Thesis Precfarming 2006

download Thesis Precfarming 2006

of 225

Transcript of Thesis Precfarming 2006

  • 7/31/2019 Thesis Precfarming 2006

    1/225

    Precision Farming in the Small Farmland in the Eastern

    Nile Delta Egypt Using Remote Sensing and GIS

    Inaugural-Dissertation zurErlangung der Doktorwrde

    der Fakultt fr Forst- und Umweltwissenschaftender Albert-Ludwigs-Universitt

    Freiburg im Breisgau

    Vorgelegt vonAbdelaziz Belal Abdel Elmontalbe Belalaus gypten

    Freiburg im Breisgau2006

  • 7/31/2019 Thesis Precfarming 2006

    2/225

  • 7/31/2019 Thesis Precfarming 2006

    3/225

    Acknowledgements I

    ACKNOWLEDGEMENTS

    I am greatly indebted to Prof. Dr. B. Koch for accepting me as a PhD student at

    the Department of Remote Sensing and Landscape Information Systems

    (FELIS), University of Freiburg and for her guide; encouragement and supportthroughout this study.

    I also want to thank Prof. Dr. R. Doluschitz for his supervision, contributions,

    and support during my studies and work at the Institute for Farm

    Management, University of Hohenheim, Stuttgart.

    My sincere thanks are also for Prof. Dr. Dr. h. c. Dieter R. Pelz for agreeing to

    act as a second examiner.

    Sincerely, I thank Prof. Dr. Georg Bareth and Dr. Rainer Laudien at the

    Institute for Farm Management, for their valuable and critical comments that

    enhanced my capability to come up with a result oriented research. I am also

    grateful to Dr. Mathias Dees and Dr. Claus Peter Gross at FELIS for their

    valuable comments during my study.

    Deep appreciation and thanks is also dedicated to Mr. Siwe Ren at FELIS for

    his friendship and help in editing the English grammar and style and to Mr.

    Markus Jochum for the translation of the German summary.

    Special Thanks to Dr. Naceur Saidani and Dr. Wahynni Ilham, at FELIS for

    their friendship and support for the continuing discussions on topics of remote

    sensing and GIS.

    I also want to express my thanks to Dr. Eva Ivits-Wasser, Dr. Gernot

    Ramminger, Mr. Oliver Diedershagen, Mr. Christian Schill, Mr. Filip Langer and

    Mr. Octavian Iercan at FELIS, for their readiness to help and fruitful

    discussions which contributed to the success of this work.

    Most importantly, I remain grateful to Mrs. Roswitha Lange and Mr. Markus

    Quinten (Network administrator) at FELIS for their always charming and

    helpful hand.

  • 7/31/2019 Thesis Precfarming 2006

    4/225

    Acknowledgements II

    I thank all the staff at Department of Remote Sensing and Landscape

    Information Systems (FELIS), University of Freiburg and Institute for Farm

    Management, University Hohenheim, Stuttgart. I feel privileged to be the

    recipient of all the many new things that I have learnt and for the wonderful

    working atmosphere, which was positively reflected in the smooth running of

    this work.

    During the field work in Egypt, many institutions and people have supported

    me. I extend my sincere thanks to them. I am much indebted to the National

    Authority for Remote Sensing and Space Sciences (NARSS) staff for providing

    me all kinds of support required during the field work. Also, special gratitude

    goes to the staff of the laboratory of the Soil Unit Department in the NationalResearch Centre (NRC) their timely analysing the soil samples, which were an

    important input data of my research work.

    Also I wish to thank the France Space Agency for making available the SPOT 5

    multi-spectral (2.5 m high resolution) data through the program Incentive for

    the Scientific use of Images from SPOT Systems (ISIS).

    Lastly, I am grateful that the Egyptian government gave me the opportunity to

    pursue the PhD study in Germany through the Egypt mission fellowship

    program.

    I would deeply like to express my respect and appreciation to all brothers and

    sister for their inspiration, encouragement and support during the execution of

    this work. I acknowledge a strong sense of appreciation to the souls of myparents for their patience and encouragement during their lives.

    At last my heartfelt love and affection goes to my beloved wife Amira Monuir Ali

    Taha Rizk and my daughters Rofaida and Eman for their inspiration, support

    and understanding of their responsibility during this work. I am deeply

    indebted to them for the time I did not spend with them.

  • 7/31/2019 Thesis Precfarming 2006

    5/225

    DEDICATEDTO

    SOULS MY PARENTS,

    MY WIFE

    AND MY DAUGTHERS (ROFAIDA AND EMAN)

  • 7/31/2019 Thesis Precfarming 2006

    6/225

    Table of contents IV

    TABLE OF CONTENTS

    ACKNOWLEDGEMENTS........................................................................................ I

    LIST OF TABLES.............................................................................................. VIII

    LIST OF FIGURES............................................................................................... X

    LIST OF APPENDICES........................................................................................ XI

    LIST OF ABBREVIATION ...................................................................................XII

    1. INTRODUCTION...............................................................................................1

    1.1.AGRICULTUREINEGYPT ................................................................................. 1

    1.1.1. BACKGROUND......................................................................................... 1

    1.1.2. AGRICULTURAL POLICY .......................................................................... 3

    1.1.3. FARM PROBLEMS....................................................................................4

    1.2.NEWAGRICULTUREMANAGEMENTTECHNOLOGY......................................... 61.3.OBJECTIVESOFTHERESEARCHWORK .........................................................7

    1.4.STRUCTUREOFTHETHESIS...........................................................................8

    2. REVIEW OF LITERATURE ..............................................................................10

    2.1.PRECISIONFARMING .................................................................................... 10

    2.2.REMOTESENSINGINPRECISIONFARMING .................................................. 11

    2.3.REMOTESENSINGANDIMAGEINTERPRETATION......................................... 14

    2.3.1. OBJECT-BASED IMAGE ANALYSIS ........................................................ 15

    2.3.1.1. Image segmentation............................................................................ 15

    2.3.1.2. Classification ...................................................................................... 18

    2.3.1.3. Accuracy assessment.......................................................................... 19

    2.3.2. POST CLASSIFICATION COMPARISON CHANGE DETECTION ................ 21

    2.4.SOILFERTILITYANDCROPPRODUCTION...................................................... 22

    2.5.CROPGROWTHMODELLING .........................................................................27

    2.5.1. STATISTICAL MODELLING (MULTIPLE LINEAR REGRESSION MODEL).. 28

    2.5.1.1. Model ................................................................................................. 29

    2.5.1.2. Assumptions ...................................................................................... 30

    2.5.1.3. Goodness-of-fit ................................................................................... 32

    2.5.1.4. ANOVA (Analysis Of Variance)............................................................. 33

    2.5.1.5. Subset selection.................................................................................. 34

    2.5.1.6. Hypothesis testing .............................................................................. 35

    2.5.2. SPATIAL MODELLING ............................................................................ 37

    3. MATERIALS...................................................................................................40

    3.1.DESCRIPATIONOFTHESTUDYAREA............................................................ 40

    3.1.1. LOCATION.............................................................................................. 40

    3.1.2. CLIMATE................................................................................................ 42

  • 7/31/2019 Thesis Precfarming 2006

    7/225

    Table of contents V

    3.1.3. GEOMORPHOLOGY................................................................................ 43

    3.1.4. GEOLOGY .............................................................................................. 44

    3.1.5. SOILS . ......................................................................................... 45

    3.1.6. CROP GROWTH AND ROTATION ............................................................ 46

    3.1.7. WATER RESOURCES - IRRIGATION AND DRAINAGE .............................483.1.7.1. IRRIGATION SYSTEMS ....................................................................... 48

    3.1.7.2. DRAINAGE SYSTEMS ......................................................................... 50

    3.2.REMOTESENSINGDATA ............................................................................... 52

    3.3.MAPS .............................................................................................................53

    3.4.SOFTWARE.................................................................................................... 53

    4. METHODS .....................................................................................................54

    4.1.REMOTESENSINGANDIMAGEINTERPRETAION........................................... 55

    4.1.1. IMAGE PRE-PROCESSING...................................................................... 554.1.1.1. Geometric correction........................................................................... 55

    4.1.1.2. Image enhancement............................................................................ 56

    4.1.2. OBJECT BASED IMAGE ANALYSIS......................................................... 57

    4.1.2.1. Image segmentation............................................................................ 58

    4.1.2.2. Classification ...................................................................................... 62

    4.1.2.3. Accuracy assessment.......................................................................... 65

    4.1.3. POST CLASSSIFICATION COMPARISON CHANGE DETECTION .............. 66

    4.2.FIELDWORK&LABORATORYANALYSIS ....................................................... 674.2.1. TEST AREA ONE ....................................................................................69

    4.2.2. TEST AREA TWO.................................................................................... 70

    4.2.3. TEST AREA THREE ................................................................................ 71

    4.2.4. QUESTIONNAIRE (INTERVIEW WITH FARMERS).................................... 72

    4.2.5 SOIL AND WATER SAMPLE COLLECTION ............................................... 73

    4.2.6. LABORATORY ANALYSIS........................................................................ 75

    4.2.6.1. Soil analysis ....................................................................................... 75

    4.2.6.2. Surface and ground water analysis ..................................................... 764.3.CROPGROWTHMODELLING .........................................................................77

    4.3.1. DESCRIPTIVE STATISTICS .....................................................................77

    4.3.2. STATISTICAL MODELLING (MULTIPLE REGRESSION MODELS)............. 78

    4.3.2.1. Model ................................................................................................. 78

    4.3.2.2. Assumptions ...................................................................................... 78

    4.3.2.3. Goodness-of-fit ................................................................................... 79

    4.3.2.4. Bivariate data analysis........................................................................ 80

    4.3.2.5. ANOVA for homogeneity test ...............................................................804.3.2.6. Subset selection.................................................................................. 80

    4.3.3. SPATIAL MODELLING ............................................................................ 81

  • 7/31/2019 Thesis Precfarming 2006

    8/225

    Table of contents VI

    5. RESULTS .......................................................................................................84

    5.1.REMOTESENSINGANDIMAGEINTERPRETATION......................................... 84

    5.1.1. OBJECT BASED IMAGE ANALYSIS......................................................... 84

    5.1.2. POST CLASSIFICATION COMPARISON CHANGE DETECTION ................ 95

    5.2.FIELDSURVEYANDLABORATORYANALYSIS................................................ 995.2.1. SOIL ANALYSIS MORPHOLOGY AND CHEMICAL PROPERTIES............ 99

    5.2.2. WATER ANALYSIS ................................................................................ 109

    5.2.2.1. Surface water ................................................................................... 110

    5.2.2.2. Ground water ................................................................................... 111

    5.2.2.3. Salinity and alkalinity hazard for surface and ground water .............. 111

    5.3.CROPGROWTHMODELLING ....................................................................... 115

    5.3.1. Descriptive Statistics ............................................................................ 115

    5.3.1.1. Cotton and rice yield in test areas one and two ................................. 1165.3.1.2. Relationship between crop yield and soil properties ........................... 118

    5.3.3. STATISTICS MODELLING (MULTIPLE LINEAR REGRESSIONS MODEL) .................................................................................. 126

    5.3.3.1. The production model for cotton crop in test area one ....................... 127

    5.3.3.2. The production model for cotton crop in test area two ....................... 127

    5.3.3.3. The production model for rice crop in test area one ........................... 128

    5.3.3.4. The production model for rice crop in test area two ........................... 129

    5.3.4. VALIDATION OF THE STATISTICAL MODELLING ................................. 129

    5.3.4.1. Cotton models validation in test areas one and two. .......................... 129

    5.3.4.2. Rice crop models validation in test areas one and two ....................... 131

    5.3.5. SPATIAL MODELLING .......................................................................... 132

    5.3.5.1 Spatial model for cotton crop in test areas one and two...................... 132

    5.3.5.2. Spatial model for rice crop in test areas one and two ......................... 135

    5.4.SUGGESTIONTHEMANAGEMENTPLAN...................................................... 137

    5.4.1. PROPOSED MANAGEMENT PLAN FOR DECISION MAKER ................... 137

    5.4.1.1. Site specific nutrient management plan ............................................ 138

    5.4.1.2 Farmer school.................................................................................... 144

    5.4.2. PROPOSED MANAGEMENT PLAN FOR THE FARMERS......................... 144

    6. DISCUSSION ................................................................................................145

    6.1.REMOTESENSINGANDIMAGEINTERPRETATION....................................... 145

    6.1.1. OBJECT BASED CLASSIFICATION ....................................................... 146

    6.1.2. POST CLASSIFICATION COMPARISON CHANGE DETECTION .............. 149

    6.2.SOILSAMPLINGTECHNIQUES..................................................................... 150

    6.3.SOILANALYSISMORPHOLOGYANDCHEMICALPROPERTIES................... 151

    6.4.SURFACEANDGROUNDWATERSAMPLESANALYSIS ................................. 154

    6.5.SALINITYANDALKALINITYHAZARDFORSURFACEANDGROUNDWATER . 155

  • 7/31/2019 Thesis Precfarming 2006

    9/225

    Table of contents VII

    6.6.STATISTICALMODELLING(MULTIPLELINEARREGRESSIONSMODEL) .... 155

    6.7.MODELSVALIDATION.................................................................................. 157

    6.8.ARCVIEWSSPATIALANALYST(MODELBUILDER)ANDSPATIALMODELING 158

    7. SUMMARY ...................................................................................................160

    8. REFRENCES ................................................................................................165

    9. APPENDICES ...............................................................................................182

    10. ZUSAMMENFASSUNG ...............................................................................211

  • 7/31/2019 Thesis Precfarming 2006

    10/225

    List of tables VIIILIST OF TABLES

    Table 1. 1: Field crops: areas, yields and returns, 2002/03 ..................................... 2

    Table 3.1: Climatologic data of the study area (monthly averages in 2004) from the

    Climatic Centre of Agriculture, Ministry of Agriculture, Giza, Egypt. ......... 43Table 3. 2: Geological formation in the study area ................................................. 45

    Table 3. 3: Three crop rotation in the investigated area..........................................47

    Table 3. 4: Two year crop rotation in the investigated area..................................... 47

    Table 3. 5: Characteristics of ETM+ data ...............................................................52

    Table 3. 6: Characteristics of TM data ...................................................................53

    Table 3. 7: Characteristics of SPOT 5 data............................................................. 53

    Table 4. 1: Variation in soil types and management processes in the three

    test areas ............................................................................................ 68Table 4. 2: Distribution of the soil profile and surface soil samples ........................74

    Table 5. 1: Accuracy assessment for land use and land cover for TM data in 1990 forstudy area ........................................................................................... 89

    Table 5. 2: Accuracy assessment for land use and land cover for ETM data in 1999for study area ...................................................................................... 90

    Table 5. 3: Accuracy assessment for land use and land cover for SPOT 5 data in2004 for study area ............................................................................. 90

    Table 5. 4: Producer and user accuracy for TM in 1990 for the study area ............. 91

    Table 5. 5: Producer and user accuracy for ETM+ in 1999 for the study area ......... 91Table 5. 6: Producer and user accuracy for SPOT 5 in the study area ....................91

    Table 5. 7: Accuracy measures for classification results (main classes) for SPOT 5 inthe test area one.................................................................................. 92

    Table 5. 8: Producer, user, overall accuracy and KAPPA statistic for the main classesin the test area one.............................................................................. 92

    Table 5. 9: Accuracy measures for classification results (subclasses) for SPOT 5 inthe test area one................................................................................ 93

    Table 5. 10: Producer, user, overall accuracy and KAPPA statistic for the subclasses in

    the test area one.................................................................................. 93Table 5. 11: Accuracy measures for classification results (main classes) for SPOT 5 in

    the test area two.................................................................................. 93

    Table 5. 12: Producer, user, overall accuracy and KAPPA statistic for the main classesin the test area two.............................................................................. 93

    Table 5. 13: Accuracy measures for classification results (subclasses) for Spot 5 in thetest area two........................................................................................ 93

    Table 5. 14: Producer, user, overall accuracy and KAPPA statistic for the subclasses inthe test area two.................................................................................. 94

    Table 5. 15: Accuracy measures for classification results (main classes) for Spot 5 inthe test area three ............................................................................... 94

  • 7/31/2019 Thesis Precfarming 2006

    11/225

    List of tables IXTable 5. 16: Producer, user, overall accuracy and KAPPA statistic for the main classes

    in the test area three ........................................................................... 94

    Table 5. 17: Accuracy measures for classification results (subclasses) for SPOT 5 inthe test area three ............................................................................... 94

    Table 5. 18: Producer, user, overall accuracy and KAPPA statistic for the subclasses in

    the test area three ............................................................................... 95Table 5.19: Land use and land cover areas by hectare (ha) and percent ..................95

    Table 5.20: Change area (hectares) and rate (%) of land use and land cover in thestudy area ........................................................................................... 96

    Table 5.21: Minimum, maximum, mean and Std. deviation for soil analysis samplesin test area one (in cotton and rice fields)........................................... 100

    Table 5.22: Minimum, maximum, mean and Std. deviation for soil analysis samplesin test area two (in cotton and rice fields) ........................................... 101

    Table 5. 23: Minimum, maximum, mean and Std. deviation for soil analysis samples

    in test area three (in cotton and rice fields)......................................... 101Table 5. 24: Relationship between SP and soil texture ........................................... 103

    Table 5. 25: The relationships between ECe and plant growth................................ 105

    Table 5. 26: Analysis of surface water samples in the test areas............................. 110

    Table 5. 27: Ground water analysis for test area one ............................................. 113

    Table 5. 28: Salinity and alkalinity hazards for the surface water in the test areas. 114

    Table 5. 29: Salinity and alkalinity hazards for the surface water in the test areas. 115

    Table 5. 30: Rotated Component Matrix (a) for soil samples in cotton field in test area

    one.................................................................................................... 124Table 5. 31: KMO and Bartlett`s test for cotton and rice field in test areas one and

    two.................................................................................................... 124

    Table 5. 32: Rotated Component Matrix (a) for soil samples in cotton field in test areatwo.................................................................................................... 125

    Table 5. 33: Rotated Component Matrix (a) for soil samples in Rice field in test areaone.................................................................................................... 126

    Table 5. 34: Rotated Component Matrix (a) for soil samples in rice field in test areatwo.................................................................................................... 126

  • 7/31/2019 Thesis Precfarming 2006

    12/225

    List of figures X

    LIST OF FIGURES

    Figure 2. 1: Nutrient availability and microbial activity as affected by soil pH; thewider the band, the greater the availability or activity. (from USDAYear book 0f agriculture, 1947-47)...................................................24

    Figure 3. 1: Location of the Study area................................................................40

    Figure 3. 2: Growth of [a] Sabkha, [b] Cotton, [c] Rice and [d] Clover in the study.41

    Figure 3. 3: Meteorological data for the study area. .............................................43

    Figure 3.4: Irrigation system flow schematic (representation of the surfaceirrigation) ........................................................................................49

    Figure 3.5: Subsurface drain systems in the study area......................................51

    Figure 4. 1: Flowchart of the methodological step applied in this research ..........54

    Figure 4. 2: Different scale parameter depending on the different imageresolution........................................................................................60

    Figure 4. 3: Scale parameters in the small area (test areas) for high resolutiondata ...............................................................................................61

    Figure 4. 4: Scatter plot for memberships of object classes separated bydiscriminant analysis [A] TM and [B] SPOT 5 ...................................63

    Figure 4. 5: Steps of the accuracy assessment.....................................................66

    Figure 4. 6: Steps Post-classification comparison change detection......................67

    Figure 4. 7: Ground water table in soil profile in test area one. ............................70

    Figure 4. 8: Hardpan in soil profile in test area two .............................................71

    Figure 4. 9: Soil profile in the test area three .......................................................72

    Figure 4. 10: Distribution of soil samples in test area one.....................................74

    Figure 4. 11: ModelBuilder diagram for cotton in the test area one .......................83

    Figure 5. 1: (A, B and C) Land use and land cover maps in the three periods 1990,1999 and 2004 respectively .............................................................86

    Figure 5. 2: (A, B and C) land use classification in the three test areas (test areaone, test area two and test area three respectively) in 2004 ..............88

    Figure 5.3: Land use and land cover change increase and decrease from 1990,1990 and 2004................................................................................97

    Figure 5. 4: (A, B and C) Land use and land cover changes maps in three different

    periods (1990-1999, 1990-2004 and 1999-2004 respectively)...........99Figure 5.5: Histogram for cotton and rice yield in test areas one and two..........117

    Figure 5. 6: (A and B) Scatter plot for the cotton prediction yield versus reportedyield fitted with the regression line for test areas one and tworespectively ...................................................................................130

    Figure 5. 7: (A and B) Scatter plot for the rice prediction yield versus reported yieldfitted with the regression line for test areas one and two respectively.....................................................................................................132

    Figure 5. 8: Cotton yield prediction map in the test area one .............................134

    Figure 5. 9: Cotton yield prediction map in the test area two .............................135

    Figure 5. 10: Rice yield prediction map in the test area one ................................136

    Figure 5. 11: Rice yield prediction map in the test area two ................................137

  • 7/31/2019 Thesis Precfarming 2006

    13/225

    List of Appendices XILIST OF APPENDICES

    APPENDIX1....................................................................................................... 182

    Field data collection (Questionnaire) ............................................................... 182

    APPENDIX 2 ......................................................................................................... 190Table 1: Soil sample analysis in the test area one ........................................... 190

    Table 2: Soil sample analysis for test area two ................................................ 191

    Table 3: Soils samples analysis for test area three........................................... 192

    APPENDIX 3 ......................................................................................................... 193

    Appendix 3a: Relationships between cotton yield and soil properties in testarea one .................................................................................. 193

    Appendix 3b: Relationships between cotton yield and soil properties in testarea two .................................................................................. 196

    Appendix 3c: Relationships between rice yield and soil properties in testarea one .................................................................................. 199

    Appendix 3d: Relationships between rice yield and soils properties in testarea two .................................................................................. 202

    APPENDIX 4 ......................................................................................................... 205

    The screeplot for the cotton soil samples in test area one ................................ 205

    The screeplot for the cotton soil samples in test area two ................................ 205

    The screeplot for rice soil samples in test area one.......................................... 206

    The Scree plot for rice soil samples in test area two ........................................ 206

    APPENDIX 5 ......................................................................................................... 207Appendix 5a: Output results for production cotton crop model in test

    area one .................................................................................. 207

    Appendix 5b: Output results for production cotton crop model in testarea two .................................................................................. 208

    Appendix 5c: Output results for production rice crop model in testarea one .................................................................................. 209

    Appendix 5d: output results for production rice crop model in testarea two .................................................................................. 210

  • 7/31/2019 Thesis Precfarming 2006

    14/225

    List of abbreviation XII

    LIST OF ABBREVIATION

    C Celsius

    Cm Centimetre

    CV Coefficient of Variation

    DEM Digital Elevation Model

    df Degree of freedom

    dSm-1 deciSiemens/meter

    ECe Electrical Conductivity extrication

    ECw Electrical Conductivity in water

    EFARP Egypt-Finland Agricultural Research Project

    ESP Exchangeable Sodium Percentage

    ET0 Evapotranspiration

    FAO Food and Agriculture Organization in the UnitedNations

    GDP Gross Domestic Product

    GIS Geographic Information Systems

    GPS Global Position Systems

    HRG High Resolution Geometric

    Hrs Hours

    HA Hectare

    IR Infrared

    LE Egyptian pound

    Kg Kilo gram

    m Meter

    m2 Meter Square

    meq/100gm soil Milliequivalents per 100 grams of soil

    meq/l milliequivalents per litremm millimetre

    MJ/m2 MegaJoule/ meter square

    Mid-IR Middle Infrared

    NDVI Normalized Different Vegetation Index

    NIR Near Infrared

    PBDAC Principal Bank for Development and AgriculturalCredit

    PCA Principle Component Analysisppm Parts per million

  • 7/31/2019 Thesis Precfarming 2006

    15/225

    List of abbreviation XIII

    PWP Permanent Wilting Point

    SAR Sodium Absorption Ratio

    Sig. Significant

    SNN Standard Nearest Neighbour

    TCT Tasseled Cap Transformation

    TM Thematic Mapper

    TTA Training or test areas

    ETM+ Enhance Thematic Mapper

    m Micro meter

    UTM Universal Transverse Mercator

  • 7/31/2019 Thesis Precfarming 2006

    16/225

    Introduction 1

    1. INTRODUCTION

    1.1. AGRICULTURE IN EGYPT

    1.1.1. BACKGROUND

    Agriculture is an important sector in the Egyptian economy providing work for

    30 % of the national labour force and contributing 16 % of the Gross Domestic

    Product (GDP). Agricultures contribution to the GDP in 1995/1996 fiscal year

    was estimated at about $7.2 billion (nine times higher than that of tourism).

    This means that the progress in agricultural productivity is crucial for the

    development of the economy.

    The total area of agricultural land in Egypt is approximately 7.3 million feddans

    (1 feddan = 0.42 ha), accounting for just 3.3 % of the total surface area. At

    present only 5.4 % of the land in Egypt is qualified as excellent for agriculture,

    while about 40 % is either of poor or low quality, due to mainly salinity, water

    logging and sodicity problems. Crops are cultivated in winter and summer at an

    average intensity of 2.3 crops a year. Farms in Egypt are generally very small.

    Nearly 50 % of farms are less than one feddan and 95 % of landowners have

    farms of less than 5 feddans. Cultivated land per farmer in Egypt is

    approximately 500 m2 ranking Egypt amongst the lowest in the world.

    Egypts agriculture is entirely dependent on irrigation. Crops are cultivated

    under irrigation with the use of organic and inorganic fertilizers. The main

    crops planted are: cotton, wheat, rice, maize and berseem (clover), or Egypts

    clover accounting for 80 % of all crops cultivated. Wheat and berseem are the

    main winter crops; in summer, cotton and rice are the important cash crops,

    while maize and sorghum are the subsistence crops. Compared with countries

    with similar agro-climatic conditions, levels of production are relatively high

    and yields have increased significantly in the last five years. There is

    considerable potential for growth, particularly in the approximately one million

    hectares of reclaimed land, which represent 25 % of the total agricultural area.

    Table (1.1) shows the statistics of main field crops production in Egypt.

  • 7/31/2019 Thesis Precfarming 2006

    17/225

    Introduction 2

    Table 1. 1: Field crops: areas, yields and returns, 2002/03

    Crop

    Area

    (000 ha)

    Yield

    (kg/ha)

    Revenue

    (LE/ha)

    Cost

    (LE/ha)

    Net return

    (LE/ha)

    Cotton 297 462 1 221 866 355

    Wheat 1 053 1 147 1 147 720 427

    Rice 650 1 659 1 152 739 413

    Maize 770 1 373 968 622 346

    LE: Egyptian pound

    Source: National Agricultural Income, 2002; Agricultural Statistics, 2003.

    Organic fertilizer (manure) includes cow, sheep, poultry and horse dung. It

    provides a slow release of nutrients as micro-organisms in the soil break down

    the organic material into an inorganic, water soluble form which the plants can

    use. The addition of organic material improves soil structure or "workability"

    immensely. It also vastly improves the water holding capacity of sandy soils.

    Manure is generally spread uniformly on the fields with great consideration

    given to the proper application rate.

    The inorganic fertilizers generally used contain one or more of the nutrients

    required for plant growth - mainly nitrogen (N), phosphorus (P), and potassium

    (K), and other essential elements (Fe, Zn, Cu, Mn). Farmers usually add the

    mineral fertilizer to the soil manually. Fertilizers are spread over the soil

    surface or applied in bands under the rows or side-dressed between planted

    rows. These fertilizers provide plant nutrients that are naturally lacking or that

    have been removed by harvesting or grazing, or by physical processes such as

    leaching or erosion (FAO, 2005).

    Water for irrigation generally comes from two sources either directly from the

    Nile in which case it is fresh or from drainage systems in which case it is re-

    used and of poor quality. This system of reusing used drainage water is

    common in the northern parts of Egypt where there is a shortage of the Nile

    fresh water.

  • 7/31/2019 Thesis Precfarming 2006

    18/225

    Introduction 3

    1.1.2. AGRICULTURAL POLICY

    Presently, the agricultural sector of Egypt has been completely liberalised. In

    the early 1960s, the government of Egypt regulated the area and the

    production of many crops including cotton, wheat, rice, sugar cane and onions.

    In addition, the farmer was obliged to deliver all or part of his harvest to the

    government at a fixed price, which was lower than the free market price. The

    government handled marketing and processing. The justification of this

    measure was that the agricultural sector is interrelated with other sectors of

    the economy. For example, a shortage in the supply of cotton would lead to

    considerable losses in the industrial sector. A "basic cropping pattern" was

    prepared by the cooperatives in each village for the agricultural year (1November to 31 October). The system also specified the crop variety as well as

    the quantity and type of fertilizers and pesticides to be supplied to farmers for

    each season. The Principal Bank for Development and Agricultural Credit

    (PBDAC) provided all agricultural inputs. Farmers were subjected to monetary

    penalties for violations of the cropping pattern (FAO, 1995 and Ender and

    Holtzman, 2002).

    These policies had negative effects on the performance of the agricultural

    sector. There were large transfers from the agricultural sector to other sectors.

    In 1980, a significant reform of these agricultural policies was introduced in the

    framework of the agricultural sector strategy for the 1980s. By 1986/87 the

    Ministry of Agriculture had pioneered an economic reform programme,

    concerning prices and marketing control, delivery quotas for the main crops

    and reduced subsidies for inputs. It encouraged private sector investment in

    crop marketing and the supply of inputs.

    By 1993, most governmental controls had either been removed or modified to

    encourage private initiatives:

    Governmental control of farms and output prices, crop areas and

    procurement quotas was removed.

  • 7/31/2019 Thesis Precfarming 2006

    19/225

    Introduction 4

    Governmental control of the private sector as regards to imports, exports

    and distribution of inputs as well as to import and export of agricultural

    crops was abolished.

    Subsidies on farm inputs were eliminated.

    The role of the PBDAC was diverted to the provision of financial services.

    Governmental ownership of land was limited.

    "New land" was sold to the private sector.

    The role of the Ministry of Agriculture was confined to agricultural research,

    extension, legislation and economic policies.

    The land tenancy system was modified.

    1.1.3. FARM PROBLEMS

    Egypt still remains one of the worlds largest food importers agricultural

    imports were $5.6 billion in 1996, representing about 40 % of total imports.

    Many factors have been documented to be responsible for the crisis affecting

    the Egyptian agricultural sector. The population of Egypt increased from about

    58 million in 1990 to 70 million in 2004. The high population growth makes

    the situation very challenging and national food security is of a high principal

    concern.

    High demographic pressure on agricultural area:

    Population densities in some areas along the Nile River are greater than 1,000

    people per square kilometre.Egypts population has increased more than six

    folds from 11 million in 1907 to almost 70 million at the beginning of the year

    2004.

    Rapid population growth is straining natural resources as agricultural land is

    being lost to urbanisation. The pressure of an increasing population combined

    with the scarcity of cultivable land, leads farmers to demand more from the

    land than it can yield. The pressure increases all the more rapidly as the

    spatial growth of human settlements, especially cities, takes a direct toll on the

    surrounding land resources: based on FAO data it has been estimated for

    instance that between 1973 and 1985 Egypt lost 13 % of its farmland to urban

  • 7/31/2019 Thesis Precfarming 2006

    20/225

    Introduction 5

    sprawl. It is commonly said that the land developed thanks to the Aswan Dam

    merely compensates for that loss to urbanisation.

    Limited agricultural land:

    97 % of Egypt is desert occupied and is therefore dependent on the Nile River

    for its existence. Only 5 % of the land area in Egypt is actually occupied and

    less than 4 % of the land is suitable for agriculture. Egypts arable and

    permanent crop land in 1993 (2.8 million hectares) was the same as in 1974-76

    and less than that in 1969-71 (FAO, 1995). The area of land per capital has

    fallen from 0.2 ha in 1907 to 0.05 ha at the beginning of the year 2004 due to

    the population increase from 11 million in 1907 to 76 million in 2004 but theincrease in agricultural land is very limited.

    Agricultural production problems:

    Many reasons account for the decreasing yield production in Egypt. Egypt is

    located in an arid region and the rainfall is insufficient to sustain agriculture.

    Therefore, agricultural land needs to be irrigated from the Nile, which is the

    main source of fresh water. Nevertheless, this water does not cover the cropwater requirement and consequently farmers use the low quality reused water

    for irrigation, which leads to degradation of the soil and low crop yield.

    Drainage problems in the north are associated with low-lying areas, clay to

    heavy clay soils with low permeability, saline to saline-sodic soils, shallow and

    salty ground water, often under artesian pressure, which tend to decrease the

    yield production (Moukhtar et al., 2004).

    Salinity and alkalinity, water logging, hard pan, urbanisation and compaction

    conditions also have adverse impacts on soil productivity, which was estimated

    to be in the range of 30-35 % of the potential productivity.

    Farm economy problems:

    There are many reasons for the decreasing farm economy in Egypt. The main

    reasons are 1- Even though new agricultural technology (harvester andvariables rate application, spray and weed controlled) has been introduced,

  • 7/31/2019 Thesis Precfarming 2006

    21/225

    Introduction 6

    farmers have very limited knowledge of this technology. Hence they continue to

    make use of manual workers that have a very high cost but also add to

    reducing of the net profit. 2- The size of the farm is small and the farmer uses

    intensive agro-chemicals to increase the yield but in an inefficient way. 3-

    Farmers have little or no knowledge about the soil reclamation, conservation

    and management. 4- Some farmers are unable to sell farm products.

    Intensive use of agrochemicals:

    Before the construction of the Aswan High Dam a few decades ago, agricultural

    productivity in the old Nile Valley and Delta was renowned for its excellent

    quality and high productivity. Application of high doses of naturally producedorganic fertilizers with few complementing chemical fertilizers were perfected

    and practised as long standing farm traditions. After the construction of the

    Aswan High Dam in the seventies, there has been a sharp reduction in the

    sediments load carried by the Nile water. This has been one of the negative

    consequences of the construction of the High Dam since these sediments that

    are rich in the nutrients are lost. The farmers use more intensive mineral

    fertilizers, insecticides and pesticides to increase productivity in the same farm

    unit especially after the liberalisation of prices of agricultural products. The use

    of pesticides increased in Egypt from 2143 tons in the fifties up to 11700 tons

    in 1990.

    Environmental pollution:

    The causes of soil and water pollution come from many sources including: the

    dumping of industrial waste water in to irrigation canals, the seepage of somesewage water with low treatment levels, chemical fertilizers and the residues of

    applied insecticides and pesticides. The pollution in the soil and water are

    directed and indirected increasing human diseases.

    1.2. NEW AGRICULTURE MANAGEMENT TECHNOLOGY

    Precision farming has emerged as a management practice with the potential to

    mitigate some of these problems and increase agricultural profits by utilising

    more accurate information about agricultural resources. For e.g. row-crop

  • 7/31/2019 Thesis Precfarming 2006

    22/225

    Introduction 7

    production which means the management of input variables, such as

    application rates, cultivation selection, tillage practices and irrigation

    scheduling. Also, precision farming is based on the management of agricultural

    fields according to soil types in different areas and on its influence on crop

    development and yield.

    The following major technological components are used for precision farming

    management practices: Geographical Information Systems (GIS), Global

    Positioning System (GPS), sensors, variable rate application, yield monitoring

    and crop growth models.

    Remote sensing are perfect tools to assess the land covers, crop situation and

    status as well as their changes.GIS for precision farming management stores data, such as land use and land

    cover map, soil type, nutrient levels, etc in layers and assigns that information

    to the particular field location. GIS can be used to analyse characteristics

    between layers to develop application maps.

    GPS for precision farming stores locations such as yield areas, boundary of

    crop type, soil sample, etc. A GPS system has the ability to return to a

    particular location repeatedly.Storing this information in a GIS makes it possible to develop site-specific

    spatial yield variability models. Spatial yield variability is a complex interaction

    of many factors, including soil properties, fertility and management. Crop

    growth models (statistics and spatial models) are excellent tools for evaluating

    these complex interactions and provide insight into the causes of spatial

    variability.

    1.3. OBJECTIVES OF THE RESEARCH WORK

    In an attempt to test the potential of using precision farming techniques to

    improve the management of the Egyptian agricultural sector, the following

    objectives were pursued in this study:

    To identify the current state of the study area including the limiting factors

    affecting crop production.

  • 7/31/2019 Thesis Precfarming 2006

    23/225

    Introduction 8

    Provide more accurate farm records through a combination of thematic

    maps, remote sensing data, local and regional data-base acquisition and

    field measurement.

    To recognise the different alternatives of the proper land management

    practices that belong to the aforementioned limiting factors.

    Suggest different schemes of land managements planning according to bio-

    physical and socio-economic parameters.

    Research questions

    What are the biophysical factors of soil that influence yield production?

    What the effect of different test area conditions on the factors affecting theyield prediction?

    Can satellite data and object oriented segmentation lead to significant

    improvement in the classification of land cover and crop status?

    1.4. STRUCTURE OF THE THESIS

    This thesis is divided in to seven chapters with the content of each summarised

    below:

    Chapter 2, Literature review: This chapter reviews the concept of precision

    farming, including remote sensing assessment methods. Pervious the different

    parameters affecting soil fertility and yield production and gives some

    fundamental information on crop growth modelling (statistical and spatial

    models).

    Chapter 3, Materials: This chapter describes the study area as well as theremote sensing and GIS data which were used in this study. The different

    software applied is also mentioned.

    Chapter 4, Methods: The methodological steps implemented to produce land

    use and land cover maps and corresponding change maps are explained. The

    planning of the field work and the subsequent collection of soil and water

    samples are documented. The laboratory analyses of the samples as well as the

    designing of questionnaires to obtain ancillary data from farmers through

  • 7/31/2019 Thesis Precfarming 2006

    24/225

    Introduction 9

    interviews are presented. Finally a detailed description of the crop growth

    model building strategy is given.

    Chapter 5, Results:This chapter is a representation of the important research

    findings. The results were pertaining to the segmentation and classification of

    the satellite imagery; a display of land use and cover maps, and land use/cover

    change maps, as well as the use of tables and graphs to interpret the changes.

    The different results of the soil and water analysis and the effect of these

    results on the soil development and yield production are elaborated. The

    ancillary data is obtained from interviews with farmers are presented. Finally

    the integration of all these findings into the crop growth models is

    disadvantageous and the output for test areas evaluated. Suggestion of amanagement plan for the different test areas.

    Chapter 6 Discussion: The implications of the results are highlighted. The

    most important factors affecting yield prediction in the test areas are discussed.

    Chapter 7 Summary: This section summarises the main topic of the research

    and draws the implications for the future research efforts.

  • 7/31/2019 Thesis Precfarming 2006

    25/225

    Review of literature 10

    2. REVIEW OF LITERATURE

    This current review includes the main effective factors on the agriculture

    production of the investigated area east north of the Nile Delta. These factors

    have been presented in the following sequence.

    2.1. PRECISION FARMING

    2.1. REMOTE SENSING IN PRECISION FARMING

    2.3. REMOTE SENSING AND IMAGE INTERPRETATION

    2.4. SOIL FERTILITY AND CROP PRODUCTION

    2.5. CROP GROWTH MODELS

    2.1. PRECISION FARMING

    Precision farming is the term used to describe the goal of increased efficiency in

    the management of agriculture. It is a developing technology that modifies

    existing techniques and incorporates new ones to produce a new set of tools for

    the manager to use (Blackmore, 1994, Rains and Thomas 2000 and

    Stombaugh, et al., 2001). It integrates a significant amount of computing and

    electronics but higher levels of control inevitably require a more sophisticatedsystem approach.

    Precision farming uses a system approach to provide a new solution to

    contemporary agricultural issues, that is, the need to balance productivity with

    environmental concerns (Shibusawa, 2002). Based on advanced information

    technology, it includes describing and modelling variation in soils and plant

    species, and integrating agricultural practices to meet site specific

    requirements. Precision farming aims at increasing economic returns whist at

    the same time reducing the energy input and the environmental impact of

    agriculture. This means, managing each crop production, input-fertilizer, lime,

    herbicide, insecticide, seed, etc. on a site-specific basis to reduce waste,

    increase profits, and maintain the quality of the environment (Mueller et al.,

    2000).

    Precision farming should be viewed as a management philosophy or approach

    to the farm and not as a definable prescriptive system (Mandal and Ghosh,

  • 7/31/2019 Thesis Precfarming 2006

    26/225

    Review of literature 11

    2000). It identifies the critical factors where yield is limited by controllable

    factors and determines intrinsic spatial variability.

    Robert, (1999) expressed that precision agriculture is the start of a revolution

    in natural resource management based on information technology that is

    bringing agriculture into the digital and information age. In fact precision

    agriculture is based on the use of revolutionary technologies such as Global

    Positioning Systems (GPS) and Geographic Information Systems (GIS).

    According to Grisso et al., (2003) use of GPS indicates the correct position of

    each soil or plant sample taken in the field and GIS eases the handling of the

    data, allowing both graphical representation of the variability of measured

    parameters and analysis. Furthermore, precision agriculture has led to thedevelopment of several technologies, e.g., sensors to measure different soil

    properties, automatic yield recording systems and variable rate, basic for the

    collection of data and the application of decisions (Roberson, 2000 and Grisso

    et al., 2002).

    According to the National Research Council (US), (1997), the potential of

    precision agriculture is limited by the lack of appropriate measurement and

    analysis techniques for agronomically important factors. Public sector support

    is needed for the advancement of data acquisition and analysis methods,

    including sensing technologies, sampling methods, data base systems and

    geospatial methods.

    2.2. REMOTE SENSING IN PRECISION FARMING

    Remote sensing has gained a lot of interest as a potential management tool for

    precision farmers (Morgan and Ess, 1997). Several applications have been

    developed to use remotely sensed data to infer both plant and soil

    characteristics (Barnes et al., 1996). Images from satellites and aerial

    photographs may allow the farmer to quickly view crops on his or her entire

    farm and decide which areas need further management without leaving the

    comfort of his/her home (Manakos et al., 2000). The information developed

    from remote sensing data must be accurate. It must be shown that measured

    reflectance can be correlated with crop properties of field conditions that will

    affect crop yield.

  • 7/31/2019 Thesis Precfarming 2006

    27/225

    Review of literature 12

    Three approaches of development appear to be emerging in the application of

    remote sensing and site-specific agriculture.

    In the first approach, multi-spectral images are used for anomaly detection;

    however, anomaly detection does not provide quantitative recommendations

    that can be directly applied to precision farming.

    A second approach involves correlating variation in spectral response to specific

    variables such as soil properties or nitrogen deficiency. In the case of nitrogen

    deficiency for example, once site-specific relationships have been developed,

    multi-spectral images can then be translated directly to maps of fertilizer

    application rates (Barnes et al., 2001). Nitrogen nutrition is known to influenceleaf chlorophyll concentration and greenness and through the use of remote

    sensing it is possible to estimate the nutritional status of a crop to assist in

    establishing more accurate side-dress nitrogen rates. The quantity of fertilizer

    nitrogen required by a crop is determined by an integration of soil and climatic

    factors and their effects on crop growth and nitrogen losses (Jennifer and

    Varco, 2004 and Bronson et al., 2005). Due to the highly transient nature of

    soil available nitrogen, there is no simple or routine soil test that can be used

    to predict availability and fertilizer needs and crop models have been shown to

    be inaccurate in predicting nitrogen nutrition requirements of cotton across

    soils (Varco et al., 2000).

    Nitrogen itself does not reflect or absorb, but significantly large amounts of

    nitrogen combine with the chlorophyll proteins in the leaves of plant. Therefore,

    the nitrogen content can be derived indirectly via chlorophyll content, which

    has characteristic absorption features in the visible wavelength region (400-700

    nm) A strong relationship exists between leaf chlorophyll and nitrogen content

    in several plant species maize and wheat for example (Oppelt and Mauser,

    2003 ).

    The third approach is converting multi-spectral data to quantitative units with

    physical meaning (such as leaf area index or temperature) and integrating this

    information into physically based growth models (Smith et al., 2005).

  • 7/31/2019 Thesis Precfarming 2006

    28/225

    Review of literature 13

    A sequence of remotely sensed images over time can provide information about

    crop growth and spatial variation within fields (Sonka et al., 1997). Detailed,

    spatially distributed multi-temporal information in visual form is not readily

    obtainable from conventional crop management systems or from site-specific

    crop management methods. Remote sensing images show spatial and spectral

    variation resulting from soil and crop characteristics.

    In general, remote sensing offers a fully pictorial representation of an area that

    improves the orientation of a farmer and is superior to traditional point

    measurements, since the farmer is able to identify heterogeneities in their

    spatial context (Jrgens, 2000).

    Remotely sensed images provide a visual method for observing the effects of

    managed inputs such as fertilizer and cultural practices such as tillage (Casady

    and Palm, 2002). They are also useful in understanding the impact of

    environmental factors such as drainage or pest infestations. In contrast to yield

    maps, which affect only future decisions, remotely sensed images may be

    collected several times throughout the growing season and allow timely

    management decisions to correct problems or deficiencies in the current crop.

    For this reason, remote sensing technology adds an important dimension to

    site-specific management of crops.

    The reflectance value for each pixel is dictated by contributions of all the

    surfaces within the pixels coverage and the characteristics of the sensor,

    namely the spectral bands (Dalsted et al., 2003). For example, a pixel could

    contain both bare soil and a growing crop within it. If this pixel encompasses

    40 % bare soil with a relative reflectance value of 20 % and 60 % vegetative

    cover with a relative reflectance value of 60 %, then the average reflectance

    value for the pixel should be about 44 % (0.4*20+0.6*60). Pixels containing two

    or more elements are classified as mixed pixels. A critical component in

    selecting remote sensing products is matching the pixel resolution to the

    amount of acceptable mixing. It is important to realise that most pixels,

    regardless of resolution, are affected by spectral mixing. Pixel mixing can

    diminish the ability to accurately map boundaries of abnormalities.

  • 7/31/2019 Thesis Precfarming 2006

    29/225

    Review of literature 14

    2.3. REMOTE SENSING AND IMAGE INTERPRETATION

    Image classification of remotely sensed data is the extraction of information

    contained in the images based on different spectral or terrestrial characteristics

    of different features on the earths surface. These earth objects register differentunique spectral signatures or and patterns in the electromagnetic spectrum

    (Murai, 1996).

    Generally, two approaches exist for the automated classification of remote

    sensing images: pixel-based classification in which the individual pixels form

    the building blocks of the classification and object-based classification in which

    the building blocks of the classification are image objects (Lee and Warner,

    2004). Traditional pixel-based approaches are based exclusively on the grey

    value of the isolated pixel. Thereby, only the spectral information is used for the

    classification. As mentioned above, the basic elements of an object-oriented

    approach are contiguous regions in an image referred to as image objects. Two

    kinds of image objects can be distinguished: objects of interest, which match

    real-world objects, e.g. the building footprints or whole agricultural parcels,

    and object primitives, which usually constitute the necessary intermediate step

    before objects of interest are formed by segmentation and classification

    processes. The smallest image object is one pixel. Image objects can be linked

    to a hierarchical network, where they are attributed with a high-dimensional

    feature space (Benz et al., 2004).

    Pixel-based classifications have difficulty to adequately or conveniently exploit

    expert knowledge or contextual information (Flanders et al., 2003). Object-

    based image-processing techniques overcome these difficulties by first

    segmenting the image into meaningful multi-pixel objects of various sizes,

    based on both spectral and spatial characteristics of groups of pixels. The

    segments (objects) are assigned to classes using often fuzzy logic and a

    hierarchical decision key.

    In landscapes, in nature, consist of patches that in an automated way can be

    detected in remote sensing imagery by the application of object basedalgorithms and therefore, ecologically, it is more appropriate to analyse objects

  • 7/31/2019 Thesis Precfarming 2006

    30/225

    Review of literature 15

    as opposed to pixels (Laliberte et al., 2004). Moreover, high- resolution satellite

    imagery such as Quick-bird can have classification problems due to greater

    spectral variation within a particular class if pixel based classification are to be

    applied.

    2.3.1. OBJECT-BASED IMAGE ANALYSIS

    Within the eCognition software the procedure is based on the so called 'Fractal

    Net Evolution' approach developed at Delphi2 Creative Technologies, which is

    an efficient method of describing complex semantics within largely self

    constructing and dynamic networks (Baatz and Schpe, 1999). For image

    analysis there are three main points resulting:

    Object orientation: The procedure first extracts image objects, which are

    later classified by means of fuzzy-logic

    Representation of the image information in different scales simultaneously:

    Each image contains different semantic levels at the same time. The basic

    strategy is therefore to build up a hierarchical network of image objects,

    which allows representing the image information content at different

    resolutions (scales) simultaneously. By operating on the relations betweennetworked objects, it is possible to classify local object context information.

    Beyond the pure spectral information this is often essential context

    information can be used together with form and texture features of image

    objects to improve classification.

    Description, processing and analysis of image information by means of

    semantic networks.

    Object oriented image classification can be divided into image segmentation,

    classification and accuracy assessment.

    2.3.1.1. Image segmentation

    Segmentation means the grouping of neighouring pixels into regions or

    segments based on similarity criteria - digital number, texture (Meinel and

    Neubert, 2004). It subdivides an image into separated regions (Baatz et al.,

    2000). Image object in remote sensing imagery are often homogenous and can

    be delineated by segmentation. As a consequence, the number of basic

  • 7/31/2019 Thesis Precfarming 2006

    31/225

    Review of literature 16

    elements for subsequent image classification is enormously reduced. The

    quality of classification is directly affected by segmentation quality. Throughout

    the segmentation procedure, the whole image is segmented and image objects

    are generated based upon several adjustable criteria of homogeneity or

    heterogeneity in colour and shape.

    There are two major approaches in image segmentation: Region-based and

    Edge-based segmentation. In the region-based approach, region growing

    algorithms cluster pixels starting with seed points and growing into regions

    until a certain threshold is reached (Jain and Farrokhnia, 1991). This

    threshold is normally a homogeneity criterion or a combination of size and

    homogeneity. A region grows until no more pixels can be attributed to any ofthe segments and new seeds are placed and the process is repeated. This

    continues until the whole image is segmented (Blaschke et al., 2002). These

    algorithms depend on a set of given seed points, but sometimes suffer from lack

    of control over the break-off criterion for the growth of a region. Common to

    operational applications are different types of texture segmentation algorithms.

    They typically obey a two-stage scheme:

    1. In the modelling stage characteristic features are extracted from the textured

    input image which includes spatial frequencies (Puzicha and Buhmann, 1998).

    2. In the optimization stage features are grouped into homogeneous segments

    by minimising an appropriate quality measure. This is most often achieved by a

    few types of clustering cost functions (Lee et al., 1992).

    Edge based segmentation is executed in two steps(Gorte, 1998). The first stepis to find segment boundaries in the image by identifying the edge pixels at

    those places where grey value change occurs. Each image region that is

    completely surrounded by edge pixels then becomes a segment. Region growing

    creates segments starting from seed pixels by iteratively augmenting them with

    surrounding pixels as long as the homogeneity criteria, which can be specified

    by the user, are satisfied (Bock and Guerra, 2001).

    Edges are regarded as boundaries between image objects and they are located

    where changes in values occur (Hoffman and Boehner, 1999). There are various

  • 7/31/2019 Thesis Precfarming 2006

    32/225

    Review of literature 17

    ways of delineating boundaries.The image segmentation is based on the

    representational values of each pixel. At first these values are calculated by a

    harmonic analysis of the values for each spectral channel. The minima in the

    matrix of representation typically arranged in pixel-lineaments represent

    spatial unsteadiness in the digital numbers. For the image segmentation, the

    vectorized minima of the representation, delimits areas consisting of pixels with

    similar spectral properties (Sonka et. al., 1998). A convergence index is

    combined with a single-flow algorithm for the vectorization of the

    representation minima.

    eCognition software first performs an automatic pre-processing segmentation of

    this imagery. This results in an abstraction of information and a knowledge freeextraction of image objects (Argialas and Tzotsos, 2004). The formation of the

    objects is carried out in a way that maintains overall homogeneous resolution

    is kept. The segmentation algorithm does not only rely on the single pixel value,

    but also on the "colour" (pixel value) and spatial continuity (Manakos et al.,

    2000). The formatted objects not only have the value and statistic information

    of the pixels of which that they consist of but also carry texture and form

    information. The user can then interact again with the procedure, and basedon statistics, texture, form and mutual relations among objects, can then

    create classes, where the classification of an object follows either the nearest

    neighbourhood method or fuzzy membership functions (Marangoz et al., 2004).

    Since an ideal object scale does not exist, objects from different levels of

    segmentation (spatially) and of different meanings (ecologically) have to be

    combined for many applications (Baatz and Schpe, 2000). The human eye

    recognises large and small objects simultaneously but not across totally

    different dimensions (Caprioli and Tarantino, 2003). From a balloon for

    instance, the impression of a landscape is dominated by land use patterns

    such as the composition of fields, roads, ponds and built up areas. Closer to

    the ground, one starts to recognise smaller pattern such as single plants while

    simultaneously the small scale pattern looses importance or can no longer be

    perceived anymore (Blaschke et al., 2004). In remote sensing, a single sensor

    highly correlates with a specific range of scales. The detect ability of an object

    can be treated relative to the sensors resolution. A coarse rule of thumb is that

  • 7/31/2019 Thesis Precfarming 2006

    33/225

    Review of literature 18

    the scale of image objects to be detected must be significantly bigger than the

    scale of image noise relative to texture. This ensures that subsequent object

    oriented image processing is based on meaningful image objects (Mitri and

    Gitas, 2002). Therefore, among the most important characteristics of

    segmentation procedure is the homogeneity of the objects. Only if contrasts are

    treated consistently can good results be expected.

    2.3.1.2. Classification

    The classification process within the eCognition programme is based on fuzzy

    logic to allow the integration of a broad spectrum of different object features

    such as spectral values, shape or texture for classification (Baatz et al., 2000).

    Utilisation of image object attributes and the relationship between networked

    image objects results in a sophisticated classification incorporating local

    context.

    Classifying an image using object-oriented approach means classifying the

    image objects either based on sample objects (training areas) or according to

    class descriptions organised in an appropriate knowledge base. The knowledge

    base itself is created by means of inheritance mechanisms (Yijun and Hussin,2003).

    eCoginition software offers two basic classifiers: a nearest neighbour classifier

    and fuzzy membership functions (Oruc et al., 2004). Both act as class

    descriptors. While the nearest neighbour classifier describes the classes to

    detect by sample objects for each class, which the user has to determine, fuzzy

    membership functions describe intervals of feature characteristics wherein the

    objects do belong to a certain class or not by a certain degree (Hofmann, 2001).

    Thereby each feature offered by eCognition can be used either to describe fuzzy

    membership functions or to determine the feature space for the nearest

    neighbour classifier (Binaghi et al. 1997). A class is then described by

    combining one or more class descriptors by means of fuzzy-logic operators or

    by means of inheritance or a combination of both. As the class hierarchy

    should reflect the image content with respect to scale, the creation of level

    classes is very useful. These classes represent the generated levels derived from

    the image segmentation and are simply described by formulating their

  • 7/31/2019 Thesis Precfarming 2006

    34/225

    Review of literature 19

    belonging to a certain level. Classes which only occur within these levels inherit

    this property from the level classes. This technique usually helps to clearly

    structure the class hierarchy (Baatz et al., 2004).

    2.3.1.3. Accuracy assessment

    Accuracy assessments of classifications of object based classification were

    undertaken using confusion matrices and Kappa statistics like pixel based

    classifications (Congalton, 1991). This is an important final step of the

    classification process. The goal is to quantitatively determine how effectively

    objects were grouped into the correct land cover classes (Jensen, 1996). The

    procedure is relatively simple; pixels are randomly selected throughout the

    image using a specified random distribution method (Sabins, 1997). The

    analyst then uses the original image along with ancillary information such as

    aerial photographs or direct field observation to determine the true land cover

    represented by each random pixel. This ground truth is compared with the

    classification map. If the ground truth and classification match, then the

    classification of that pixel was accurate. Given that enough random pixels are

    checked, the percentage of accurate pixels gives a fairly good estimate of the

    accuracy of the whole map (Lillesand and Kiefer, 2000). A more rigorous and

    complicated estimate of accuracy is given by the kappa statistics, which are

    obtained by a statistical formula that utilises information in an error matrix.

    An error matrix is simply an array of numbers indicating how many pixels were

    associated with each class both in terms of the classification and the ground

    truth (Congalton, 1991, Congalton and Green, 1999).

    There are two primary components of error in thematic maps; position errorand thematic error. In a map with poor position error the shape and size of a

    particular feature, such as a lake, might be correct but the placement on the

    map could still be incorrect. Thematic error occurs when a feature is

    misidentified (Horning, 2004). For example, if an area labelled shrub on the

    map was actually grassland then the thematic error of the map would increase.

    In most cases both of these error components work together. For example,

    when trying to delineate the boundary between two kinds of ground cover types

  • 7/31/2019 Thesis Precfarming 2006

    35/225

    Review of literature 20

    that gradually change from the one land cover type to the other it is difficult to

    accurately draw that the line that divides these two cover types.

    Overall accuracy: Overall accuracy is the proportion of all reference pixels,

    which are classified correctly (in the sense that the class assignment of the

    classification and of the reference classification agree). It is computed by

    dividing the total number of correctly classified pixels (the sum of the elements

    along the main diagonal) by the total number of reference pixels (Congalton,

    1991).

    Overall accuracy is a very coarse measurement. It gives no information about

    what classes are classified with good accuracy [Overall accuracy = (totalnumber correct) / (total reference or total classified)* 100].

    Producers accuracy: Producers accuracy is a reference-based accuracy that

    is computed by looking at the predictions produced for a class and determining

    the percentage of correct predictions. Producers accuracy estimates the

    probability that a pixel, which is of class A in the reference classification, is

    correctly classified (Fitzpatrick-Lins, 1981). It is estimated with the reference

    pixels of class A divided by the pixels where classification and reference

    classification agree in class A. Producers accuracy tells us how well the

    classification agrees with reference classification and Producers accuracy is a

    reference-based accuracy that is computed by looking at the predictions

    produced for a class and determining the percentage of correct predictions

    (Producers accuracy = number correct / reference total).

    Users accuracy: Users accuracy is a map-based accuracy that is computed bylooking at the reference data for a class and determining the percentage of

    correct predictions for these samples. Users accuracy is estimated by dividing

    the number of pixels of the classification result for class A with the number of

    pixels that agree with the reference data in class A (Story and Congalton,

    1986). Users accuracy predicts the probability that a pixel classified as class A

    actually belongs to class A (Users accuracy = number correct / classified total).

    Kappa Statistics:The Kappa analysis is a discrete multivariate technique used

    in accuracy assessment for statistically determining if one error matrix is

  • 7/31/2019 Thesis Precfarming 2006

    36/225

    Review of literature 21

    significantly different to another (Cohen, 1960). The result of performing a

    Kappa analysis are KHAT statistics (actually K, an estimate of Kappa), which

    are another measure of agreement or accuracy. This measure of agreement is

    based on the difference between the actual agreements in the error matrix

    (Congalton et al., 1983).

    2.3.2. POST CLASSIFICATION COMPARISON CHANGE DETECTION

    Change detection is used to compare and contrast two images with symmetrical

    positions, and to apply image-handling techniques to analyse the reformed

    area. Singh, (1989) defined change detection as a process of identifying

    differences in the state of objects or phenomena by observing them at different

    times.

    Several land cover change detection approaches have been developed such as

    traditional post-classification cross-tabulation, cross-correlation analysis,

    neural networks and knowledge-based expert systems with most of the

    procedures relying on pixel-based methods (Muttitanon and Tripathi, 2005).

    These methods are widely used with moderate resolution images thus restricted

    by the high spatial resolution of new remote sensing images (Civco et al.,2002).

    The post-classification change detection is widely used and the direction of

    change is also provided for other methods (Peterson et al., 2004).

    It involves classifying the rectified images separately from two pe

    riods of time, giving appropriate marks to different particles on the surface of

    the ground. The classified images are then compared and analysed from the

    two periods to determine the change-detecting matrix, and finally construct the

    change map (Churchill, 2003). According to Jensen, (1996) the post-

    classification comparison technique compares, on a pixel-by-pixel basis,

    multiple maps created from remotely sensed data collected at different times.

    This was the classification that rectification errors influence the change

    detection result.

  • 7/31/2019 Thesis Precfarming 2006

    37/225

    Review of literature 22

    With the advent of high spatial resolution remote sensing data and recent

    innovations in image interpretation, this situation is different. The

    interpretation of the change results can be improved by an object-oriented

    post-classification of the changed pixels (Niemeyer and Canty 2001). Defining

    different object classes of the change pixels helps to distinguish between the

    different changes (man-made, vegetation etc.). By means of semantic relations

    between the object classes of changes and other classes it is possible to exclude

    shadow affected regions and to concentrate on specific areas of interest (Walter,

    2004).

    Object-based change detection is much more a GIS task than remote sensing

    tasks if one considers post-classification analysis (Blaschke, 2005). It involves amuch smaller number of objects compared to pixels, which are subjected to

    change detection. However, in contrast to geometrically predefined pixels one is

    faced with a much more complex task. While a usual real world data set based

    on IKONOS-like to Landsat-like resolutions typically results in the range of 105

    to 106 objects the combination of the two data sets would result in 107 objects.

    2.4. SOIL FERTILITY AND CROP PRODUCTION

    Characteristics and variability of soil parameters have been analysed and

    documented in many scientific studies (Bergmann, 1992; Mills and Jones,

    1996; Jones, 1998; Logsdon et al., 1998; Havlin et al., 1999). The parameters

    documented in the various literatures are generally based on the primary

    objectives laid down in the respective research studies. Based on the objectives

    that were pursued in this study, only soil parameters influencing crop yield

    were analysed. These include: organic matter (O.M.), available nitrogen (N),available phosphorous (P), exchangeable cations, cation exchangeable capacity

    (CEC), iron (Fe), zinc (Zn), manganese (Mn), copper (Cu), pH and electrical

    conductivity (EC).

    Organic Matter (OM)

    Soil organic matter is carbon-rich material that includes plant, animal, and

    microbial residue in various stages of decomposition(Herrick, 2002).

  • 7/31/2019 Thesis Precfarming 2006

    38/225

    Review of literature 23

    Live soil organisms and plant roots are part of the carbon pool in soil but are

    not considered soil organic matter until they die and begin to decay. Soil

    organic matter enhances soil functions and environmental quality because it

    binds soil particles together into stable aggregates, thus improving porosity,

    infiltration, root penetration and reducing runoff and erosion (Christensen and

    Johnston, 1997). Furthermore it enhances soil fertility and plant productivity

    by improving the ability of the soil to store and supply nutrients, water, and air

    (Zanen and Koopmans, 2005). Additionally soil organic matter provides habitat

    and food for soil organisms; sequesters carbon from the atmosphere; reduces

    mineral crust formation and runoff; and, reduces the negative water quality

    and environmental effects of pesticides, heavy metals, and other pollutants by

    actively trapping or transforming them. The amount of organic matter in the

    soil is a balance between additions of plant and animal materials and losses

    through decomposition and erosion (Herrick et al., 2001).

    Soil pH

    Soil pH is a measure of soil acidity and alkalinity. A pH of 7 is neutral, a pH

    below 7 is acid, and a pH above 7 is alkaline.

    Many factors have an effect on soil pH: fertilizers, rain, organic matter, soil

    texture, soil microorganisms, etc. Soil pH is normally increased or decreased

    using agricultural lime or gypsum respectively (Wang et al., 1998). Soil pH is an

    important chemical property because it affects the availability of nutrients to

    plants and the activity of microorganisms in the soil (Brady, 1990). A pH

    measurement is therefore an important part of a soil testing program. The

    effect of pH on microbial activity and nutrient availability in mineral soils is

    shown in figure 2.1.

  • 7/31/2019 Thesis Precfarming 2006

    39/225

    Review of literature 24

    Figure 2. 1: Nutrient availability and microbial activity as affected by soil pH; the wider

    the band, the greater the availability or activity. (from USDA Year book 0f

    agriculture, 1947-47)

    Cation Exchange Capacity (CEC)

    Cations held on the clay and organic matter particles in soils can be replacedby other cations; thus, they are exchangeable. For instance, potassium can be

    replaced by cations such as calcium or hydrogen, and vice versa.

    The total number of cations a soil can hold or its total negative charge is the

    soil's cation exchange capacity. The higher the CEC, the higher the negative

    charge and the more cations that can be held (Stewart and Hossner, 2001).

    CEC is measured in milliequivalents per 100 grams of soil (meq/100gm soil). A

    meq is the number of ions which total a specific quantity of electrical charges.

    In the case of potassium (K+), for example, a meq of K+ ions is approximately 6

    x 1020 positive charges. With calcium, on the other hand, a meq of Ca2+ is also

    6 x 1020 positive charges, but only 3 x 1020 ions because each Ca ion has two

    positive charges (Brady and Weil, 2002)

  • 7/31/2019 Thesis Precfarming 2006

    40/225

    Review of literature 25

    Nitrogen (N)

    N is the most frequently deficient nutrient in crop production therefore, most

    non-leguminous cropping systems require N inputs (Havlin et al., 1999). Many

    N sources are available for use in supplying N to crops. Plants normally

    contain 1 to 5 % N by weight and absorb N as both Nitrate (NO3-) and

    ammonium (NH4+) (Mulvaney and Khan, 2001).The total N content of soils

    ranges from less than 0.02 % in subsoil to more than 2.5 % in peat. All N

    forms are mobile in plants, consequently N deficiency symptoms first appear on

    older leaves (Mills and Jones, 1996). When plants are deficient in N, they

    become stunted and yellow in appearance. The loss of amino acid N from

    chloroplasts in older leaves produces yellowing. Under N shortage, plants growslowly and are weak and stunted (McCauley, et al., 2003). Leaves are small, the

    foliage colour is light-green to yellow, and older leaves often fall prematurely.

    Root growth is reduced and branching restricted; yet, there is usually an

    increase in the root/shoot ratio. Yield and quality are significantly reduced

    (Dobermann and Fairhurst 2000).

    Phosphorus (P)

    Phosphorus (P) does not occur as abundantly in soil as N and K (Havlin et al.,

    1999). Total P in surface soils varies between 0.005 and 0.15 %. Phosphorus

    concentration in plants ranges between 0.1 and 0.5 %; considerably lower than

    N and K. Plants absorb either H2PO4- or HPO42- orthophosphate ions (Johnston

    and Sten, 2000). Absorption of H2PO4- is greatest at low pH values, whereas

    uptake of HPO42- is greatest at higher values of pH. Slow growing, weak, and

    stunted plants that may be dark green in colour with older leaves showing a

    purple pigmentation are symptomatic of P deficiency (Jones, 1998).

    Iron (Fe)

    Soil solution concentration and availability of Fe to plants is predominantly

    governed by the organic fraction in soils (Havlin et al., 1999). Fe is absorbed by

    plant roots as Fe2+ and Fe3+. Fe concentration in soil varies widely, from 0.7 to

    55 %. Most of this soil Fe is found in primary minerals, clays, oxides, and

    hydroxides. Iron deficiency affects many crops with a common symptom lime

  • 7/31/2019 Thesis Precfarming 2006

    41/225

    Review of literature 26

    chlorosis occurring frequently in alkaline soils (Jones, 1998). A typical

    deficiency symptom is the interveinal chlorosis of younger leaves with the

    chlorosis spreading to older leaves as the deficiency becomes severe.

    Zinc (Zn)

    Zn content of soil depends on the nature of the parent material, organic matter,

    texture and pH. The most quoted range for total Zn in soils is 10 to 300 ppm

    (Hodges and Crozier, 1996). Thesufficiency range for Zn in leaves is between

    15 to 50 mg.Kg-1