tocolonistsmoving intoeachcommunity.TheserecordsareheldbyINC.Thetransportationnetworkandthelandpropertygridcombinetoprovidethespatialstageonwhichcolonistsactouttheirlivelihoods,whilesimultaneouslyspatiallyconstrainingtheiractivities.

7.2.2 Methods

Tounderstand thespatialand temporal relationshipsbetween thedifferentagentsof change, one of us (Bradley) conducted detailed surveys in three communitiesbetween2000and2003.34SalientdetailsofeachcommunityarelistedinTable7.1.In each community the following research was undertaken to develop an under-standingofthespatialandtemporaldynamicsoflandcoverchange:

1.ThelandpropertygridforeachcommunitywasobtainedfromINCandtheownersofeachpropertyidentified.

2.Permissionwasobtainedfromthecommunitysindicatotointerviewlandowners/managers.Subsequentlythoseinterviewedwereselectedrandomly.

tAble 7.1salient Information Concerning the three Communities studied

Community Area (ha)Altitude (m.a.s.l.)

number of properties

Year of first settlement

Prevailing economic activities 2000–2003

Arequipa 1,220 250 60 1983 Banana,blackpepper,cassava,heartofpalm,andricecultivation

Bogotá 3,196 250–350 90 1972 Cattlerearing:beefanddairy

Caracas 1,745 220 110 1963 Banana,mandarin,andorangecultivation

42963_C007.indd 123 10/29/07 10:49:02 AM

124 Land Use Change

3.LandcovermapsofeachcommunitywerecreatedusingLandsatMSS,TM,andETM+imageryacquiredinthedryseasonsof1975,1976,1986,1992,1993,1996,and2000usingclassificationalgorithmsinERDASImagine(fulldetailsofwhichcanbefoundinBradley34).Fieldverificationwascar-riedoutonthemapsderivedfromimageryacquiredin2000(Figure7.2).Thesemapswerethensimplifiedintobinaryforestandnonforestcovers.

4.Eachlandowner/managerselectedwasshownthetimesequenceofforest/nonforestmapsfortheirpropertyandaskedtorecallaspectsofforestclear-anceandwhatcropshadbeengrownatthetimestheimageswereacquired.Thiswasdoneusingparticipatoryruralappraisalmethods,themostinfor-mativeofwhichwastowalkeachfarmer’spropertywithhim.Thisenabledfarmerstoverifytheirrecallofwhathadbeengrownatparticulartimes,andalsoenabledgeolocationoftheseobservationsusingaglobalposition-ingsystem(GPS)receiverinnondifferentialmode.

5.For each property surveyed, a forest/nonforest map—a property forest/nonforest map—wasannotatedwiththeowner/manager’sobservations.

Propertiesaretypically20hainareasofcultivationand50hainareasoflive-stockrearing.Fifteen,13,and17propertiesweresurveyedindetailforArequipa,Bogotá,andCaracas, respectively(Table7.1).Thenamesof thecommunitiesandthefarmersweinterviewedhavebeenmadeanonymousinaccordancewithnormalsocial science survey practices, and because some of the farmers have illegallygrowncocainthepast.Theobservationsmadeaboutfarmsandfarmer’sresponsesto questions were used in two ways. First, to understand the drivers of land usechangeinChaparefromthe1970stothepresenttime,34andsecond,toverifytheforest/nonforestmapsofeachcommunity—community forest/nonforest maps—thattheproperty forest/nonforest mapsofeachpropertysurveyedwereextractedfrom.Theverifiedcommunity forest/nonforest mapswerethenusedtomapareasofforestandnonforestforeachcommunity.

7.3 A ConCePtUAl MoDel oF FRAgMentAtIon

Bycomparingtheprogressivedevelopmentofspatialpatternsofforestfragmenta-tionbetweenthethreecommunitieswedevelopedaconceptualmodelofforestfrag-mentationthathassixphases(Figure7.3).Thefirstorplanningphaseoccursbeforecolonization,and,consequently,noforesthasbeenclearedatthistime(Figure7.3a).However, thisstage is importantfor it isat this timethat thegeneralspatialcon-figuration of forest and agriculture patches that will ultimately populate thisgeographicalspaceisdetermined.Thisisbecausethepropertygridissurveyedandlaidout,andthe(unimproved)accessroadsandtracksareconstructed.Thisphaseexistsforashortperiodoftimebeforecolonistsarrive.Inthesecondphase—earlycolonization—olonists clear forest at the primary ends of each property, therebycreatingasimplepatternoffragmentationoneithersideoftheprimaryaccessroad(Figure7.3b). In the three communities we researched intensively, all propertieswereoccupiedalmost immediately,and theareasofeachplotcleared(calculatedfromcommunity forest/nonforest maps)were similar because the colonists eitherarrivedtogetherorwithinafewmonthsofeachother.Moreover,theirmotivations

42963_C007.indd 124 10/29/07 10:49:03 AM


TM B

and

4 im

age:

11th

Apr

il 19

86TM

Ban

d 4:

18th

July

199

3ET

M B

and

4: 1

4th Ju

ly 2

000

FIg

UR

e 7.

2 S

eque

nce

of im

ages

sho

win

gpr

ogre

ssiv

ede

fore

stat

ion

that

is s

pati

ally

con

stra

ined

wit

hin

ane

twor

kof

pri

mar

yan

dse

cond

ary

feed

erro

ads

in b

etw

een

whi

ch a

rere

ctan

gula

rlan

dhol

ding

s. T

heim

ager

yis

fro

me

aste

rn C

hapa

reb

etw

een

1986

an

d20

00 a

ndis

10

by 1

0km

in a

rea;

all

imag

esa

re a

tmos

pher

ical

lya

ndg

eom

etri

call

yco

rrec

ted

TM

/ET

M+

Ban

d4.

(a)

Show

sth

ele

astf

ores

t cle

aran

cea

nd w

as a

cqui

red

on 1

1A

pril

198

6. T

he d

ark

gray

tone

sth

atd

omin

ate

the

imag

ear

edi

ffer

entt

ypes

of

low

land

trop

ical

fore

st.T

he m

ain

Coc

haba

mba

-San

taC

ruz

bise

cts

the

imag

ean

dfo

rest

is c

lear

eda

bout

500

mo

nei

ther

sid

eof

th

ero

ad. A

net

wor

kof

pri

mar

yfe

eder

road

sca

nju

st b

ese

enin

the

fore

sted

are

as o

nei

ther

sid

eof

the

mai

nro

ad, a

ltho

ugh

ther

eis

ver

yli

ttle

cle

aran

ce in

any

of t

he c

omm

unit

ies

thes

ero

ads

serv

e. T

heb

lack

sin

uous

pix

els

are

smal

lriv

ers

that

bis

ectC

hapa

re

and

eman

ate

from

the

And

es f

ooth

ills

to th

eso

uth

of th

eim

ages

.(b)

Im

age

was

acq

uire

don

18

July

199

3. T

heli

ght g

ray

area

sal

ong

the

prim

ary

feed

er r

oads

are

are

as o

fcl

eara

nce.

(c)

Imag

ew

as a

cqui

red

on 1

4Ju

ly2

000

and

show

sfu

rthe

rde

fore

stat

ion.

D

iffe

rent

siz

efo

rest

pat

ches

tha

t hav

ebe

enis

olat

ed b

yde

fore

stat

ion

fron

tsc

oale

scin

gfr

omd

iffe

rent

dir

ecti

ons

can

be s

een

in

the

cent

ero

fth

eim

age.

The

cas

tell

ated

nat

ure

ofd

efor

esta

tion

inth

isty

peo

fco

loni

zati

ons

chem

eca

nbe

cle

arly

see

n.

42963_C007.indd 125 10/29/07 10:49:06 AM

126 Land Use Change

18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

1 2 4 5 6 7 8 9 10 11 12 13 14 15 16 17 3 1 2 4 5 6 7 8 9 10 11 12 13 14 15 16 17 3

18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

BA

C

C

1 2 4 5 6 7 8 9 10 11 12 13 14 15 16 17 3

18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

BA

C

1 2 4 5 6 7 8 9 10 11 12 13 14 15 16 17 3

18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

C

BA

1 2 4 5 6 7 8 9 1011121314151617 3 1 2 4 5 6 7 8 9 1011121314151617 3

Stream

Property grid

Primaryaccessroad

Secondaryaccessroad(s)

(a)

(b) (e)

(f )(c)

(d)

FIgURe 7.3 Six-phaseconceptualmodelof forest fragmentationbasedonacommunitywith34plots(numbered1to34)ofequalarea.Aprimaryaccessroad(black)runsthroughthecenterofthecommunityandastreamcutsthroughproperties1to7.Imagesrepresent:(a) the planning phase, (b) the early colonization phase, (c) the illicit coca phase, (d) theimprovedaccessphasewithsecondaryroads(grayroads),(e)thecomplexclearancephase,and(f)theplotexhaustionphase.Thelightgraycellsinphases(e)and(f)representsecondaryregrowthforest.

42963_C007.indd 126 10/29/07 10:49:08 AM


forclearanceweresimilar,thatis,toclearforestforlandtoplantsubsistencecrops(followedbycashcropsinsubsequentyears)andtoacquireconstructionmaterialsforhouses.Inaless-detailedexaminationoftheimageseriesforallofChapare,wesawthisphasereplicatedinallcommunities.

Morecomplexspatialpatternsof forest fragmentationestablish themselves inthethirdphaseofthemodel—theillicitcocacultivationphase(Figure7.3c).BecausecocacultivationinthelowlandsofBoliviahasalwaysbeenillegal(incomparisontocultivationforchewinginthesubtropicalmontaneforestswhereitislegal),farmersgenerallyadoptedstrategiestocultivatecocathatfragmentedforestsinparticularways. However, in the 1970s when coca cultivation and cocaine production wasbarelycontrolledbythegovernment,cocawasgrownopenlyat theprimaryendsofmanyproperties.Asgovernmentcrackdownsoncocagrowingtookeffect,manyfarmersgrewlegalcashandsubsistencecropsattheprimaryendsoftheirplotsandretreatedintotheremainingforestontheirpropertiestoclearsmallareastogrowcoca.Thiswasthemaincauseofforestperforation,andtheextentofperforationdepictedinthismodel(Figure7.3c)ishighbecauseofillegalcocacultivationandisprobablygreater than itwouldbe inothercolonizationareas.Thisassumptionhasyet tobe tested.Thecolonist footprintmodeldevelopedbyBrondizioetal.20predictshighratesofforestclearanceatthisstageasfarmerspreparelandtoplantperennialcashcrops.Butourevidenceindicatesthatalthoughafewfarmersclearedforestatmuchfasterratesthanothers(e.g.,property28,Figure7.3c),thiswasexcep-tionalbecausethevastmajorityoffarmersonlyhadtoclearsmallareastocultivatetheperennialcropofchoice—coca—which,becauseithasahigh-sellingreturn,isconservativeinitslandrequirements.Differencesinforestclearanceratesbetweenindividualpropertiesoccuratthisstagebecausefewfarmerscultivatedland-hungryperennial crops at this time, as predicted by Brondizio et al.,20 rather than coca.Thesedifferencesleadtothecastellatedpatternoftheforest/nonforestboundariesthatcharacterize“fishbone”deforestation.

Thedevelopmentofsecondary,unimprovedfeederroadsatsomepointintimeduringcolonizationistypicalofmostcommunitiesinChapare.Roadsandtracksareconstructedalongtheboundariesofcommunitiestoconnectwiththeroadsthatwereconstructedinitially.Wehavecharacterizedthisphaseasimprovedaccess,andinFigure7.2dsecondaryfeederroadsaredrawnalongthesecondaryendofproperties1to11and18to27.Theestablishmentofsuchroadsandtracksallowsplotstobecultivatedfrombothends,buttheactualreasonsfortheirconstructionareunclearatpresent.Ourinterviewssofarsuggesttheymaybeconstructedtoconsolidatecom-munityboundaries,buttheymaysimplyimproveaccess.Whateverthereason,theycanbeusedtosplituppropertiestosatisfyactualandpotentialdisputesoverinheri-tance,orallowfarmers tocultivatemorefertilesoilsatoneendof theirpropertywhileallowingrecoveryofvegetationandsoilattheotherendoftheplot.

Inthefifthphaseofthemodel—complexclearance—asignificantamountofthelandinacommunityisundersometypeofcultivation(Figure7.3e).Thetermcomplexarisesbecausemanylandscapeecologymetricsattaintheirhighestvaluesduringthisphase.Theformationofbothforestpatchesandtheextensionoftheforestperimeterare due to differential rates of clearance between farmers, and the continuation offorestclearancefromboththeprimaryandsecondaryendsofsomeproperties.The

42963_C007.indd 127 10/29/07 10:49:08 AM

128 Land Use Change

formationofaforestpatchduetothedifferencesinratesofforestclearanceisshowninFigures7.3dand7.3e.Thefarmersinproperties21and23haveclearedforestatfasterratesthanthefarmerinproperty22.Asaconsequenceapatchofforest(B)onproperty22,whichonceshieldedacocaclearing(A),isnowsurroundedbyagricul-turalland.ThismethodofpatchformationiscommonplaceinChapare,andoccursbecause of the intersection of government coca eradication policies (which causesperforation-styleclearanceof forestdeep inproperties),differences incropchoicesbetweenfarmers(whichleadstodifferentlandrequirementstogrowparticularcrops),and differences in household circumstances and aspirations. The creation of forestpatchConproperty10isavariantonthewayinwhichforestpatchBwasformed.Inthiscasetheratesofforestclearancebetweenproperties9,10,and11arenotonlydif-ferentintheamountsofforestclearedannually,butalsothedirectionsofclearancearedifferentbecauseoftheinfluenceofthesecondaryaccessroadonproperty9.

Wehavetermedthefinalphaseplotexhaustion(Figure7.3f).Bythiswemeanthatmostoftheforesthasbeencleared.Someisolatedpatchesremain,andtherearealsopatchesofsecondaryregrowthforestandforestsinareasthataredifficulttoclearorarelocatedonlandthatcannotbecultivated(e.g.,theriparianforestinproperties1to7).

We have evidence that properties are already changing hands by the timethepenultimateandfinal stagesof themodelare reached.Althoughwehavenotrecordedlandbeingsoldinthe45propertieswehavesurveyedindetail,wehavecomeacrossthisonfarmswehavevisitedinothercommunities.Somepropertiesare being sold to new owners and some wealthy farmers purchase adjacent plotstoincreasetheircontiguouslandholdings.Wehaveindicatedthisinthemodelbycombiningproperties28and29inFigure7.3f.

7.4 beHAVIoR oF lAnDsCAPe MetRICsThe conceptual model outlined above is based on detailed observations made inthreecommunities,andtoevaluateitsutilityforanalyzingtheecologicalimplica-tionsoffragmentationweusedmetricscommonlyemployedbylandscapeecologistsand conservation biologists. We calculated proportional forest cover, the numberofforestpatches,andtheforest/nonforestedgelengthforeachphaseinthemodel(Table7.2).ThesedataarevisualizedinFigure7.4.

Lambin11postulatedthatlandscapemetricsusedtocharacterizefragmentationwouldfollowaparticulartrajectoryastropicalforestlandscapeschangedfromthosethatwere entirely forested, through landscapesof agricultural patches in a forestmatrix,toentirelyagriculturallandscapes(i.e.,afewforestpatchesinanagricul-turalmatrix).Hedidnotquantify thispostulatedbehavior,buthypothesized thatthemetricswouldattainpeakvaluesintheheterogeneous,intermediatelandscapesand would be low for homogenous forest or agricultural landscapes. Trani andGiles40 simulated deforestation of a hypothetical forest and calculated metrics atvariouspoints alongadeforestation/fragmentation trajectory.Threemetrics fromtheir analysis—mean forest patch size, the forest/nonforest edge length, and themeannearestneighbordistancebetweenforestpatches—areshowninFigure7.5.We calculated the same landscape ecology metrics as Trani and Giles40 for eachcommunitywestudiedusingFragstats.41Asthemetricsfollowedsimilartrendsineachcommunity,weonlyillustratethemetricsforCommunidadArequipainthis

42963_C007.indd 128 10/29/07 10:49:09 AM


tAble 7.2selected landscape Metrics Calculated for the six Phases of the Conceptual Model

Phase in conceptual model

Proportional forest loss

(%)

number of forest

patches

number of cultivation

patches

Mean patch size (nominal

units)

Forest edge length

(nominal units)

Planning 0 1 0 306 0

Earlycolonization 10.8 2 1 137 37

Illicitcoca 29.7 4 20 71 140

Improvedaccess 36.6 6 15 49 151

Complexclearance 58.8 12 7 11 203

Plotexhaustion 83.3 20 2 3 135

100

350

300

250

200

150

100

50

80604020

Mea

n pa

tch

size

100

250

200

150

100

50

80604020

Tota

l for

est e

dge l

engt

h

Proportion of forest cleared (%)(b)

Proportion of forest cleared (%)(a)

FIgURe 7.4 Metricscalculatedfortheconceptualmodel:(a)meanforestpatchsize,and(b)totaledgelength.Bothareinnominalunits.

42963_C007.indd 129 10/29/07 10:49:10 AM

130 Land Use Change

10080604020

Mea

n pa

tch

size (

km2 ) 1000

800

600

400

200

Tota

l for

est e

dge l

engt

h (k

m)

MN

N d

istan

ce b

etw

een

fore

st p

atch

es (k

m)

10080604020

250

200

150

100

50

10080604020

1.2

1.0

0.8

0.6

0.4

0.2



Proportion of forest cleared (%)(c)

FIgURe 7.5 SelectedmetricsfromTraniandGiles40: (a)meanforestpatchsize inkm2,(b)totaledgelengthinkm,and(c)meannearestneighbordistancebetweenforestpatchesinkm.

42963_C007.indd 130 10/29/07 10:49:12 AM


chapter(Figure7.6)and,asforestlosshadonlyreached46%by2000forthiscom-munity,thedatadonotextendtoveryhighproportionalforestlosses.Comparingthe limited number of metrics in Figures7.4, 7.5, and 7.6 provides an initial testof the robustnessof theconceptualmodel.Adecline in forestcoverover time intheconceptualmodelisclearinTable7.2.ThisallowsthesuccessivephasesofthemodeltobeparameterizedasproportionalforestlossesinthegraphsinFigure7.4.BothLambin11andTraniandGiles40usedforestcoverintheirgraphs.

Mean patch size declines in a consistent manner (Figures7.4a, 7.5a, and 7.6a).Initially thedecline is rapid,more so in reality (inCommunidadArequipa) than inthemodelorthesimulation.Ataroundhalftheareadeforested,therateofdecreaseinmeanpatchsizedeclinesandthentherateofdecreaseinforestpatchsizetapersoff.Totaledgelengthinitiallyincreasesasforestsbegintobecleared,andtheedgelengthisatitsgreatestatintermediateforestcoversandthendeclines.Thisisevidentintheconceptualmodel, the simulation, and in the realdata,Figures7.4b,7.5b, and7.6b,respectively.Therearedifferencesinthepeakvaluesoftotaledgelength,whichsuggestthevariationinthepeakmayberelatedtothespatialconfigurationoffragmentation.WhereasTrani andGiles40 simulateda somewhat randompatternof fragmentation,thatinCommunidadArequipaandthesimulationmodelareforregularlystructuredlandscapes,which,becausepropertiesarelongandthin,hasatendencytohavehighedge lengths at intermediate forest losses compared to lower edge lengths in morerandomlyfragmentedforests.Thethirdmetric—meannearestneighbor(MNN)dis-tancebetweenforestpatches—hasonlybeencalculatedfortheactualdatafromCom-munidadArequipaandextractedfromTraniandGiles’s40simulation.Forthismetricthere isadifference inbehavior.There is relatively littlevariation in theMNNdis-tanceinCommunidadArequipa,andthedistancesovertherangeofforestcoversinthe community describe a shallow U-shape. However, Trani and Giles’s simulationdescribesanupturned-UdistributioninMNNdistanceasforestisprogressivelylost.

Insummary,ouranalysisoflandscapemetricssuggeststhat:

1.Meanpatchsize(MPS)andtotaledge length(TEL)areveryrobustandconsistentmeasuresoffragmentation.AlthoughinallthreecasestherearesimilartrendsinMPSandTELwithprogressivedeforestationandincreas-ingforestfragmentation,therearedifferencesintheprecisenatureofthecurves,whichmaybeafunctionoftheeffectthatthespatialconfigurationof land properties have on fragmentation. From the viewpoint of modelvalidationthelatterpointisnotthatimportant,butitdoeslendweighttotheargumentthatunderstandinglandowner’sdecisionsatthesmallscaleisimportantfordevelopingourunderstandingoffragmentationandfutureplanningofcolonizationzones.

2.Thebehaviorofthemeannearestneighbordistancebetweenforestpatchesdoes, however, vary between the studies and is due either to the differ-encesinspatialscalebetweenthetwostudiesorthenatureofforestloss.InCommunidadArequipathespatialimprintofthecadastralgridandthewaysinwhichpeopleclearforestwithinthegridmayleadtoarestrictedset of possible distances between forest patches and, as these have onlybeenobservedintheearlystagesoffragmentation,theymaychangesig-nificantlyasmoreforestisclearedandfragmentationproceeds.

42963_C007.indd 131 10/29/07 10:49:12 AM

132 Land Use Change

10080604020

Mea

n pa

tch

size (

km2 )

MN

N p

atch

es (k

m2 )

60

50

40

30

20

10

Tota

l for

est e

dge l

engt

h (k

m)

10080604020

120

100

80

60

40

20

10080604020

60

50

40

30

20

10



Proportion of forest cleared (%)(c)

FIgURe 7.6 MetricscalculatedforCommunidadArequipa:(a)meanforestpatchsize in km2, (b) total edge length in km, and (c) meannearest neighbor distancebetweenforestpatchesinkm.

42963_C007.indd 132 10/29/07 10:49:13 AM


3. In theanalysisof landscapemetricsdescribed in thischapter the resultsfromTraniandGiles’s studymaydiverge fromourconceptualmodelattheplotexhaustionstageinwaysthatwedonotyetunderstand.ThismayoccurbecausewhereasthedeforestationsimulatedbyTraniandGileswasfor progressive deforestation, what happens in colonization schemes isthattherecanbesignificantregrowthinthemediumtolaterstageinthesequenceofdeforestation.

7.5 DIsCUssIon

Wesee the conceptualmodelwehavedeveloped fromChapareas apreliminaryattempt to thickenourunderstandingofwhyparticularpatternsof fragmentationoccur in a so-called fishbone colonization scheme. The regularity of the spatialimprintsofroadsandpropertygridsinsuchareasmadetheminterestingcandidatesforthisinitialinvestigation(Figure7.7).Thisstudyis,however,limited,partlybytherelativelysmallnumberofcommunitieswehaveresearchedintensively,whichmight lead tocontext specificgeneralizations,42 andpartlybecausewehaveusedretrospective analyses of decision making by colonists. While taking a differentapproachtoPerzandWalker25—byfocusingoncolonist’sresponsestochangesineconomicconditionsandanti-cocapoliciesandthelossofprimaryforest—wejointhemintheirclarioncallforresearcherstofocusontheforestoutcomesofsmall-farmcolonists.Theystatethat“most land use models … do not take account of land taken out of production and left to fallow”(p.1009),andwewouldarguethatmostlandusemodelsdonottakeaccountoflandtakenoutofproduction,lefttofallow,or how land under primary and secondary forest is spatially configured. Onlyifweresearchalongtheselineswillwebeabletoapplymethodssuchasthoseadvocatedtoplancolonizationinforestareas43orbeabletorestorefragmentedareas.44

Thefieldthenisopenforfurtherresearch,andwearguethefollowinglinesofinvestigationareneededtodeepenourunderstandingfurther:

1.More communities in Chapare could be investigated, particularly thosewithphysicalandsocioeconomiccharacteristicsotherthanthoseoutlinedinTable7.1.However,amoreprofitablelineofinvestigationwouldbetoresearchclustersofadjacentcommunities.Thiswouldallowinteractionsbetweencommunitiesattheirboundariestobeinvestigatedandmightalsorevealtheextenttowhichcooperationbetweencommunitiestakesplace,orgaugethepotentialofcooperationinthefuture.

2.ComparativeresearchbetweenChapareandsimilar—fishbone—coloniza-tionschemesintheAmazonBasinwouldenableustofurtherconsidertherobustnessofmanyaspectsoftheconceptualmodelwehavedevelopedandenableustomovetowardageneraltoolthatcouldbeusedforbasin-wideintegratedplanningincolonizationschemes.

3.So far,ourmodel reliesona retrospectiveanalysisofdata.However, itsutilityforplanningliesinitsabilitytointegrateconservationintheplan-ningoffuturecolonizationintheAmazonBasin.Therefore,researchintodecisionmakingbyindividuallandownersandsindicatos underdifferent

42963_C007.indd 133 10/29/07 10:49:14 AM

134 Land Use Change

futureeconomic,political,andenvironmentalscenariosisbothattractivetoresearchersandessentialtoplanners.

4.Aswenotedin the introduction, thisstudyisgroundedinoneclearancetypology—planned settlement leading tofishbonepatternsof land clear-ance.Othertypologiesrequiresimilarresearch.

Basedonourobservationsanaggingquestion remains: is itpossible tobuildstructuresforconservationinplansforcolonizationschemes?Therequirementisto

a)

b)

c)

FIgURe 7.7 (a)Chapareroadbuilding.TheextensionoftheroadnetworkinChaparecon-tinues.Thisphotographwas taken inAugust2003,andshowsa roadbeingpusheddeepintorelativelyintactlowlandtropicalforestalongthelineofaformerfootpathwhichlinkedsomeisolatedsettlementstotheformerroadhead.(b)IntheeastofChaparecattlerearingisthemainfarmingactivity.Landparcelshereare50hainarea,comparedto20hainsettle-mentsdominatedbycultivation.(c)BananasareoneofthemainalternativecropstococainChapare.Theyoftenresultinlargeareasbeingclearedtocompensateforlossofcocaincomewhichleadstolargeareasofnonforestmonoculture.Theroadcuttingthroughthisbananaplantationisanunimprovedprimaryfeederroad.

42963_C007.indd 134 10/29/07 10:49:15 AM


setasidesomelandwithinthecolonizationschemesthatwilleitheractas“sinks”or“reservoirs”withthecolonizationzone,ortoplanwildlifecorridors—eithercon-tiguous forest corridors or stepping-stones of forest patches for migration acrossand within the colonization landscapes. Theoretically this is possible, but wheredoes thelandcomefrom?InChaparetheoriginalcadastrehadlandthatwasnotassignedtosettlers;oftenstripsofforestalongriversor, to thenorth,forests thatare inundated for many months each year. In other words wildlife corridors andforest patches were “planned by accident” but were not afforded protection untilforestsalongwatercourseswerespecificallyprotectedinthenewForestryLawthatwaspassedin1996.45Disappointinglyfromtheviewpointofconservation,butnotsurprisinglygiventhedemandsonlandandthelaissezfaire attitudetospontaneouscolonization, many of these unallocated lands have been occupied, the exceptionbeingtheinundatedforests.Forexample,adjacenttoCommunidadArequipaastripofforestlandalongariverhasbeenillegallycolonized,andanunallocatedforestareaadjacenttoCommunidadCaracaswasaddedtothecommunityasitexpandedaftertheyhadpetitionedINC.

Ifwe fail to thickenourunderstandingofhow landmanagersmake landusedecisionsinlandscapesofcolonization,primaryforestswillcontinuetodisappearinwayswedonotcomprehend,secondaryforests(whichhavedifferentecologicalproperties and conservation values to primary forest) will come and go, and thespatialoutcomesmaycontinuetosurpriseus.Theselandscapeswilllosetheirabilitytoallowfaunalandfloralinterchangebetween“intactforestblocks”asthedeforesta-tionfrontstheyrepresentclosedown.Newdeforestationfrontswillopenup.Ifweareunabledevelopa “thick”understandingand inject it intoplanningprocesses,thediseaseof“thin”understandingwillcontinuetoprevailandinalllikelihoodthescenariosoutlinedforroadcorridorsinBrazilbyFearnside29andLaurenceetal.30,31willbecomecommonplaceinAmazonia.

ACKnoWleDgMents

PartofthisresearchwasfundedbytheEuropeanUnion(ContractERBIC189CT80299).AndrewBradley’sPhDresearchwaspartlyfundedbythisgrant,aswellasaSlawsonAwardfromtheRoyalGeographicalSocietywiththeInstituteofBritishGeographers,andfundingfromtheUniversityofLeicester.WearegratefultoRichardJ.AspinallandMichaelHill for invitingus topresentourresearch, toChristianBrannstromforcriticallycommentingonanearlyversionofthischapter,andtoFelixHuancaViracaforaccompanyingusinthefield.

ReFeRenCes

1. Godron,M.,andForman,R.T.T.Landscape Ecology.JohnWiley&Sons,NewYork,1986. 2. Fahrig,L.Effectsofhabitatfragmentationonbiodiversity.Annual Review of Ecology

and Systematics34,487–515,2003. 3. Ries,L.etal.Ecologicalresponsestohabitatedges:mechanisms,modelsandvariability

explained.Annual Review Ecology, Evolution and Systematics 51,491–522,2004. 4. Turner,I.M.Specieslossinfragmentsoftropicalrainforest:Areviewoftheevidence.

Journal of Applied Ecology33,200–209,1996.

42963_C007.indd 135 10/29/07 10:49:16 AM

136 Land Use Change

5. Debinski,D.M.,andHolt,R.D.Asurveyandoverviewofhabitatfragmentationexperi-ments.Conservation Biology14,342–355,2000.

6. Bierregaard,R.O.,Jr.,andLovejoy,T.E.Thebiologicaldynamicsoftropicalrainforestfragments.Bioscience42,859–866,1992.

7. Laurence,W.F.,andBierregaard,R.O.Fragmented tropical forests.Bulletin of the Ecological Society of America77,34–36,1996.

8. Laurence,W.F.,andBierregaard,R.O.,eds.Tropical Forest Remnants.UniversityofChicagoPress,Chicago,1996.

9. Laurance, W. F. et al. Ecosystem decay of Amazonian forest fragments: A 22-yearinvestigation.Conservation Biology16,605–618,2002.

10. Husson,A. et al.Studyof forestnon-forest interface:Typologyof fragmentationoftropicalforest.TREESSeriesB,ResearchReportNo.2,EuropeanCommissionEUR16291EN,Brussels,1995.

11. Lambin, E. F. Modelling and monitoring land-cover change processes in tropicalregions.Progress in Physical Geography21,375–393,1997.

12. Mertens,B.andLambin,E.F.SpatialmodellingofdeforestationinSouthernCameroon.Spatial desegregation of diverse deforestation processes. Applied Geography 17,143–162,1995.

13. Imbernon,J.,andBranthomme,A.Characterizationoflandscapepatternsofdeforestationintropicalrainforests.International Journal of Remote Sensing 22,1753–1765,2001.

14. Hargis,C.D.,Bissonette,J.A.,andDavid,J.L.Understandingmeasuresoflandscapepattern.In:Bissonette,J.A.,ed.,Wildlife and Landscape Ecology: Effects of Pattern and Scale,Springer-Verlag,NewYork,231–261,1997.

15. Rudel,T.,andRoper,J.Thepathstorainforestdestruction:Crossnationalpatternsoftropicaldeforestation,1975–90.World Development25,53–65,1997.

16. Geist,H.J.,andLambin,E.F.Whatdrivestropicaldeforestation?LUCCReportSeriesNo.4,CICAO,Louvain-la-Neuve,2001.

17. Lambin,E.J.,Geist,H.J.,andLepers,E.Dynamicsandland-useandland-coverchangeintropicalregions.Annual Review of Environment and Resources28,205–241,2003.

18. Alves,D.Space-timedynamicsofdeforestationinBrazilianAmazonia.International Journal of Remote Sensing23,2903–2908,2002.

19. Arima, E. Y. et al. Loggers and forest fragmentation: Behavioral models of roadbuildingintheAmazonBasin.Annals of the Association of American Geographers95,525–541,2005.

20. Brondizio, E. S. et al. The colonist footprint: Toward a conceptual framework ofdeforestation trajectories among small farmers in frontier Amazonia. In: Wood C.,andPorro,R.,eds.,Deforestation and Land Use in the Amazon.UniversityPressofFlorida,Gainesville,Fla.,133–161,2002.

21. Walker,R.T.MappingprocessandpatterninthelandscapechangeoftheAmazonianfrontier.Annals of the Association of American Geographers93,376–398,2003.

22. Walsh,S.J.etal.Characterizingandmodelingpatternsofdeforestationandagricul-turalextensificationintheEcuadorianAmazon.In:Walsh,S.J.,andCrews-Meyer,K.,eds.,Linking People, Place and Policy.Kluwer,Boston,187–215,2002.

23. Moran, E. F. et al. Integrating Amazonian vegetation, land-use and satellite data.BioScience44,329–338,1994.

24. Pichón,F.J.Theforestconversionprocess:adiscussionofthesustainabilityofpre-dominantlandusesassociatedwithfrontierexpansionintheAmazon.Agriculture and Human Values13,32–51,1996.

25. Perz,S.G.,andWalker,R.T.HouseholdlifecyclesandsecondaryforestcoveramongsmallfarmcolonistsintheAmazon.World Development30,1009–1027,2002.

26. Reid, J. W., and Bowles, I. A. Reducing the impacts of roads on tropical forests.Environment39,10–17,1997.

42963_C007.indd 136 10/29/07 10:49:16 AM


27. Holdsworth,A.R.,andUhl,C.FireinAmazonianselectivelyloggedrainforestandthepotentialforfirereduction.Ecological Applications7,713–725,1997.

28. Nepstad,D.etal.Roadpaving,fireregimefeedbacks,andthefutureofAmazonforests.Forest Ecology and Management154,395–407,2001.

29. Fearnside, P. M. Avança Brasil: Environmental and social consequences of Brazil’splannedinfrastructureinAmazonia.Environmental Management30,735–747,2002.

30. Laurence, W. F. et al. The future of the Brazilian Amazon. Science 291, 438–439,2001.

31. Laurence,W.F.etal.DeforestationinAmazonia.Science304,1109–1111,2004. 32. Nepstad,D.etal.AvançaBrasil:Os Custos Ambientais para Amazônica.Institutode

PesquisaAmbientaldeAmazônica(IPAM),Belém,Brazil,24pp,2000.(Availableat:http://www.ipam.org.br/avanca/politicas.htm).

33. Soares-Filho,B.etal.Simulatingtheresponseofland-coverchangestoroadpavingand governance along a major Amazon highway: The Santarém-Cuiabá corridor.Global Change Biology 10,745–764,2004.

34. Bradley,A.V.Land-UseandLand-CoverChangeintheChapareRegionofBolivia.PhDthesis,UniversityofLeicester,UK,2005.

35. Henkel, R. The Chapare of Bolivia: A Study of Tropical Agriculture. PhD thesis,UniversityofWisconsin,Madison,Wisc.,1971.

36. Henkel,R.Coca(Erythroxylum coca)cultivation,cocaineproduction,andbiodiversitylossintheChapareregionofBolivia.In:Churchill,S.P.,Balslev,H.,Forero,E.,etal.,eds.,Biodiversity and Conservation of Neotropical Montane Forests.TheNewYorkBotanicalGarden,Bronx,N.Y.,551–560,1995.

37. Millington, A. C., Velez-Liendo, X., and Bradley A. V. Scale dependence in multi-temporal mapping of forest fragmentation in Bolivia: Implications for explainingtemporal trends in landscape ecology and applications to biodiversity conservation.Photogrammetry and Remote Sensing57,289–299,2003.

38. Steininger,M.C.etal.TropicaldeforestationintheBolivianAmazon.Environmental Conservation 28,127–134,2002.

39. MACA:MinisteriodeAsuntosCampesinosyAgropecuarios.ModelsofExploración–Tipo,parasieteSubregionesdelsubtropicohumidodeCochabamba.IBTA,Chapare,Bolivia,1991.

40. Trani,M.K.,andGilesR.H.,Jr.Ananalysisofdeforestation:Metricsusedtodescribepatternchange.Forest Ecology and Management114,459–470,1999.

41. McGarigal,K.,andMarks,B.Fragstats.Spatialpatternanalysisprogramforquantifyinglandscapestructure.ForestScienceDepartment,OregonStateUniversity,Corvallis,1994.

42. Fujisaka,S.,andWhite,D.Pastureorpermanentcropsafterslash-and-burncultivation?Land-usechoiceinthreeAmazoncolonies.Agroforestry Systems42,45–59,1998.

43. Venema, H. D., Calamai, P. H., and Fieguth, P. Forest structure optimization usingevolutionary programming and landscape ecology metrics. European Journal of Operational Research16,423–439,2005.

44. Lamb, D. J. et al. Rejoining habitat remnants: Restoring degraded forest lands. In:Laurence,W.F.,andBierregaard,R.O.,eds.,Tropical Forest Remnants.UniversityofChicagoPress,Chicago,1997.

45. GovernmentofBolivia.LeyForestal1700.LaPaz.Bolivia,1997.

42963_C007.indd 137 10/29/07 10:49:17 AM

42963_C007.indd 138 10/29/07 10:49:17 AM

139

8 Urban Land-Use Change, Models, Uncertainty, and Policymaking in Rapidly Growing Developing World Cities: Evidence from China

Michail Fragkias and Karen C. Seto

Contents

8.1 Introduction................................................................................................. 1398.2 ModelingUrbanLandUseChange,Policymaking,andUncertainty........ 140

8.2.1 ModelingUrbanLandUseChange................................................. 1408.2.2 PolicyMaking.................................................................................. 1428.2.3 TheIntersectionofModelingandPolicyMaking........................... 1438.2.4 PolicyEvaluationandUncertainties................................................ 146

8.3 ANewApproachinModelingUrbanGrowthinDataSparseEnvironments.............................................................................................. 1478.3.1 MechanicsoftheModel.................................................................. 1478.3.2 ApplicationtoThreeCitiesofthePearlRiverDelta,China........... 150

8.4 DiscussionandConclusions........................................................................ 154References.............................................................................................................. 159

8.1 IntroduCtIon

Projectionssuggestthatastheworld’surbanpopulationwilljumpto61%by2030(from today’s50%mark),mostof thisurbangrowthwilloccurprimarily in lessdeveloped countries, and in Asia in particular.1 Much interest already exists inmegacities—citieswithpopulationsof10millionormore—onwhichasignificantamountofinformationisbeingcollected.Ithasbeennotedthoughthatthemajorityofurbangrowthwilloccurinmedium-sizedcities.2Giventhaturbangrowthisa

42963_C008.indd 139 11/6/07 7:12:47 AM

140 Land Use Change

major component of global environmental change3,4 and the danger of potentialundesirableenvironmentalandsocialeffectscausedbyhighratesofgrowthisever-present,therelativeimportanceofstudyingmedium-sizedcitiesversusmegacitiescitiesinthenextcenturyishigh.Furthermore,developingworldcitieshavelimitedhumanandfinancialresourcesemployedinvariousaspectsofpolicymaking.Con-sequently,thecollectionofreliabledataandtheuseofmoreadvancedmethodsinplanningpracticeandpolicymakingbecomesextremelydifficult(Figure8.1).

Policymakersindevelopingworldcitiesareincreasinglyfacedwithpressuretoassesstheimpactoftheirlandusestrategiesandpolicies5ashighpopulationgrowthtrends are predicted for at least the next 25 years. Potential socioeconomic andenvironmentalimpactsofpoliciescanbeassessedwithquantitativemodels.Giventhe number and underlying motives of different approaches to modeling, policy-makers,especiallythoseindevelopingworldcities,couldbenefitfromassistanceinchoosingthemostappropriatemodel.Iscurrenttechnologyormethodologyadvance-mentbasedoncurrentandrecentfuturerealitiesofmedium-sizeddevelopingworldcities?Prosandconsofdifferentmodelingapproachesforlandusepolicymakingneed to be evaluated given the particularities of such cities (e.g., the problem ofincompleteandscarceinformation).Thesuccessofsustainabledevelopmenteffortsreliessignificantlyontheidentificationofbetter(asaccurateaspossible)forecastingschemesregardingratesandpatternsoffutureurbandevelopmentthatalsoconnectbetterwiththeprocessofpolicymaking.Thus,thischapterprovidesaninquiryintoquestionsandtradeoffsapolicymakerfaceswhenitcomestothechoiceofcontext-specific suitable modeling tools and the establishment of guidelines assisting thedecision-makingprocess.

Thepurposeofthischapteristhreefold.First,itdiscussesissuesofurbanlandusechangemodelingandexplorestheintersectionoflandusemodelingwithurbanpolicymakingatdifferentscalesinthecontextofdevelopingworldcities.Second,it discusses the effects of uncertainties in the data sources, theories, and modelsmethodsofaddressingtheseissues.Third,itreviewsapredictivemodelofrapidurbantransformationthatrelatesastandardmodelingtraditiontoanexplicituncertainty-reducingpolicy-makingframeworkusingChinesecitiesasacasestudy.

8.2 ModelIng urban land use Change, PolICyMakIng, and unCertaInty

8.2.1 Modeling Urban land Use Change

Urbanareas,andtheirformandfunction,havebeenstudiedinthecontextsofurbanplanning,urbaneconomics,urbangeography,andurbansociology,muchofwhicharegroundedinthespatiallandusemodelsofvonThünen.6Aneedforquantita-tiveanswersregardingtheeffectsofextent,rateofchange,andpatternsofglobalurbanlandusechangehasledtothedevelopmentofurbanlandusechangemodels(ULCM).AULCMisasimplificationofreality,anditssuccessliesinretainingthefundamentalcharacteristicsofthesystembysimplifyingrealityasmuchasneces-sary(butnotbeyondthat).Thoughtofasatool,aULCMtargets“usefulness”;unfor-tunately,thiscapacityisnotalwayssuccinctlystatedordemonstrated.

42963_C008.indd 140 11/6/07 7:12:47 AM

Urban Land-Use Change, Models, Uncertainty, and Policymaking 141

-180

-60

Robi

nson

Pro

ject

ion

Cent

ral M

erid

ian

0.00

Sour

ce: E

SRI D

ata &

Map

s CD

Kilo

met

ers

0

1,45

0

2

,500

5,50 0

8,70

0

11,6

00

C

ount

ry B

ound

ary

POPU

LATI

ON

7500

00 –

1000

000

1000

001

– 300

0000

3000

001

– 500

0000

5000

001

– 100

0000

0

1000

0001

– 23

6200

00

Anta

rctic

Circ

le

Trop

ic of

Cap

ricor

n

Equa

tor

Pacifi

c Oce

an

Trop

ic of

Can

cer

Indi

an O

cean

Pacifi

c Oce

an

Atla

ntic

Oce

an

–180

–160

–140

–120

–100

–80

–60

–40

–20

020

4060

8010

012

014

016

018

0

–160

–140

–120

–100

–80

–60

–40

–20

–60

–40

–20

20

40

60

80

0

–40

–2020

40

60

80

0

020

4060

8010

012

014

016

018

0

Arcti

c Cir

cle

FIg

ur

e 8.

1 (S

ee c

olor

ins

ert

follo

win

g p.

132

.) Po

pula

tion

dis

trib

utio

n of

the

wor

ld’s

urb

ana

gglo

mer

atio

nsw

ith

750,

000

peop

leo

r m

ore

arou

ndy

ear

2000

.

42963_C008.indd 141 11/6/07 7:13:05 AM

142 Land Use Change

Theusefulnessofanurbanlandusechangemodelisjudgedinconnectiontothegoalofthemodelingexercise;thesegoalscanbetightlyorlooselyconnectedwithgoalsofpolicymakers.InthischapterwediscusstwodistinctfunctionsforaULCM:explanationandprediction/forecasting(thatleadstoprescription).Initsfirstfunc-tion,ithelpstheresearcherorthepolicymakerimprovehisorherunderstandingofprocessesthatleadtochangeandshedlightonelementsofcausalityguidedbyandtestingalternativetheoriesofurbangrowth.Initssecondfunction,itcandescribeandpredictthetypesoflandusechangethatoccur(type,amount,rate,pattern,andtimingofchanges)andmorepromptlyleadtoprescription.

Therearenowdozensoflandusemodelsavailable;areviewandtypologyof(urban)landusechangemodelshasbeenpresentedindetailelsewhere.7,8,9,10Manynew“flavors”ofmodeling arebeingdeveloped.11,12This proliferation reflects themethodologicalprogress in theattempt tounderstandorpredict thenatureof thelandscape, the types of changes occurring, the causal structure connecting theunderlyingfactorsofchange,andthehypothesestobetested.Alternativeclassifica-tionsofurban landusechangemodels includea three-dimensionalcontinuumofspatial scale, time scale, andhumandecision-making,11 overlapping categoriesofequation-basedsystem,statistical technique,expert,evolutionary,cellular,hybrid,andagent-based,13anddistinctcategoriesof large-scale, rule-based, state-change,andcellularautomata.14

ULCMsoftenclaimpolicyrelevancebutlackacleardefinitionofthedegreeofthisrelevance.Landusechangemodelingiscurrentlyweaklycoupledwithlandusepolicymaking.Althoughwedonotclaimaneedforaverystrongcoupling(duetotheadverseresourceandpolitical realityforsucha task indevelopingcountries),wesuggestthatitneedstobestrengthenedforoptimalknowledgeutilizationinthepolicy-makingprocess.Thiscanbeachievedbyexplicitlyintroducingmechanismsformodel uncertainty reduction and a policy-makingmodule in landuse changemodels.Itisveryimportantthattherelevanceofmodelsismoreclearlyunderstoodandfuturedirectionsreevaluated.Inwhatfollows,weaddressissuesexistingatthemodeling–policy-makinginterface.

8.2.2 PoliCy Making

Policy-relevant landusechangemodelsmay target avarietyof types, levels, andstagesofpolicy-makingactivitythatheavilyinfluencesobservedlandusepatterns.Some facets of urban land use change derive in part from policies implemented(synchronously or asynchronously) at different administrative unit levels: at localmunicipal,county,state,prefecture,andregionallevels.Nationalmacroeconomic,regional, and local policies have dramatic direct and indirect effects on agents’choicesofcurrentandfuturelanduse.Policymakersattheselevelsincludearangeofpublicofficials,suchasurbanandenvironmentalplanners,andvariousadminis-tratorsatlocalgovernmentagencies.

Atthelocalandnationallevels,concernsregardingsocialwelfaremeasuredinlevelsofconsumption,productiveactivity, cityamenitiesanddisamenities, exter-nalitiesandideasofsustainabilityguidepolicy-makingeffortsintargeting—amongothergoals—an“optimal”urbanareasize,shape,andpopulationmix.Localurban

42963_C008.indd 142 11/6/07 7:13:06 AM


andexurbangovernmentsconsistentlyutilizezoning,growthcontrols,andtaxation/subsidiestodriveurbangrowthandregulate,distribute,orredistributegainsfromurban development (while implicitly targeting that the maximization of propertyvaluesinurbanareas).Increasingly,environmentalconcernsregardingtheimpactofurbanlanduseconversionsalsodirectpolicymaking.Atthegloballevel,institution-allydesignedpoliciesinfluenceprocessesofurbanlandusechangeinamultitudeofwaysthroughtheestablishmentofincentiveschemesandstructuraladjustmentpro-grams.Closemonitoringofurbanpopulationtrendssuggeststheheightenedinterestofinternationalfinancialandotherinstitutions(suchastheUnitedNationsandtheWorldBank)thatdriveglobalchange.

Manytheoriesofpolicyformationexist,withdifferentassumptionsregardingknowledgeutilizationwithintheformationprocess.Mosturbangrowthmodelsarenotusuallyexplicitontheirassumptionsregardingthepolicy-makingprocess;themostwidelyadoptedviewofthepolicy-makingprocessisthatoftherationallinearprocessoragendasettingtheory.Aswithmosttechnicalanalysisenteringthepolicyrealm,thepolicyrelevanceofalandusemodelisofamoreinformationalnatureratherthanaconcretepolicydrivernature.

Fromtherationalpolicymaker’sstandpoint,theuseofalandusechangemodelinvolvesasequenceofdecision-makingstepsandactionsthatrequires(i)theexami-nationofavailablemodelingoptions,(ii)thechoiceofmodelevaluationcriteriaandtheirweights(dependingonthepreferencesofthepolicymakerandtherealitiesofthepolicy-makingsetting), (iii)evaluating themodelby theselectedcriteria,and(iv)deriving the overall evaluation through the collection of individual weighedcriterionevaluations.5Criteriafortheselectionofamodelingprocessmayincludetheemphasisonpredictionversusexplanation, thedata sparsenessor richnessofthepolicy-makingenvironment,levelsofuncertaintyinthequalityofthedata,theemphasis on probabilistic versus heuristic/mechanical approaches, the flexibilityof the model to alternative variable specifications, the sophistication in accuracyassessment(validation)ofpredictions,theneedofdeepversusbasicunderstandingof themodelingapproach, theneed forweakversusstrongcouplingofmodelingwiththeprocessofpolicymaking,themodel’scapacitytoinformaboutavarietyofpolicy-makinggoalsandatdifferentlevelsofpolicymaking,andtheemphasisonthetheoreticalfoundationofthemodelingapproach.Severalofthesecriteriaarediscussedinmoredetailbelow.

8.2.3 The inTerseCTion of Modeling and PoliCy Making

Recently we have witnessed a scarcity of application of ULCMs for developingworldcities.ThisreflectsanunderestimationofthepotentialeffectsofurbangrowthinLDCs,theproblematicnatureofempiricalworkinLDCs,alackofunderstandingofwhatcouldbethebestmodelingoptionavailabletoadecisionmakerinadevel-opingworldcity,andadearthofapplicableULCMs.Throughourworkwearguethatpresentpressingpredictiveneedselevate theimportanceofstatisticalmodelsthatutilizeaminimalinputscheme.Modelswithsimpleinputrequirementscanfindwiderapplicationinaddressingcurrentandfutureneedsinthesecities.DatasetsinLDCsarescarceandinmanycasesinexactduetoinstitutionalfactorsandlimited

42963_C008.indd 143 11/6/07 7:13:07 AM

144 Land Use Change

resources. A future increased allocation of resources toward the collection ofdetailedgeoreferencedsocioeconomicdatabythegovernmentsofthesecountriesisnotcertain,andalthoughdataarebeingcollectedataninternationallevel, thisoccursataveryslowpaceandataquiteaggregatedlevel.Furthermore,problematicmeasurementcanbecatastrophicforthepredictivepowerofmodelsthataresuc-cessfulincapturingthetruedata-generatingprocess(DGP).

Understanding the importance of knowledge utilization in decision making,thequestionoftherelativeimportanceofexplanationversuspredictionforpolicy-makersarises.Whendoesapolicymakerneed(i)predictionsregardingthelocationandtimingoflandusechangeunderalternativescenariosand/or(ii)theknowledgeof whether theoretical hypotheses stand up to statistical tests and of magnitudesoftheexpectedchangesassociatedwithshiftsinavarietyofpolicyleveragesandviceversa?Itisnotclearifthepolicymakeralwaysneedsadeeperunderstandingofprocessesandknowledgeofthecausationchain.Possibly,theanswertosuchaquestiondependsontheactualpolicy-makingformationprocessandthetype/levelofgovernmentor institution responsible for thedecision.Variousauthors suggestthat,ataminimum,policymakersshouldbeabletounderstandthefoundationsofamodelingapproachoratleastbeabletoidentifyhowtheresultsaregenerated.10,15Giventhenumberandlevelofcomplexitiesofalternativemodelingapproaches,thismaybeanunrealistictarget.Ourexperiencewithdevelopingworldcitiesshowsthatpolicymakersaredefinitelymoreinterestedinknowinghowshiftsinpoliciesaffectoutcomes;theymaynotwanttoknowtheinnerworkingsofthemodel.

Policymakerpreferencesoveroutputdefines if andwhen thepolicymakerhasa stake in the choice of methodology (e.g., process-based or mechanistic models).Althoughsocioeconomicprocessesgeneratetheobservedlandscapeoutcomes,modelsthatbelongtoarule-basedapproachmayinfactresultinbetterpredictionsthanpro-cess-basedmodelsutilizingsocioeconomicdata.Thiscanbepartlyattributedtodataimperfection:variablescapturingthesocioeconomicprocessescanbeinaccurateorsimplytheseprocessesmightbehardorimpossibletoquantify.Mechanisticmodelsusedataconstructs thatarebasedonproximate(rather thanunderlying)causesoflandusechange.Unfortunately,thesemodelsarealsomoresensitivetoomittedorinexistentinformation,afactthatcanpotentiallymisguidepolicymaking.Process-basedmodelscanstillbesuccessful tovariousdegrees for forecasting,dependingonthegeographicallocation,methods,andaerialunitlevelofanalysisemployedforprediction.Suchmodelswithprovenhighpredictivepowerarealsousuallybasedonproximateratherthanunderlyingcauses.Naturally,successesinpredictivecapabilityofrule-basedmodelsdonotvoidthesearchforaDGP.

Models—asopposed to theories—of landusechange theoryhavebeenmorepopulartoolsforpolicymakingandhavebeendevelopedmoreforbothsubstantiveandpracticalreasons.10Substantivereasonsincludethecomplexityofthelandusechangephenomenaandthecomplexinterrelationsofvariousinstitutional,cultural,political,economic,andsocialchangedeterminants in theoreticalwork.Practicalreasonsincludetheavailabilityofresourcesandthe“demandsoftheofthedecisionmaking‘clientele’.”Inshort,solidquantitativeresultsthataremarketable,visually

42963_C008.indd 144 11/6/07 7:13:08 AM


powerfulandreadyforuseastoolsforawiderangeofdecisionmakersarevaluedmorehighly—andmodelsproduceresultsmuchbetterthantheory.�

Established approaches in different scientific disciplines and pressures regard-ingpeeracceptanceandcareeradvancementalsodrivemethodology-relatedchoicesforurbanlandusechangemodelsandarepartlyresponsibleforlooseconnectionsofmodelswithpolicymaking.Theevidenceforthisisanecdotal,astheauthorshavebeenexposedtosuchcomplaintsinpersonaldiscussionswithotherresearchers.Inshort,theproducer’s(anacademicresearcher)incentivesforconsiderableoutputintheformofjournalpublicationsmayleadtomodelslooselyorvaguelyconnectedtothepracticeofpolicymaking.Thisissueisadmittedlydifficulttoresolveunderthecurrentpractices.

Awarenessofthetheoreticalfoundationofanapproachmayalsobeimportantforthepolicymaker’schoice.Urbancellularautomata(CA)models,forexample,haveatheoreticalgroundingonideasofcitiesasself-organizedandemergentphenomenain bottom-up complex systems and fail to capture urban growth in the top-downpoliticaldimension.Unfortunately,thesearestill“largelyabstractarguments.”16Evencutting-edgeadvancedmultiagentsystemCA(MAS/CA)simulatingcities“atthefinescaleusingcells,agents,andnetworks”arefornowfarfrombeingreadyforanyprac-ticaluseor“largely…pedagogic”value.16

Asthedecision-makingclienteleofurbanlandusechangemodelstargetsavari-etyofgoals,agoodmodelshouldaccommodatesuchavariety.Policymakerscarefordifferentsizeadministrativeareasdependingonwhethertheyareemployedatalocal,provincial,ornationallevel.Modelsshouldbeabletoaddressneedsofeachlevelofdecisionmakinganddistillresultsderivedatthehighestleveltothelowestlevel and vice versa. Connected to this issue is the capacity to address single ormultipleneighboringurbanareasinthesamemodel.Singleurbanmetropolitanareaanalysisisnotinclusiveofsurroundingregionalspatialdynamics(immigrationandout-migrationflows,tradeflows,andsoforth),animportantlimitationsincecitiesareinterconnectednodeswithinanetworkofflows,aswellascomponentsofasystemofcentral(urban)places.

Whenmodelsareweaklyinformedbytheory,anadvantageofaULUCmodelisitsflexibilityinallowingtheusertomakedecisionsonmodelspecification(althoughthedangerofmodelselectioniseverpresent—thisisaddressedinthenextsection).Currentdesignofmechanicalrule-basedmodelsshowssomeinflexibilitytoalterna-tivespecifications,witharesulting“onesizefitsall”/“cookie-cutter”feel.Finally,animportantconsiderationisthelimitationsinspatialrepresentationsofalternativescenariosimposedbytheULCM,assumingthatquantifiableinformationonpoten-tial alternative directions in local, regional, or national policies can be provided.Policiessuchaszoningorgrowthcontrolsaretheeasiesttorepresent,whileopen-nesstoin-migrationorothereconomicinformationsuchasmarketconditionsmaynot be easily incorporated into models. Highly stylized (input-restricted) modelsareonlyabletoincorporatepoliciesreflectingroaddevelopmentand“off-limitstodevelopment”zoning;thisisalimitingfactorinthecapacityofthemodelstoincludeotherformsofpolicymaking.

�Pre-1981literatureonthepoliticsofmodeluseindecisionandpolicymakingisreviewedinBriassoulis10(1999,chap.5).

42963_C008.indd 145 11/6/07 7:13:09 AM

146 Land Use Change

8.2.4 PoliCy evalUaTion and UnCerTainTies

Model and expert knowledge utilization targets the reduction in the uncertaintyofoutcomepredictions and the consequent effectsof theseoutcomes.Statisticaldecision theory provides quantitative tools for the reduction of uncertainty inoptimalpolicymaking.17Giventhatanyurbanpolicysimulationresultsaredepen-dent on models, how is the “best” model defined and how is it chosen amongall possible models? Amodel is a single representation of reality, and, althoughstatisticalcriteriacanbeusedtoidentifythe“best”one,itrepresentsjustoneofmanypossibledatageneratingprocesses;thus,modelselectionshouldbeavoided.Modelselectionhasbeencriticizedasbeingaweakbasisforpolicyevaluationandderivationoffutureprescriptions;thesearchforasinglebestpredictivemodelismisguidedwhenitcomestopolicy-relevantmodels.Robustnessacrossmodels,ontheotherhand,isbeingadvancedforpolicy-relevantwork.Unfortunately,modelselectionignoresthefundamentaldimensionof“modeluncertainty,”butmethod-ologiesforrobustnessofthepolicyprescriptionacrossalternativemodelspecifica-tionscanbealternativelyutilized.

Methodologiesaddressingtheissuesoftheoryandmodeluncertaintyarenowavailableforincorporationinpolicy-relevantresearch.17,18Uncertaintyovercompet-ing theories results fromuncertaintyaboutwhich theoryofurbangrowth shouldbeutilizedduetoinstitutionalandculturalfactorsaffectinglandmarketsindevel-opingcountriesordifferingassumptions regardingagentdecisionmaking; it canleadtomodelsthatarenotwellinformedbytheory.Modeluncertaintyresultsfromuncertaintyover functional formspecification for statisticalmodelsand isdue tosubjective perceived relevance and endogeneity issues as well as the question ofappropriatespatialandtimelags,proxyvariables,andsoforth.Incompleteknowl-edgeregardingthebestmodelofasystemshouldforcetheresearchertoexplorethesensitivityofmodelingapproaches toalternative specifications.18This frameworkpartiallysolvestheproblemsofsubjectivityandadhocspecificationsinuncertainenvironmentsregardingthecapacityofavarietyofproxiestocapturetheeffectofvariablesenteringthedatageneratingprocess.

An applied statistical framework of policy-relevant urban growth modelingthataccounts formodeluncertaintymakesanexplicit reference toapolicymaker(PMhereafter)whoexaminesasetofurban–growth-relatedpoliciesPforadmin-istrativeunits(e.g.,sub-citydistricts,cities,counties,provinces)throughtheselec-tion of a single policy p.17,18 The PM utilizes data d about a metropolitan area’sland-use and transportation systems (realizations of a process), and the choice isconditionalonamodelmoftheurbaneconomy(mcanconstitutealternativetheoriesandstatisticalspecifications).ThePMminimizestheexpectedvalueofanobjective(loss)functionl(p,θ)whereθisatheexogenousstateofnature(notcontrolledbythePMbutaffectingtheinfluenceofpontheloss,l)�.

�ConsiderX,avectoroftargetedurbangrowthratesperadministrativeunitpertimeperioddefinedbythePM.Differentpolicieswilldrivedifferentratesandpatternsofurbangrowth.Thedeviationofthesegrowthratesfromthetargetedgrowthratescanbeexpressedasalossfunction.Eachadminis-trativeunitisweightedaccordingtothepreferencesofthePM.Theweightsareassignedaccordingtotheimportanceofconvergencetothetargetforeachadministrativeunit:thegreatertheimportanceofmeetingthetarget,thehighertheweight18.

42963_C008.indd 146 11/6/07 7:13:10 AM


Unknownexogenousfactorsthatinfluencelandusedecisions(thestateofnaturein a decision-theoretic framework) such as monetary or fiscal macroeconomicpoliciesleadtotheminimizationoftheexpectedvalueofthelossfunctionbythePM.Usuallyprobabilitiesofthestatesofnatureareconditionedonexistingdataandselectedmodels:μ(θ|d,m).Accounting formodeluncertainty, theyarenot condi-tionedonmsincethePMunderstandsthatthereisprobablyno“best”modelfortheurbanlandusesystem.Theprobabilitydensityfunction(pdf)forθ,μ(θ|d),isassumedconditionalontheexistingdatadonly(andnotonm).Withawell-definedlossfunc-tionandpdf, theoptimalpolicyis theonethatminimizes theexpectedloss—thesolutiontotheoptimizationproblem.WethusarguethataULCMshouldgenerateaprobabilitydistributionofestimatesandpredictionsaswellastheirdistributioncharacteristics/properties for eachpixel for each timeof change.AmathematicalformulationofthemodelcanbefoundinFragkiasandSeto18andBrocketal.17Thefollowingsectiondescribestheworkingsofanurbangrowthmodelthattakesintoaccounttheaboveconsiderations.

8.3 a new approach in Modeling urban growth in data sparse environments

8.3.1 MeChaniCs of The Model

Wepresenthereanapproachinlandusechangemodelingthatcanexplicitlyevaluatepoliciesunderuncertaintieswithinaspatialsocioeconomicenvironmentbyincorpo-ratingamethodologythataddressesissuesoftheoryandspecificationuncertainty.ThemodelproposedbyFragkiasandSeto18 isahybridspatiallyexplicitmodelofurbanlandusechangewithafoundationoneconomicandstatisticaldiscretechoicemodelsoflandusechange,8adjustedforuseindata-sparseenvironments.

Themodelfirstreadstheinputdataavailableprovidedthroughremotesensinganalysisandothersourcesatadefinedspatialresolution.Initscurrentimplementationtheseare:urban/non-urbanland-usemaps,atransportationnetwork,areasexcludedfromdevelopment,andacentralbusinessdistrictlocation.Itprocessesthisdatapro-ducing new images/matrices such as new urban growth between examined years(abinaryimagefromwhichweextractinformationonadependentvariable)andacollectionofnewimagesfromwhichthemodelextractsindependentvariables.�

Next, themodelemploysstatisticalanalysisfor thecalibrationstage.Separat-ing the study area into two parts (its East and West half), it performs a randomsamplingofdevelopablepixelsoftheinitialEasthalfoftheurban/non-urbanimage(attimet0).Itcreatesacalibrationdatasetwithasingledependentvariableandmulti-ple setsof independent variables ready for regression analysis.Using two (t0, t1)calibrationimages,themodelrunsmultiplelogisticregressionswithbinarydepen-dentvariableyas‘changetourbanornochange’betweentimeperiodst0andt1.The

�Inthismodelwefocusonpredictionofchangesanddonotutilizedataonsocioeconomicprocessesthatresultinlandusepatterns.Inanefforttodevelopamodelwithminimaldatarequirementsweusespatialdensityanddistancevariablestocheckthepredictiveaccuracyofamodel.Themodelalsoutilizesdistrictdummyvariables,eachrepresentingone(oracollection)ofthedistrictsofeachurbanareainthestudy.FragkiasandSeto18presentsamoredetaileddescriptionofthemodel.

42963_C008.indd 147 11/6/07 7:13:10 AM

148 Land Use Change

modelingapproachpresentedinthiscasestudysystematicallyincorporatesavarietyofmodels(orspecifications)andaccountsformodeluncertaintyinlandusechangerelatedpolicymaking�.Themultiple specificationmodel runs reflect theneedsoftheemployedmodelaveraging technique.Fornexplanatoryvariables thatcanbeselectedforthemodels,atotalof2nsetsofalternativespecificationexistandareutilizedintheanalysisasalternativeregressorsets.

Pseudo-Bayesian model averaging is then performed using the calibrationsample.Each2nlogitmodelrungeneratespredictedprobabilitiesofchange(fittedvaluesofthedependentvariable)foreachsamplepoint,andaweightedaverageofthepredictedprobabilitiesiscalculated;the2nsetsoffittedvaluesareweightedbytheirrespective(normalized)pseudo-R2statistic�.Aseriesofbinarysamplesetsofpredictedurban/non-urban land are then createdutilizing an arrayofprobabilitycut-offpoints (thresholdvalues) that range from0 to1.Themodel compares theseriesofpredictedurban/non-urbanvalueswith theactualrealizationof landuseduringthetimeperiodunderstudyandselectsthe“optimal”thresholdlevelforthecalibrationperiod(thecut-offpointthatgeneratestheminimumdifferencebetweenpredictedurbanlandandactualurbanland).�

Modelvalidationoccursattwospatialscales:theindividualpixel(throughPCPvalidation)andachosenadministrativeunitlevel(byaggregatingpixellevelinfor-mationandvalidationthroughsampleenumeration).Thevalidationsampleisderivedthroughasecondrandomspatialsamplewithinthesecond(West)halfofthestudyarea.All2nsetsofindependentvariablevaluesareextractedforthenewvalidationsample.Togetherwiththeestimatedsetsofvariablecoefficientsfromthecalibrationstagetheyareappliedtothefittedprobabilitylogitformulaforeachmodel.Thisgen-eratespredictedprobabilitiesofchangeforthesampleddevelopablepixelsfortimeperiodt1utilizingtheaverageofthefitted/predictedprobabilitiesofallthemodels(weightedbythenormalizedpseudo-R2scorethateachmodelachieves).

Thefirsttypeofvalidationoccursthroughthegoodness-of-fitmeasureof“percentcorrectlypredicted”(PCP)�.Typically,inPCPvalidation,choice(orprediction)isdefinedasthealternativewiththehighestpredictedprobability.TheseclassificationsarethencomparedwithactualchangesandthePCPmeasureiscalculated.Apartfromtheintuitivebinarycut-offvalueof0.5,anyprobabilitythresholdvaluecanbesetforthegenerationofbinarypredictedchangevalues.Weautomatetheselectionofthisthresholdinthecalibrationstageutilizingthecriterionof“bestgrowthrate

�Probabilisticmodelsaresensitivetotheproblemsofpredictivebiasandlackofcalibration:predictivebiasisaproblemofbalanceor“the systematic tendency to predict on the low side or the high side”(p.391)19,andaveragingmodelswithalternativespecificationsincreasesthechancesofaveragingouttheproblem;lackofcalibration,is“a systematic tendency to over- or understate predictive accuracy”(p.391)19andisalsoanegativefactorforvalidationthroughthresholdingduetoanincreasedsensitivitytoit.

�Thestandarddeviationsofthepredictedprobabilityofchangeestimatesarealsocalculated(butonlyatthestageofthefullimageapplication).

�Thisisaformofanexternalimpositionofanurbangrowthratescenarioonthemodel.Athresholdcanalsobeselectedinsuchawaythatanalternativeurbangrowthscenarioisportrayed.

�PCPvalidationoccursintheformofseparationbyspacewithinanout-of-samplemodelingframework.20

42963_C008.indd 148 11/6/07 7:13:11 AM


matching.”�Thisprobabilitythresholdvalueisappliedinthenewvalidationsample,theclassificationisperformed,andthePCPmeasureiscalculated.

Thesecondtypeofvalidationoccursatalargerscalethantheindividualpixelthroughsampleenumeration.Thetechniqueofsampleenumerationsumspredictedprobabilities over a set of agents or observationswith thegoal of generatingcon-sistentestimationsofaggregateoutcomes.21Themodelutilizesaspatiallyexplicitversionofthisaggregationmethodforthepurposesofvalidationandforecasting,summingupprobabilitiestothedistrictlevel.Itemployssampleenumerationforthevalidationdatasetandcomparestheaggregateestimateofurbanchangetotheactualaggregatechangeinthesample.Predictiveaccuracyisdefinedbythesuccessofthesummedaveragedpredictedprobabilitiesinaccuratelycapturingaggregatechangeattheselectedadministrativeunit�.

Themodelpredictsland-useintwoways.First,wethresholdthepredictedprob-abilities.Predictionresultsarepresentedinitiallyforthepopulationofdevelopablepixelsint2andpotentiallyforanydiscretenumberoffutureiterations(t3,t4,andsoforth).Eachiterationaccountsforthepassingofasingle—equivalenttothelogitmodels—timeperiod.Thepredictionsutilizetheestimatedsetsofregressioncoeffi-cientsand(potentiallyforeachiteration)apartiallynewsetofindependentvariables(reflectinglandscapeevolution).Theoptimalcalibrationthresholdisappliedagainforthe“translation”fromprobabilitytopredictedchange.Second,predictionoccursthroughthesampleenumerationtechniqueforforecastingataggregateadministra-tiveunitlevelsutilizingahypotheticalscenariodataset.

Amultiplescenarioexaminationisusuallyanintegralpartofapolicydecision-makingprocess.Giventhenatureofthecurrentlyincorporatedvariables,asimula-tionmodulecanbeusedfortheexaminationofpolicy-relevantalternativescenariosregarding simple policy leverages: first, the researcher can define new areas ofundevelopedlandthatareexcludedfromdevelopment;second,alteredtransporta-tionroutescanbedesignedaccordingtoexistingplansofroadorrailwayexpansion;third,theusercancreatepatchesofnewdevelopedlandofhighintentionality(e.g.,anewairport)capturingspill-overeffectsofsuchdevelopmentsthatwouldotherwisebedifficulttopredict.Theuser/decisionmakercanfeedthemodelacollectionofscenarioimagesanddefinealossfunctionthatconnectspredictionsofurbangrowthwith the objective the decision maker is trying to achieve (e.g., minimization ofagriculturallandloss).Themodelcanprovidepredictionsofchangeandassociated

�Thisautomatedcalibrationprocessthoughhasadisadvantage.Sinceitisbasedonanadhoccriterionselectedbytheresearcher,theselectionofaprobabilitythresholdover0.5issubjective.

�TheobviousvalueofPCPvalidationistheeaseofvalidationatthepixellevelandanyaggregatedlevelofgroupsofpixels(forexample,withinadministrativeboundaries).Unfortunately,theapplica-tionofprobabilitythresholdsinstatisticalmodelsiscountertothenotionofapredictedprobabilityinsuchmodels.Limited informationdue to theunobservablecomponent in the formulationof thechoiceprocessforbidsthepredictionofthechoiceofanalternativeforasingleunitofobservation.Thenatureofthepredictedprobabilitiesindiscretechoicemodelshasastandardstatistical“largesamplerepetition”interpretation:thealternativewiththehighestprobabilityisnottheunit’schoiceeachtime.Thisinterpretationmakesamodeldesignthatattemptsacrossfromprobabilityspacetochoicespaceamorecomplexanddifficulttask.21Thus,PCPmeasuresmaynotprovidethebestwaytovalidateastatisticalmodel.

42963_C008.indd 149 11/6/07 7:13:12 AM

150 Land Use Change

lossesforeachscenario,thusgivingthedecisionmakerthechoiceofthepolicythatminimizesthelossaccountingformodeluncertainty.

8.3.2 aPPliCaTion To Three CiTies of The Pearl river delTa, China

While urban land use change is a worldwide phenomenon, it is most dynamic inAsia—andparticularlyinChina—whereunprecedentedratesofurbangrowthhaveoccurredoverthelasttwoandahalfdecades.1Theinteractionofcompellinglocal,regional,andglobalfactorshasresultedinremarkablyvariedurbanconfigurations.22OurstudyareaiscomprisedofthreeofthemostdevelopedcitiesinthePearlRiverdelta(PRD)regioninthecoastalsoutheastChina,Shenzhen,Guangzhou,andFoshan(Figures8.2and8.3);theseandothercitiesinthePRDhaveexperienceddramaticurbanlandgrowthratesinthepasttwodecades.Duringan11-yearperiod—from1988to1999—urbanlandgrew�inthePRDby451.6%orapproximately16.5%ayear23.TheDeltageneratesmorethan70%oftheprovincialGDPandishometo21millionpeople, nearly one-third of the province’s official population.23 Previous researchshowsthattheregionisundergoingrapidurbantransformation.24Althoughthecitiesareincloseproximity,theydifferinhistory,demographics,andeconomics.18,25,26

Themodel is runfor the threecitiesat twopixel resolutions,30mand60m.Usinga randomcalibrationsample fromthedevelopablepixelsof the initial1988urban/nonurbanimage�themodelexploresallpotentialcombinationsoflogitmodelscreatedbyallowingfivevariablestoenteralternativespecifications,whichresultsin32models.�Afterderivingallthesetsofestimatedcoefficientsandcalculatingthemodel-averagedpredictedprobabilities,thecalibrationstageidentifiesthethresholdsthatbestfittheobservedurbangrowthratefortheperiod1988to1996forallcities.

Validation occurs with the use of the second random sample and the model-averaged predicted probabilities for all cities and specified resolutions. In a caseoftwodiscretestatesoflanduse(urbanandnonurban),onefindstwocaseswherepredictions are accurate (correctly predicted urban and nonurban) and two caseswherepredictionsarewrong(wronglypredictedurbanandnonurban).InthecaseofPCPvalidationthetotalpercentageofwronglypredictedpixelsrangesbetween23%27%.Themodelconsistentlygeneratesadisproportionatelylargerpercentageofwronglypredictednonurbanpixelsrelativetothepercentageofwronglypredictedurbanpixels,amanifestationofthepredictivebiasproblem:weobserveasystematictendencyofsmallerprobabilityvalues.Themodelalsoreportscomparisonsbetweentheaggregate—atthedistrictlevel—actualchange,theaggregatepredictedchangethroughsampleenumeration,and theaggregatepredictedchange through thresh-olding for all cities and resolutions.The sample enumeration techniqueperformsconsistentlybetter in thevalidationof theaggregatecountsofchange.Acrossallcities, we observe that the vector distances between the aggregate actual changewithaggregatesampleenumerationpredictedchangeandtheaggregatethresholding

�In particular, the metropolitan areas of cities such as Shenzhen, Guangzhou, and Foshan grew by132.3%(8%annually),247.6%(12%annually),and140%(8.3%annually),respectively

�Forthe30mresolution,thisnumberiscloseto14,000pixelsintheShenzhenstudyarea,3,300pixelsintheFoshanstudyarea,and10,200pixelsintheGuangzhoustudyarea(Figure8.3).

�Excludingdistrictdummyvariableswouldprovideanadditional32models.

42963_C008.indd 150 11/6/07 7:13:13 AM


predictedchangedropsignificantlyasweloweredtheresolutionfrom30mto60m.The performance of aggregate prediction through thresholding improves relativetoaggregatepredictionthroughsampleenumerationbutisstillnotsatisfactory.Inshort,thereexistsaclearfailureofspatiallyaccuratepredictionwhenusingthresholdprobabilitiesfortransitiontochoices.Thepixellevelpredictionscanbesuccessfully

Kilometers6030150

KilometersFoshan study areaGuangzhou study areaShenzhen study areaGraticule (5 degrees)Province BoundariesGuangdong Province

25012562.50

Urban land use Urban land 1988 Growth 1988–1996

20N

25N11

0E

115E

SOUTH CHINA SEA

GUANGXI PROVINCEGUANGDONG PROVINCE

HUNAN PROVINCE JIANGXI PROVINCEFUJIAN PROVINCE

PEARL RIVER DELTA

N

FIgure 8.2 (See color insert following p. 132.) The Pearl River Delta in southeastChinaandtheShenzhen,Foshan,andGuangzhoustudyareas(urbanlandin1988,newurbanlandbetween1988and1996).

42963_C008.indd 151 11/6/07 7:13:15 AM

152 Land Use Change

usedthroughaggregationtoalargeradministrativeunitsuchastheneighborhoodorthetownshipthroughtheprocessofsampleenumeration.

Atthepredictionstage,themodelaveragingprocessthatgeneratesthepredictedprobabilityvaluesisrepeatedforthefullpopulationofobservations;thet1period(1996)dataprovide thevaluesof independentvariables,anda full imageofpre-dictedprobabilitiesofdevelopmentisgenerated.Eachpixelisassignedtheaveragedpredicted value from the logit formula using the sets of estimated coefficients.Figures8.4,8.5,and8.6mapthesepredictedprobabilitiesofdevelopmentbetween2004and20012andtheassociatedstandarddeviationsofthepredictions.

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

100

200

300

400

500

600

100

200

300

400

500

600

0.12

0.1

0.08

0.06

0.04

0.02

0

100 200 300 400 500 600 700 800 900

100 200 300 400 500 600 700 800 900

FIgure 8.3 (See color insert following p. 132.) Predicted probability of change tourbanareasbetween2004and2012andstandarddeviationofpixelpredictedprobabilitiesforShenzhen(60mresolution;valuesinpercentagepoints).

42963_C008.indd 152 11/6/07 7:13:16 AM


0.35

0.3

0.25

0.2

0.15

0.1

0.05

0.16

0.14

0.12

0.1

0.08

0.06

0.04

0.02

0

50

100

150

200

250

300

350

400

450

500

550

50

100

150

200

250

300

350

400

450

500

550

100 200 300 400 500 600

100 200 300 400 500 600

FIgure 8.4 (See color insert following p. 132.) Predicted probability of change tourbanareasbetween2004and2012andstandarddeviationofpixelpredictedprobabilitiesforFoshan(30mresolution;valuesinpercentagepoints).

42963_C008.indd 153 11/6/07 7:13:18 AM

154 Land Use Change

For predictions based on thresholding, the model generates images of urban/nonurbanlanduseforanyfuturetimeperiod;Figure8.7plotstheimagefor2012.Amountsofurban landusechange foreachdistrictbetweenanysetofyearsarepredictedthroughthetechniqueofsampleenumeration.

8.4 discussion and Conclusions

Accountingformodeluncertaintyinlandusechangerelatedpolicymakingthroughtheemployedmethodologyreducesuncertaintiesforadecisionmakerthatareinher-entintheuseofmodelsinthedecisionmakingprocess.Theproposedmethodology

0.6

0.5

0.4

0.3

0.2

0.1

0

0.12

0.1

0.08

0.06

0.04

0.02

0

50

100

150

200

250

300

350

400

50

100

150

200

250

300

350

400

100 200 300 400 500 600 700

100 200 300 400 500 600 700

FIgure 8.5 (See color insert following p. 132.) Predicted probability of change tourbanareasbetween2004and2012andstandarddeviationofpixelpredictedprobabilitiesforGuangzhou(60mresolution;valuesinpercentagepoints).

42963_C008.indd 154 11/6/07 7:13:19 AM


9008007006005004003002001000

7006005004003002001000

6005004003002001000

0

100

200

300

400

500

600

0

100

200

300

400

500

0

50

100

150

200

250

300

350

400

FIgure 8.6 Shenzhen,Guangzhou,andFoshanurban/nonurbanpredictionfor2012.

42963_C008.indd 155 11/6/07 7:13:21 AM

156 Land Use Change

FIg

ur

e 8.

7 C

ity

ofS

henz

hen

–M

arch

200

2.

42963_C008.indd 156 11/6/07 7:13:23 AM


potentiallyreducestheproblemsofpredictivebiasandlackofcalibration.Throughtheabovemethodology,aPMisabletocalculateavarietyofcharacteristicsofprob-abilitydistributionsofoutcomes (suchas themeanandvariance) thatmayaffectpolicy-makingchoices.Furthermore,sincethemethodologyenhancesthecapacityfor convergence of the true value of underlying parameters and the estimates ofthoseparameters,aresearchermayworrylessaboutpossiblestatisticalmodelbiasesandcanfeelmoresecure in thedecreasedcapacityofmanipulationof themodelby policy stakeholders toward particular answers. Unfortunately, though, movingfromthissimplesinglepolicyleveragemodeltomultiplesimultaneousleveragelevelchoicesbyPMsincreasessignificantlythecomplexityofthemodel.Furthermore,thisrationaldecision-makingframeworkisalsofarfromaperfectdepictionoftheactualpolicy-makingprocess;itignoresissuessuchastheconflictsoverideology,powerdifferentials,problemsincommunicationandcollaboration,andtheinfluenceofbureaucraciesandspecialinterestgroups.

Theconceptofuncertaintyreductionthroughmechanism/modeldesign—andtherelevanceofthismethodology—issupportedbymodernpoliticaleconomy.Sincearealisticviewofgovernmentsupportsthatpoliticians,bureaucrats,andpolicymakerswithin national and local governments are a mix of actors with public-good andprivate-gain motivation, policies should be designed with the assumption that allpolicymakersoperateunderthehomoeconomicusmodel(thatis,personalgainmoti-vationdrivesactionandisahugefactorinanydecision-makingprocess27).Giventhatmodelsandmodelselectioncanbeemployedforthesupportofprivateinterests,amodelingapproachthatprovidesadefenseagainstmodelselectionisdesirable.

Themodelinputaccuracyissue(“junkin—junkout”)naturallyaffectsland-usechangemodelstoo.Socioeconomicdatacanbenonexistent,incomplete,inaccurate,unreliable,oralltheaboveinadevelopingworldsetting.Evendataaccuraciesofafundamentalvariablesuchasurbanizationanditsforecastsareseverelycriticized.28Resultsofmodelingapproachesthataresensitivetoomittedornonexistentinforma-tionarepotentiallymisguidingforpolicymaking.Wefindthatenrichingdatasetswithremotelysensedinformation,incombinationwithmodelsofastatisticalnature,isapositivestepinanymodelingexerciseregardingdevelopingworldcities.Onthe one hand, given data inaccuracies in socioeconomic data, statistical modelsprove tobe lesssensitive to inputdata imperfections,since theyassignerrors inmeasurementinthestochasticpartof themodelandproduceunbiasedestimateseven when errors plague the data (if those errors are not systematic—randomlydistributed).Ontheotherhand,theuncertaintyovertheaccuracyofinputscanbecontrolled/measuredmoreeasilyfordataderivedfromsatelliteimageryratherthatsocioeconomiccensussources(assumingtheminimizationofuncertaintyandnobiasinsatelliteimageryclassification).

The application of probability thresholds in statistical models is problematicdue to the notion of a predicted probability in probabilistic models; pixel-levelvalidationtechniquesmaybemisguidingwhenusedinstatisticalmodels.Thestan-dardstatistical“largesamplerepetition”interpretationofprobability(whichappliestoourstatisticalmodel)conflictswiththeideaofPCPmeasuresofgoodness-of-fitdue to thisfundamentalcharacteristicofprobabilisticdiscretechoicemodels.An

42963_C008.indd 157 11/6/07 7:13:23 AM

158 Land Use Change

improvementtostandardthresholdingtechniques(whichalsoautomaticallydiscardinformationprovidedbytheneighboringpixels’predictedprobabilities)shouldbeofferedby the research community and is a researchgoal for the authorsof thischapter.Wesuggestthatuntilamoreadvancedvalidationmechanismisproposed,statisticalmodelsshouldbeeffectively limited tomainly thenonspatiallyexplicitsampleenumerationvalidationandprediction.Theseestimatescouldpotentiallybecoupledwithamoreadequatemoduleoperatingasanpixelallocationmechanismofpredictedgrowth.

Theaforementionedproblemofpredictivebiasmanifests itself in themodelthrough an imbalance in the percentages of wrongly predicted nonurban/urbanpixels. The model presented here partially corrects for this limitation throughpseudo-Bayesianmodelaveraging.Still,wefindthatasignificantlylimitingchar-acteristicofpolicy-relevant landusechangemodelsis that theydonotallowtheuser to define which type of error is more important.� Future versions of urbangrowthmodelsshouldincorporateamechanismthatwouldallowaPMtocontroltheamountofdifferenttypesoferrors(optimally,bypolicyregion),whentherela-tivecostsofwrongpredictionsbymodelsaresignificantlydifferent,andthusattainhigherpolicyrelevance.

Weidentifyseveralkeypointsfromtheaboveanalysis.First,theuserofproba-bilisticchoicemodelsshouldprimarilyfocushisorherattentiononthepredictedprobabilitymapsso thatprobablehotspots fordevelopmentare identifiedratherthan generating predicted maps of choices through simple thresholding mecha-nisms(changeornotchangeinabinarysetting).Second,theadvantageofutiliza-tionofavarietyofmodelsismanifestedinthefactthatmapsofstandarddeviationsfor the predicted probability maps convey important information regarding thespatiallyexplicitagreementofthevariousmodels.Third,inlightofscarceinfor-mationanddifficultyofdatasetenrichment,predictedprobabilitymapsshouldbecoupledwithadditionalinformationaboutdevelopmenttrendsthatmaybederivedfromsurveyswithimportantstakeholdersinlandusechangedecisionmakingandlocalknowledge.

Generally,similarlytoothermodelsinitsclass,thesuccessofthemodeliscon-strainedbytheavailabilityofquantifiableproxies.Similartoothermodelsofthistype,scenariobuildingrequirestheuseofvariablesforwhichfuturevaluescanbeprovided.Futureimplementationsofthemodelwillconsider(i)validationthroughseparationbytime;(ii)themoreimmediateincorporationofcalibrationschemestodifferentlocalities/districtsthatcalibratethemodelandcaptureinmoredetailtheurbangrowthrates(thequantityofchange)atthedistrictlevel;(iii)thecouplingofthismodelwithothermodelsoperatingatotheraggregationlevels,sheddingmorelightintotherealisticprospectsofdevelopmentofanymodel-identifiedhotspot;and(iv)thecollectionandanalysisofscenariosandpolicymakerpreferencesincollabo-rationwithChinesecityplanningdepartments.

�Instatistics,forexample,whentestingahypothesisaboutapopulation,theresearchercancontroltheprobabilityoftwodistincttypesoferrors,typeIand.typeIIerrors,basedontheseverityofapossiblemistake.Thedilemmaofwhethertoadoptahighorlowsignificancelevelisstrongerwhenthestakesarehigher.

42963_C008.indd 158 11/6/07 7:13:24 AM


reFerenCes

1. UnitedNations.WorldUrbanizationProspects:The2003Revision.UNPress,NewYork,2004.

2. UNCHS.TheStateoftheWorld’sCitiesReport2001.UnitedNationsCentreforHumanSettlements(Habitat),NewYork,2002.

3. Lambin,E.F.etal.Thecausesofland-useandland-coverchange:Movingbeyondthemyths.GlobalEnvironmentalChange11,261,2001.

4. National Research Council. Human Dimensions of Global Environmental Change:Global Environmental Change: Research Pathways for the Next Decade. NationalAcademyPress,Washington,D.C.,293pp,1999.

5. Boulanger, P.-M. and Bréchet, T. Models for policy-making in sustainable devel-opment:Thestateoftheartandperspectivesforresearch.EcologicalEconomics.55,337,2005.

6. von Thünen, J. H. Der isolierte staat in beziehung auf landwirtschaft und national-ökonomie. Gustav Fisher, Stuttgart; translation by C. M. Wartenburg (1966), TheIsolatedState.OxfordUniversityPress,Oxford,U.K.,1826.

7. Brown,D.G.,etal.Modelinglanduseandlandcoverchange.In:Gutman,G.etal.,eds.,LandChangeScience:Observing,Monitoring,andUnderstandingTrajectoriesofChangeontheEarth’sSurface.KluwerAcademicPublishers,Dordrecht,Netherlands,395,2004.

8. Irwin,E.G.,andGeoghegan,J.Theory,data,methods:Developingspatiallyexpliciteconomicmodelsoflandusechange.AgricultureEcosystemsandEnvironment85,7,2001.

9. EPA.ProjectingLand-UseChange:ASummaryofModelsforAssessingtheEffectsofCommunityGrowthandChangeonLand-UsePatterns.U.S.EnvironmentalProtectionAgency,OfficeofResearchandDevelopment,Cincinnati,Ohio,2000.

10. Briassoulis, H. Analysis of land use change: Theoretical and modeling approaches.In:LoveridgeS.,ed.,The Web Book of Regional Science.RegionalResearchInstitute,WestVirginiaUniversity,Morgantown,W.Va.,2000(Availableat:http://www.rri.wvu.edu/regscweb.htm).

11. Agarwal,C.etal.A Review and Assessment of Land-Use Change Models: Dynamics of Space, Time, and Human Choice.U.S.DepartmentofAgricultureForestService,NortheasternForestResearchStation,UFSTechnicalReportNE-297.Burlington,Vt.,2002.

12. Brown,D.G.etal.Pathdependenceandthevalidationofagent-basedspatialmodelsoflanduse.International Journal of Geographic Information Systems19,153,2005.

13. Parker,D.C.etal.Multi-agentsystemsforthesimulationofland-useandland-coverchange:Areview.Annals of the Association of American Geographers93,314,2003.

14. Klosterman, R. E., and Pettit, C. J. Guest editorial. Environment and Planning B: Planning and Design32,477,2005.

15. Szanton,P.Not Well Advised.RussellFoundationandtheFordFoundation,NewYork,1981.

16. Batty, M. Editorial. Environment and Planning B: Planning and Design 31, 326,2004.

17. Brock,W.A.,Durlauf,S.N.,andWest,K.D..Policyevaluationinuncertaineconomicenvironments.Brookings Papers on Economic Activity1,235,2003.

18. Fragkias,M., andSeto,K.C.Modelingurbangrowth indata sparseenvironments:Anewapproach.Environment and Planning B: Planning and Design34,858,2007.

19. Hoeting,J.A.etal.Bayesianmodelaveraging:Atutorial.Statistical Science14,382,1999.

42963_C008.indd 159 11/6/07 7:13:25 AM

160 Land Use Change

20. Pontius, R. G., Huffaker, D., and Denman, K. Useful techniques of validation forspatiallyexplicitland-changemodels.Ecological Modelling179,445,2004.

21. Train,K.Discrete Choice Methods with Simulation.CambridgeUniversityPress,NewYork,2003.

22. Webster, D. R. On the edge: Shaping the future of peri-urban East Asia. StanfordUniversityAsia/PacificResearchCenterUrbanDynamicsDiscussionPaper,2002.

23. Seto,K.C.,andKaufmann,R.K.ModelingthedriversofurbanlandusechangeinthePearlRiverDelta,China:Integratingremotesensingwithsocioeconomicdata.Land Economics79,106,2003.

24. Seto,K.C.etal.Monitoringland-usechangeinthePearlRiverDeltausingLandsatTM.International Journal of Remote Sensing23,1985,2002.

25. Seto,K.C.,andFragkias,M.Quantifyingspatiotemporalpatternsofurbanland-usechangeinfourcitiesofChinawithtimeserieslandscapemetrics.Landscape Ecology,20,871,2005.

26. Seto,K.C.UrbangrowthinSouthChina:WinnersandlosersofChina’spolicyreforms.Petermanns Geographische Mitteilungen148,50-57,2004.

27. Brennan,G.,andBuchanan,J.M.The Reason of Rules: Constitutional Political Economy.CambridgeUniversityPress,Cambridge,U.K.,1985.

28. Cohen, B. Urban growth in developing countries: A review of current trends and acautionregardingexistingforecasts.World Development32,23,2004.

42963_C008.indd 160 11/6/07 7:13:25 AM

Part III

Synthesis and Prospect

42963_S003.indd 161 8/17/07 2:29:27 PM

42963_S003.indd 162 8/17/07 2:29:28 PM

163

9 Synthesis, Comparative Analysis, and Prospect

Michael J. Hill and Richard J. Aspinall

Contents

9.1 Introduction................................................................................................. 1639.2 ASummationoftheChapters..................................................................... 1649.3 OverallMessages........................................................................................ 169

9.3.1 RealizingtheFullPotentialfromRemoteSensing......................... 1699.3.2 ApplicationofIntegratedMethods.................................................. 1719.3.3 PlacingAnalysisinContext............................................................. 173

9.4 Conclusions................................................................................................. 174References.............................................................................................................. 176

9.1 IntRoDUCtIon

This book has examined the issue of integrated analysis of spatial structure andspatiotemporal processes related to landuse change in terrestrial coupledhumanenvironmentsystems.TheconsequencesoflandusechangehavebeentotransformalargeproportionofthelandsurfaceoftheEarth,andthesechangesareinfluenc-ing the global carbon cycle, regional climate, water quality and distribution, andbiodiversitythroughhabitatloss.1Thebookpresentsanumberofcasestudiesthatinclude urban, wilderness, wet tropical forest, and arid desert-like environments,humanpopulationdensitiesfromhightoverylow,andthosethatdemonstratebothdirectandindirecthumaninfluences(Table9.1).Thecasestudiesincludeseveralthatfocusonanalysisofclearingoftropicalforestenvironments(Chapters4,5,6,and7;Table9.1).Theseareenvironmentsofhighsignificancetoglobalcarbonstocks,bio-diversity,regionalclimate,andAfrican,Asian,andSouthAmericaneconomies.2,3However,bothintensiveagriculturallands(Chapter3;Table9.1)andextensivegraz-inglands(Chapter2;Table9.1)arealsocovered,whilespecificattentionispaidtothelate20th-centuryphenomenonoftheriseofmegacities(Chapter8;Table9.1).

Thesecasestudiesandother literature1,2,3,4,5 suggest thata trade-offapproachisprobablytheonlypragmaticoptiontomoderatingtheimpactofhumansontheterrestrial system.Thepaceofmodificationof the landsurface showsnosignofslowingwhilehumansmodifytheirapproachesinresponsetochangesindemandandpriceforproductandtoavoiddetectionandcontrol.6Inthischapterwebrieflysummarizethemainpointsfromeachchapterandthenexamineinturnthemain

42963_C009.indd 163 11/6/07 7:11:47 AM


messagesdescribedaboveandlookatthepotentialfordeliveryofsocietalbenefitfromaholisticunderstandingof,andapproachto,landusechange.

9.2 A sUMMAtIon oF tHe CHAPteRs

Theapproachinthisbookcanbesummarizedintermsoftheproblemcontext,meth-odological approaches, and geographical-system context (Table9.1). The sciencecontextforthethemeofthebookhasbeenoutlinedinChapter1.Fivebasicsciencequestionswereidentified:dynamicsofchangeinspaceandtime;integrationoffeed-backsbetweenlandscape,climate,socioeconomicandecologicalsystems;resilience,

tAble 9.1A summary of the Key Issues, Contexts, and Methods examined in each Chapter of the book

Chapter Problem contextMethodological

approach(es)Geographical-system

context

1. Aspinall Technicaloverviewofdynamics,scale,accuracy,uncertainty,patternandprocess

Models—conceptual,GIS,RS,CA,MAS,simulation,statistical,empirical,visualization,space-timescaling

Internationalresearchframeworks—GLP,bio-complexity,etc.

2. Hill Transformationofspatiotemporaldataandrelationshipstosimpleindexes

Integrationoftime-seriesanalysis,spatialanalysis,numericalandheuristicmethods

Savannaandgrasslandbiomes/livestockasagents

3. Byron/Lesslie Socialsurveysofattitudestonaturalresourcemanagementissues

Assignmentofrelationshipsbetweenlandholderperceptionandopinionandbiophysicalfeaturesormanagementpractices

RuraleasternAustralia

4. Babigumira,Müller,andAngelsen

Tropicalforestclearing Spatiallyexplicitlogisticmodels

Africantropicalforests

5. EtterandMcAlpine

Tropicalforestclearing Regressiontreeandregressionmodelsofdeforestation/regeneration

Amazoniantropicalforest

6. Crews-Meyer Fragmentationofforest Patchpanelmetrics—pixel-patchhistories

Thailandtropicalforestandregrowth

7. MillingtonandBradley

Tropicalforestclearing Spatialimprintofcadastralgrids;differentialbehavioroffragmentationmetricsasforestcoverdeclines

Amazoniantropicalforest

8. FragkiasandSeto

Urbanexpansion/megacities

Multiplelogisticregression,pseudo-Bayesianmodelaveragingtogetpredictedprobabilitiesofchange

CitiesofPearlRiverDelta—Shenshen,Guangzhou,andFoshan,China

42963_C009.indd 164 11/6/07 7:11:48 AM

Synthesis, Comparative Analysis, and Prospect 165

vulnerability,andadaptabilityofcoupledhumanandnaturalsystems;scaleissues;andaccuracyanduncertaintyissues.Approachesincorporatingalloftheseconsider-ationscanbegroupedunderthegeneralheadingof“models,”butthisincludesadiver-sityoftypesfromconceptualthroughtoempirical(Table9.1).Abroadercontextforincorporatingthedifferentelementsofcasestudiesoflandusechangeistoconsidera rangeof integrating frameworks (Figure9.1).These include theanalytical struc-turefortheGlobalLandProject7; theHumanEcosystemModel8; theverygeneralframeworkofexplicitfocusonlinkagesbetweenthedynamicsofhumanandnaturalsystemsoftheU.S.NationalScienceFoundationBiocomplexityintheEnvironmentprogram(http://www.nsf.gov/geo/ere/ereweb/fund-biocomplex.cfm.),and,withaviewto a more design-oriented approach to landscape change, the Landscape DesignResearch Framework.9 All of these frameworks seek interdisciplinary definitionandfocusonkeyquestionswithinthebroadscopeoflandusechangeandlinkstomanagementofchangeincoupledhumanenvironmentsystems.

InChapter2,theroleofmultiplecriteriaandtrade-offanalysisisdiscussedinthecontextofmethodsandapproachesforcaptureandtransformationofcomplexprocesses.Thechapterproposessimpleindex-basedcomparativeframeworksinaninteractiveenvironmentthatassistdecisionmaking,butretainthelegacyandcriti-calinformationcontentneededforcompleteappreciationofissuesassociatedwithland use change. The analysis seeks to balance evidence-based science and soft-systemsapproachesbyintegrationofharddatawithvaluejudgment,publicopinion,andpolicy andmanagement goals.Temporal analysis is important for entraininglegacyandhistorical factors indecisionmaking; spatial analysis is important forappreciationofsocial,economic,andbiophysicalimpactsoutsidetheboundaryofthedirectlyaffectedareaorgeographicallocationofinterest.Theapproachseekstomeasureandaggregatetheperformanceofalternativeoptions.Itrequiresahighlysystematicandtransparentapproachtomanagementofinformation.

Theapplicationofmethodsthataddressthescienceissuesidentifiedabove,andaggregationofdiverseinformationanddatasourcesintoframeworkstoassistdecisionmaking,mustbemediatedbythepolicycontextfortheanalysisandmodeling.IntheIntroductionwesuggestedthatsustainabilityscience,andspecificallythetonguemodelofPotschinandHaines-Young,10offersacontextand“choicespace”forlinkingscienceanddecisionmakinginpolicyandmanagement.Thecasestudiesprovideamixofanalysisacrosssocial,economic,andbiophysicalperspectivesnecessarytodevelopanunderstandingoflandusechangerelevanttosustainability.

InChapter3,asocialsurveyapproachisusedtodevelopunderstandingoftheattitudes and perceptions of rural communities. The work introduces importantmethodological issues surrounding relationships between point survey data andspatiallyexplicitbiophysicalfeaturesofalandscape.Theanalysislooksatrelation-shipsbetweenanswers to surveyquestionsabout landmanagementpracticesandundesirablelandscapefeaturesbasedonadistanceanalysis.Thisimmediatelyintro-duces interestingquestions abouthowhumansperceive their spatial environmentandhowtheirspatialsensitivitiesandawarenessdiffer.Italsorequiressomeatten-tiontothelegacyeffectsofhistoricalexperienceandviewsofthelandscape,rootedinhistoricparadigms,and the impactofaspatialviews,norms,andmedia issues

42963_C009.indd 165 11/6/07 7:11:48 AM


DecisionMaking

EcosystemsServices

Distu

rbance

How

shou

ld th

e la

ndsc

ape

be d

escr

ibed

?

How

doe

s the

land

scap

e op

erat

e?

Is th

e cu

rren

t lan

dsca

pe w

orki

ng w

ell?

How

mig

ht th

e la

ndsc

ape

be a

ltere

d?

Wha

t pre

dict

able

diff

eren

ces m

ight

the

chan

ges c

ause

s?

How

shou

ld th

e la

ndsc

ape

be ch

ange

d?

Dat

a In

form

atio

n Cu

ltura

l kno

wle

dge

Repr

esen

tatio

n M

odel

s

Proc

ess M

odel

s

Eval

uatio

n M

odel

s

Chan

ge M

odel

s

Impa

ct M

odel

s

Dec

isio

n M

odel

s

Ecol

ogic

alSy

stem

sBi

ogeo

chem

istry

Biod

iversi

tyW

ater

Air

Soil

Soci

alSy

stem

s

T1. D

ynam

ics of

land

–sys

tem

sT2

. Con

sequ

ence

s of l

and–

syste

m ch

ange

T3

. Int

egra

ting a

nalys

is an

d m

odell

ing f

or la

nd su

stain

abili

ty

Land

Use

&M

anag

emen

t

Popu

latio

nSo

cial/E

cono

mic

struc

ture

Polit

ical/I

nstit

utio

nal

regim

esCu

lture

Tech

nolo

gy

Earth

Syste

m

Land

Syste

ms

T2.

1

T2.

2

T2.4

T3.1

Crit

ical

pat

hway

s of c

hang

eT3

.2 V

ulne

rabi

lity a

nd re

silien

ce of

Lan

d–Sy

stem

sT3

.3 E

ffect

ive go

vern

ance

for s

usta

inab

ility

T2.

3

T1.

1

T1.

2

T1.3

Nat

ural

Cultu

ral

Soci

o–ec

onom

ic

Soci

alIn

stitu

tions

Soci

alCy

cles

Soci

alO

rder

Cri

tical

Res

ourc

esH

uman

Soc

ial S

yste

m

Nat

ural

Sys

tem

Hum

an S

yste

m

(a)

(b)

(c)

(d)

FIG

UR

e 9.

1 (S

ee c

olor

inse

rt f

ollo

win

g p.

132

.) A

var

iety

of

inte

grat

ing

fram

ewor

kst

hat

seek

int

erdi

scip

lina

ryd

efini

tion

and

foc

uso

nke

yqu

es-

tion

sw

ithi

nth

ebr

oad

scop

eof

man

agem

ent o

fla

nd u

se c

hang

ein

cou

pled

hum

an e

nvir

onm

ents

yste

ms.

(a)

Ana

lyti

calf

ram

ewor

kfo

rth

eG

loba

l Lan

dP

roje

ct o

f IG

BP

and

IHD

P.(

From

GL

P, 2

005,

wit

hpe

rmis

sion

.)(b

) The

Hum

an E

cosy

stem

Mod

el. (

From

Mac

hlis

eta

l.,8 1

997,

wit

hpe

rmis

sion

.)(c

) The

U

.S. N

atio

nalS

cien

ceF

ound

atio

npr

ogra

mo

nB

ioco

mpl

exit

yin

the

Env

iron

men

t(ht

tp://

ww

w.n

sf.g

ov/g

eo/e

re/e

rew

eb/f

und-

bioc

ompl

exit

y.cf

m);

(d)

The

L

ands

cape

Des

ign

Res

earc

hFr

amew

ork.

(Fr

omS

tein

itz

eta

l.,9

2003

,wit

hpe

rmis

sion

.)

42963_C009.indd 166 11/6/07 7:12:23 AM


insociety.Thechapterisimportantbecauseitbeginstoaddressspatialanalysisofopinions,perceptions,andattitudes.

In Chapter 4, an economic analysis centered on the concept of land rent asadriverof landusechange is applied toassessmentofdeforestation inUganda.Aneconometricmodel(binary—logistic)incorporatingexplanatoryvariablesthatdescribesocioeconomic,spatial,andinstitutionalcontextsisusedtoestimatetheprobability of deforestation. The analysis tests a set of hypotheses and seeks todefine reasons for change. This analysis is also interesting for what it does notinclude:socialsurveyofattitudes(suchasdescribedinthepreviouschapter)couldaddimportantmissingmotivationalinformationonagentbehaviors.Theanalysisimplicitlyassumes thatmotivationwill solelybebasedonmaximizationof landrent. This is one of the main issues for analysis of coupled human environmentsystems:iseconomicreturnafullyadequatedriverforlandusechangeandcon-sequentlandcoverchangeinmostcircumstances?Anassessmentofthefirsttwocase studies might conclude that each would benefit if both of their approacheswerecombined(i.e.,bothsocialsurveyandeconometricmodelingwereapplied).However, evidence from the other case studies shows that further improvementmightbeobtainedifmoresophisticatedspatialandtemporalanalysiswasappliedinconjunctionwithsocialsurveyandeconomicmodeling.

Thenextthreechaptersalladdressdeforestationintropicalforests.InChapter5,regressiontreeanalysisandlogisticregressionareappliedtoanalysisofdeforesta-tionandregeneration.Socialandeconomicprocessesareimportant,asislocalandregionalcontext,anddeforestationfollowsatemporallyexplicittrajectorydescribedbyasigmoidalcurve.Patternsare linked toprocessatdifferentscales:national,regional,and local.Althoughdistances to roadsand towns,proximate factors inthe terminology of Geist and Lambin,11,12 are the best predictors at the nationallevel, deforestation and regeneration occur at local hot spots at a regional level.However,atalocallevelmoreexplicitrelationshipsareobtainedwithaccessibilityand soil type, sincedeforestation rates are strongly related to a spatialmetric—forestedgedensity.Thisprovidesanexampleofspatiallyexplicitanalysisderivingametricwithdirectmeaninginrelationtodeforestationpotentialassociatedwithaccessibility.Hence,spatialanalysisandcalculationofpatternmetricscanbeusedtogenerateindicatorsoflikelihoodofdeforestationatscalesinwhichpatternandprocessaredirectlyconnected.

ThisthemeofpatternmetricsasdescriptorsorindicatorsoflandscapeprocessesiscontinuedfurtherinChapter6.Here,changesinpatternmetricsareanalyzedforlandscapepatchesthroughanumberoftimestepsusingremotelysensedimagery.Inparticulartheinterspersion-juxtapositionindexandmeanpatchsizeindexpro-videmeasuresoffragmentation.Theseindicesgivetemporalprofilesfordifferentlandcoverclassessuchasforest,savanna,andriceagriculture.Theanalysisinthechapter combines the definition of landscape change in terms of pattern metrics,forexample,fluctuations through time in the interspersionof landuse/landcoverclasses,withassignmentofmeaningtothechangesinpatternmetrics,forexample,reductioninforestinterspersionasforestpresencedeclineswiththespreadofriceagriculture.Explanationofregionaldifferencesisbasedonuseofcontextualinfor-mation,forexample,proximitytoareasofmilitaryinstabilitydeterringagricultural

42963_C009.indd 167 11/6/07 7:12:23 AM


encroachment.Thischapterillustratestheapplicationofsophisticatedspatiotemporalanalysis combined with the explanatory contextual information, as discussed inChapters1and2.Thequalityofthisanalysisishighlydependentonthesensitivityandaccuracyofchangedetectionfromremotesensing.

The assessment of forest fragmentation is continued in Chapter 7 where thegoal is to increase understanding of relationships between road construction andforestfragmentationinAmazonia.Here,thefocusisonthebehavioroftheagentsofchangeratherthanonthestructureofthelandcover.Hence,ifChapter6approachesthe issue with a pattern to process orientation, Chapter 7 is examining the sameissuewithaprocess topatternorientation.13Asix-phaseconceptualmodelof thedevelopment of forest fragmentation is described. The approach uses a combina-tion of initial context (colonization) and resulting spatial arrangement of distur-bance(theroadsystem)asthefoundationtodeveloptheconceptualmodel,whichateachphasehasasocio/econo-politicalcontextthatdrivesestablishmentofmoreelaboratespatiallyexplicitlandscapestructure,leadingtoincreasedfragmentation.Patternmetricssuchasmeanpatchsizeandtotaledgelengtharegooddescriptorsoffragmentation,aligningwellwiththeedgemetriccorrelationwithdeforestationin Chapter5, and the interspersion and patch size metric relating to agriculturalexpansioninChapter6.However,thegoaltothickenunderstandingofthehumandimensionofthechangeleadstothedevelopmentofamodelwithanemphasisoncontextinthepredictiveprocess.Thislandscapeisoneofaparticularpattern(i.e.,herringboneclearancepattern),asaresultofthecolonizationcontext,whichestab-lishesthespatialskeleton,thatisthefoundationforthefinalfragmentationpattern.

Thethreechaptersthatdealwithspatialanalysisoftropicaldeforestationandfragmentation provide a powerful case for combination of spatial analysis andmetricationofspatialpatternsindisturbedlandscapes;analysisofchangesinspatialmetrics through timeas indicatorsofparticularprocessesandparticular trajecto-riesinlandscapestructuretowhicheconomicandbiophysicalfunctionalitycanbeascribed; and application of detailed analysis of socio/econo-political contexts inorder to explain evolution of spatial patterns and regional differences in patternsdescribedbyspatialmetrics.

ThefinalcasestudyinChapter8addressestheissueofrapidurbantransforma-tion.Adecisiontheoryframeworkispresentedthatusespolicies,economicdata,an economic model, and an optimization procedure that minimizes an objectivefunction to produce probability maps of predicted urban expansion. There is nobest or true outcome; the output is a probability surface. The expected rates ofgrowthfortheglobalmegacities—therewerealready19citieswithpopulationsinexcessof10millionin2000—introducesanurgencytodevelopmentofpredictivemodelsforplanningduetoexpandedneedforenergy,sanitation,transport,educa-tion,emergencymanagement,healthandsafety,andcleanairandwater.Captureofsocial,psychological,andeconomicdriverswithinacomplexspatialcontextisevenmoreimportant.Manycitiesarealready“landscapesoffate”thathavemostoftheundesirablepropertiesdescribedlaterinthischapter,anddependuponwealthgener-ationandgentrificationofpooreroruglierareasforsignificanttransformationbacktoamore“desirable”landscape.Therefore,themodelingdescribedinChapter8isofparamountimportanceinprovidingspatiallyexplicitprobabilitiesindicatingareas

42963_C009.indd 168 11/6/07 7:12:24 AM


fordevelopmentthatmayenablesomebalancetobeattainedbetweendesirableandfatefulurbanlandscapesforhumanhabitation.Theurbancaserequiresmoreatten-tiontospatiallyexplicitattributionofsociospatialpropertiesandmeasuresofqualityofbuiltspatialhabitatsforhumanactivitiesinordertobalancethepowerfulinflu-enceofcitylandvaluesandcity-basedcommerceandfinancialenterprise.14

9.3 oVeRAll MessAGes

TheframeworkoftheGLPprovidesausefultemplatewithinwhichtoexplorethekeymessagesarisingfromboththeoverviewandcasestudychapterspresentedinthisbook.WehaveusedtheelementsfromFigure2oftheGLPSciencePlanandImplementationStrategy7(“Thecontinuumofstatesresultingfromtheinteractionsbetweensocietalandnaturaldynamics”)toillustratethekeyenablingtechnologies,methods,andapproachesneededtoproviderealsocietalbenefitfromanalysisandstudyoftheseinteractions(Figure9.2).Simplyput,fourofthemajormessagesfromthecasestudiesare:

1. Inordertodetectandaccuratelymeasurechange,remotesensingdataandassociated methodologies must deliver the highest possible informationcontentandaccuracyinchangediscrimination.

2.Remotely sensed data should be complemented with detailed householdand other socioeconomic data from field and census surveys to addressdecision-makingprocesses indetailandgainabetterunderstandingandcapacitytomodelhumanandothersocialandeconomicprocessesinflu-encinglandusechange.

3.Usingthebasicremotesensingandsurveydataresources(numbers1and2) together with spatially explicit descriptions of social, financial, juris-dictional,political,andpsychologicalunitsandinfluences,thereneedstobeconcertedandintegratedapplicationofavarietyofnumerical,heuristic,spatial,andtemporalmethodstoderivethehighestlevelsofunderstandingandquantificationofdependencybetweenpatternsandprocesses.

4.Theanalysisfromnumber3shouldbeplacedinapragmaticcontextthroughcloser relations between science with management and policy related tolanduseandlandusechange.

9.3.1 Realizing the Full Potential FRom Remote SenSing

The implementation plan for the Global Earth Observation System of Systems(GEOSS)15hasdefinedninekeytargetsfordeliveryofsocietalbenefitoverthenext10 years, including improving the management and protection of terrestrial andcoastal ecosystems; supporting sustainable agriculture and combatingdesertifica-tion;andunderstanding,monitoring,andconservingbiodiversity.Alargenumberofobservationalrequirementshavebeendefinedforecosystems,biodiversityassess-ment, and agricultural monitoring. These include particular properties associatedwithlandusechangesuchasburnedareas,landdegradation,speciesdistribution,alien species, extent and location of ecosystems and habitat types, fragmentationof ecosystemsandcommunitycomposition, cultivationandclearing, andgrazing

42963_C009.indd 169 11/6/07 7:12:25 AM


Land

use

Hum

an co

ntro

l

Dyn

amic

land

tran

sitio

nsLa

nd co

ver

Biop

hysic

al co

ntro

l

Enab

ling t

echn

olog

ies,

met

hods

and

ap

proa

ches

• H

ighe

st p

ossib

le in

form

atio

n co

nten

t and

chan

ge

disc

rimin

atio

n fro

m re

mot

e sen

sing

• C

once

rted

and

inte

grat

ed ap

plic

atio

n of

num

eric

al,

heu

ristic

, spa

tial a

nd te

mpo

ral m

etho

ds•

Con

nect

ion

and

harm

oniza

tion

of p

arad

igm

s

bet

ween

econ

omic

, soc

ial, p

sych

olog

ical

and

b

ioph

ysic

al do

mai

ns

Dec

ision

–mak

ing

Choi

ces,

know

ledg

e, va

lues

, pre

fere

nces

and

thei

r soc

ial a

nd p

oliti

cal c

onte

xt

Soci

al ch

alle

nges

Pove

rty

Confl

ict

Soci

al ju

stic

eM

igra

tion

Cons

umpt

ion

Hea

lth

Soci

etal

Ben

efit

Ecol

ogic

al ch

alle

nges

Pollu

tion

Dise

ases

Food

/fibe

r/fu

el sh

orta

geO

verc

row

ding

Clea

n w

ater

supp

ly

Ecos

yste

m g

oods

and

serv

ices

Clea

n ai

rCl

ean

wat

erW

aste

recy

clin

gFo

od/fi

ber/

fuel

Recr

eatio

n

Soci

al sy

stem

sPo

pula

tion

Soci

al/e

cono

mic

stru

ctur

ePo

litic

al/in

stitu

tiona

l reg

imes

Cultu

reTe

chno

logy

Ecol

ogic

al sy

stem

sBi

ogeo

chem

istry

Biod

iver

sity

Air

Wat

erSo

il

FIG

UR

e 9.

2 A

dia

gram

of

the

maj

ore

lem

ents

of

the

coup

led

hum

an e

nvir

onm

ent

syst

em(

afte

rG

LP,

200

5;

Figu

re2

)au

gmen

ted

wit

hke

yen

abli

ngt

echn

olog

ies,

met

hods

and

dat

a, a

ndp

arad

igm

sfo

ran

alys

is o

fco

uple

dhu

man

env

iron

men

tsys

tem

san

dde

live

ryo

fso

ciet

alb

enefi

ts.

42963_C009.indd 170 11/6/07 7:12:26 AM


impacts.Animprovement in thequalityandcoverageofobservationsof the landsurfacefromremotesensingisneededtorealizetheserequirements.

Allofthecasestudiesherethataddresstropicalrainforestclearance(inSouthAmerica,Asia,andAfrica)dependtoalargedegreeonremotesensingasaprimarysourceofdataforbasicchangedetection.CrewsinChapter6discussestheprob-lemswithmultiplicativeerrorswhenusingimagesfrommultipledates,evenwithhigh accuracies for individual classifications. The increasing availability of veryhigh-resolution imagery from space (down to 60 cm pixel resolution) means thatverydetaileddefinitionoflandcoverboundariesandindividualvegetationunitsispossibleatspecificlocations.However,highimagecostandsmallimagefootprintsmeanthatthisapproachisstillimpracticalforwidespreadchangedetection.Recentresearchhasshownthatdetectionofchangesinforestsystems,previouslylimitedbytheinsensitivityofmultispectralinstrumentssuchasLandsattosmallchangesinspectralsignaturesinheavilyfoliatedforestsystems,canbegreatlyimprovedusingspectralunmixingwith time seriesofmultispectraldata16 andwithhigh spectralresolutionspace-bornesensorssuchasHyperion.6,17Aglobalhyperspectralimagingsystemwithsufficientsignaltonoise,moderatepixelresolution(40to60m),andhighcycleforglobalcoverage(15to30days)coulddramaticallyincreaseaccuracyandsensitivityoflandcoverchangedetectionduetolandusepractices.Itisarguablethatthenaturalconclusiontothedevelopmentofremotesensingtechnologyisthecapabilitytoundertakespectroscopyofbiosphericsurfacetargetstodeliverquanti-tativevaluesforkeysurfacestructuralandbiogeochemicalproperties.18

9.3.2 aPPlication oF integRated methodS

TheapplicationofintegratedmethodsdependsonthecomprehensiveaddressingofthebasicsciencequestionsoutlinedbyAspinallinChapter1.Manyaccuracyandscaleissuescanbeaddressedbymaximizingtheinformationcontentfromremotesensing.Theinformationcontentfromremotesensingismaximizedbyprovidinga synergistic mix of imagery with full spectral fidelity and calibration, completeglobalcoverageandhightemporalfrequencyatmediumresolution,andfullspectralfidelityandcalibrationwithveryhighspatialresolutionsuchthatradiometricandspatialscalingisoptimized.ThisremotesensingsystemwouldsatisfymanyoftheinitialneedsofGEOSS,15andtheproductsofthissystemwouldprovideabench-marklevelofreliable,spectrallycomprehensive,temporalandspatialcoveragewithexplicitquantitativeuncertaintyestimates.

This improved information content in remote sensing addresses the needfor better capture of system dynamics in time and space in the future. However,historicalanalysiscanbegreatlyenhancedjustbycomprehensiveanalysisofglobalLandsat MSS and Landsat TM/ETM archives. The Australian government sup-portedan innovativeprogram toanalyzemore than30yearsofLandsatdata forAustraliainordertomonitorandmeasurelandcoverchangetosupportanationalcarbon accounting system.19,20,21 A large archive exists for the United States, forexample,butafull-timeseriesanalysisof thishasyet tobeundertaken(researchhascommencedtoexamineforestdisturbanceusingthesedatabutonlyataregionalscale;SamGoward,personalcommunication).Theprecedingisnotintendedtoover-

42963_C009.indd 171 11/6/07 7:12:27 AM


emphasizetheimportanceofremotesensinginintegratedanalysis,merelytohigh-lightthecriticalroleitcanandshouldplay.

The important biophysical processes are described in detail by many mod-els of varying complexity.22,23 Often these models, developed for site-basedapplication, are difficult to supply with spatially explicit parameters and inputs,resulting in insufficient information to constrain model parameters and provideeffectivemodelpredictions.24Thecross-fertilizationbetweendisciplines,initiatedbyEarthsystemconcernsandfocus,hasprovidedmanynumerical,quantitative,andheuristicmethodsforfindingoptimalmodelfitstoobservationsusingdiversedatawithdifferentspatialandtemporalproperties.Thesemathematicaltechniquesarecollectivelyreferredtoasmultiple-constraintsmodel-dataassimilation.24Themethodsarediverseandhavemigratedfromdiversedisciplinarydomainssuchasnumericalweather prediction25 and economicoptimization andmathematics.26,27These approaches, specifically identified in the Global Land Project strategicimplementationplan(Figure9.1),andintheGEOSS10-yearimplementationplan15mayprovidewaystocombineinformationfromallavailabledatasetstocapturespatiallyexplicitsurfacesprocessesthatdirectlyinfluencelandsystemsandrepre-sentmanyoftheconsequencesoflandusechange.

Thecaptureofthesebiophysicalprocessesisonepartofthepuzzleneededtounderstandthefeedbacksincoupledhumanenvironmentsystems.28Theotherpartinvolvestheintegrationandlinkingofthespatiallyandtemporallyexplicitdynamicsfromobservationandbiophysicalmodelingwithsocialandeconomicmeasuresandmetrics,andwithculturalviewsandhumanperceptionsandopinions.Theresearcheffortsto“socializethepixel”29havebeenparalleledbydigitizingofparcelinforma-tioninpartsoftheUnitedStates,forexample(therebyprovidingdetailedownershiphistoriesandfinescalelandstatistics30),andincreasedabilitytoconsiderthespatialorganizationofpopulations, leading todevelopmentof spatiallyexplicitdata setscontaining survey information on attitudes and behaviors with concomitant con-cernsforconfidentiality.31Inadditionobservationofurbanecosystemsisexplicitlyconsideredwith,forexample,theBaltimoreandPhoenixsitesintheU.S.LongTermEcologicalResearchnetwork.32

Integrationofallthisinformationandunderstandingofdynamicsmaybepro-vided by land use change models33 that range from cellular automata types,34 tostatisticalor simulationmodels,33agent-basedmodels,35and integratedecologicaland economic modeling.36 These models must accommodate dynamics acrossscales, represent thedrivingforces,capturespatial interactionsandneighborhoodeffects,capture the temporaldynamics,and thenbecapableof integrationacrossdisciplinarydomains.33Somecautionisrequiredhere,sincewehavebeendownthispathwithverycomplexprocess-basedbiologicalandecosystemmodels.Wemayinfactneeda“qwerty”solution,apragmaticcombinationofmethodsandapproachesthatisgoodenoughandthatinterfaceswiththehumandecisionframework,butthatis not necessarily the optimum in science or computation—something judged byitseffectivenessratherthanelegance.However,thereissomereasonforoptimism,expressed earlier, that computational power, process understanding, and diversemethodswilldeliverhighlysophisticatedandeffectiveanalysisofcomplexcoupledhuman environment systems. A range of tools that enable humans to interface

42963_C009.indd 172 11/6/07 7:12:27 AM


meaningfullywiththeinformationneededfordecisionsarealsorequired.Onedis-cussedbrieflyinChapter2ismulticriteriaanalysiswithaspatialinterfacedesignedforinteractivework-shopping;37however,agent-basedmodelsandotherinteractiveparadigmsmayworkjustaswell.

Throughincrementalresearchovermanyyears,crossfertilizationofmethodsbetweendisciplines,digitalcaptureofmorespatiallyexplicitbiophysicalandsocialdata,andcontinuedcomputationaladvance,theprospectsforsophisticatedanalysisofcomplexcoupledhumanenvironmentsystemsleadingtomuchbetterinformeddecisionmakinghavebecomerealizable.OneexampleofthisprocessfromAustraliaseespolicymakersandscientistsdrivingahighlysophisticated landusemappingprocess, accompanied by a sophisticated approach to definition of land manage-mentpractices38anddevelopmentofarobustecosystemscience-basedclassificationsystemforvegetationdisturbance.39,40However,theanalysisofthecoupledhumanenvironmentsystemmustalwaysbeplacedincontext.Thetrade-offbattlebetweenanthropocentricandenvirocentricviewsofthefutureoftheearthsystemwillbeamajorbattlegroundinthe21stcentury.Socontextmatters.

9.3.3 Placing analySiS in context

Therequirementfordeliveryofsocietalbenefitthathasbeenemphasizedinmanyforums7,15hasplacedanimperativeontheintegrationandharmonizationofanalysis,modeling,anddecisionmakingandlinkageofsciencetomanagementandpolicymakingbysociety.Emphasisisplacedonthecontextforproblemsandwhethertheyareimportant(orperceivedasimportant).Thisoftendependsoncommunicationofissuesinalanguageorframeworkthatimpactsdirectlyonsocietymembers.

Althoughnotwithoutcontroversy,thesimplesocietalmodelofLuhmann41pro-videsonewaytoconceptualizetheproblem(Figure9.3).Luhmannviewssocietyasacenterlesssetof“functionsystems”thatconstrainbothwhatcanbecommunicatedand how it is communicated. He labels economy, law, science, politics, religion,andeducation as themost important function systems in contemporary society.42Thesefunctionsystemshavedifferent time intervals forexternalcommunication,ranging fromdailydiscourse in science and religion,monthly courtprocesses inlaw,quarterlyteachingandreportingcyclesineconomicsandeducation,toannualelectioncyclesinpolitics(averageoveralllevelsofgovernment).Informationfromone function system only becomes active in another function system when it istranslated into the code of that function system. Hence, environmental informa-tiondoesnothaveanimpactoneconomicorpoliticalprocessuntilitistranslatedintothecodeoftheeconomicorpoliticalfunctionsystem.Globalclimatechangeis a science/environment issue that has crossed between function systems and iswidelyconsideredinthepoliticalfunctionsystem.However,thereisstillaprocessofcontinuedtransferofimprovedscientificinformationandacontinueddemandforinformationinaformthatcanbeusedtomakepoliticaldecisions.

If we therefore return to the choice between alternate future landscapes,10the“landscapesofdesire”and the“landscapesof fate,”even if informationfromenvironmental sciencehasbeen translated into thecodeof thepolitical andeco-nomicsystemstosaythat,forexample,“ifwecontinuewithacertainformofman-agement, certain consequences will ensue”, there remains a need to provide the

42963_C009.indd 173 11/6/07 7:12:28 AM


methodsforexcellentanalysisofdifferentscenariosandoutcomes(Figure9.4).Forexample,thenumberofcitieswithonemillionpeoplehasgrownfrom16in1900stomorethan400in2000,andthenumberofcitieswith10millionpeoplehasgrownfromonein1950to19in2000(seeChapter8).Thereisaneedtohaveverysophis-ticatedmethodsthatcanexplorethewiderangeofinteractionsandconsequencesensuingfromsuchrapidurbangrowth.ThesemethodsmustaddressthekeysciencequestionsfromChapter1inprovidingthebestpossibleanalysisandscenariosfordecisionmaking,sincethiswillinevitablyinvolve“trade-offsbetweenimmediatehumanneedsandmaintaining thecapacityof thebiosphere todelivergoodsandservicesinthelongterm.”1Sincethesedecisionspaceswillbehighlycontested,thequalityandtransparencyofdata,methods,andassumptionswillbeofparamountimportance.Inaddition,regionallandusechangewithnegativeimpactsmayhaveglobalconsequencesthroughclimate-surfaceinteractions.43

9.4 ConClUsIons

Thisbookdocumentsthedevelopmentofanalysisoflandusechangethroughpre-sentationofarangeofcasestudies,placedincontext throughreviewof landusescience,integrativemethods,andframeworksforaddressingcomplexityandinter-relationships.Anabidingthemethatlingerssubliminallythroughoutthisbookisthetime-limiteddecisionspaceofPotschinandHaines-Young10andthetrajectoriesofSteinitzandcolleagues.9Bythesemeanshumandecisionsleadto“landscapesoffate”(perhapsacompletelyurbanizedworld),andhumanaspirationscrave“landscapesofdesire”(perhapsafairytaleland).Thechallengeistousethesophisticatedsciencemarriedtosocialandsoftsystemsparadigmstoweaveapathtoalandscapethatpreservesthefulldimensionsofhumandesireandaspirationwhileretainingafullyfunctionalearthsystem.

Data Information Cultural Knowledge

Politics

SocietalSub–systems

Active – translatedinto code of

functionsubsystem

Inactive – not yettranslated into thecode of function

subsystem

Externalcommunication

Information

Economics

Education

Law

Religion

Science

Translation

Annual

Quarterly

Monthly

Daily

FIGURe 9.3 Connecting the paradigms. Modeling coupled natural human systems.Luhmann41viewssocietyasacenterlesssetof“functionsystems”thatconstrainbothwhatcanbecommunicatedandhowitiscommunicated.Helabelseconomy,law,science,politics,religion,andeducationasthemostimportantfunctionsystemsincontemporarysociety.42

42963_C009.indd 174 11/6/07 7:12:29 AM


Lan

dsca

pe

Posi

tive

prop

ertie

se.

g., A

esth

etic

ally

ple

asin

gPa

stor

alE

xciti

ngPr

oduc

tive

Mul

ti-fu

nctio

nal

Upl

iftin

gR

elax

ing

Bal

ance

d ec

osys

tem

Rem

ote

sens

ing

for

chan

gede

tect

ion

Pol

icy

cont

ext,

impo

rtan

ce,

impa

ct, e

ffec

tof

cha

nges

Lan

dsca

peof

des

ire

Cur

rent

land

scap

e

Man

agem

ent

New

land

scap

e

Neg

ativ

e pr

oper

ties

e.g.

, Ugl

yIn

dust

rial

Bor

ing

Art

ific

ial

Uni

-fun

ctio

nal

Dep

ress

ing

Stre

ssin

gD

egra

ded

ecos

yste

m

Div

erse

spa

tio-

tem

pora

lm

etho

ds f

oran

alys

is o

fpa

ttern

and

proc

ess

Mul

tiple

crite

ria

and

trad

e-of

fap

proa

che.

g.,

ecos

yste

mse

rvic

e

Lan

dsca

peof

fat

e

Prop

ortio

n de

term

ined

by

rela

tive

desi

rabi

lity

orun

desi

rabi

lity

Mix

FIG

UR

e 9.

4 T

he c

onte

xtf

or s

pati

ala

nd t

empo

ral

anal

ysis

in

dete

rmin

ing

the

mix

of

land

scap

eso

ffa

te

and

desi

re.

42963_C009.indd 175 11/6/07 7:12:31 AM


ReFeRenCes

1. Foley,J.A.etal.Globalconsequencesoflanduse.Science309,570–574,2005. 2. Lewis,S.L.Tropicalforestsandthechangingearthsystem.Philosphical Transactions

of the Royal Society B361,195–210,2006. 3. Soares-Filho,B.S.etal.Modellingconservation in theAmazonbasin.Nature440,

520–523,2006. 4. Silva,J.F.etal.Spatialheterogeneity,landuseandconservationinthecerradoregion

ofBrazil.Journal of Biogeography33,536–548,2006. 5. Defries,R.S.,Asner,G.P.,andHoughton,R.Trade-offsinlandusedecisions:Towards

a framework for assessing multiple ecosystem responses to land-use change. In:Defries,R.S.,Houghton,R.,andAsner,G.P.,eds.,Ecosystems and Land Use Change.GeophysicalMonographSeries153,AmericanGeophysicalUnion,2004.

6. Asner, G. P. et al. Ecosystem structure along bioclimatic gradients in Hawaii fromimagingspectroscopy.Remote Sensing of Environment96,497–508,2005.

7. GLP.Global Land Project. Science Plan and Implementation Strategy.IGBPReportNo.53/IHDPReportNo.19.IGBPSecretariat,Stockholm,64pp,2005.

8. Machlis,G.E.,Force,J.E.,andBurch,W.R.Thehumanecosystem,PartI:Thehumanecosystemasanorganizingconceptinecosystemmanagement.Society and Natural Resources10,347–367,1997.

9. Steinitz,C.etal.Alternative Futures for Changing Landscapes: The Upper San Pedro River Basin in Arizona and Sonora.IslandPress,Washington,D.C.,202pp,2003.

10. Potschin,M.,andHaines-Young,R.“Rio+10”,sustainabilityscienceandLandscapeEcology.Landscape and Urban Planning75,162–174,2006.

11. Geist,H.,andLambin,E.F.Proximatecausesandunderlyingdrivingforcesoftropicaldeforestation.Bioscience52,2,143–150,2002.

12. Geist,H.J.,andLambin,E.F.Dynamiccausalpatternsofdesertification.Bioscience54(9),817–829,2004.

13. Laney,R.Aprocess-ledapproachtomodellinglandusechangeinagriculturalland-scapes:AcasestudyfromMadagascar.Agriculture, Ecosystems and Environment 101,135–153,2004.

14. Daniel, T. C. Whither scenic beauty? Visual landscape quality assessment in the21stcentury.Landscape and Urban Planning54,267–281,2001.

15. GEO.Global Earth Observation System of Systems. 10-year Implementation Plan and Reference Document.B.Battrick,ed.EuropeanSpaceAgencyPublicationsDivision,Netherlands,209pp,2005.

16. Asner,G.P.etal.SelectiveloggingintheBrazilianAmazon.Science310,480–482,2005.

17. Asner,G.P.etal.DroughtstressandcarbonuptakeinanAmazonforestmeasuredwithspaceborneimagingspectroscopy.Proceedings of the National Academy of Sciences101,6039–6044,2004.

18. Hill,M.J.,Asner,G.P.,andHeld,A.A.Hyperspectralremotesensingofvegetationincoupledhuman–environmentsystems—societalbenefitsandglobalcontext.Journal of Spatial Sciences32,49–66,2006.

19. Richards, G. P. The FullCAM carbon accounting model: Development, calibrationand implementation for the National Carbon Accounting System, National CarbonAccountingSystem. TechnicalReportNo.28,AustralianGreenhouse Office,Canberra,2001.

20. Furby, S. Land cover change: Specification for remote sensing analysis. NationalCarbon AccountingSystem,TechnicalReportNo.9. AustralianGreenhouseOffice,Canberra,2002.

42963_C009.indd 176 11/6/07 7:12:31 AM


21. Furby,S.,andWoodgate,P.W.,eds.Remotesensinganalysisoflandcoverchange-pilottestingoftechniques.NationalCarbonAccountingSystem, TechnicalReportNo.16,AustralianGreenhouse Office,Canberra,2002.

22. Plummer, S. E. Perspectives on combining ecological process models and remotelysenseddata.Ecological Modelling129,169–186,2000.

23. Nightingale,J.M.,Phinn,S.R.,andHeld,A.A.Ecosystemprocessmodelsatmultiplescales formapping tropical forestproductivity.Progress in Physical Geography 28,1–41,2004.

24. Barrett,D.J.etal.Prospectsforimprovingsavannacarbonmodelsusingmultiplecon-straintsmodel-dataassimilationmethods. Australian Journal of Botany55,689–714,2005.

25. Todling,R.EstimationtheoryandatmosphericdataassimilationIn:Kasibhatla,P.etal.,eds., Inverse Methods in Global Biogeochemical Cycles. American GeophysicalUnions,Washington,D.C.,49–65,2000.

26. Whitley, D. An overview of evolutionary algorithms: practical issues and commonpitfalls.Information and Software Technology43,817–831,2001.

27. Yang,R.L.Convergenceofthesimulatedannealingalgorithmforcontinuousglobaloptimisation.Journal of Optimization Theory and Applications104,691–716,2000.

28. Verberg,P.H.Simulatingfeedbacksinlanduseandlandcoverchangemodels.Land-scape Ecology21,1171–1183,2006.

29. Rindfuss,R.R.etal.Developingthescienceoflandchange:challengesandmethod-ologicalissues.Proceedings of the National Academy of Sciences101,13976–13981,2004.

30. Aspinall,R.J.Modellinglandusechangewithgeneralizedlinearmodels—amulti-modelanalysisofchangebetween1860and2000inGallatinValley,Montana.Journal of Environmental Management72,91–103,2004.

31. VanWey,L.K.etal.Spatialdemographyspecialfeature:confidentialityandspatiallyexplicit data: Concerns and challenges. Proceedings of the National Academy of Sciences102,15337–15342,2005.

32. Grimm, N. B. et al. Integrated approaches to long-term studies of urban ecologicalsystems.Bioscience50,571–584,2000.

33. Verburg,P.H.etal.Landusechangemodeling:currentpracticeandresearchpriorities.GeoJournal61,309–324,2004.

34. White, R., and Engelen, G. High resolution integrated modelling of the spatialdynamicsofurbanandregionalsystems.Computers, Environment and Urban Systems24,383–400,2000.

35. Barreteau,O.,andBousquet,F.SHADOC:Amulti-agentmodeltotackleviabilityofirrigatedsystems.Annals of Operations Research94,139–162,2000.

36. Vionov,A.etal.Patuxentlandscapemodel:integratedecologicaleconomicmodellingofawatershed.Environmental Modelling and Software14,473–491,1999.

37. Hill,P.,Cresswell,H.,andHubbard,L.SpatialprioritisationofNRMinvestmentintheWestHumearea(MurrayCMAregion).TechnicalReportNo.2006,CSIROWaterforaHealthyCountryNationalResearchFlagship,Canberra,2006.

38. Lesslie, R., Barson, M., and Smith, J. Land use information for integrated naturalresources management—a coordinated national mapping program for Australia.Journal of Land Use Science1,45–62,2006.

39. Thackway,R.,andLesslie,R.Vegetationassets,statesandtransitions(VAST):Account-ingforvegetationconditionintheAustralianlandscape.BRSTechnicalreport,BureauofRuralSciences,Canberra,2005.

40. Thackway, R., and Lesslie, R. Reporting vegetation condition using the vegetation,assets, states and transitions (VAST) framework. Ecological Management and Restoration7,s53–s62,2006.

42963_C009.indd 177 11/6/07 7:12:32 AM


41. Luhmann, N. Ecological Communication. Translation Bednarz, J. Jr. University ofChicagoPress,Chicago,187pp,1989.

42. Grant,W.E.,Peterson,T.R.,andPeterson,M.J.Quantitativemodellingofcouplednatural/human systems: simulation of societal constraints on environmental actiondrawingonLuhmann’ssocialtheory.Ecological Modelling158,143–165,2002.

43. Asner, G. P., and Heidebrecht, K. B. Desertification alters ecosystem-climate inter-actions,Global Change Biology10,1–13,2004.

42963_C009.indd 178 11/6/07 7:12:32 AM

179

Index

AACLUMP see Australian Collaborative Land Use

Mapping ProgramAdvanced Very High Resolution Radiometer

(AVHRR), 58AEM see Airborne electromagneticsAEZ see Agroecological zones databaseAgricultural landscape metrics, 128Agroecological zones (AEZ) database, 70Airborne electromagnetics (AEM), 59ALUM see Australian Land Use Management Amazon basin forest fragmentation, 119–134 case study and methods, 122 Chapare, 122 conceptual model, 124–127 landscape metrics behavior, 128–132 methods, 123Amazon Colonization Fronts, 90–91Amazon region of Colombia, 84 forests, 95 global deforestation, 83Analysis in context, 173Animal impact, 31Applied science, 9–11 decisions, 10 human and environmental responses to change,

11 scientific knowledge interpretation and

communication, 11Arequipa forest fragmentation, 123 metrics, 132Asia urban populations, 139Australian Collaborative Land Use Mapping

Program (ACLUMP), 46Australian Land Use Management (ALUM), 48 classification, 49Australian rangelands, 23–32, 35 biophysical attributes, 30 ecosystem assessments, 44 grazing piospheres, 25 landscape structure, 26 MCA framework, 23 representation scale, 29 scale of representation, 29–30 spatial interrelationships, 25 spatial patterns and relationships, 25–26 spatiotemporal inputs to rangeland MCA,

30–31

temporal patterns and influences, 26–29 vegetation types, 36AVHRR see Advanced Very High Resolution

Radiometer

BBasic and applied land use science, 3–11Basic science, 5–8 ecological system, 7 landscape and climate, 7 land systems resilience, vulnerability and

adaptability, 8 scale issues, 8 socioeconomic system, 7 space and time change dynamics, 5–6 uncertainty, 8Binary forest maps, 86Binary non forest maps, 86Biophysical attributes Australian rangelands, 30 land cover change, 85 land use science, 54–59Biophysical data, 50–52 data source integration, 48–49 land use science, 43–60 mapping relationships, 53–54Biophysical integration, 44–59 data and land mapping correspondence, 48–53 future directions, 59 implications, 54–58 regional scale mapping, 45Bogotá forest fragmentation, 123Bolivia see Chapare region of BoliviaBorrow pits, 114

CCA see Cellular automata modelsCaracas forest fragmentation, 123Caribbean soil fertility, 88Carrasco, 123Case studies, 10Cellular automata (CA) models, 145Central America, 94Chapare region of Bolivia, 120, 122–123, 133 Amazon basin forest fragmentation, 122 land use change, 124, 127–128

42963_Index.indd 179 11/5/07 1:51:16 PM

180 Index

progressive deforestation, 125 road building, 134China city planning departments, 158 Pearl River Delta, 150–153 urban land use change, 150Chinese urban land-use change, 139–154 modeling, 140–146 modeling and policy making intersection,

143–145 new modeling approach, 147–150 policy evaluation and uncertainties, 146 policy making, 142Clearance morphologies, 120Clearance typologies, 120Colombia Amazon region, 84 economy, 84 forested ecosystems, 84 land cover change, 83 lowland forests, 92 non forest ecosystems, 84 population, 83Colombian Amazon forest cover change, 93Colonization forest maps, 90Colonization front-line Amazon, 90Communidad Arequipa metrics, 132Community forest maps, 124Comparative analysis and prospect, 163–173Complex spatial patterns, 23, 24Complex spatial relationships, 24Complex spatial signals, 24Complex temporal patterns, 23, 24Complex temporal relationships, 24Complex temporal signals, 24Conceptual model of forest fragmentation, 127Conservation values, 55Correspondence assessment spatial methods analysis, 50 approach, 50 context, 50

DDasymetric mapping, 21Data-generating process (DGP), 144Data source integration, 48–49 Glenelg Hopkins Landholder Survey, 49 natural environment categorized, 49 salinity discharge sites, 49Decision making process models, 154Deforestation, 66–69 agents, 64 conceptual framework, 68–69 data sources, 69–70 definition, 66–67 distribution, 66

econometric results, 74–75 economic model, 70–71 economic models, 64 factors, 68 forest cover, 67 fronts, 93, 135 global land cover, 82 hot spots, 89 hot spot spatial location, 90 institutional context, 76–77 institutional intervention, 76 methodological issues, 71–72 methods, 69–72 modeling unplanned land cover change across

scales, 91 national threat, 88–90 patterns, 91, 94 poor households, 77 prediction, 88, 89 processes, 68 progressive, 125 reduction, 78 results and discussion, 72–77 socioeconomic context, 75–76 spatial context, 76 trend analysis, 91 Uganda, 66, 72, 73, 77 Western Uganda, 63–78Developing world city policy makers, 140DGP see Data-generating processDynamics land use science, 5–7

EEarth Resource Technology Satellite (ERTS), 100Earth surface phenomena, 6Earth system land use, 3Ecological deterioration, 119Economic development of Uganda, 67Economic growth rate of Uganda, 65Economic model deforestation, 70–71Economic statistics, 21Ecosystem assessments of Australia, 44Ecosystem management, 45Ecosystem services improvement strategies, 34 MCA, 34Ecuador land use models, 94Environmental crisis in Uganda, 76Environmental responses to change, 11ERTS see Earth Resource Technology SatelliteErythroxylum coca, 85

FFAO see Food and Agriculture OrganizationFFT see Fourier transform

42963_Index.indd 180 11/5/07 1:51:16 PM

Index 181

Final phase plot exhaustion, 128Fishbone deforestation, 127Food and Agriculture Organization (FAO), 65Forest(s) Amazon region of Colombia, 95 fragmentation patterns, 121 logistic pattern, 92Forest conversion management interventions, 122 planning, 122 spatial patterns, 85Forest cover Colombian Amazon, 93 decline, 92 deforestation, 67 transformation process, 92Forested ecosystems of Colombia, 84Forest energy in Uganda, 68Forest fragmentation Amazon Basin, 119–135 Arequipa, 123 Bogotá, 123 Caracas, 123 case study, 122–124 conceptual model, 124–128, 127 lowland forests, 121 progressive development of spatial patterns,

124 spatial patterns, 127 spatial relationships, 123 temporal relationships, 123 tropical lowlands, 122 tropics, 120Forest landscape metrics, 128Forest loss, 119Forest maps, 90Forest presence, 88Forest regeneration, 90Forestry sector background, 64 Uganda, 65–66Fourier transform (FFT), 27Fragmentation metrics used, 128 patterns, 121Fragstats, 103, 128 landscape, 104

GGeographical information systems (GIS), 6, 18 Glenelg Hopkins region, 48 landscape scales, 45 platforms, 21Geographic Information Science community

(GISc), 102Geology, 6

Giles metrics, 130GIS see Geographical information systemsGISc see Geographic Information Science

communityGlenelg Hopkins Landholder Survey, 48 data source integration, 49Glenelg Hopkins region, 58 GIS, 48 location, 48Glenelg Hopkins Salinity Plan, 50Global deforestation, 82 Colombian Amazon region, 83Global human population, 81Global land cover deforestation, 82Grazing piospheres, 25Green Revolution, 105Groundwater flow systems, 48–49Guangzhou nonurban area predications, 155 urban area predictions, 155

HHumidity, 90

IInstantaneous field of view (IFOV), 102Integrate methods application, 171–172Interspersion/juxtaposition index (IJI), 111

LLand cover, 5 Amazon Colonization Fronts, 90–91 biophysical characteristics, 85 characteristics of theory, 4 Colombia, 83 data sources, 87 deforestation patterns, 91–92 discussions and conclusions, 92–96 emerging trends, 10 GIS, 86 global human population, 81 human activities, 81 human-induced changes, 82 human intervention, 10 integration of theory, 4 local level, 85, 86 methods, 85–88 national level, 86 regional level, 84–85, 86 study region, 82–85 unplanned modeling, 81–96

42963_Index.indd 181 11/5/07 1:51:17 PM

182 Index

Land managers attitudes, 52 conservation value characteristics, 56–57 conservation values, 55 distance to salinity discharge, 51 landscapes of colonization, 134 land use change, 54 perceptions, 48–54 practices, 48–54 relationships, 53–54 salinity awareness, 53, 57 salinity discharge sites, 50 surveys, 45 values, 48–54Land mapping correspondence correspondence assessment spatial methods,

50–52 data source integration, 48–49 mapping relationships, 53–54Landsat Thematic Mapper images, 69Landscape analysis, 133 Australian rangelands, 26 behavior, 128–133 climate, socioeconomic, and ecological

systems integration, 7 ecology, 102, 119 fragmentation, 119 Fragstats, 104 GIS, 45 impacts, 96 level, 100, 101 LULCC, 101 metrics, 128–133 phase calculations, 129 plot exhaustion, 133 scales, 45 seasonality impacts, 114 structure, 26 tropical forest landscapes, 96Landscape dynamism, 99–115 extension to pattern metrics, 103 local lessons, 106–111 panel approach, 102–103 paneled pattern metrics, 112–115 pattern and structure, 101–102 pattern metrics, 103 process and function, 101–102 temporal frequency, 100–101 Thai testing grounds, 104–107Land systems empirical studies, 4 frameworks for study, 7 human systems, 8 interdisciplinary approaches, 7 natural systems, 7 resilience, vulnerability and adaptability, 8

societal benefits, 8 temporal dynamics, 6 transdisciplinary approaches, 7Land use, 5 conservation, 49–50 conversions, 143 dynamics, 6 Earth system, 3 emerging trends, 10 human intervention, 10 information, 11, 49 natural processes, 5 regional mapping, 45–48 social processes, 5 spatial methods, 50–53Land use change, 5 areal change, 46 biology, 20 color figure, 47 concept, 19 context of role, 70 dynamics, 46 economic statistics, 21 importance, 46 information transformations, 18 land managers, 54 lower Murray Region, 46–48 measurement approaches, 46 modeling, 142 multicriterion decision analysis, 17–37 pattern-led analysis, 21 policy decision making, 19 prediction, 46 process-led analysis, 21 related policy making, 148 scale-dependent effects, 20 social statistics, 21 socioeconomic context, 70 space-time domain, 17 spatially applied multicriterion analysis, 18 theories, 144 transformation, 46 transformation domains and methods, 21–23 transformation issues, 19–21 types, 46, 142 Uganda, 77 water supply rivers, 47Land use land cover (LULC), 106 classes, 109 classification, 105 cloud-free imagery, 106 digital archives, 113 mapped products, 113 MPS, 110 pattern metric change, 110 rice agriculture, 109 Thailand, 108

42963_Index.indd 182 11/5/07 1:51:17 PM

Index 183

Land use/land cover change (LULCC), 100 community problems, 103 constant fluctuation, 112 digital archives, 113 drivers, 112 landscape level, 101 landscapes, 115 panel approach, 103 panel method, 115 research, 101Land use models Central America, 94 Ecuador, 94 measurement tools, 149 policy decision making process, 149Land use related research communication, 11 interpretation, 11Land use science applied science, 9–11 basic and applied, 3–11 biophysical data, 54–59 biophysical data integration, 43–60 definition, 3 dynamics, 5–7 feedback with various systems, 7–8 future directions, 59 integration with various systems, 7–8 scales, 8 science questions, 4 social data, 54–59 social data integration, 43–60 theoretical foundations, 4–5 uncertainties, 8Land use systems scientific knowledge interpretation, 11 time-related factors, 6Lowland forests, 121LULC see Land use land coverLULCC see Land use/land cover change

MMagdalena soil fertility, 88Mapping relationships, 53–54 analysis, 54 approach, 53 biophysical data and land mapping

correspondence, 53–54 context, 53MAS/CA see Multiagent system CAMature forests, 91MAUP see Modifiable areal unit problemMCA see Multicriteria analysisMean nearest neighbor (MNN), 131Mean patch size (MPS), 105 LULC, 110 metrics, 131

Metrics Communidad Arequipa, 132 Giles, 130 MPS, 131 Trani, 130Migrants in Uganda, 76MNN see Mean nearest neighborModeling unplanned land cover change across

scales, 81–96 Amazon colonization fronts, 90 deforestation patterns at local level, 91 local level, 85, 88 methods, 85–88 national deforestation threat hot spots, 88–89 national level, 86 regional level, 84, 86–87 results, 88–91 study region, 83–85Moderate Resolution Imaging Spectroradiometer

(MODIS), 59Modifiable areal unit problem (MAUP), 20MODIS see Moderate Resolution Imaging

SpectroradiometerMPS see Mean patch sizeMultiagent system CA (MAS/CA), 145Multicomponent analysis framework, 32Multicriteria analysis (MCA), 18 Australian rangelands, 23 ecosystem services, 34 environments, 32 framework, 23 transformation schema, 35Multicriterion decision analysis, 17–37 Australian rangelands, 23–30 concept, 19 example landscape context, 23–30 multicomponent analysis framework, 32 procedures and models, 17–37 transformation domains and methods, 21–31 transformation issues, 19–20

NNational Biomass Study (NBS), 65National threat, 88–90Natural environment categorized, 49Natural process land use, 5Natural resource manage, 44Natural resource management, 45Natural systems, 7NBS see National Biomass StudyNDVI see Normalized difference vegetation

indexNet ecosystem productivity (NEP), 28, 29 scaling of effect, 30Nonbiophysical time series, 28Non forest ecosystems of Colombia, 84

42963_Index.indd 183 11/5/07 1:51:18 PM

184 Index

Nonurban land-use maps, 147Normalized difference vegetation index (NDVI),

18

PPanel approach, 102–103 LULCC, 115Paneled pattern metrics approach, 115 construction, 104 landscape dynamism, 112–114 research, 103Parque Nacional (PN) Carrasco, 123Pattern-led analysis, 21Pattern metric change, 110Pattern metrics extension, 103PCA see Principle components analysisPCP see Percent correctly predictedPCT see Percentage landscapePearl River Delta (PRD), 150–153Percentage landscape (PCT), 112Percent correctly predicted (PCP), 148Plot exhaustion, 133PN see Parque Nacional CarrascoPolicy decision making, 19Policy makers preferences, 144Policy making local levels, 142 national levels, 142 process, 143Policy relevant land use change models, 142Population Asia, 139 Colombia, 83 growth, 65 Uganda, 65PRD see Pearl River DeltaPrinciple components analysis (PCA), 29Private land management decisions, 10–11Process-led analysis, 21Progressive deforestation, 125Pseudo-Bayesian model, 148Public land management issues, 10–11

RRainfall in Uganda, 76Regional scale mapping land use change importance, 46 land use change in Murray region, 46–47 land use change types, 46Remote sensing, 169–170Rice agriculture, 109Road building, 134

SSalinity awareness, 57Salinity discharge, 51 groundwater flow systems, 48–49 maps, 58 sites, 48–49Scale-dependent effects, 20Scale issues, 8Scale-pattern-process, 99Seasonality impacts on landscape, 114Shenzhen nonurban area predictions, 155 urban area predictions, 155Shuttle Radar Topographic Mission (SRTM), 70Social and biophysical integration, 44–59 biophysical data and land mapping

correspondence, 48–53 future directions, 59 implications, 54–58 regional scale mapping, 45Social data integration, 43–60 land use science, 54–59Social processes, 5Social statistics, 21Societal benefits, 8SOI see Southern oscillation indexSoil fertility Caribbean, 88 Magdalena, 88Soil salinity, 105Southern oscillation index (SOI), 26Space and time change dynamics, 5–6Space-time domain, 17Spatial auto correlation, 71Spatial context Uganda, 71 Western Uganda deforestation, 76Spatial interrelationships, 25Spatially applied multicriterion analysis, 18Spatial methods analysis, 50 approach, 50 context, 50 land use, 50–53Spatial patterns Australian rangelands, 25–26 forest conversion, 85 forest fragmentation, 127Spatial relationships Australian rangelands, 25–26 forest fragmentation, 123Spatial structure, 102Spatiotemporal inputs to rangeland MCA, 30–31Spatiotemporal multicriteria frameworks, 22

42963_Index.indd 184 11/5/07 1:51:19 PM

Index 185

SRTM see Shuttle Radar Topographic MissionSynthesis, comparative analysis and prospect,

163–173

TTEL see Total edge lengthTemporal dynamics, 6Temporal patterns and influences, 26–28Temporal relationships animal impact, 31 forest fragmentation, 123Temporal signals, 31Thailand, 108Thai testing grounds, 104–106Thematic Mapper (TM) images, 69Time-related factors, 6Time-series land use mapping, 47TM see Thematic Mapper imagesTotal edge length (TEL), 131Trani metrics, 130Transdisciplinary approaches, 7Transformation domains and methods, 21–31Tropical deforestation, 5, 63, 67Tropical forest landscapes, 96Tropical forests fragmentation, 120 humidity, 90Tropical lowlands, 122

UUganda background, 64–66 community wealth, 75 cumulative deforestation, 73 deforestation, 72, 77 distribution of deforestation, 66 economic development, 67 economic growth rate, 65 environmental crisis, 76 FAO, 66 forest energy, 68 forestry sector, 65–66 institutional context, 71 land use change, 77 migrants, 76 model results of deforestation, 75 population growth, 65 rainfall, 76 spatial context, 71ULCM see Urban land use change modelUNEP see United Nations Environmental ProgramUnited Nations Environmental Program (UNEP),

69

Universal Transverse Mercator (UTM), 70Urban areas predicted probability of change,

153–154Urban cellular automata models, 145Urban land use change China, 150 data sparse environments, 147 decision-making clientele, 145 developing world cities, 139–158 issues, 140 issues of theories, 146 modeling, 140–147 policy evaluation, 146–147 policy making, 142–143 policy relevant model, 146 threshold predictions, 154 uncertainty reduction, 157 zoning, 145Urban land use change model (ULCM), 140 accuracy issue, 157 probability of threshold application, 157 scientific approaches, 145 scientific disciplines, 145 usefulness, 142Urban land-use maps, 147Urban populations in Asia, 139UTM see Universal Transverse Mercator

VVegetation types, 36von Thünen development process, 73

WWater and soil salinity, 105Watershed-based approach, 44Western Uganda deforestation, 63–78 background, 64–65 conceptual framework, 68 data and methods, 69–71 data sources, 69 defined, 66 descriptive statistics, 72–73 econometric model, 70 econometric results, 74–76 good or bad, 67 institutional context, 76 methodological issues, 71 results, 72–76 socioeconomic context, 75 spatial context, 76World’s urban population, 139 distribution, 141

42963_Index.indd 185 11/5/07 1:51:19 PM

42963_Index.indd 186 11/5/07 1:51:19 PM

(a) (b)

(c) (d)

PCA classes12345678910111213141516171819202122

St. Dev. NEP0.004 – 0.0270.027 – 0.050.05 – 0.0730.073 – 0.0960.096 – 0.1180.118 – 0.1410.141 – 0.1640.164 – 0.1870.187 – 0.210.21 – 0.233

NEP 8593< –2–2 – –1–1 ––0.5–0.5 – –0.2–0.2 – 00 – 0.20.2 – 0.50.5 – 1> 1

NEP 9400< –2–2 – –1–1 ––0.5–0.5 – –0.2–0.2 – 00 – 0.20.2 – 0.50.5 – 1> 1

Figure 2.4 (a) The spatial pattern of temporal signals may be grouped by applying prin-cipal components analysis to the time series and creating a classification based on the major principal components. (b) The temporal metrics may be calculated on the spatial times series to create maps of, for example, standard deviation of net eco system productivity (NEP). (c) A running integration, trend, or wavelet analysis may define periods of distinct behavior in the time series that can then be summarized by metrics such as an integral of NEP for periods of decline and increase. Shown for 1985–1993 and 1994–2000 here.

42963_Art_Color.indd 2 11/14/07 3:40:20 PM

(a) (b)

(c) (d)

Cattle distribution

VRD Vegetation polygonsGrassland classes Other forest and bush Sorghum woodland – low SUR Sehima woodland – moderate SUR Triodia woodland – low utilisation potential Dicanthium grassland – high SUR Triodia grassland – low utilisation potential Mitchell grassland – very high SUR Aristida woodland – Imoderate SUR Not rangeland

VRDNo. threatened birds 0 1 2 3 4

VRDMines 0 1 2 5

VRDPFE ($) NA –1 0 1 2 3 4

CattleNot grazed

Water pointsVictoria RiverDistrictRoads/Rivers

Figure 2.5 Temporal signals are usually based on biophysical or human phenomena that operate at a large scale (e.g., climate, interest rates). Demographic changes at a fine scale may have scale limitations due to level of aggregation in reporting. Temporal signals and indica-tors are filtered by spatial variation. The Victoria River District in the Northern Territory of Australia is highly productive. (a) Cattle are distributed of freehold-leasehold land but confined by water points. (b) Both productivity and ecological impact vary with vegetation type, which is associated with soils, topography, and rainfall gradient. (c) Costs are low and enterprises are profitable but the increment is small on a per hectare basis. (d) Mining with major physical disturbance occurs sporadically across the area. There are threatened bird species in the region and these may be ground nesting and impacted by grazing.

42963_Art_Color.indd 3 11/14/07 3:40:22 PM

Pote

ntial

pro

duct

ivity

for g

razin

gRa

infa

ll reli

abili

tyRa

infa

ll rel.

(w/s

p)

Rain

fall r

el. (a

nn)

Fora

ge p

oten

tial

Gro

wth

Folia

ge p

roj c

over

Soil n

utrie

nt

Acce

ssib

ility

(ARI

A)

Soil P

Soil N

Soil c

arbo

nN

DVI

mea

nN

PP m

ean

soil_

fert

Fig

ur

e 2.

7 C

ogni

tive

map

ping

int

erfa

ce s

uita

ble

for

com

bina

tion

of

dive

rse

spat

iote

mpo

ral m

etri

cs, i

ndic

es,

and

data

laye

rs d

escr

ibin

g m

eani

ngfu

l pro

pert

ies

of a

sys

tem

und

er a

naly

sis.

Thi

s ex

ampl

e sh

ows

the

cons

truc

tion

of

a c

ompo

site

inde

x to

rep

rese

nt p

oten

tial

gra

zing

pro

duct

ivit

y fr

om r

ange

land

s.46

,58

42963_Art_Color.indd 4 11/14/07 3:40:23 PM

Figure 3.1 Land use change in the Barmera, Berri, and Renmark areas of South Australia.

Statistical Local AreasIrrigated horticulture first mapped prior to 1990 (mapped in 1988 for SA)Irrigated horticulture first mapped in 1995 and between 1990–1995Irrigated horticulture first mapped between 1995–1999Irrigated horticulture first mapped in 2001Irrigated horticulture first mapped in 2003

Irrigated horticulture data provided bySA Department of Environment and Heritage

Loxton

BerriBarmera

Renmark

42963_Art_Color.indd 5 11/14/07 3:40:25 PM

N

S

W E

Kilometers

Land managers who did not identify salinity

0 – 0.0456258270.045625827 – 0.0912516530.091251653 – 0.1368774800.136877480 – 0.1825033070.182503307 – 0.2281291340.228129134 – 0.2737549600.273754960 – 0.3193807870.319380787 – 0.3650066140.365006614 – 0.4106324400.410632440 – 0.456258267

Glenelg Hopkins regionNo (land manager identified salinity)Yes (land manager identified salinity)Salinity discharge sites

Land managers who identified salinity

Std. Dev = 6674.11Mean = 6369N = 702.00

Std. Dev = 4146.41Mean = 2536N = 290.00

36,000 – 38,000

32,000 – 34,000

28,000 – 30,000

24,000 – 26,000

20,000 – 22,000

16,000 – 18,000

12,000 – 14,000

8,000 – 10,000

4,000 – 6,000

0 – 2,000

36,000 – 38,000

32,000 – 34,000

28,000 – 30,000

24,000 – 26,000

20,000 – 22,000

16,000 – 18,000

12,000 – 14,000

8,000 – 10,000

4,000 – 6,000

0 – 2,000

Distance to nearest salinity discharge site (meters)

Distance to nearest salinity discharge site (meters)

Freq

uenc

y

0 30 60 120

300

200

100

0

Freq

uenc

y

300

200

100

0

Distance to nearest salinity discharge site<VALUE>

Legend

Figure 3.3 Land managers’ perception of salinity and mapped salinity discharge sites.

42963_Art_Color.indd 6 11/14/07 3:40:27 PM

N

S

W E

Kilometers

0 – 0.0124435330.012443533 – 0.0248870660.024887066 – 0.0373305980.037330598 – 0.0497741310.049774131 – 0.0622176640.062217664 – 0.0746611970.074661197 – 0.0871047300.087104730 – 0.0995482620.099548262 – 0.1119917950.111991795 – 0.124435328

Glenelg Hopkins regionFarmerNon–farmerConservation of natural environments

Std. Dev = 1595.02Mean = 1951N = 640.00

Std. Dev = 1249.48Mean = 1276N = 355.00

8000 – 8500

7000 – 7500

6000 – 6500

5000 – 5500

4000 – 4500

3000 – 3500

2000 – 2500

1000 – 1500

0 – 500

8000 – 8500

7000 – 7500

6000 – 6500

5000 – 5500

4000 – 4500

3000 – 3500

2000 – 2500

1000 – 1500

0 – 500



Non–farmer

Farmer

Freq

uenc

y

0 30 60 120

Freq

uenc

y

140

120

100

80

60

40

20

0

140

120

100

80

60

40

20

0

Distance to nearest area of high conser vation value<VALUE>

Legend

Figure 3.4 Land managers who manage properties near areas of high conservation value.

42963_Art_Color.indd 7 11/14/07 3:40:30 PM

Fig

ur

e 5.

3 P

redi

cted

de

fore

stat

ion

hot

spot

s ob

tain

ed

by

com

bini

ng

area

s pr

edic

ted

to

have

th

e hi

ghes

t pr

obab

ilit

y of

fo

rest

con

vers

ion

(>70

%)

from

the

bes

t m

odel

(th

e re

gion

-spe

cific

cla

ssifi

ca-

tion

tre

e) w

ith

the

area

s w

ith

grea

ter t

han

2% r

ural

po

pula

tion

gr

owth

ra

te

(198

5–19

93).

Red

dep

icts

th

e de

fore

stat

ion

hot s

pots

(a

reas

wit

h >7

0% p

roba

-bi

lity

of

fore

st c

onve

rsio

n an

d >2

% r

ural

pop

ulat

ion

grow

th).

Ora

nge

and

red

depi

ct

area

s w

ith

>70%

pr

obab

ilit

y of

for

est

con-

vers

ion.

G

reen

de

pict

s fo

rest

ed a

reas

, gr

ay r

ep-

rese

nts

clea

red

fore

sted

ar

eas,

an

d w

hite

re

pre-

sent

s no

nfor

este

d ar

eas.

W

hite

cir

cled

are

as i

ndi-

cate

cur

rent

hot

spo

ts o

f de

fore

stat

ion,

whi

ch a

re a

lso

area

s of h

igh-

valu

e bi

o div

ersi

ty v

alue

: (1)

Qui

bdó-

Tri

bugá

, (2

) Fa

rallo

nes-

Mic

ay, (

3) P

atía

-Mir

a, (

4) F

ragu

a-Pa

tasc

oy , (

5) A

lto D

uda-

Gua

yabe

ro,

(6)

Mac

aren

a, (

7) G

uavi

are,

and

(8)

Per

ijá. B

lack

line

is t

he A

ndea

n re

gion

, and

lig

ht

gree

n li

nes

are

nati

onal

par

ks. (

From

Ett

er e

t al.50

Wit

h pe

rmis

sion

.)

N S

WE

Kilo

met

ers

Majo

r roa

dAl

l –yea

r roa

dW

ater

bod

y

Yes

No

050

2510

015

0Lege

nd

Tanz

ania

Rwan

da

Keny

a

Suda

n

Cong

o, D

RC

Kam

pala

North

Central

East

West

Def

ores

tatio

n

Fig

ur

e 4.

1 U

gand

a st

udy

area

sho

win

g th

e di

stri

buti

on o

f de

fore

stat

ion

wit

hin

the

wes

tern

reg

ion

of th

e co

untr

y.

N

Km50

025

001

2

3

45

6

7

8

42963_Art_Color.indd 8 11/14/07 3:40:33 PM

Fore

st co

ver z

one (

%)

Clea

red

1989

(a)

(b)

1996

1999

2002

Fore

st

0–10

10–2

020

–30

30–4

040

–50

50–6

060

–70

70–8

080

–90

90–1

00

Fig

ur

e 5.

4 Fo

rest

map

s of

the

col

oniz

atio

n fr

ont f

or e

ach

stud

y da

te: (

a) e

xten

t of

fore

st c

over

(bl

ack

= fo

rest

); an

d (b

) pe

rcen

tage

fo

rest

cov

er a

t 10%

incr

emen

t zon

es. (

From

Ett

er e

t al.39

With

per

mis

sion

.)

42963_Art_Color.indd 9 11/14/07 3:40:34 PM

LEG

END

Rapi

d de

fore

statio

nRa

pid

fore

st re

gene

ratio

nM

inor

chan

ges

Tow

nsRi

vers

Road

s

1999

–200

2 m

ilita

ryex

clusio

n zo

ne

1989

–199

619

96–1

999

1999

–200

2

Fig

ur

e 5.

5 Sp

atia

l lo

cati

on o

f th

e lo

cal

hot

spot

s of

def

ores

tati

on (

red)

and

reg

ener

atio

n (g

reen

) fo

r th

e th

ree

tim

e pe

riod

s of

st

udy.

(Fr

om E

tter

et a

l.39 W

ith

perm

issi

on.)

42963_Art_Color.indd 10 11/14/07 3:40:36 PM

IJI = 14.2MPS = 8.1

IJI = 17.9MPS = 4.7

IJI = 23.7MPS = 2.2

IJI = 19.2MPS = 3.1

(2)

(2a)

(2b)(3a)

(3)

(1)

Figure 6.1 The panel process, conducted at both the pixel and patch levels: (1) four multi-spectral satellite images are each categorized into a thematic LULC classification; (2) pattern metrics are run on each of the four LULC classifications, each producing a set of patch, class, and landscape statistics (here the interspersion/juxtaposition index [IJI] and mean patch size [MPS] are shown) as well as an output image of the delineated patches; (2a) pattern metric output for each of the four times is used to calculate three piecemeal change maps for each pattern metric and each consecutive pair of images (e.g., showing fluctuations in IJI or MPS between two time periods) as per Crews-Meyer11,27; (2b) three pattern change maps are stacked into one panel of all structural change for each given metric (e.g., showing fluctuation in IJI or MPS through all time periods) as per Crews-Meyer11,27; (3) three thematic change maps are created for each of the time periods represented by the four classifications; (3a) the three thematic change maps are stacked to represent the full record of all thematic change across the four classifications as per Crews-Meyer.3

42963_Art_Color.indd 11 11/14/07 3:40:37 PM

Fig

ur

e 6.

3 (a

) L

UL

C in

the

grea

ter

stud

y ar

ea in

the

1972

/197

3 w

ater

yea

r; (

b) 1

985;

and

(c)

1997

.

42963_Art_Color.indd 12 11/14/07 3:40:38 PM

(a)

(b)

(c)

Figure 6.6 (a) The change in configuration from 1972/1973 to 1975/1976, reveal-ing that the entire subset area has experienced a greater than 10% increase in interspersion/juxtaposition index (IJI) scores (due to increased fragmentation and concomitant interdigitation). Note that most of the area experienced the same type of change. (b) Illustration of a different trend between 1975/1976 and 1979, whereby increases, decreases, and relative stability in IJI vary spatially. More upland areas (most central in the subset) experienced a consolidation on the landscape, while peripheral areas remain relatively stable in terms of configuration with notable exceptions on the southeastern perimeter. (c) Illustration of the continued spatial heterogeneity in IJI, with lowland/peripheral areas under going continued fragmen-tation, while the less accessible, upland areas appear to have leveled off in terms of larger-scale fragmentation or consolidation but continue to experience small pockets of fragmentation throughout.

42963_Art_Color.indd 13 11/14/07 3:40:39 PM

-180

-60

Robi

nson

Pro

ject

ion

Cent

ral M

erid

ian

0.00

Sour

ce: E

SRI D

ata &

Map

s CD

Kilo

met

ers

0

1,45

0

2

,500

5,500

8,70

0

11,6

00

C

ount

ry B

ound

ary

POPU

LATI

ON

7500

00 –

1000

000

1000

001

– 300

0000

3000

001

– 500

0000

5000

001

– 100

0000

0

1000

0001

– 23

6200

00

Anta

rctic

Circ

le

Trop

ic of

Cap

ricor

n

Equa

tor

Pacifi

c Oce

an

Trop

ic of

Can

cer

Indi

an O

cean

Pacifi

c Oce

an

Atla

ntic

Oce

an

–180

–160

–140

–120

–100

–80

–60

–40

–20

020

4060

8010

012

014

016

018

0

–160

–140

–120

–100

–80

–60

–40

–20

–60

–40

–20

20

40

60

80

0

–40

–2020

40

60

80

0

020

4060

8010

012

014

016

018

0

Arcti

c Cir

cle

Fig

ur

e 8.

1 Po

pula

tion

dis

trib

utio

n of

the

wor

ld’s

urb

an a

gglo

mer

atio

ns w

ith

750,

000

peop

le o

r m

ore

arou

nd y

ear

2000

.

42963_Art_Color.indd 14 11/14/07 3:40:48 PM

Kilometers6030150

KilometersFoshan study areaGuangzhou study areaShenzhen study areaGraticule (5 degrees)Province BoundariesGuangdong Province

25012562.50

Urban land use Urban land 1988 Growth 1988–1996

20N

25N

110E

115E

SOUTH CHINA SEA

GUANGXI PROVINCEGUANGDONG PROVINCE

HUNAN PROVINCE JIANGXI PROVINCEFUJIAN PROVINCE

PEARL RIVER DELTA

N

Figure 8.2 The Pearl River Delta in southeast China and the Shenzhen, Foshan, and Guangzhou study areas (urban land in 1988, new urban land between 1988 and 1996).

42963_Art_Color.indd 15 11/14/07 3:40:51 PM

0.6

0.5

0.4

0.3

0.2

0.1

0 0.12

0.1 0.08

0.06

0.04

0.02

0

50 100

150

200

250

300

350

400 50 100

150

200

250

300

350

400

100

200

300

400

500

600

700

100

200

300

400

500

600

700

Fig

ur

e 8.

5 P

redi

cted

pr

obab

ility

of

ch

ange

to

ur

ban

area

s bet

wee

n 20

04 a

nd 2

012

and

stan

dard

dev

i-at

ion

of p

ixel

pre

dict

ed p

roba

bilit

ies

for

Gua

ngzh

ou

(60

m r

esol

utio

n; v

alue

s in

per

cent

age

poin

ts).

0.6

0.5

0.4

0.3

0.2

0.1

0 0.12

0.1 0.08

0.06

0.04

0.02

0

50 100

150

200

250

300

350

400 50 100

150

200

250

300

350

400

100

200

300

400

500

600

700

100

200

300

400

500

600

700

0.35

0.3

0.25

0.2

0.15

0.1

0.05

0.16

0.14

0.12

0.1

0.08

0.06

0.04

0.02

0

50 100

150

200

250

300

350

400

450

500

550 50 100

150

200

250

300

350

400

450

500

550

100

200

300

400

500

600

100

200

300

400

500

600

0.35

0.3

0.25

0.2

0.15

0.1

0.05

0.16

0.14

0.12

0.1

0.08

0.06

0.04

0.02

0

50 100

150

200

250

300

350

400

450

500

550 50 100

150

200

250

300

350

400

450

500

550

100

200

300

400

500

600

100

200

300

400

500

600

Fig

ur

e 8.

4 P

redi

cted

pro

babi

lity

of c

hang

e to

urb

an a

reas

bet

wee

n 20

04 a

nd 2

012

and

stan

dard

dev

iatio

n of

pix

el p

redi

cted

pro

babi

li-

ties

for

Fosh

an (

30 m

res

olut

ion;

val

ues

in p

er-

cent

age

poin

ts).

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

100

200

300

400

500

600

100

200

300

400

500

600

0.12

0.1

0.08

0.06

0.04

0.02

0

100

200

300

400

500

600

700

800

900

100

200

300

400

500

600

700

800

900

Fig

ur

e 8.

3 P

redi

cted

pro

babi

lity

of c

hang

e to

urb

an a

reas

bet

wee

n 20

04 a

nd 2

012

and

stan

-da

rd d

evia

tion

of p

ixel

pre

dict

ed p

roba

bilit

ies

for

Shen

zhen

(60

m r

esol

utio

n; v

alue

s in

per

-ce

ntag

e po

ints

).

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

100

200

300

400

500

600

100

200

300

400

500

600

0.12

0.1

0.08

0.06

0.04

0.02

0

100

200

300

400

500

600

700

800

900

100

200

300

400

500

600

700

800

900

42963_Art_Color.indd 16 11/14/07 3:41:03 PM

DecisionMaking

EcosystemsServices

Distu

rbance

How

shou

ld th

e la

ndsc

ape

be d

escr

ibed

?

How

doe

s the

land

scap

e op

erat

e?

Is th

e cu

rren

t lan

dsca

pe w

orki

ng w

ell?

How

mig

ht th

e la

ndsc

ape

be a

ltere

d?

Wha

t pre

dict

able

diff

eren

ces m

ight

the

chan

ges c

ause

s?

How

shou

ld th

e la

ndsc

ape

be ch

ange

d?

Dat

a In

form

atio

n Cu

ltura

l kno

wle

dge

Repr

esen

tatio

n M

odel

s

Proc

ess M

odel

s

Eval

uatio

n M

odel

s

Chan

ge M

odel

s

Impa

ct M

odel

s

Dec

isio

n M

odel

s

Ecol

ogic

alSy

stem

sBi

ogeo

chem

istry

Biod

iversi

tyW

ater

Air

Soil

Soci

alSy

stem

s

T1. D

ynam

ics of

land

–sys

tem

sT2

. Con

sequ

ence

s of l

and–

syste

m ch

ange

T3

. Int

egra

ting a

naly

sis an

d m

odell

ing f

or la

nd su

stain

abili

ty

Land

Use

&M

anag

emen

t

Popu

latio

nSo

cial/E

cono

mic

struc

ture

Polit

ical/I

nstit

utio

nal

regim

esCu

lture

Tech

nolo

gy

Earth

Syste

m

Land

Syste

ms

T2.

1

T2.

2

T2.4

T3.1

Crit

ical

pat

hway

s of c

hang

eT3

.2 V

ulne

rabi

lity a

nd re

silien

ce of

Lan

d–Sy

stem

sT3

.3 E

ffect

ive go

vern

ance

for s

usta

inab

ility

T2.

3

T1.

1

T1.

2

T1.3

Nat

ural

Cultu

ral

Soci

o –ec

onom

ic

Soci

alIn

stitu

tions

Soci

alCy

cles

Soci

alO

rder

Cri

tical

Res

ourc

esH

uman

Soc

ial S

yste

m

Nat

ural

Sys

tem

Hum

an S

yste

m

(a)

(b)

(c)

(d)

Fig

ur

e 9.

1 A

var

iety

of

inte

grat

ing

fram

ewor

ks t

hat s

eek

inte

rdis

cipl

inar

y de

fini

tion

and

foc

us o

n ke

y qu

esti

ons

wit

hin

the

broa

d sc

ope

of m

anag

e-m

ent o

f la

nd u

se c

hang

e in

cou

pled

hum

an e

nvir

onm

ent s

yste

ms.

(a)

Ana

lyti

cal f

ram

ewor

k fo

r th

e G

loba

l Lan

d P

roje

ct o

f IG

BP

and

IHD

P. (

From

GL

P,

2005

, w

ith

perm

issi

on.)

(b)

The

Hum

an E

cosy

stem

Mod

el.

(Fro

m M

achl

is e

t al

.,8 19

97,

wit

h pe

rmis

sion

.) (c

) T

he U

.S.

Nat

iona

l Sc

ienc

e Fo

unda

tion

pr

ogra

m o

n B

ioco

mpl

exit

y in

the

Env

iron

men

t (h

ttp:

//ww

w.n

sf.g

ov/g

eo/e

re/e

rew

eb/f

und-

bioc

ompl

exit

y.cf

m);

(d)

The

Lan

dsca

pe D

esig

n R

esea

rch

Fram

ewor

k. (

From

Ste

initz

et a

l.,9

2003

, wit

h pe

rmis

sion

.)

42963_Art_Color.indd 17 11/14/07 3:41:19 PM

Land Use Change

Documents

Transcript of Land Use Change