J. Great Lakes Res. 33 (Special Issue 3):13–26Internat. Assoc. Great Lakes Res., 2007

Methods for Generating Multi-scale Watershed Delineations for IndicatorDevelopment in Great Lake Coastal Ecosystems

Thomas P. Hollenhorst1,*, Terry N. Brown1, Lucinda B. Johnson1,Jan J.H. Ciborowski2, and George E. Host1

1Center for Water and the EnvironmentNatural Resources Research Institute

University of Minnesota Duluth5013 Miller Trunk Highway

Duluth, Minnesota 55811-1442

2Department of Biological SciencesUniversity of Windsor

Windsor, Ontario N9B 3P4

ABSTRACT. Watersheds represent spatially explicit areas within which terrestrial stressors can bequantified and linked to measures of aquatic ecosystem condition. We delineated thousands of GreatLakes watersheds using previously proven and new watershed delineation techniques. These were used toprovide summaries for a variety of anthropogenic stressors within the Great Lakes. All delineation tech-niques proved useful, but each had applications for which they were most appropriate. A set of watersheddelineations and stressor summaries was developed for sampling site identification, providing relativelycoarse strata for selecting sites along the U.S. Great Lakes coastline. Subsequent watershed delineationswere used for high-resolution site characterization of specific sites and characterizing the full coastalstressor gradient. For these delineations we used three general approaches: 1) segmentation of theshoreline at points midway between adjacent streams and delineation of a watershed for each segment;2) specific watershed delineations for sampled sites; and 3) a Great Lakes basin-wide, high-resolutionapproach wherein sub-basins can be agglomerated into larger basins for specific portions of the coast.The third approach is unique in that it provides a nested framework based on hierarchies of catchmentswith associated stressor data. This hierarchical framework was used to derive additional watersheddelineations, and their associated stressor summaries, at four different scales. Providing anthropogenicstressor metrics in such a format that can quickly be summarized for the entire basin at multiple scales,or specifically for particular areas, establishes a strong foundation for quantifying and understandingstressor-response relationships in these coastal environments.

INDEX WORDS: Scale, watersheds, indicators, coastal, aquatic ecology, Great Lakes.

INTRODUCTION

Watersheds have become widely recognized ascritical functional and ecological management units(Diana et al. 2006). Long recognized for their im-portance in flood control and management, water-sheds are now also recognized for their importancein water quality management and protection ofaquatic habitats. Watersheds also provide a sum-mary unit for monitoring and reporting on totalmaximum daily loads (TMDL) (Tong and Chen

*Corresponding author. E-mail: thollenh@nrri.umn.edu

2002) currently used by the U.S. EnvironmentalProtection Agency and state water managementagencies (http://www.mda.state.mn.us/agdev/impairedwater). This focus on watersheds has led tomany advances in hydrologic science. The U.S.Army Corps of Engineers’ Hydrologic EngineeringCenter in Davis, California developed several mod-els (HEC-RAS, HEC-HMS) using watershed datato simulate precipitation-runoff processes and floodconditions while analyzing water quality within wa-tersheds (http://www.hec.usace.army.mil/). SeveralGIS software companies have developed GIS-based

13

14 Hollenhorst et al.

tools that help delineate and analyze watershedsusing standard elevation data. Others have begun touse NEXRAD radar to map precipitation patternsand estimate precipitation amounts in a spatially(specific to each watershed) and temporally explicitmanner (hourly, daily, and monthly measures)(Knebl et al. 2005). States within the U.S. have alsostarted to incorporate these technologies into theirmonitoring programs, as described by Magner andBrooks (2007). The Environmental Systems Re-search Institute (ESRI), along with key state, na-tional, and international contributors, developed theArcHydro data model (Maidment and Morehouse2002) to better manage and process watershed in-formation and watershed delineation methods. TheArcHydro data model is unique in that it provides aframework for delineating small sub-catchmentsand linking them along the stream network so thattheir attributes can be accumulated as a point of ref-erence moves downstream through the watershed.This provides a system for developing standard sub-catchment characterizations that cover the entirebasin, but can be queried and summarized to pro-vide information for specific sites throughout thebasin.

The importance of developing cross-jurisdic-tional comprehensive management plans for theGreat Lakes has also become more widely recog-nized (International Joint Commission [IJC] 1995).This is particularly true as watersheds, the bound-aries of which are shaped by geomorphic and physi-cal processes rather than political borders, havebecome more accepted as functional ecologicalunits. At the same time, Great Lakes managementagencies and non-government organizations havebegun to focus on the socioeconomic and ecologicalimportance of coastal margins, where terrestrial in-fluences and stressors meet and mix with lakewardinfluences and stresses (IJC 2005). Because manyterrestrial stressors are delivered to the coastal mar-gin through hydrologic processes, quantifyingstresses at a watershed scale may contribute to amechanistic understanding of ecosystem conditionat the coastal margin. An understanding of the rela-tionship between anthropogenic stresses and thecondition of ecosystems is a fundamental require-ment for development of effective environmentalindicators, and a key precursor to the creation andimplementation of management strategies (Niemiand McDonald 2004, Niemi et al. 2004). Sincethese relationships span multiple temporal and spa-tial scales, scalable summary units (preferablybased on watershed boundaries) are essential to ac-

curately quantify stressor-response relationships.Quantification of stressors within hierarchicallystructured watersheds would enable landscape sum-maries to be conducted over the entire extent of astudy region, then queried and reconstituted forsmaller watersheds.

Also, ecological patterns and processes arethought to have characteristic scales at which theirproperties can best be described and studied (Wuand Li 2006). Unfortunately, the characteristic (in-trinsic) spatial and temporal scales that regulate keyprocesses are usually not known; therefore, multi-ple observations made at different scales are neces-sary (Bloschl and Sivapalan 1995, Niemi andMcDonald 2004). In this paper, we illustrate the de-velopment of a hierarchical watershed frameworkthat allows for testing various stressor-response re-lationships at multiple scales throughout the GreatLakes.

Watershed Delineation Approaches

Here we describe six types of watershed delin-eations, resulting from three different approaches,defining relevant spatial units of the landscape forsummarizing stressors which could impact adjacent,downstream systems. These watershed delineationsform the basis of indicator development for a broadvariety of Great Lakes coastal margin indicators(Danz et al. 2007, Reavie et al. 2006, Morrice et al.2007). Initial watershed delineations were devel-oped to span the entire U.S. Great Lakes in supportof site selection and specific sample site characteri-zation needs. A intermediate delineation methodwas developed to assess particular groups of sitessampled in the field by project researchers. Experi-ence gained from each of these delineations led to athird, alternative approach; a high resolution, GreatLakes-wide coverage of watersheds that providesfor agglomerating the smallest spatial units up todifferent aggregate levels, depending on the purposeand need. In this third approach, watersheds weredelineated for every stream draining an area ofmore than 2.7 km2 toward the Great Lakes. Water-sheds were ordered sequentially along the coast, en-abling each drainage basin and its associatedstressor summaries to be agglomerated to representlarger portions of the coast.

The objective of the Great Lakes EnvironmentalIndicator (GLEI) project was to quantify stressor-response relationships for novel and existing indica-tors, develop predictive models to infer ecologicalstatus at the coastal margin, and derive integrative

Multi-scale Watershed Delineations for Coastal Indicator Development 15

metrics of that ecological status among several indi-cator groups (birds, amphibians, fish, macroinverte-brates, algae, and water quality). To develop andtest these indicators, we sampled multiple bioticand abiotic components of ecosystems across theU.S. side of the Great Lakes (Niemi et al. 2007).Sample sites were selected using a probabilisticsampling design stratified across five geomorphictypes and relative amounts of anthropogenic stress(Danz et al. 2005).

Watershed Delineation Approach 1—Segment-Sheds

One of the highest priorities in the developmentof a comprehensive sampling program is site selec-tion. Since robust indicators must be capable ofidentifying conditions along the full gradient of an-thropogenic disturbance, quantifying stressorsacross the U.S. side of the entire Great Lakes basinwas essential. The initial delineation approach ad-dresses two problems. First, there was no singlespatial unit that encompassed the coastal areas ofthe Great Lakes for summarizing stressors. Existingwatershed delineations were too coarse, and further,did not include the coastal areas between rivermouths. Second, although many different data sets(and variables) existed that could be used to de-scribe the dominant stressors, only limited time wasavailable to bring them into a common mappingframework. We chose to describe, map, and stratifystressors at a broad scale for site selection purposes,while reserving higher resolution summaries forlater site characterization and analysis of areas actu-ally sampled. Anthropogenic stressor data werecompiled and summarized for unique contributingareas called “segment-sheds.”

We defined a segment-shed as the land areadraining to the shoreline beginning and ending half-way between a second- or higher-order stream andthe adjacent streams on either side (Fig. 1) on theU.S. side of the Great Lakes (Fig. 2). Seven hun-dred sixty-two segment-sheds were delineated,using the National Elevation Data Set (NED) andwatershed delineation tools available in ArcInfoVersion 9.1 (ESRI 2000). The positions of segmentshed-boundaries were visually assessed for errors,and edited when necessary, using ancillary mapdata including streams, aerial photos, and USGSscanned quad maps (digital raster graphics 1:24,000and 1:250,000). We then compiled 207 differentstressor variables from 19 data sources (Danz et al.2005), and summarized them for each coastal seg-ment-shed. Stressors representing human activities,

including land cover, population density, point-sources, and agricultural land use, were assembledfor the U.S. side of the Great Lakes basin. Themulti-dimensional stressor gradient was representedby principal components (PCs) calculated withinseven categories of stress (Danz et al. 2005).

The segment-sheds were effective units for siteselection, and also proved useful as an intermediateproduct for site stressor characterization. Becausethe segment-sheds covered the entire U.S. side ofthe Great Lakes basin, stressor summaries representthe full gradient of stress, ranging from the mostpristine (reference) to most degraded. This informa-tion allows us to gauge the absolute position ofsampled sites along the stressor gradient.

Watershed Delineation Approach 2—Complex Specific Watersheds

An additional challenge was to precisely linkareas sampled by research teams to the delineated-segment sheds and accompanying stressor sum-maries. This was complicated by the fact that thelocation, timing, duration, and spatial extent ofsampling within segment-sheds varied greatlyamong GLEI research teams, reflecting the habitat,characteristics, and sampling strategies unique to

FIG. 1. Example of second-order and higherstreams along the coast of western Lake Superiorwith arrows indicating the boundary between adja-cent shoreline segments halfway between eachstream mouth.

16 Hollenhorst et al.

the biota of interest. Unique sample units (localepolygons) were identified for each research team bycombining GPS coordinates from each sample loca-tion. A total of 1,179 locale polygons was identifiedand delineated. These locale polygons were used togenerate “complex” polygons, which encompassedthe spatial boundaries of each unit sampled by allthe groups (n = 530; hereafter called complexes).Watersheds draining to the sampled wetlands andembayment complexes were delineated (n = 344;wetlands and embayments only) using the NED andwatershed delineation tools available in ArcInfoVersion 9.1 (ESRI 2000). The delineations of the344 complex-specific watersheds were visually as-sessed for errors, and edited when necessary, usingancillary map data including streams, aerial photos,and USGS scanned quad maps (digital raster graph-ics 1:24k and 1:250k).

This approach established a site-specific basis fordefining topographic contributing areas for coastalwetlands and embayments receiving landscape-derived stressors via riverine flow paths. Subse-quently, the 207 anthropogenic stressor variablespreviously compiled for segment- sheds were sum-

marized across each complex watershed and ana-lyzed using principal component analysis (PCA).Details of how GIS tools were used to support indi-cator development from field measurements of a va-riety of biota are provided in Johnston et al. (2006).

Watershed Delineation Approach 3—Hierarchically Nested Watersheds

The approach of using complex-specific water-sheds worked well for geomorphic types that re-ceive flow directly from the adjacent, upstreamwatershed (e.g., riverine wetlands and embay-ments), but was less applicable to lacustrine wet-lands or high-energy areas where the boundaries oftopographic contributing areas are not as obvious.For these areas, an alternative to the classic topo-graphically-defined watershed was needed to quan-tify stressors that might be delivered from bothnearby terrestrial areas as well as from the lake it-self, via long-shore currents, wind, and waves.

The ArcHydro data model (Maidment and More-house 2002) delineates hierarchically structuredsub-catchments for each steam segment (reach) be-

FIG. 2. Segment-sheds for the U.S. side of the Great Lakes.

Multi-scale Watershed Delineations for Coastal Indicator Development 17

tween stream confluences. Stream reaches are num-bered in sequence such that each sub-catchment in-cludes a unique identification label and the“next-down” identification of the catchment intowhich it flows (Fig. 3). This allows stressor sum-maries to be calculated for each sub-catchment; forthese values to be summed cumulatively down thehydrologic network, and for the entire watershed.This data structure better accommodates analyses ofsystems exhibiting different hydrology amongdrainage basins, and permits future analyses that in-corporate different scales of influence.

The ArcHydro data model also allows for water-

shed delineations that incorporate existing maps ofstreams to create hydrologically corrected elevationmodels (Maidment 1997). We used map data repre-senting connected stream networks of the NationalHydrologic Data Base (http://nhd.usgs.gov) com-bined with National Elevation Data (http://ned.usgs.gov). Using ArcHydro, we also locatedand “filled” sinks (areas in the elevation data thatare lower in elevation than their surroundings) toensure that flow continuity and flow direction wereproperly delineated along the course of a drainagebasin. Flow accumulation was calculated from thesecorrected delineations. Areas with high flow accu-

FIG. 3. Network connectivity of ArcHydro catchments. Each stream seg-ment is identified both by its own unique numerical label (Hydro ID) and bythe number of the “next-down” segment into which it flows (next-down ID).Sub-catchments (shown faintly) for each stream segment are also assignedthe same IDs.

18 Hollenhorst et al.

mulation were then designated as streams. To beoperationally identified as a stream, an area had toreceive flow from an upstream area of at least 3,00030 × 30 m pixels (2.7 km2), to roughly coincidewith streams mapped at 1:24,000. After the streamswere delineated, ArcHydro catchments were delin-eated for stream lengths between stream conflu-ences.

Connected networks of streams flowing to theGreat Lakes coast were individually identified, andall streams and catchments within that networkwere given a network-level identification number.Discrete watersheds flowing to the coast were thenuniquely identified and merged to form a hierarchi-cal network of highly detailed watersheds flowingto the Great Lakes coast.

Drainage basins narrow at their downstream ter-minus where their rivers discharge into the lake,and the land areas between river-mouths drainingdirectly to the shoreline are not accounted for.These areas, referred to as coastal “interfluves”(Gilliam et al. 1997) were delineated by intersect-ing the coastal watershed boundaries with a poly-gon representing the terrestrial portion of the GreatLakes basin, derived from satellite imagery (Wolter

et al. 2006). These coastal interfluves (n = 1,818),were then added to the network of stream water-sheds (n = 1,773), creating a total of 3,591 coastalwatersheds and interfluves (Fig. 4).

Watershed delineation approach 3a—streaminterfluves delineation. After coastal interfluvepolygons and stream watershed polygons werecombined into a single map layer, they wereordered and numbered along the coast from theU.S./Canada border in western Lake Superior,counter-clockwise to the easternmost polygonwhere Lake Ontario empties into the St. LawrenceRiver. Like the previous watershed delineations, theboundaries were visually assessed for errors andedited when necessary, using ancillary map dataincluding streams, aerial photos, and USGS-scanned quad maps (digital raster graphics 1:24,000and 1:250,000). Ordering the watersheds withsequential identification numbers along the coastprovided a framework for combining smallwatershed units and their related stressors along theentire U.S. Great Lakes shoreline. Watersheds andassociated stressors adjacent to an area of interestare easily identified by their consecutive ID

FIG. 4. The immediate watersheds (stream and interfluve) adjacent to ahigh-energy fish study site.

Multi-scale Watershed Delineations for Coastal Indicator Development 19

numbers (Fig. 5). These unique areas consisting ofthe shoreline and adjoining riverine watershedsprovide a means to summarize and assess theeffects of watershed-scale anthropogenic stressesdelivered to specific coastal ecosystems.

The relative contribution of stressor effects flow-ing from adjacent land areas along the coast variesgreatly based upon seasonal currents, storm andwind events, nearshore topography, and other localconditions. Spatial ordering of watersheds providesthe means to account for stressor effects contributedfrom outside a particular ecosystem’s immediatewatershed (e.g., by long-shore currents) (Fig. 5). Asstressor delivery mechanisms become better under-stood and mapped, this scalable watershed frame-work can be used to differentially weight thecontributions to a specific site from nearby water-sheds, based on their proximity and direction ofprevailing currents, to represent the lakeward deliv-ery of sediments, contaminants, and other water-borne stressors to these coastal areas. Thisframework might also be applied to circulationmodels for embayments and harbors or large lakes,providing an ordered link between stressors in thewatershed and the receiving body of water.

Watershed delineation approach 3b—agglomeratedstream delineation. Coastal interfluve areas arenumerous and generally quite small (85% are < 5km2 in area), yet important because they represent a

large percentage of the shoreline (all coastlinebetween stream mouths). Not wanting to discountor over-weight stressors contributed by theseinterfluve watersheds, we agglomerated them withstream watersheds to provide a spatially relevantunit over which stressors could be characterized.Ideally, we would have divided the interfluvewatersheds along a boundary, beginning and endinghalf way between stream mouths that would ascribea hydrologically-relevant portion of the interfluveto each stream watershed in much the same way asthe segment-sheds. However, there was no practicalway to accomplish this in an automated fashion,due to lack of a side-to-side topographic gradient.Therefore, we calculated area-weighted stressorsummaries for the interfluve stressors and assigned50% to each adjacent stream watershed. A total of1,773 agglomerated stream interfluve watershedunits resulted from this process.

Watershed delineation approach 3c—core high en-ergy. Since wind- and wave-exposed, high-energycoastal locations are subject to both land- and lake-based stressors, an alternative to traditional watersheddelineations is needed to account for stressor effectsthat originate from the land, but may be transportedby nearshore currents. To summarize stressors likelyto affect high energy sites, we agglomerated multiplewatersheds and their stressor scores, using only water-sheds immediately adjacent to each “core” high-energy site. Like the other watershed agglomerationsdescribed above, area-weighted means of the immedi-ately adjacent watershed stressors were used to calcu-late the combined stressor score for each “core” highenergy site.

Watershed delineation approach 3d—minimum areathreshold. For a broader-scale characterization ofstressors associated with high-energy sites, we alsoagglomerated multiple watersheds and their stressorvalues, until a minimum area threshold was reached(Fig. 6). The stream and interfluve watersheds adja-cent to the high-energy sample sites were agglomer-ated until a minimum area threshold was reached.Although any threshold size can be used, we set thethreshold at twice the area of the median size of the762 original segment-sheds (approach 1) (9 km2),providing a slightly broader watershed delineationfor each high-energy sample site. Watershed stres-sor summaries (area-weighted means) were com-

FIG. 5. Ordered and numbered stream andcoastal interfluve watersheds along the GreatLakes coastline near a high-energy site (HE).

20 Hollenhorst et al.

piled for these minimum area watersheds in thesame way as the core watersheds.

RESULTS AND DISCUSSION

Management of coastal ecosystems presents aparticular challenge compared to lakes or rivers, be-cause material and energy are transported to thecoast from both the land and the lake (or sea).Newly developed and now readily available tech-niques have enabled us to overcome some of the de-ficiencies of traditional watershed delineationmethods to facilitate studies of interactions betweenlandscapes and coastal ecosystems. Six differenttypes of coastal watersheds, delineated using threemain approaches, were found to be useful for quan-

tifying stressors affecting coastal ecosystems: 1)segment-sheds, 2) complex-specific site watersheds,3) stream/interfluve watersheds, 4) agglomerated-stream watersheds, 5) “core” high-energy area wa-tersheds, and 6) minimum area thresholdhigh-energy area watersheds. Three of these delin-eations provided watershed measures comprehen-sively across the entire U.S. Great Lakes coastline(methods 1, 3, and 4 above). To date the other twomethods have been applied to watersheds specificto sites sampled by investigators of the GLEI pro-ject, although these methods could be used to char-acterize additional sites in the future. Theadvantages and limitations of each watershed delin-eation approach are shown in Table 1.

Because the segment-shed and the merged

FIG. 6. Core watershed for immediate high-energy site watersheds andagglomerated watershed based on a minimum area threshold for a high-energy site.

Multi-scale Watershed Delineations for Coastal Indicator Development 21

stream/interfluve watershed approaches cover thefull U.S. side of the Great Lakes basin, the derivedstressor scores represent the entire sampling uni-verse of possible values for the U.S. Hence, the ex-treme values of the stressor distributions canpotentially be used to identify reference (definedhere as least disturbed; Host et al. 2005, Stoddard etal. 2006) as well as degraded “most disturbed”areas.

Size Frequency Distributions

Because large watersheds are likely to functiondifferently than small watersheds, we calculated thesize frequency distribution of watershed sizes foreach of the six approaches for the Great Lakescoastal areas. Across the six watershed delineationtypes, watershed sizes varied widely from small (<1 ha) coastal interfluve watersheds to large riverinewatersheds, e.g., the Maumee River basin of west-ern Lake Erie (16,837 km2). In general, the rangesof watershed size for each of the delineation typeswere similar, except for the minimum-area high en-ergy types. High-energy “minimum area threshold”watersheds had the largest median size, whereas the

stream/interfluve and complex specific watershedshad the smallest median size (Fig. 7). Segment-sheds and complex-specific watersheds were fairlycomparable in size and seemed to fall evenly be-tween the smallest and largest watershed types (Fig.7). This was likely due to the fact that segment-sheds were defined by second-order and higherstreams, and selection criteria for river-influencedwetland sites included second-order and higherstreams.

All three of the delineation techniques that en-compass the entire basin had many outliers in thesize frequency distribution, particularly on the largeside of the distribution (Outliers were defined asvalues that are greater than 1.5 times the middle50% range or either the bottom 25th or top 75thquantile). Stream/interfluve watersheds had thegreatest number of small drainage basins; bothlarge and small outlier values were observed. Theagglomerated-stream approach, which combinedthe coastal interfluve watersheds with the nearbystream watersheds, eliminated all outliers on thesmall side of the size distribution, leaving a limitednumber of large watersheds as outliers in a log-nor-

TABLE 1. Advantages and disadvantages of each watershed delineation approach.

Approach Type Extent Advantages Limitations Best use

Standard Segment-sheds Basin-wide Easy to calculate; Coarse resolution Site selection;watershed (n = 762) basin-wide generalized stressdelineation gradient, reference/

degraded designation

Standard Complex-specific Sampled sites High resolution Hard to calculate; Site characterizationwatershed (n = 344) not basin-widedelineation (sample dependent)

ArcHydro Stream/interfluve Basin-wide Easy to calculate; Produces many Site characterization;(n = 3,591) fine resolution; small units reference/degraded

scalable; basin- designationwide

ArcHydro Agglomerated- Basin-wide Easy to calculate; Small units Site characterization;stream fine resolution; reference/degraded(n = 1,773) scalable; basin- designation

wide

ArcHydro Core high-energy Sampled sites Easy to calculate; Not basin-wide Site characterization(n = 47) scalable (sample dependent)

ArcHydro High-energy min. Sampled sites Scalable; addresses Moderately difficult Site characterizationarea threshold long shore currents to calculate; not (n = 47) nearby tributaries basin-wide

etc. (sample dependent)

22 Hollenhorst et al.

mal distribution. These large watersheds could be asource of bias in some comparisons, since water-shed size can affect the delivery of anthropogenicstresses to coastal ecosystems. Trends in the rela-tionships between watershed size and the mecha-nisms by which stressors are delivered from thewatershed to the coastal ecosystem in question are atopic of continuing study.

Fortunately, the third approach described above

provides a method by which large watersheds canbe disaggregated into smaller units to address thisscale question. This could be accomplished by sub-dividing a large watershed into component tributarywatersheds and interfluves that flow to the river’smain stem. A similar technique could be used todisaggregate watersheds for embayments and har-bors or large lakes. The ability to scale watershedunits up or down in size is necessary because largewatersheds are likely to function differently thansmall watersheds. Watershed size affects base-flows, flow velocities, and fluvial processes (Bilbyand Ward 1989), which in turn affect the delivery ofanthropogenic stresses to coastal ecosystems.Future indicator development should consider therelationships between watershed size and the mech-anisms by which stressors are delivered from thewatershed to the coastal ecosystem in question.Watershed indicators of anthropogenic stress shouldalso be verified at multiple scales.

Wiens (1989) identified several ecosystem char-acteristics that he suggested depend upon the scaleat which they are measured (fine or broad). Theseinclude decreasing variability with increasing dataresolution, the number of variables important incorrelations, the potential for deriving generaliza-tions, and the form of the model (empirical, correla-tive) most appropriate for characterizing the system

FIG. 7. Average watershed size (ln) for the sixmethods of watershed delineation.

TABLE 2. General characteristics of various attributes of ecological systems andinvestigations at fine and broad scales of study. “Fine” and “broad” are definedrelative to the focus of a particular investigation and will vary between studies(Wiens 1989).

Scale Fine Broad

Number of variables important in correlations Many FewRate of processes or system change Fast SlowCapacity of system to track short-term

environmental variations High LowPotential for system openness High LowEffects of individual movements on patterns Large SmallType of heterogeneity Patch Landscape mosaicResolution of detail High lowSampling adequacy (intensity) Good PoorEffects of sampling error Large SmallExperimental manipulations Possible DifficultReplication Possible DifficultEmpirical rigor High LowPotential for deriving generalizations Low HighForm of models Mechanistic CorrelativeTestability of hypotheses High LowSurveys Quantitative QualitativeAppropriate duration of study Short Long

Multi-scale Watershed Delineations for Coastal Indicator Development 23

TA

BL

E3.

Wat

ersh

edde

lin

eati

onap

proa

ches

use

dby

Gre

atL

akes

Env

iron

men

tal

Indi

cato

rs(G

LE

I)pr

ojec

tin

vest

igat

ion

s.S

eete

xtfo

ra

desc

ript

ion

ofth

ede

lin

eati

onm

eth

ods.

Geo

mor

phic

type

sar

ew

etla

nds

(W),

emba

ymen

ts(E

B),

and

hig

h-e

ner

gysh

orel

ines

(HE

),an

dop

enw

ater

(OW

).

Str

ess

sum

mar

yav

aila

ble

Ind

ivid

ual

stre

ssor

sX

XX

XX

XP

Cs*

XX

XM

axre

l**

XX

XX

Inth

isis

sue

Str

esso

rty

pe

Res

pon

seva

riab

leG

eom

orp

hic

typ

es

Pet

erso

net

al.

XP

Cs

Sta

ble

Nis

otop

es(p

lank

ton

and

bent

hos)

W,E

B,H

EB

razn

eret

al.

XX

Indi

vidu

alst

ress

ors

Bir

ds,a

mph

ibia

ns,f

ish,

mac

roin

vert

ebra

tes,

Wdi

atom

s,ve

geta

tion

Tre

bitz

etal

.X

XP

Cs

Wat

erqu

alit

yW

Rea

vie

XX

XP

Cs

Dia

tom

s(d

iato

min

ferr

edw

ater

qual

ity)

W,E

B,H

EH

owe

etal

.X

PC

sB

irds

WK

iret

aet

al.

XX

XP

Cs

Dia

tom

s(d

iato

min

ferr

edw

ater

qual

ity)

W,O

WK

ang

etal

.X

PC

sM

acro

inve

rteb

rate

sW

,EB

,HE

Pri

ceet

al.

XP

Cs

Am

phib

ians

WB

haga

tet

al.

XP

Cs

Fis

hW

Oth

erpu

blic

atio

ns

Del

inea

tion

app

roac

h1

23a

3b3c

3dS

tres

sor

typ

eR

esp

onse

vari

able

Geo

mor

ph

icty

pes

Bha

gat

2005

XX

PC

sF

ish

W,E

B,H

EB

ourd

aghs

and

John

ston

2006

XP

Cs

Veg

etat

ion

WR

eavi

eet

al.2

006

XX

PC

sD

iato

ms

(dia

tom

infe

rred

Ww

ater

qual

ity)

,wat

erqu

alit

yB

razn

eret

al.2

007

XH

uman

dist

inde

xB

irds

,am

phib

ians

,fis

h,m

acro

inve

rteb

rate

s,W

diat

oms,

vege

tati

onD

anz

etal

.200

7X

PC

sW

ater

qual

ity,

fish

,bir

dsW

,EB

,HE

Han

owsk

iet

al.2

007

XB

irds

WM

orri

ceet

al.2

007

XP

Cs

Wat

erqu

alit

yW

Rea

vie

etal

.200

7X

XX

PC

s,m

axre

lD

iato

ms

(tax

onch

arac

teri

stic

s,di

atom

W,E

B,H

Ein

ferr

edw

ater

qual

ity,

mul

tim

etri

cin

dex)

Sgr

oet

al.2

007

XX

PC

sD

iato

ms

(dia

tom

infe

rred

wat

erqu

alit

y)W

Yur

ista

and

Kel

ly20

07X

PC

sW

ater

qual

ity,

plan

kton

OW

Bha

gat

etal

.in

prep

XX

XX

PC

,ind

ivst

ress

ors

Fis

hW

,EB

,HE

Bra

dyet

al.i

npr

epX

XX

XP

C,i

ndiv

stre

ssor

sM

acro

inve

rteb

rate

sW

,EB

,HE

John

son

etal

.in

prep

XX

XX

Max

-rel

,PC

s,F

ish,

habi

tat

W,E

B,H

Ein

div

stre

ssor

s

*D

anz

etal

.200

5;**

Hos

tet

al.2

005

Delineation type

1-Segment-sheds

2—Complex specificwatersheds

3a-Stream/interfluvedelineation

3b—Agglomeratedstream delineation

3c-Core high-energy

3d—Minimum areathreshold

24 Hollenhorst et al.

(Table 2). Although watersheds delineated at differ-ent scales may or may not exhibit these characteris-tics exactly, it is informative to consider them. Forinstance, small watersheds may be more subject tooutside influences (openness) than large water-sheds. Are the effects of sampling error larger withsmall watersheds than with large watersheds?Wiens (1989) suggested there may be domains ofscale wherein patterns change slowly or not at all.These domains are then separated by transitions inwhich system dynamics seem chaotic and patternschange quickly. The availability of tools to createflexible, multiple-scale, coastal watershed delin-eations may bring us one step closer to identifyingthe appropriate scales needed to best understand therelationship between environmental stress and eco-logic response.

The watershed delineations described above andtheir subsequent stressor summaries have been usedto evaluate stress response relationships for differ-ent taxa in a variety of coastal ecosystems. Many ofthose efforts are described in the following paperspresented in this special issue. Table 3 summarizesthose studies by taxonomic emphasis and indicatesthe watershed delineations used. Although mostgroups applied the segment-shed delineations forsite selection and many used the complex-specificwatershed delineations for site-specific characteri-zations, work continues using the detailedstream/interfluve delineations and agglomerationsfor predicting water quality, plankton, fish, and in-vertebrate communities.

SUMMARY

We used three approaches to delineate the contri-bution of water and anthropogenic materials fromland to aquatic habitats at Great Lakes coastal mar-gins. They have all proven useful in facilitating thedevelopment of indicators for the Great Lakes, par-ticularly in providing a means for a priori site se-lection across a stressor gradient, and incharacterizing the locally derived stresses on sitesthat had been sampled to estimate biological condi-tion. These alternative approaches have muchhigher resolution than our initial segment-shed de-lineations, and permit the stressor data to be queriedand scaled to a particular area of interest using area-weighted means. Because of their scalability, webelieve stream/interfluve watersheds, and their as-sociated stressor summaries will be useful for manyapplications, in addition to indicator development,including setting conservation priorities for mini-

mally-disturbed areas, developing restoration end-points, and prioritizing management actions.

A limitation of the work to date is that it ad-dresses only the U.S. side of the Great Lakes. Scal-able stream/interfluve watersheds are currentlybeing derived for the Canadian portion of the GreatLakes and will provide for an integrated binationalset of detailed watersheds for the entire Great Lakesbasin. A complete basin-wide inventory of GreatLakes watersheds, combined with unified landcover and stressor summaries, will ultimately pro-vide a high resolution “scalable” and consistentcharacterization of the entire Great Lakes anthro-pogenic stressor gradient. A data set that quantifiesthe stressor gradient seamlessly across the entirebasin will greatly facilitate our ability to develop bi-national management for the largest freshwater sys-tem in the world.

ACKNOWLEDGMENTS

We would like to thank Nick Danz, Connie Host,Paul Meysembourg, Jennifer Olker, and GeraldSjerven who helped compile and analyze essentialdata sets for this project. This research has beensupported by a grant from the U.S. EnvironmentalProtection Agency’s Science to Achieve Results Es-tuarine and Great Lakes program through fundingto the Great Lakes Environmental Indicators project(U.S. EPA Agreement EPA/R-8286750) and theReference Condition project (U.S. EPA AgreementEPA/R-82877701-0). Although the research de-scribed in this article has been funded in part by theU.S. EPA through the cooperative agreement -, ithas not been subjected to the agency’s required peerand policy review and therefore does not necessar-ily reflect the views of the agency and no officialendorsement should be inferred. This is contribu-tion number 470 of the Center for Water and theEnvironment, Natural Resources Research Institute,University of Minnesota Duluth.

REFERENCESBhagat, Y. 2005. Fish indicators of anthropogenic stress

at Great Lakes coastal margins: multimetric and mul-tivariate approaches. M.Sc. thesis, University ofWindsor, Ontario, Canada.

———, Ciborowski, J.J.H., Johnson, L.B., Uzarski, D.G.,Burton, T.M., Timmermans, S.T.A., and Cooper, M.J.2007. Testing a fish index of biotic integrity forresponses to different stressors in Great Lakes coastalwetlands. J. Great Lakes Res. 33 (Special Issue3):224–235.

Multi-scale Watershed Delineations for Coastal Indicator Development 25

Bilby, R.E., and Ward, J.M. 1989. Changes in character-istics and function of woody debris with increasingsize of streams in western Washington. T. Am. Fish.Soc. pp. 368–378.

Bloschl, G., and Sivapalan, M. 1995. Scale issues inhydrological modeling: a review. Hydrol. Process.9:251–290.

Bourdaghs, M., and Johnston, C.A. 2006. Floristic qual-ity index in Great Lakes coastal wetlands. Wetlands26:718–735.

Brazner, J.C., Danz, N.P., Trebitz, A.S., Niemi, G.J.,Regal, R.R., Hollenhorst, T.P., Host, G.E., Reavie,E.D., Brown, T.N., Hanowski, J.M., Johnston, C.A.,Johnson, L.B., Howe, R.W., and Ciborowski, J.J.H.2007a. Responsiveness of Great Lakes wetland indi-cators to human disturbances at multiple spatialscales: A multi-assemblage assessment. J. GreatLakes Res. 33 (Special Issue 3):42–66.

———, Danz, N.P., Niemi, G.J., Regal, R.R., Trebitz,A.S., Howe, R.W., Hanowski, J.M., Johnson, L.B.,Ciborowski, J.J.H., Johnston, C.A., Reavie, E.D.,Brady, V.J., Sgro, G.V. 2007b. Evaluation of geo-graphic, geomorphic and human influences on GreatLakes wetland indicators: a multi-assemblageapproach. Ecol. Indicators 7:610–635.

Danz, N.P., Regal, R.R., Niemi, G.J., Brady, V.J., Hol-lenhorst, T.P., Johnson, L.B., Host, G.E., Hanowski,J.M., Johnston, C.A., Brown, T.N., Kingston, J.C.,and Kelly, J.R. 2005. Environmentally stratified sam-pling design for the development of Great Lakes envi-ronmental indicators. Environ. Monit. Assess.102:41–65.

——— , Niemi, G.J., Regal, R.R., Hollenhorst, T.P.,Johnson, L.B., Hanowski, J.M., Johnston, C.A.,Axler, R.P., Ciborowski, J.J.H., Hrabik, T., Brady,V.J., Kelly, J.R., Brazner, J.C., and Howe, R.W. 2007.Integrated measures of anthropogenic stress in theU.S. Great Lakes basin. Environ. Manage. 39:631–647.

Diana, M.J., Allan, J.D., and Infante, D.M. 2006. Theinfluence of physical habitat and land use on streamfish assemblages in southeastern Michigan. In Land-scape influences on stream habitats and biologicalassemblages, R.M. Hughes, L. Wang, and P.W. Seel-bach, eds. Bethesda, Maryland: American FisheriesSociety.

ESRI. 2000. ArcGIS, Environmental Systems ResearchInstitute. Redlands, California: ESRI Press.

Gilliam, J.W., Osmond, D.L., and Evans, R.O. 1997.Selected agricultural best management practices tocontrol nitrogen in the Neuse River basin. North Car-olina Agricultural Research Service technical bulletin311, North Carolina State University, Raleigh, NorthCarolina.

Hanowski, J.M., Danz, N.P., Howe, R.W., Niemi, G.J.,and Smith, C.R. 2007. Considerations of scale and

space in the development of Great Lakes coastal wet-land bird indicators. EcoHealth 4:194–205.

Host, G.E., Schuldt, J., Ciborowski, J.J.H., Johnson,L.B., Hollenhorst, T.P., and Richards, C. 2005. Use ofGIS and remotely sensed data for a priori identifica-tion of reference areas for Great Lakes coastal ecosys-tems. Int. J. Remote. Sens. 26(23):5325–5342.

Howe, R.W., Regal, R.R., Hanowski, J.M., Niemi, G.J.,Danz, N.P., and Smith, C.R. 2007. An index of eco-logical condition based on bird assemblages in GreatLakes coastal wetlands. J. Great Lakes Res. 33 (Spe-cial Issue 3):93–105.

International Joint Commission. 1995. Practical steps toimplement an ecosystem approach in Great Lakesmanagement. A workshop co-sponsored by U.S. Envi-ronmental Protection Agency and EnvironmentCanada, in cooperation with the International JointCommission and Detroit, MI, Wayne State Univer-sity. http://www.ijc.org/php/publications/html/toc.html. Accessed 25 Sept 07.

———. 2005 A discussion paper on the internationalwatersheds initiative. Second report to the govern-ments of Canada and the United States .http://www.ijc.org/php/publications/pdf/ID1582.pdf.Accessed 25 Sept 07.

Johnston, C.A., Brown, T.P., Hollenhorst, T.P., Wolter,P., Danz, N.P., and Niemi, G.J. 2006. GIS in supportof ecological indicator development. In Manual ofgeographic information systems, M. Madden, ed.Bethesda, Maryland: American Society for Pho-togrammetry and Remote Sensing, in press.

Kang, M., Ciborowski, J.J.H., and Johnson, L.B. 2007.The influence of anthropogenic disturbance and envi-ronmental suitability on the distribution of the non-indigenous amphipod Echinogammarus ischnus atLaurential Great Lakes coastal margins. J. GreatLakes Res. 33 (Special Issue 3):198–210.

Kireta, A.R., Reavie, E.D., Danz, N.P., Axler, R.P.,Sgro, G.V., Kingston, J.C., Brown, T.N., and Hollen-horst, T.P. 2007. Coastal geomorphic and lake vari-ability in the Laurentian Great Lakes: Implications fora diatom-based monitoring tool. J. Great Lakes Res.33 (Special Issue 3):136–153.

Knebl, M.R., Yang, Z.-L., Hutchison, K., and Maidment,D.R. 2005. Regional scale flood modeling usingNEXRAD rainfall, GIS, and HEC-HMS/RAS: a casestudy for the San Antonio River basin summer 2002storm event. J. Environ. Manage. 75:325–336.

Magner, J.A., and Brooks, K.N. 2007 Integrating sen-tinel watershed-systems into the monitoring andassessment of Minnesota’s (USA) waters quality.Environ. Monit. Assess. http://www.springerlink.com/content/11686v2t1v734520/fulltext.pdf. Accessed 1Oct 07.

Maidment, D.R. 1997. AGREE-DEM surface recondi-tioning system. http://www.ce.utexas.edu/prof/

26 Hollenhorst et al.

maidment/gishydro/ferdi/research/agree/agree.html.Accessed 25 Sept 07.

——— , and Morehouse, S. 2002. Arc hydro: GIS forwater resources. Redlands, California: ESRI Press.

Morrice, J.A., Danz, N.P., Regal, R.R., Kelly, J.R.,Niemi, G.J., Reavie, E.D., Hollenhorst, T.P., Axler,R.P., A.S. Trebitz, A.S., Cotter, A.M., and Peterson,G.S. 2007. Human influence on water quality in GreatLakes coastal wetlands. J. Environ. Manage. in press.

Niemi, G.J., and McDonald, M.E. 2004. Application ofecological indicators. Annu. Rev. Ecol. Evol. Syst.35:89–111.

———, Wardrop, D., Brooks, R., Anderson, S., Brady,V., Paerl, H., Rakocinski, C., Brouwer, M., Levinson,B., and McDonald, M. 2004. Rationale for a new gen-eration of ecological indicators for coastal waters.Environ. Health Perspect. 112(9):979–986.

———, Kelly, J.R., and Danz, N.P. 2007. Environmentalindicators for the coastal region of the North Ameri-can Great Lakes: Introduction and prospectus. J.Great Lakes Res. 33 (Special Issue 3):1–12.

Peterson, G.S., Sierszen, M.E., Yurista, P.M., and Kelly,J.R. 2007. Stable nitrogen isotopes of plankton andbenthos reflect a landscape-level influence on GreatLakes coastal ecosystems. J. Great Lakes Res. 33(Special Issue 3):27–41.

Price, S.J., Howe, R.W., Hanowski, J.M., Regal, R.R.,Niemi, G.J., and Smith, C.R. 2007. Are anurans ofGreat Lakes coastal wetlands reliable indicators ofecological condition? J. Great Lakes Res. 33 (SpecialIssue 3):211–223.

Reavie, E.D. 2007. A diatom-based water quality modelfor Great Lakes coastlines. J. Great Lakes Res. 33(Special Issue 3):86–92.

———, Axler, R.P., Sgro, G.V., Kingston, J.C., Kireta,A.R., Brown, T.N., Danz, N.P., Hollenhorst, T.P., andFerguson, M.J. 2006. Diatom-based weighted-averag-ing transfer functions for Great Lakes coastal waterquality: relationships to watershed characteristics. J.Great Lakes Res. 32:321–347.

———, Kereta, A.R., Kingston, J.C., Sgro, G.V., Danz,

N.P., Axler, R.P., and Hollenhorst, T.P. 2007. Com-parisons of simple and multimetric diatom-basedindices for Great Lakes coastline disturbance. J. Phy-col. in press.

Sgro, G.V, Reavie, E.D., Kingston, J.C., Kireta, A.R.,Ferguson, M.J., Danz, N.P., and Johansen, J.J. 2007.A diatom quality index from a diatom-based totalphosphorus inference model. Environ. Bioindicators2:15–34.

Stoddard, J.L., Larsen, D.P., Hawkins, C.P., Johnson,R.K., and Norris, R.H. 2006. Setting expectations forthe ecological condition of running waters: the con-cept of reference condition. Ecol. Appl. 16(4):1267–1276.

Tong, S.T.Y., and Chen, W. 2002. Modeling the rela-tionship between land use and surface water quality.J. Environ. Manage. 66:377–393.

Trebitz, A.S., Brazner, J.C., Cotter, A.M., Knuth, M.L.,Morrice, J.A., Peterson, G.S., Sierszen, M.E., Thomp-son, J.A., and Kelly, J.R. 2007. Water quality in GreatLakes coastal wetlands: Basin-wide patterns andresponses to an anthropogenic disturbance gradient. J.Great Lakes Res. 33 (Special Issue 3):67–85.

Wiens, J. 1989. Spatial scaling in ecology. Funct. Ecol.3:385–397

Wolter, P., Johnston, C.A., and Niemi, G.J. 2006. Landuse land cover change in the Great Lakes basin1992–2001. J. Great Lakes Res. 32:607–628.

Wu, J., and Li, H. 2006. Concepts of scale and scaling.In Scaling and uncertainty analysis in ecology: meth-ods and applications, eds. J. Wu, K.B. Jones, H. Li,and O.L. Loucks. Dordrecht, The Netherlands:Springer.

Yurista, P.M, and Kelly, J.R. 2007. Spatial patterns ofwater quality and plankton from high-resolution con-tinuous in situ sensing along a 537-km nearshore tran-sect of western Lake Superior, 2004. Aqua. Ecosys.Health Manage. in press.

Submitted: 13 October 2006Accepted: 3 October 2007Editorial handling: John R. Kelly