Characterising Historical Land Cover Change, and Understanding Trends

CHARACTERISING HISTORICAL LAND COVER CHANGE,

AND UNDERSTANDING TRENDS IN THE GOUKOU

CATCHMENT, WESTERN CAPE, SOUTH AFRICA

Gcobani Nzonda

Submitted in fulfilment of the academic requirements for the degree of

Master of Science in the Discipline of Geography in

School of Agriculture, Earth and Environmental Sciences

College of Agriculture, Engineering and Science

University of KwaZulu-Natal

Pietermaritzburg

South Africa

27 September 2016

PREFACE

The research contained in this thesis was completed by the candidate while based in the Discipline of Geography, School of Agricultural, Earth and Environmental Sciences of the College of Agriculture, Engineering and Science, University of KwaZulu Natal, Pietermaritzburg, South Africa. The research was financially supported by South African National Biodiversity Institute and the National Research Foundation Global Change Grand Challenge Research Grant 78648.

The contents of this work have not been submitted in any form to another university and, except where the work of others is acknowledged in the text, the results reported are due to investigations by the candidate.

______

Signed (Supervisor): Prof G.F. Midgley

Date: 27 September 2016

DECLARATION: PLAGIARISM

I, Gcobani Nzonda, declare that:

i. The research reported in this dissertation, except where otherwise indicated or acknowledged, is my original work;

ii. This dissertation has not been submitted in full or in part for any degree or examination to any other university; iii. This dissertation does not contain other persons’ data, pictures, graphs or other information, unless specifically acknowledged as being sourced from other persons; iv. This dissertation does not contain other persons’ writing, unless specifically acknowledged as being sourced from other researchers. Where other written sources have been quoted, then:

a. Their words have been re-written but the general information attributed to them has been referenced;

b. Where their exact words have been used, their writing has been placed inside quotation marks, and referenced;

v. Where I have used material for which publications followed, I have indicated in detail my role in the work; vi. This dissertation does not contain text, graphics or tables copied and pasted from the Internet, unless specifically acknowledged, and the source being detailed in the dissertation and in the References sections.

______

Signed: Gcobani Nzonda

Date: 27 September 2016

DEDICATION

To my Mother

Nolwandle Florence Nzonda

iii

ABSTRACT

Land cover transformation is a phenomenon that is spreading in extent and significance, both globally and in South Africa. The phenomenon includes conversion of natural vegetation for intensive agricultural and forestry uses, and urbanisation. These uses result in changes to natural habitats through processes that are strongly influenced by human activities, and can lead to over-utilisation of ecosystems and resulting environmental change, such as encroachment of alien plant species, altered hydrological behaviour of catchments and related flooding risk, and loss of soil and biodiversity. In the past two centuries, the influence of human activities on the land surface has increased enormously. This has altered landscapes worldwide, and ultimately is influencing the earth’s nutrient and hydrological cycles, as well as regional and global climate. The main empirical focus of this study was to document the types, geographical distribution, and rates of land cover change in a well circumscribed South African rural catchment that has seen significant land cover change over several decades. The ultimate objective has been to provide a basis for future work on the impacts of these changes on ecosystem services, and possible adaptation measures. The thesis thus also speculates briefly on the trends observed, possible driving forces behind these changes, their likely consequences for ecological infrastructure and ecosystem services such as water quality and quantity, and implications for resilience to climate change.

The Goukou River catchment was selected for study as it is well circumscribed, of an appropriate size, and has an historical land transformation trend that is well captured by Black and White aerial photography at roughly decadal time steps, and more recently by more modern imagery. Thirty- three land cover types were defined priori for the catchment based on the Biodiversity Geographic Information System (BGIS) land cover map, for potential digitisation from available remote-sensed imagery. The digitisation was conducted by using advanced Geographic Information Systems and Remote Sensing tools to enhance grey-scale aerial photographic images (Black and White Aerial Photographs (BWAP)), for the time periods 1940, 1954, 1967, 1974, 1985 and 1991. More recently acquired colour infrared images (CIR) were processed and analysed for the time periods 2006 and 2010. The grey-scale aerial photographic images were processed and analysed using textural analysis, which comprises the use of spatial filters over kernel windows. The CIR images were visually processed and analysed using on-screen digitisation. The original 33 land cover classes were consolidated into 11 classes, mainly because of a difficulty in discerning

iv between them in the BWAP, to result in a final consistent mapping of clearly identifiable land cover classes. Ground truthing was also performed during field visits to allow matching of images to the finally selected 11 land cover classes.

Results indicated that, as expected, land cover in the Goukou catchment has undergone considerable change over the study period (1940 to 2010), revealing a significant intensification of land use practices. Probability matrices derived using Markov chain analysis allowed some inference of the likely drivers in the most recent decadal time period. The 10 to 15 year interval periodicity available provides fairly regular snapshots in time of the major land cover change over the past six decades. Land cover change did not proceed linearly over the full period, and it appears that certain drivers can change trends significantly and rapidly at the decadal time scale.

Results show that there was a rapid decrease in the area of natural vegetation, including wetlands and bare surfaces, within the study period centred on the 1960s. The total cultivated area was 12% of the entire catchment in 1940 and increased only slightly to 15% in 1954. However, a period of more rapid increase in cultivated field area occurred by 1967, where the total cultivated area increased to 21%. This was followed by a period of stasis, with a possible slight decrease by 1985 to 19%, although the image for 1985 was the poorest of all BWAP and thus cannot provide definitive indication of a reversal. By 1991, conversion of land for cultivation appeared to increase again, up to 26%, which was the highest increase in the rate of cultivation of land in the Goukou catchment in this period post 1960. This may have been followed by a slight decrease by 2006 and 2010 to 23% and 24%, respectively, but the switch to the use of colour images precludes a definitive conclusion except to confirm the increase observed by 1991. Wetland cover was 1.2% in 1940 and by 1954 had remained essentially constant at 1.3% but decreased by 1967 to 0.9%. Wetland surface area decreased to 0.6% by 2010, which is a reduction of 50% from the 1.2% cover of 1940. There were also noticeable increases in disturbed areas (degraded areas characterised by eroded surfaces), alien vegetation and dams constructed over the entire time period. Disturbed surfaces increased from 19% of the total catchment in 1940 to 32% of the catchment area by 2010; area under constructed dams increased sevenfold from 0.03% in 1940 to 0.2% of the total catchment area by 2010, and alien vegetation increased from 0.6% in 1940 to 2% by 2010, generally concentrated close to the river courses.

Overall, this study concludes that there has been a decline in the cover of natural vegetation, focused mainly around the 1960s and to a lesser extent towards the 1990’s. This indicates strong drivers to convert natural vegetation at these time periods, which included some apparently injudicious land conversion characterised by resulting erosion. Wetlands were also degraded overwhelmingly due to the conversion of natural habitat by intensive agriculture. Conversion via invasive alien species encroachment, wetland degradation and soil erosion appear to be minor components of land cover change relative to the large changes distant from the watercourses. However, these water course-related changes may have effects that are out of proportion with their surface area due to their likely impacts on hydrology and ecosystem services. Some relevant potential responses are presented and discussed, but implementation requires quantitative modelling efforts to identify sustainable options.

Key words

Land cover, land use, ecosystem services, GIS, Remote Sensing, Colour Infrared Images, Black and White aerial images

ACKNOWLEDGEMENTS

I would like to thank you all the people who played a role and made it possible for me to accomplish this research. I would like to thank you, my supervisor, Prof G.F. Midgley and my co-supervisor Z. Jonas and also T. Oliphant, M. Tshindane, K. Dyambothi, M. Tshikalange, F. Ramwashe, S. Mlambo, S.I. Mfenyane, S. Mayekiso, S. Bantwini, T. Mxokozeli, Dr D. Mapisa, S. Synders, my family for their support and National Geospatial Information for providing me with the data.

vii

TABLE OF CONTENTS PREFACE ...... i

DECLARATION: PLAGIARISM ...... ii

DEDICATION...... iii

ABSTRACT ...... iv

ACKNOWLEDGEMENTS ...... vii

UNITS OF MEASURE ...... xx

CHAPTER 1: INTRODUCTION ...... 1

1.1 Problem statement ...... 4

1.2 Hypotheses ...... 4

1.3 Main objective ...... 5

1.4 Key questions ...... 5

1.5 Thesis outline ...... 5

1.6 Description of study area...... 6

1.6.1 Location of the Goukou Catchment ...... 6

1.6.2 River tributaries ...... 7

1.6.3 Climatic conditions ...... 8

1.6.4 Ecological state ...... 8

1.6.5 Land use types and practices ...... 8

1.6.6 Geological structure ...... 9

1.6.7 Current land use related pressures ...... 10

CHAPTER 2: LITERATURE REVIEW ...... 13

2.1 Land cover classification systems ...... 13

2.2 Historical aerial photographs importance, uses and challenges ...... 14

2.3 Land use change, drivers and impacts ...... 15

2.4 Impacts of land cover change on ecosystem services ...... 17

2.5 Impacts of land cover change on water quantity and quantity ...... 19

viii

2.6 Impacts of land cover change on ecological infrastructure ...... 21

2.7 Land degradation ...... 23

2.8 Impacts of land cover change on climate change ...... 24

2.10 Summary ...... 27

CHAPTER 3: METHODS AND DATA ...... 29

3.1 General overview of methods ...... 29

3.2 Description of data ...... 30

3.3 Software utilised ...... 32

3.4 Data pre-processing ...... 32

3.4.1 BWAP pre-processing ...... 33

3.4.1.1 Geometric corrections ...... 33

3.4.1.2 Textural analysis ...... 35

3.5 Principal Component Analysis (PCA) ...... 39

3.5 Classification ...... 41

3.5.1 Unsupervised classification ...... 43

3.6 Image smoothing and generalisation ...... 44

3.7 Post-classification...... 45

3.8 Accuracy assessment ...... 45

3.8.1 Accuracy assessment for BWAP ...... 46

3.9 CIR images ...... 47

3.9.1 Preliminary pre-processing ...... 47

3.9.2 Manual digitising ...... 48

3.10 Change detection analysis ...... 52

3.11 Markov chain analysis for future land use change ...... 53

CHAPTER 4: RESULTS, DISCUSION AND RESEARCH LIMITATIONS ...... 56

4.1 Summary ...... 56

4.2 Accuracy assessment results ...... 56

4.3 Classification results ...... 62

4.1.4 Trend analysis results ...... 66

4.1.5 Rainfall results ...... 69

4.1.6 Change detection analysis ...... 74

4.1.7 Markov chain analysis results...... 78

4.2 Discussion ...... 81

4.2.1 Trends ...... 82

4.2.3 Rainfall ...... 84

4.2.4 Change detection ...... 85

4.2.5 Future projections ...... 86

4.3 Research limitations ...... 87

CHAPTER 5: CONCLUSIONS AND PRELIMINARY RECOMMENDATIONS ...... 89

5.1 Introduction ...... 89

5.2 Evaluation of the research questions and the objectives ...... 93

5.2.1. What are the historical geographical patterns of land cover change in this catchment? ...... 93

5.2.2. What are the main changes between land cover types and what are their rates? ... 94

5.2.3. What are the potential implications of these changes? ...... 94

5.2.4. What are the potential future patterns of land cover change assuming historical trends continue? ...... 95

5.2.5. What recommendations could have value in reducing the impacts of land use and land cover change in this catchment? ...... 95

5.4 Preliminary recommendations...... 96

5.4.1 Water abstraction from the river ...... 96

5.4.2 Monitoring and maintenance of water canals ...... 96

5.4.3 Alien vegetation ...... 97

5.4.4 Awareness about wetlands and water ...... 97

5.4.6 Flood structures ...... 97

5.4.7 Dam construction and expansion ...... 98

5.5 Conclusions ...... 99

REFERENCES ...... 102

APPENDICES ...... 133

Appendix A. Missing data as represented by DRID-reference ...... 133

Appendix B. Accuracy assessment Errors and Accuracies, and formulas ...... 134

Appendix C. The 1974 Confusion matrix results including errors and accuracy scores in (%). Errors are: Commision (CE) and Ommision (OE), and accuracy scores are Producer’s accuracy (PA) and User’s accuracy (UA) ...... 134

Appendices D. The 1991 Confusion matrix results including errors and accuracy scores in (%). Errors are: Commision (CE) and Ommision (OE), and accuracy scores are Producer’s accuracy (PA) and User’s accuracy (UA) ...... 135

Appendix E. Land cover trends from 1940 to 2010 in (ha) ...... 136

Appendix F. Landscape changes over years (1: wetlands converted to agriculture, 2: natural hazards and 3: aliens species)...... 137

Appendix G. Contributors to land cover change of the Goukou catchment from 1940 to 2010 ...... 138

Appendix H. Overlay maps of 1940 to 2010 ...... 140

Appendix I. Percentage of variance explained by Eigen values of Principal Component Analysis ...... 141

Appendix J. Graphs for Gains and Losses from 1940 to 2010 as obtained by using change detection analysis in Markov...... 147

LIST OF TABLES

Table 1: Current land use related pressures in the Goukou catchment, as determined by C.A.P.E Estuary Programme Report 2007...... 10

Table 2: Key information regarding datasets used in this study. Information is provided about the spatial scale of the photography, camera used, lens, format, principal distance in (mm) and data acquired, for years 1940, 1954, 1967, 1974, 1985, 1991, 2006 and 2010...... 31

Table 3: The percentages of variance explained by Eigenvalues of Principal Component Analysis. The shaded area represents the number of images that were used to run PCA, and sum of basic stats, Eigen values and percentages...... 40

Table 4: Initial 33 land use types as obtained by digitising 2006 and 2010 CIR images...... 50

Table 5: The procedure taken to merge the 33 initial classes from CIR images...... 51

Table 6: The Goukou River catchment final land use classes as obtained by digitising 2006 and 2010 CIR images...... 52

Table 7: Confusion matrix results including errors and accuracy scores in (%) for the classification derived from the 1940 Black and White Aerial Photography images. Errors are Comission (CE) and Omission (OE) and accuracy scores are Producer’s accuracy (PA) and User’s accuracy (UA). Natural vegetation is represented by Nv, Cultivated fields by Cf, Wetlands by W, Disturbed surfaces by Ds, Plantations by P, Roads by Ro, Bare surfaces by Bs, River by R, Dams by D, Aliens by A and Cleared plantations by Cp...... 57

Table 8: The 1974 Confusion matrix results including errors and accuracy scores in (%). Errors are: Commision (CE) and Ommision (OE), and accuracy scores are Producer’s accuracy (PA), User’s accuracy (UA), and Overall accuracy and Cohen’s Kappa Coefficient of 1974...... 58

Table 9: The 1974 Confusion matrix results including errors and accuracy scores in (%). Errors are: Commision (CE) and Ommision (OE), and accuracy scores are Producer’s accuracy (PA), User’s accuracy (UA), and Overall accuracy and Cohen’s Kappa Coefficient of 1974...... 58

xii

Table 10: Land cover of major classes (ha) in the Goukou catchment from 1940 to 2010. These areas are calculated from images analysed by digitising and classifying Colour Infrared Images and Black and White Aerial Photographs images taken roughly decadal between 1940 and 2010 for the Goukou catchment, Southern Cape...... 65

Table 11: Land cover of minor classes (ha) in the Goukou catchment from 1940 to 2010. These areas are calculated from images analysed by digitising and classification of Colour Infrared Images and Black and White Aerial Photographs images taken roughly decadal between 1940 and 2010 for the Goukou catchment, Southern Cape...... 65

Table 12: Land cover trends and average land cover trends of the Goukou catchment from 1940 to 2010 in (%). Calculated as change in area (ha/year) occupied by each land use types over change in time (years) between two consecutive decades...... 67

Table 13: Monthly, annual totals and average rainfall in (mm) for the town of Heidelberg in the Goukou catchment from 1940 to 2010. Calculated from rainfall data collected from South African Weather Service, the data station is identified by the South African Weather Service as [0009815x] Heidelberg C/K...... 69

Table 14: Seasonal and average seasonal rainfall per year in (mm) for the Goukou catchment from 1940 to 2010. Calculated from rainfall data collected from South African Weather Service, the data station is identified by the South African Weather Service as [0009815x] Heidelberg C/K...... 72

Table 15: Net gains and losses in (ha) and in (%) from 1940 to 2010. These are as calculated by overlaying images for two consecutive decades in IDRISI...... 76

Table 16: Transition probability matrices (area-based) used for projections of land use and land cover in 2025 (2006-2010). This table contains the number of pixels which are expected to change from each land use and land cover type to each other land use and land cover type over the specified time period (Eastman, 2012). Natural vegetation is represented by Nv, Cultivated fields by Cf, Wetlands by W, Disturbed surfaces by Ds, Plantations by P, Roads by Ro, Bare surfaces by Bs, River by R, Dams by D, Aliens by A and Cleared plantations by Cp...... 79

xiii

Table 17: Conditional probability of changing land cover and land use classes. Transition probability matrices in (% cover) used for projections of land use and land cover in 2025 (2006-2010). This reports the probability that each land use and land cover type would be found at each pixel after the specified time period (Eastman, 2012). Rows represent initial state and columns represent future state. Natural vegetation is represented by Nv, Cultivated fields by Cf, Wetlands by W, Disturbed surfaces by Ds, Plantations by P, Roads by Ro, Bare surfaces by Bs, River by R, Dams by D, Aliens by A and Cleared plantations by Cp...... 80

xiv

LIST OF FIGURES

Figure 1: The location of the Goukou water catchment in the Southern Cape in relation to the Western Cape as a whole. The map indicates the primary and secondary drainage lines, and the sub-catchments are indicated by brown lines while major towns are indicated by red dots and rivers are indicated by blue lines. The upper part of the catchment is represented by catchment numbers H90A, H90B and H90C, the middle part of the catchment is represented by catchment number H90D, while catchment number H90E represent lower reaches of the catchment...... 7

Figure 2: Main land cover classes in the Goukou catchment, as determined by the 2009 National Land Cover product (BGIS, 2009). The upper catchment is indicated by A, while B indicates the middle parts of the catchment and C indicates the lower parts of the catchment. 9

Figure 3: Geo-referenced BWAP single tile of 1940 by using 2006 as a reference layer. .... 34

Figure 4: Subset BWAP single tile of 1940...... 34

Figure 5: Subsetted and mosaicked BWAP for 1974...... 35

Figure 6: Example of filters and kernel window sizes used during textural analysis: Extracted portion of BWAP of 1940 with kernel window sizes and filters of 3x3 skewness, 3x3 variance, 5x5 MED, 7x7 MED, 13x13 MED and 25x25 kurtosis ...... 38

Figure 7: Pseudo colour image of stacked spatial filters and kernel windows of 1940 BWAP. As obtained by combining spatial filters of skewness, variance, kurtosis and MED, and kernel window sizes of 3x3, 5x5 5, 7x7, 13x13 and 25x25. These were stacked together using ENVI 4.4...... 38

Figure 8: The output results of four PCA images of 1967. This is pseudo colour image obtained by running PCA on the twenty textured images and one original image of 1967. ... 39

Figure 9: Formula to calculate Cohen’s Kappa coefficient as expressed by Congalton (1991). Where r is the number of rows in the matrix, xkk is the number of observations in row i, and column j, and xk+ and x+k are the marginal totals for row i, and column j and N is the total number of observations...... 47

Figure 10: Examples of sites visited during study area visits on September 2013, their land cover types, and their coordinates...... 49

Figure 11: Examples of 2006 CIR digitised land use classes...... 49

Figure 12: Maps of final land cover classes for the Goukou catchment, which were obtained from classified and digitised images of 1940, 1954, 1967, 1974, 1985, 1991, 2006 and 2010. The insert is the location of the Goukou catchment relative to South Africa as a whole and the legend represent classes present on the map...... 60

Figure 13: Land cover of major classes (ha) in the Goukou catchment, Southern Cape, between 1940 and 2010. These areas are calculated from images analysed by digitising and classifying Colour Infrared Images and Black and White Aerial Photographs images taken roughly decadal from 1940 to 2010 for the Goukou catchment. Natural vegetation is represented by Nv, Cultivated fields by Cf, Wetlands by W, Disturbed surfaces by Ds, Plantations by P, Roads by Ro, Bare surfaces by Bs, River by R, Dams by D, Aliens by A and Cleared plantations by Cp...... 63

Figure 14: Land cover of minor classes (ha) in the Goukou catchment from 1940 to 2010. These areas are calculated from images analysed by digitising and classification of Colour Infrared Images and Black and White Aerial Photographs images taken roughly decadal between 1940 and 2010 for the Goukou catchment, Southern Cape. Natural vegetation is represented by Nv, Cultivated fields by Cf, Wetlands by W, Disturbed surfaces by Ds, Plantations by P, Roads by Ro, Bare surfaces by Bs, River by R, Dams by D, Aliens by A and Cleared plantations by Cp...... 64

Figure 15: Average rate of land cover change in (ha/year) from 1940 to 2010. As calculated by using average change for each combination of years...... 68

Figure 16: Graphical representations of the Goukou catchment land cover trends from 1940 to 2010 in (ha/year)...... 68

Figure 17: Seasonal rainfall in (mm) for four seasons of the year from 1940 to 2010, for the Goukou catchment, Southern Cape, obtained by using a “Moving Average” of three intervals. Rainfall data was collected from South African Weather Service, the data station is identified

xvi by the South African Weather Service as [0009815x] Heidelberg C/K...... 71

xvii

GLOSSARY

Climate Change: “A change in the state of climate that can be identified by changes in the mean and or the variability of its properties and that persists for an extended period, typically decades or longer” (IPCC, 2012)

Ecological Infrastructure: Natural functioning ecosystems that deliver services to people, such as naturally-cleansed water and climate regulation, soil formation and disaster risk reduction (SANBI, 2014)

Ecosystem services: Ecological process or function having monetary or non-monetary values to individuals or society at large. These include supporting services, provision services, regulating services and cultural services (IPCC, 2012)

Ecosystem: Community of living organisms in conjunction with the non-living components of their environment interacting as a system (IPCC, 2012)

Gabion: Constructed cage filled with rocks and sometimes with sand and soil, used to control erosion by flowing water

Grid: Orthogonal reference lines on a map permitting exact positioning of any location

Kurtosis: Measure of the relative degree of sharpness of the peak on a frequency distribution curve (Brown, 2014)

Land cover: Existing vegetative, structural or other surface cover, usually resulting from a given land use

Land use: How land is used by human beings

Levéé: An embankment built to prevent the overflow of a river

Mean Euclidean distance: Ordinary distance between two points in Euclidean space (Himmelstein et al., 2010)

Reservoir: Natural or artificial surface water store and potential source of water supply

Skewness: Measure of the distribution of intensity of the asymmetry in the image (Jensen, J. R., 1996)

xviii

Variance: Measure of the distribution of variability in the image (Jensen, J. R., 1996)

Water Quantity: The amount of available water that is accessible for use

Weir: A low wall built across a river to raise the level of water upstream or regulate its flow

xix

UNITS OF MEASURE

: Minutes

'': Seconds

%: Percentage

0 : Degree

0C: Degrees Celsius ha: Hectares km: Kilometres km2: Square kilometres m: Meters m3: Cubic metres m2: Square metres mm: Millimetres

CHAPTER 1: INTRODUCTION

Land use change is a widespread, accelerating and very significant process due to human activities (Loveland et al., 1999). Over centuries, human populations have increased, shifted their distribution and changed the way in which they have used natural terrestrial resources to sustain themselves. These human transformations of the land surface have changed the earth’s ability to support human well-being especially through their effects on ecosystem service delivery (Cohen et al., 2000).

The way that the land surface is used, via the transformation of its original cover, lies at the heart of sustainable socio-economic development. According to Brown and Duh (2014) “land use” and “land cover” are two related words but not synonymous, in that the same land cover may be subject to different land uses. Changing land cover provides a window into understanding changing land use, but interpretation is required to determine the drivers behind that land cover change, and the effects on sustainable socio-economic growth. Land cover transformation has been driven by increasing mechanisation since the rise of intensive agricultural and cultural practices, and especially over the past 50 years globally (Steffen et al.,2004), nationally (Hoffman, 2010) and locally in many parts of South Africa.

Human activities have transformed natural ecosystems, and managed or semi-managed systems over an ever-increasing fraction of the earth’s surface (Foley et al., 2012). In pre- agrarian times, humans were depended on hunting wild animals and accumulate wild plants for their survival (Hancock, 2012). Farming was established at least 10,000 years ago and it has undergone extensive development since the time of the initial adoption of cultivation (Gorlinski, 2012). In the previous century, agricultural activities have been characterised by enhanced efficiency, the substitute of work by manufactured fertilizer and pesticides, selective breeding, mechanisation, and water contamination (FAO, 2011).

Lately, there has been a response against the extreme ecological impacts of conventional agriculture, leading to the organic movement (Gorlinski, 2012). Since the mid-20th century, new anthropogenic climatic and atmospheric compositional changes have also begun to drive land cover change (Meiyappan and Jain, 2012). It is increasingly recognised that human society may be approaching limits to development, also termed “planetary boundaries” (Steffen et al., 2015). The planetary boundary approach is centred on the concern that

1 increased human disturbances of critical Earth System processes escalate the risk of threatening the Earth System and challenging the environmental framework within which human societies have developed (Steffen et al., 2015).

It is crucially important to understand the drivers and the impacts of land use and land cover change in South African landscapes, especial as many parts of the country approach what appear to be limiting resource conditions. These are due to full allocation of limited surface water resources (DWAF, 2009) and agriculturally productive land (FAO, 2000a). This is not only important information for the development of policy and planning in the face of climate change and future development demands, but also for allowing better assessment of the potential role of natural or semi-natural ecosystems in providing resilience to future climatic trends. Furthermore, the issue of land use and land ownership is central in addressing the inequities in South African society due to the well-known legacy of apartheid policies (White Paper on South African Land Policy, 1997).

It is interesting and important to study a catchment containing multiple land cover states, ranging from heavily transformed to a semi-natural and natural condition. Such a study allows the assessment of human impacts on ecosystem services and further interpretation of which pathways of development could be sustainable versus unsustainable. The Goukou catchment represents such an area, with strong transformation driven by socio-economic factors and market demands within the South African political landscape. Furthermore, the Goukou allows reconstruction of the process of land transformation through black and white aerial photographs that have been captured roughly decadal since 1950..

There are currently large knowledge gaps with reference to the drivers and the effects of land cover change in South Africa, and catchments such as those of the Goukou River, but it seems that much progress could be made if recent land cover change trends and their drivers could be better understood, and if the relative costs and benefits of these changes could be better quantified (SAEON, 2011). A coherent and consistent system of land cover classification is crucial for increasing our understanding of decision making on land cover and land use change and the impacts of land management as well increasing our understanding of the processes by which changes occur, also it will assist in understand the dynamics and trends in land use and land cover change (Lambin et al., 2003). Systems of land cover are dependent on spatial scale, so at smaller scales the system needs to be more “tailor-made”, but must be translatable to a national and even larger scale system (FAO,

2000). The portrayal of class limits ought to be clear, precise, and maybe quantitative and based upon targeted criteria (FAO, 2000).

Regardless of the fact that there are various classification systems that exist throughout the world, there is no single worldwide putative land cover classification system (Duhamel, 1995; Di Gregorio et al., 1997; Di Gregorio, 2005). Due to this FAO and UNEP have developed the Land Cover Classification System (LCCS), within the framework of the Africover Program which allows comparison of land cover classes irrespective of mapping scale, land cover type, data collection method or geographical location (UNEP/FOA, 2005). Usually, land cover classes are developed for a particular reason and are inappropriate for other tasks (e.g. statistical needs), the scale is related to a specific task and information is typically out dated (Anderson et al., 1976; Di Gregorio et al., 1997; Di Gregorio, 2005). An example would be the UNESCO Vegetation Classification, designed to serve mainly for vegetation maps at a scale of 1:10 000 000, which consider only natural vegetated areas, while all other areas such as agricultural activities and urban development are ignored. Moreover, the use of factors in the classification systems brings about an undesirable mixture of possibilities and actual land covers (Townshend, 1992).

Once a classification system is established and converted into a useable source of data, such as a system of land cover characterisation can provide a baseline to observe land use and land cover change at a catchment scale. It can likewise help with adding to the understanding that permits models to be created for anticipating land use and land cover change into the future. Such a database would likewise be suitable for an enhanced capacity to couple both land use and land cover with research in climate change adaptation, ecosystem service delivery, carbon stocks and hydrological cycles through developing scenarios.

Such a system would also provide a valuable service to related projects within South African National Biodiversity Institute, and the broader land cover change community that are exploring these areas of concern. This study aims to bridge the scientific or implementation gap through providing a better platform for observing changes in the land use and land cover that in turn affects the service delivery of both ecological and hydrological ecosystems services.

The Goukou River catchment faces a number of natural and unnatural environmental issues. These issues arisen due to different factors such as agricultural increase, increasing alien invasive species, wetland degradation and other such factors. The catchment therefore

3 provides an ideal focus for study of the associated effects that these transformations have on the environment, with the potential to quantify past change and its effects, and project their future changes and effects.

1.1 Problem statement

Land cover change is a strategically important issue for planning and development that underpins societal and ecosystem well-being. As human pressure on the environment grows, a greater demand for goods and services drives land cover change. A basic issue facing South African culture is to find the most sustainable ways to meet developing human needs for living space, sustenance, fuel, and different materials while supporting ecosystem services and biodiversity. These demands occur within the context of historical changes that may have altered the potential for future sustainable land use. It is crucial to understand how historical land use change and its drivers have affected the potential for future development in South Africa, how they could continue to affect this if trends continue, and how more sustainable trajectories could be achieved. The Goukou catchment in the southern Cape provides a valuable test bed to explore this problem.

1.2 Hypotheses

The hypothesis is that changes in land cover and land use due to human activities have significantly altered catchment surface conditions, and these are detectible and interpretable through analysis of remotely sensed Black and White Aerial Photographs and Colour Infrared Images. I intend to test this hypothesis by quantifying changes to the surface area of identifiable land cover classes based on the BWAP and CIR since the 1950s. I will also use change projection via Markov chain analysis of processed and classified land cover maps over the full time period represented by the decadal intervals. These are amongst the appropriate post analysis techniques for detecting and projecting future changes of land cover (Mirkatouli et al., 2015). By using simple quantitative change detection methods, the amount of gains and losses, and thus the net change can be observed. However, the drivers of changes of each land use and land cover category can only be inferred. Markov chain analysis produces transition probability matrices that give the future probable distribution of each land cover type in a defined time period, providing a projection of future land cover assuming that the same drivers continue to operate. This projection may be useful for considering

4 interventions to achieve alternative land cover change trajectories that may be more sustainable.

1.3 Main objective

The main objectives of this study were (i) to determine the main types of land cover and their geographical distributions, (ii) to determine historical changes in land cover and their rates, and (iii) to develop a Markov model of the observed land cover change, and use this to project continued land cover transformation assuming that recent historical trends continue.

1.4 Key questions

The following questions were asked in order to address the hypothesis.

1. What are the historical geographical patterns of land cover change in this catchment?

2. What are the main changes between land cover types and what are their rates?

3. What are the potential implications of these changes?

4. What are the potential future patterns of land cover change assuming historical trends continue?

5. What preliminary recommendations could have value in reducing the impacts of land use and land cover change in this catchment?

1.5 Thesis outline

This thesis is divided into five chapters. Chapter one introduces the study and provides a description of the study area.

Chapter two focuses on the literature review of the related work to this study. Literature is consulted and reviewed that focuses on land cover classification systems, the importance, uses and challenges of using historical aerial photographs, land use change, drivers and impacts. Furthermore, literature is reviewed that focuses on the impacts of land cover change on ecosystem services, impacts of land cover change on water quality and quantity, impacts of land cover change on ecological infrastructure, land degradation and impacts of land cover change on climate change.

Chapter three deals with the materials and methods used in this study to examine and quantify the impacts, rate and geographical distributions of land cover and land use change in the Goukou catchment.

Chapter four addresses the research questions. It provides detailed results and discussions on supervised land cover classification, accuracy assessment, trend analysis, change detection, and the results of Markov chain. This chapter also highlights limitations to the study.

Chapter five provides an overarching conclusion to the study, including preliminary recommendations that might be considered to reduce the adverse impacts of ongoing land cover and land use change in the Goukou catchment.

1.6 Description of study area

1.6.1 Location of the Goukou Catchment

The Goukou catchment is a secondary catchment under the DWAF quaternary catchment system, number H9, covering quaternary catchments H90A to H90E. The Goukou catchment is located in the south-west of South Africa in the Western Cape, under the municipality of Hessequa, in Eden district. The Goukou catchment is about 30 km South of Riversdale at 340 22' S (lat) and 210 25' E (long). The catchment drains the Langeberge Mountains and flows south to the coast, west of Mossel Bay in the South Cape. The following figure (Figure 1) shows the location of the Goukou catchment in relation to the Western Cape as a whole.

1.6.2 River tributaries

There are five main tributaries draining into the Goukou River, namely the Soetmelks, Naroo, Brak, Vet and Kruis rivers. The main water supply to the catchment is from major dam called Korente-Vet dam, which is about 8.1 million m3, and it supplies Korente-Vette River irrigation canals, north-east of Riversdale, and the water needs of the town of Riversdale (Pitman et al., 1981). The length of the river from the source to the sea is approximately 64 km.

1.6.3 Climatic conditions

The mean annual precipitation for the overall Goukou catchment is 482 mm while that of the upper catchment is 634 mm (Estuaries of the Cape: Report 34, 1990). The Goukou River lies within a climatic region which receives rain almost uniformly spread throughout all seasons with peaks in autumn and spring. The mean annual temperature in the area is approximately 26 0C in summer and 16 0C in winter, with the extreme temperatures in summer that reach 42 0C and 32 0C, respectively prior to 1990 (Estuaries of the Cape: Report 34, 1990).

1.6.4 Ecological state

The recent ecological state report for the Goukou River catchment indicates that the river condition deteriorates rapidly from the source to the sea (CSIR, 2011). The upper reaches of the Goukou River are reasonably pristine wetland areas. The middle reaches, however, are infested with black wattle and hyacinth, while the lower reaches are impacted by low flows from over-abstraction and have little to no freshwater inputs.

1.6.5 Land use types and practices

The most recent existing land use data of the Goukou catchment has been provided by Harrison et al. (2001). Land use within the catchment comprises both dryland and irrigated agriculture and commercial forestry. Intensive agriculture accounts for about 35% of the current land cover. The main crop types under irrigation are vineyards, lucerne, and pasture. About 2% of the catchment is degraded shrub land, while natural land cover comprises about 63% and comprises of shrubs, grassland, and grass with some water bodies and wetlands (Harrison et al., 2001; CSIR, 2011).

In the upper part of river catchment land is held in privately owned farms, or is in State hands. In this region cover is natural or semi-natural, and some mixed farming is practised; this includes cultivation of small grain cereals (mainly wheat) and dry land pasture in combination with wool sheep and dairy cattle. Urban development is limited, limited to middle and lower catchment, and mainly comprises residential area and industrial developments associated with the coastal settlement of Stilbaai and the inland town of Riversdale, and accounts for less than 1% of the catchment. Land is significantly transformed

8 by intensive agriculture in the middle catchment, and much less so in the upper and lower catchments (See Figure 2).

1.6.6 Geological structure

The Goukou River originates in the Langeberge Mountains of the Cape Fold Belt formed of the Table mountain sandstone of the mid-Palaeozoic Cape super group. The river passes through 10 km of high erosive cretaceous sedimentary rocks of the Enon formation followed by about 40 km of Palaeozoic Bokkeveld shale. The Goukou River enters the coastal zone in a gentle slopping surrounding of tertiary aeolinites (dune rocks) and coastal sand, with a few outcrops of Table Mountain Group sandstones (e.g. Morris point) (CSIR, 1990; Reed, 2014).

1.6.7 Current land use related pressures

Table 1 shows the current land use related pressures that the catchment is facing, which can be divided into four categories, namely environmental hazards, land cover change, infrastructural development, and degrading water quality and quantity.

Table 1: Current land use related pressures in the Goukou catchment, as determined by C.A.P.E Estuary Programme Report 2007.

Category Types or Activity Impacts on the ecosystem services

Land cover Livestock grazing Natural wetlands are damaged by domestic animal grazing. This leads to reduced productivity, habitat destruction and ultimately bank erosion. Agricultural return Pollutants from farming activities in the catchment flow and surrounding environs pose a threat to the catchment ecosystem (water quality). Alien vegetation The fast growing of alien invasive species (mainly Acacia mearnsii, Acacia saligna and Sesbania punicea) shades the river, which results in reduction in sunlight available for the aquatic biodiversity (indigenous plants, fishes and invertebrates). In addition, they utilise or extricate large amounts of water. Environmental Floods Riparian vegetation and wetlands naturally protect hazards and reduce floods against the catchment. However; removal and degradation of wetland and riparian vegetation makes the catchment vulnerable to floods and reduces soil fertility.

Category Types or Activity Impacts on the ecosystem services

Droughts Many living organisms (plants and animals) that depend on or require water for their survival will eventually die, once this happens; the land becomes barren and uninhabitable. Droughts deteriorate the water quality and natural areas and thereby affecting the way the land will be utilised in the future. Climate change Flow changes, sea level rises and increased storminess pose long-term threats to the ecosystem and livelihoods in the area. Clearing of Through gaining access to high yielding soils, riparian vegetation clearing of riparian vegetation leaves the catchment vulnerable to erosion, while burning of reeds and sedges for grazing often poses a similar risk. Infrastructural Road infrastructure Existing road infrastructure encroaches on the development Goukou estuary and floodplain reducing its resilience to development pressures. Riparian Continual degradation of saltmarsh and natural infrastructure riparian vegetation by low-lying developments and (fences and low- infrastructure reduces the protective effect that lying natural vegetation produces against wave action developments) (caused by tidal action) and floods. In-stream In-stream infrastructure interferes with the natural infrastructure hydrodynamics under high flow conditions. Artificial bank stabilisation associated with in- stream infrastructure such as jetties introduces foreign habitats to the system. Water quality Wastewater The quality of water is important, as poor water and quantity quality poses a risk to the natural environment and human wellbeing (pollutants from wastewater discharge). Water abstraction The over-allocation of water resources in the

Category Types or Activity Impacts on the ecosystem services

(direct abstraction, catchment deprives the Goukou Estuary of the groundwater and freshwater necessary to sustain a healthy ecosystem. fountains) Over-exploitation of groundwater resources (from fountains) could cause these fountains to cease providing a habitat that nurses eels and other wild species that are seep-dependent.

CHAPTER 2: LITERATURE REVIEW

2.1 Land cover classification systems

For several centuries, globally, many research bodies have been collecting information about land surface, but have operated largely independently and without coordination (Brown et al., 2005). Frequently, this has implied duplication of effort, or it has been observed that information gathered for a particular objective may not be applicable for other uses at a later stage (Büttner et al., 2002). The aim of a land cover classification systems should be a complete, reliable system, designed to meet individual requirements for specified users, and applicable for mapping that is autonomous of the scale or means used to gather the information (UNEP/FAO, 2005). Classification systems have been developed throughout the country for current land use and land cover, on a foundation that is uniform in classification, and is generalised at higher levels of class grouping. Classification system needs also to be receptive to data from satellite and aircraft remote sensor information (Agarwal et al., 2002).

An Interagency Steering Community on Land Use Information and Classification was developed early in 1972 in USA due to the necessity of management agencies to develop a far reaching review of national land use and land cover trends, patterns and environmental standards (Anderson et al., 1976). The principle goal of the board was to create a national classification system that would be responsive to inputs of information from both customary sources and remote sensors on high altitude aircraft and satellite platforms, and that would in the meantime shape the structure into which detailed classes of land use studies by region, state and local agencies could be tailored and aggregated upwards for less detailed small scale use at the country wide level (Darwin et al., 1996).

The information defining present land cover is a significant input for planning and modelling, but the quality of such information plays a major role in determining the consistency of the simulation outputs (Belward, 1996). It is essential for a modern nation to record and have adequate information on various, composite and interrelated aspects of its land cover activities in order to make appropriate decisions. Land use is only one such aspect but information regarding land use and land cover change has become more important as the nations’ plans to overcome the issues of overarching development, deteriorating

13 environmental quality, loss of prime agricultural land, destruction of important wetlands, loss of aquatic life and wildlife habitats (Sims, 1995).

Current land cover classification systems can be described as either vegetation classification systems, broad land cover classification or systems related to the description of the functional use of features e.g. agricultural areas. Consequently; they are limited in their ability to describe the entire variety of possible land cover classes (UNEP/FAO, 2005). An example of this is the UNESCO Vegetation Classification, intended to serve mainly for vegetation maps at a scale of 1: 1 000 000, which considers only natural vegetation, whereas other vegetated areas such as agricultural zones and urban vegetated areas are not dealt with (UNEP/FAO, 2005).

There are two main types of basic classification systems, namely hierarchical and non- hierarchical (Di Gregorio, 1991; UNEP/FAO, 2005). Hierarchically structured classification systems give greater consistency because of their ability to accommodate various types of data, and as a result most classification systems are hierarchically organised (Di Gregorio and Jansen, 1997). Hierarchical systems may be less useful, because, as stated by Otara (2012): “A non-hierarchical method generates a classification by partitioning a dataset, giving a set of generally non-overlapping groups having no hierarchical relationships between them”. Many classification systems are not useable for many mapping purposes and subsequently for monitoring purposes (Di Gregorio and Jansen, 1996c). It is unlikely that one ideal classification system of land use and land cover could ever be developed (FAO, 2000). Nevertheless; individual classification system are often needed to meet the requirements of specific users (Eiten, 1968).

Regardless of the apparent need for a uniform classification system, none of the existing land cover classifications has been internationally accepted (Duhamel, 1995; Di Gregorio et al., 1997). This is because, as discussed above, most land cover classes have been developed for a specific purpose or scale and are hence not appropriate in other contexts. Since there is no generally accepted or applicable system, FAO and UNEP have developed a generic Land Cover Classification System (LCCS) which allows comparison of land cover classes irrespective of mapping scale, land cover type, data collection method or geographical location (UNEP/FOA, 2005).

2.2 Historical aerial photographs importance, uses and challenges

Historical aerial imagery offers the analyst a more extensive historical perspective on geographic processes such as land use and land cover change, urban growth and development, and vegetation patterns, by providing a unique opportunity to study long-term patterns of land use and land cover change (Kadmon and Harari-Kremer, 1999; Awwad, 2003; Alhaddad et al., 2009). Greater perspective on historical changes such as agricultural activities and loss of natural vegetation over time can be gained by utilising such imagery (Foody et al., 1995; Biediger, 2012). In this regard, aerial photographs serve as a fundamental source of historical data on vegetation cover and landscape conditions (Fensham and Fairfax, 2002).

This study used grey-scale images from 1940 to 1991, which are referred to as historical aerial photographs or images (Ceraudo, 2013). Aerial black-and-white photography yields analogue images, as opposed to modern digital imagery, because it has been directly imaged onto film (Wetzstein et al., 2011). It may also be referred to as either panchromatic or grey scale imagery (Beidiger, 2012). Aerial photographs remain the primary important source of information on land cover before the accessibility of satellite imagery, even today (Cost- Folch et al., 2007).

Historical grey-scale imagery provides only one band of data, which complicates the interpretation and analysis of the data (Siljander, 2010). Land use and land cover classification usually involves visual interpretation and manual digitising, and the difficulty of using digital image processing techniques with the historical images is one of the major challenges when working with Black and White Aerial Photographs (BWAP). Foody et al., (1995) argues that, “the ability to facilitate digitising of land cover types on historical aerial imagery would make it a more usable resource to study long-term land use and land cover transformations over time”.

2.3 Land use change, drivers and impacts

Changing land cover is a phenomenon that is growing in extent and importance, both globally and in South Africa. This includes conversion of natural habitats to agricultural crop and forest plantations, and that of natural vegetation through bush encroachment and overgrazing (Midgley et al., 2012). These land use types have very different but important impacts on ecosystems service delivery (Schaffartzik et al., 2015). Demands for natural resources will increase due to continuing population growth, and consumption growth (Godfray et al.,

2010). Humans have made major modifications in their environment. This has occurred over multiple centuries and in virtually every part of the inhabited earth. Most of major metropolitan areas face the growing problem of urban development, loss of natural habitats and open space, and an overall deterioration in the magnitude and affinity of wetlands and wildlife habitat (USGS, 1999). Despite international goals to advance its sustainable management, human pressure on land has increased during recent years (UNEP, 2012). Human activities are the main causes of biodiversity loss and habitat fragmentation as well as rapid changes in the ecosystems services and landscapes around the world (Vitousek et al., 1997).

One important result is that, increasing human demands for biologically productive land to provide food and fuels limits humanity’s ability to preserve biodiversity (Lambin et al., 2011). Intensive agricultural systems are concurrently degrading land, water, biodiversity and climate change on a global and local scale, while growing human population and consumption are placing exceptional demands on agriculture and natural resources (Foley et al., 2011). Increasing human population in developing countries is putting pressure on their limited land resources and causing land degradation (FAO, 1995). Lack of capability in the scientific, technical and application domains limits the ability to address these environmental issues, and this is illustrated by land degradation in dryland ecosystems (UNEP, 2012). Environmental degradation created by human activities, driven by their choices, alters the delivery of ecosystem goods and services (Cohen, 2003).

The dominant driver of global environmental change has been increasingly recognised as human activities on land (e.g. Moran, 2001). Resource depletion, ecosystem degradation, and competing pressures on land are mounting: deforestation rates are still high in the tropics, wetland destruction continues, environmental impacts of droughts have intensified and productivity of drylands is decreasing, these issues rise as the results of continuing consequences of land use and land cover drivers (UNEP, 2012). Interrelated effects of land use and land cover drivers need to be understood for proper planning and decision making processes (Ali et al., 2011).

Vulnerability of places and societies to natural disasters such as erosion and floods is caused by modifications in land cover and land use, which in part affects the capability of biological systems to support human needs (de Sherbinin, 2002). Land cover and land use changes often indicate major effects on biodiversity, especially those changes that show loss of natural

16 habitats due to urban sprawl and cultivation (SANBI, 2009). There is a need to wisely manage the increasing pressures and competing demands on land to avoid the intensification of land degradation (UNEP, 2012). Information on current land use patterns and changes in the land use through time is one of the primary prerequisites for better use of land (Anderson et al., 1976).

Before the 1950s, land cover in the Goukou catchment was predominantly natural vegetation, with few land use limited to scattered agricultural activities that were happening mostly in the mid-catchment. Wetlands were pristine and covered by palmiet vegetation. After the 1950’s, agricultural activities can be seen to have increased to occupy about a quarter of the catchment by the year 2010.

Dam construction increased and the land became susceptible to natural hazards, such as floods and erosion. Moreover; the ecological state of the river catchment deteriorated. The upper reaches of the river are now affected by wetland degradation, while the mid-reaches are heavily infested by alien vegetation, and affected by over-extraction of water and polluted return flows from agricultural activities. The lower reaches are impacted by reductions is fresh water due to irrigation and return flows from agriculture. The details of how these changes occurred over time are not clear.

2.4 Impacts of land cover change on ecosystem services

Ecosystems provide ecosystem services for the benefit of people. One of the greatest challenges of our time is to maintain these services and other goods while maintain ecosystem functions and biodiversity that support their sustainable source (Lawler et al., 2014). These services are categorised according to their relevance of use; (a) Provisioning services are widely used and can fit in any discipline as it deals with poverty alleviation, examples are food and water. (b) Regulating services are more useful in the agriculture sector where small scale farmers are hit the hardest when these services are not properly managed, examples are flood and disease control. (c) Cultural services play a vital role at society level where communities rely on these services for spiritual and cultural benefits. (e) Supporting services such as nutrient cycling are more important for sustainable management of our ecosystems for the benefit of future generations and ensuring their resilience to climate change and land cover change (Leadley et al., 2010). Socio-economic characteristics, land

17 use, biodiversity and climate determine the future capability of the ecosystem to provide these services (Metzger, et al., 2006).

Land use change can significantly modify the delivery of ecosystem services globally, the transformation of natural grasslands, forests and wetlands into croplands, tree plantations and developed ones, has led to vast increase in production of food, timber, housing and other commodities, but at the cost of reductions in many ecosystem services and biodiversity (Joshua et al., 2013).The increasing demand for the provision of natural ecosystems, the effects of land use changes on ecosystem services and their economic value have become a focus of concern for scientists, policy makers and stakeholders over the last decade (Chen et al., 2014). The influences of land use change on ecosystem service are complex and interdependent (Schutte, 2012), and further influenced by human population pressures (Turner et al., 1995).

Due to a desire for greater short term benefit and a lack of knowledge of ecosystem services value, people have globally developed many natural ecosystems into cropland and building land in the past, which resulted in the alteration and destruction on the functions of ecosystem goods and services to society (Chen et al., 2014). Changes in private returns to landowners owing to land use change can be often be measured, but modifications in the supply and importance of ecosystem services and the delivery of biodiversity maintenance have been harder to measure, while changes in magnitude and composition of plantations, grasslands, wetlands and other natural environments have great effects on the delivery of ecosystem services, biodiversity conservation and returns to landowners (Polasky et al., 2010).

Changes in land cover such as draining of wetlands, clearing of land for agricultural practices, felling of forests for timber production and pollution of environment and fragmentation occurs via the modification and destruction of habitats which has ultimate adverse impacts on ecosystem services (de Sherbinin, 2002). One the greatest significant driver of the change in ecosystem service delivery is land cover change; however information on the effects of land cover change for ecosystem and human well-being at local scale is largely absent (O’Farrell et al., 2009). Transforming the landscape can have a serious environmental impact that affects ecosystem services (Hu, et al., 2007). Land use change often leads to ecosystem degradation, which may be slow or abrupt (Schutte, 2012), and also can lead to utilised, replaced or completely removed ecosystems (Hobbs, 2000).

Direct impacts on biotic diversity, biological systems and their ability to support human needs and environmental risks such as flooding, water pollution and climate change occurs as the results of land use and land cover change which is a direct driver of change in the ecosystem services (Verburg et al., 2013). These occur with a number of consequences such as direct impact on biotic diversity, biological systems and its ability to support human needs and environmental risks such as flooding, water pollution and climate change (Daniel, 2008). Human drivers of changes in the terrestrial surface hold wide-ranging significance for the structure and function of the ecosystems within the earth system (IPCC, 2014), with equally far-reaching for human well-being (Turner et al., 2008).

The ability for an ecosystem to deliver services for human well-being is largely affected by changes in land cover which affected the biological diversity (Hussain and Miller, 2014). Ecosystem services play a central role in both adaptation and mitigation of climate change (OECD, 2008). Provision of ecosystem services, biodiversity conservation and return to land owners are largely affected by changing the extent and composition of forest, grassland, wetlands and other ecosystem services (Polasky et al., 2010).

Having basic material for good life, health, good societal relations and freedom of choice and action is associated with human safety (Proenc et al., 2011). Ecosystem services influence many of these factors, playing an important part in providing resources for a good life, health, safe access to resources and safekeeping from disasters (Fisher et al., 2014).

2.5 Impacts of land cover change on water quantity and quantity

Not only in developing countries like India and many other African countries that freshwater scarcity is an issue, but also in countries and regions traditionally considered as water rich countries like United States of America (GEO, 2000). Water scarcity is a critical issue because about 18.3% people lack access to safe, clean and freshwater (WHO, 2003), whilst agricultural food production consumes more than an average of about 90% of global fresh water (Shiklomanov, 2000).

For South African society to lessen the cost of cleaning its water resources, it is necessary to first address the sources of pollution that contaminate the rivers and its tributaries, this means fixing old infrastructure, such as sewerage plants, and preventing pollution effects from agricultural runoff and failing storm water systems and also repair and safeguard the wetlands

19 and water catchments that feed the river’s systems that supplies the nation's water needs (Maze and Joubert, 2014). One of the main threats and challenges humanity faces today is water pollution together with the loss of biodiversity, climate change and socio-economic issues (Hughes et al., 2012). The value of the quality and quantity of freshwater is essential for sustainable development (Gleik et al., 2014). Human, ecological health and the use of water are directly affected by water quality, therefor, the quality of water is imported (Chin, 2006). Surface water quality management is very important as surface water bodies are the potential receivers of the impurities contained in the surface run-off from their catchment (Groppo et al., 2008). Nevertheless, human activities often adversely affect water resources (Gulbaz et al., 2012). Overuse and pollution, due to inadequate consideration of environmental integrity has resulted in drastic impacts on water quality and quantity (Obata, 2010).

Forthcoming land use alterations must consider possible impacts on hydrological systems in order to meet ecosystem and human requirements, especially trade-offs between water, salt and nutrient balances, to improve sustainable water resources (Scanlon et al., 2007). The amount of available freshwater resources for both ecosystems and people is reduced by water contamination (Ganoulis, 2009). Water pollution due to human activities and natural hazards can become a risk to human well-being, risk to natural ecosystems even risk to nation’s economy (Groppo et al., 2008).

A combination of low rainfall, high evaporation rates and growing population, whose geographical demands do not conform to the distribution of exploitable water supply, has resulted in the water supply crisis that South Africa is facing today (O’Keeffe, 1998). Water is a limited natural resource and with the country’s rapidly growing population and thriving economy, water requirements will soon exceed natural availability (Mojtaba, 2012). The challenge of ensuring fresh, clean and good water quality and quantity are becoming increasingly prominent as population increases (CRS, 2005). The fact is that water is the main component in the balanced functioning of natural habitat and as such, it must be managed in combination with other natural resources (Dent, 2001).

Major determinants of water quality and quantity are land use and different kinds of human activities (Obata, 2010). The effects on water quality and quantity by land use and land cover change are poorly identified (Chu et al., 2013), therefore, the role of land use types on water

20 sustainability should be investigated to avoid present and future problems like floods and water pollution (Gulbaz et al., 2012).

Naturally vegetated areas are increasingly converted to agriculture, recreation, and urban cover, so it is essential to look at the influences of these changes on water quality and quantity (Lohani et al., 2007). Effects of land use change on water resources, particular those associated with human activities (e.g. agricultural activities), may rival or exceed those associated with the climate change, but the focus has been placed on the effects of climate change on water resources (Vorosmarty et al., 2000). Past land use modifications have greatly impacted global water resources, with often contrasting effects on water quality and quantity (Scanlon, 2007). Original vegetation that is being replaced, vegetation that is replacing it, whether it is a permanent or temporary change and associated land management practices and related water and nutrient application, determines the effects that land use has on water resources (Scanlon, 2007).

Both ecosystem services and local communities rely on catchment for water supply and other goods services (Stipinovich, 2005). Many catchments are threatened and impacted upon by land degradation through poor resource management, as well by the spread of invasive plant species (Hughes et al., 2012). Huge quantities of water are unsuitable for various uses, this is due to human activities and human-related substances and wastes introduced into rivers, lakes, groundwater, aquifers and the oceans that modifies environmental water quality (WMO, 2013). This consequently affects terrestrial and aquatic ecosystems where clean fresh water is a prerequisite for life; it does not only affect human related uses such as drinking, bathing, agricultural irrigation and industrial production (Vorosmarty et al., 2010).

2.6 Impacts of land cover change on ecological infrastructure

Ecological infrastructure refers to naturally effective ecosystems that brings valuable services to people, such as healthy mountain catchment, rivers, wetlands, coastal dunes, nodes and corridors of natural habitats, which altogether form an interconnected structural elements in the landscape, therefore, it is the benefit from which a variety of valuable services flow (SANBI, 2014). Because ecological infrastructure is largely free, its value is rarely captured in market transactions and we tend to under-invest in it (Asian Development Bank, 2012). South Africa has abundant ecological infrastructure, providing opportunities to support the development and unlock economic potential (Asian Development Bank, 2012).

Improving resilience to climate change, reducing risks of natural hazards and improve food and water security is every country’s goal, this can be achieved by managing and monitoring ecological infrastructure which can be a cost-effective approach that will result in achieving those goals (DESA, 2013), all of which contribute directly to poverty alleviation, sustainable livehoods and job creation (UNEP, 2012). Planning the investment in ecological infrastructure extends the life of current infrastructure and can lessen the necessity for additional built infrastructure; often with significant cost savings (Asian Development Bank, 2012).

Growing populations will require an increase in the capacity of existing ecological infrastructure if it is continuing to produce the range of ecosystem services necessary for our present living standards to be maintained and improved (Bristow et al., 2010). Anthropological effects on the ecological system are often not seen for many years, because of the time it takes for a given action to propagate through the components of the system (Dale and Hacuber, 2001). Landscape pattern is a product of the interaction between ecological and socio-economic processes (BenDor et al., 2014), and thus revealing of this relationship.

Provision of key ecosystem services such as water (both quality and quantity), soil formation and pollution is undermined by competing land uses that contribute to the overexploitation of natural resources which consequently degrade natural resources resulting in an overall adverse impacts on ecological infrastructure (Asian Development Bank, 2012). Socio- economic factors influence land use decision making which strongly affect ecological dynamics in human-influenced landscapes (Berry et al., 2011). The significant ecological consequences of land cover (Turner et al., 1997) and land use can persist for many decades and it is not yet clear if the ecological damage caused by land use change can be reversed (Palmer et al., 2010). In many ways the effects of land cover and land use change on ecological processes remain poorly understood (Tyler et al., 2008).

Important components of ecological infrastructure are located in rural areas, such as natural catchments, corridors, tracts of natural vegetation (SANBI, 2012). Overgrazing in these catchments, poor planning, and the construction and maintenance of roads has damaged wetlands and river ecosystems causing severe erosion, floods and land degradation (Ayivor and Gordon, 2013). One of the greatest threats on freshwater resources has been identified as human-induced changes on natural landscapes (Palmer et al., 2010). Sediments, hydrologic

22 and nutrient regimes are influenced by land use change, which in turn influences aquatic biota and ecological processes in freshwater (Palmer et al., 2010). Rivers play a significant role in the delivery of water for domestic, agriculture and industrial purposes (Ayivor and Gordon, 2013). Nevertheless, land use dynamics continue to affect river catchments in ways that have adverse effects for river health (Ayivor and Gordon, 2013).

From a development management and land use planning point of view, a regional scale allows a strategic view on land use planning and policy to sustain the functional interaction and spatial coherence of ecological infrastructure under the pressure of increasing land use intensity. Understanding the purpose and the structure of landscapes, mainly in terms of anthropological impacts, requires integration of biological and socioeconomics knowledge. Natural resources managers, in particular, need this integration to successfully assess the social and environmental consequences of alternative management scenarios (Berry et al., 2011).

Inadequate investment in ecological infrastructure has led to a worsening environmental crisis (Harris, 2013), in which critical ecosystem services have been lost in many regions across the globe (Bristow et al., 2010). Understanding the relationship between current and future patterns of ecological infrastructure is vital for sustainable development (Bowman et al., 2009). If land use and natural land cover can be spatial organised to create and maintain a semi-natural region landscape matrix, then the cumulative effects of land cover change can be reduced (Tyler et al., 2008).

2.7 Land degradation

Maitima et al., (2009) defines land degradation as the lack of capability of the land to provide ecosystem goods and services and thereby assures properly functionality over a period of time for the beneficiaries of these. Land degradation is a worldwide concern that is also affecting South Africa (Magaba, 2012). There are several of factors that causes land degradation such as extreme weather conditions, particular droughts and human activities which affect soils and utility thereby have negative impact on the production of food, livelihoods and the production and delivery of other ecosystem services and goods (Maitima et al., 2004). Concerns about the modifications of land use and land cover has emerged globally, due to the understanding that land surface activities impact climate and those alterations influences the provision of ecosystem services and goods (Lambin et al., 2003).

The impacts of land use on biological diversity, soil degradation and the ability of biological systems to support human needs has been of major concern (Maitima et al., 2004a).

The intensification of agricultural activities and livestock production, urbanisation, clearing of forests and extreme weather events such as droughts and coastal surges has accelerated land degradation during the 20th century (Campbell et al., 2003). Large areas and many people in dryland regions are affected by land degradation (Campbell et al., 2003). Increased human population pressure and extreme human development into drylands during wet periods leave a cumulative number of people stranded during dry periods (Bajocco et al., 2012). World’s arable lands and pastures that are crucially important for the provision of food, water and quality air are under stress due to land degradation (FAO, 2006).

Socio-economic problems, reduction in land productivity, mitigation, limited development, loss to ecosystem service delivery including uncertainty in food security, are all effects of land degradation (GEO-4, 2007). Relocation of populations is not always by choice, sometimes it is due to pressures that land degradation places on them, for an example when land is degraded in some areas, food production is reduced, water resources dry up, deserts increase as results populations are forced to relocate to more friendly areas (Kaihura and Stoking, 2003; WHO, 2015). Increased demands for land by various sectors of the economy have resulted to conflicts about the usage of land (Maitima et al., 2009). The encroachment of humans into the protected areas raised a conflict between livestock keepers, wildlife conservationists, cultivators, individual land users and governments (Hoare, 1999).

In the last few decades, some of the most significant ecological problems globally have been identified as climate change, soil deterioration, and land use and land cover changes and land degradation (Millennium Ecosystem Assessment, 2005). If land is severely degraded, it may no longer be able to provide a variety of ecosystem services and this would result in a loss of goods and several other potential environmental, social, economic and non-material benefits that are critical important for humanity and development. Furthermore, it is costly to reclaim degraded land (FAO, 2006). Without taking into consideration the hidden cost of increased fertilisers use, loss of biodiversity and loss of unique landscapes, the cost of land degradation is estimated to be US $40 billion yearly global Campbell et al., 2003).

2.8 Impacts of land cover change on climate change

The integrity of the world’s natural ecosystems is being challenged by global warming and climate change, which is a major result of greenhouse gas emissions by human activities and natural climate change (Ometto et al., 2013). Moreover, by changing land use and thus land cover, people further affect climate (Brown et al., 2014, Solomon 2009). Indeed, changes in atmospheric greenhouse gases composition solely due to land use and land cover changes affect climate on a global scale (Boysen et al., 2014). The replacement of forests by agricultural activities and grazing lands which has occurred during the industrial era is estimated to have caused about half of all anthropogenic land cover change (Pollack, 2004; USGS, 2014).

Land surface changes affect local precipitation and temperature via feedback effects (Miller and Cotter, 2013). The contributions from increasing greenhouse gases due to land use and land cover may rival the climate response to gases emitted by industrial activity (Dirmeyer et al., 2010). Land cover modifies the physical properties of the land surface which in turn affect near-surface climate, mainly on the local to regional scale (Boysen et al., 2014), especially through albedo changes (USGS, 2014). Cloud formation and precipitation can be influenced by vegetation patterns and soil composition through their impacts on evaporation and convection (de Sherbinin, 2002). One of the most challenging aspects of projecting human effects on climate change are the influences of future land use and land cover changes on a regional and global basis (Trail et al., 2013).

Extensive deforestation has been associated with reduced rainfall, reduced cloud formation and enhanced shortwave radiation and temperature (Stone, 2009). Vegetation has a significant bio-geophysical influence on local climate regulation (Conditions et al., 2014); however the extent of the influence depends on the biome and geographical location (Bagley, 2011). Air temperature and near-surface moisture can be changed in an area where a large enough area of natural vegetation is converted to agriculture (Fall et al., 2010; Karl et al., 2012). As natural vegetation is removed and replaced with pastures and crops, a shift in the surface and atmospheric energy balances occur (Badger and Dirmeyer, 2015). These shifts in the energy balance modify near surface climate state variables such as water vapour and temperature (Bagley, 2011). Surface albedo and transpiration modifies surface temperature and latent heat flux and ultimately can cause changes in regional temperature and precipitation patterns due to the effects of land cover change on the exchange of water, energy and gases within the atmosphere (Dale, 1997)

While the conversion and management of land has released carbon to the atmosphere for centuries (Pongratz et al., 2014), the direct effects of land use change on climate change and climate variability did not receive priority attention for many years in the climate change discussion and policy making process (Pielke, 2005; Sulle, 2014). The significant changes in the global land use patterns in the past 50 years (Singh and Shi, 2014) have directly or indirectly exerted significant influence on global climate change (Deng et al., 2013). Land cover change is thought by some to have had the largest total impact on global climate of any human activity to date (Bagley, 2011; Mahmood et al., 2010), especially in the tropics (Seymour and Busch, 2014). A number of alterations in the regional and global climate and land system have likely resulted from land cover changes. These include increasing vulnerability to floods, modified atmospheric CO2 uptake, changing surface albedo, changing surface evapotranspiration, modification of wind patterns, and exposure to heat-waves and resilience to high temperature events (van den Hurk et al., 2012).

The realisation that land surface processes directly influence climate change has led to land use and land cover change emerging in the research agenda on global environmental change (Lambin et al., 2003). The land cover types present in a specific region determines the effects of land cover change on climate change (CIMA, 2006; ELC, 2013). Decisions to adapt to or mitigate the effects of climate change can be developed by people through control of land use and the resulting land cover (Brown et al., 2014). Land cover and land use practices determine the cover, and subsequently, the effects of that cover on weather and climate (Mukhopadhyay et al., 2013).

Demand for agricultural land and urban expansion could result in continuation of trends such as such as conversion of peat lands to agricultural land, which results in substantial emission of CO2 and NO2, despite the opportunity to mitigate climate change by protecting wetlands (UNEP, 2012). Understanding land use and land cover is thus crucial to understanding climate change (IPCC, 2007). Quantitative analysis of the effects of land use and land cover changes on surface climate is one of the fundamental scientific concerns to quantitatively analyse the impacts of land use and land cover change on the climate, so as to scientifically comprehend anthropological influences on climate change (Deng et al., 2013).

To summarise briefly, terrestrial vegetation influences climate change on a variety of scales through biophysical fluxes of water and energy, and thus widespread transformation of land cover has played an essential role in the development of recent climate to the state it is today

(Bagley, 2011). To date, impacts of biogeophysical effects due to land cover perturbations have been largely neglected in key simulations of future climate for policy makers, with the focus instead being on biogeochemical impacts (Davin and Noblet-Docoundre, 2010).

2.9 Theoretical and policy implications of land use and land cover change

Land is used to meet a variety of human needs and to serve multiple purposes. Desirable and undesirable impacts on other natural resources occurs as a consequence, due to the decisions that land users take to in employing this resource for different purposes (e.g. Gibbs, 2000). The analysis of land use is the analysis of the relationship between people and land, thereby providing key information on the drivers of land use change.

In South Africa, the Development Facilitation Act 67 of 1995 (Land Use Management Act) emphasises the need for sustainable use of land, and states the necessity for “a coordinated land use management system within national, local and other tiers of government. This requires a coordinating capacity to manage land use planning at a national government level, as well as inter-governmentally” (White Paper, 2010). Sustainable land use serves the requirements for food, energy, housing, recreational, ecosystem service goods and services, for all the societies living on earth today and into the future, and taking into account the limits and the resilience of ecological systems (Kaphengst, 2014).

Despite the development of theories and policy management strategies which have attempted to reduce the adverse impacts of land use change on natural resources and natural habitats, land use change remains an ongoing global challenge. Land cover and land use has adverse impacts on the environment if development is not properly managed. Land use management theory and strategy needs to be implemented effectively, supported by scientific assessment not only at a global scale, but also at a local scale for proper land use planning and development. The pursuit of greater land use benefits results in increased risk of environmental degradation (Wang et al., 2008). Land use strategy should therefore be based on maintaining the resilience of the ecological system.

2.10 Summary

The phenomenon of land cover change has been described as conversion of natural habitats through human activities (Townshend, 1992). The main contributing aspects of land cover change over the past few decades are anthropogenic, due both to population growth and

27 increasing demands for goods and services. This summary of the literature indicates that land cover and land use change are major contributing factors to a global deterioration in ecosystem service delivery, water quality and quantity, ecological infrastructure and to ongoing climate change.

The reviewed literature mainly focused on revealing adverse impacts of altering natural habitats on the environment. These are impacts on ecosystem services, and include water quality and quantity, and ecological infrastructure, with feedbacks relevant to regional and even global climate change. The conversion of natural vegetation such as wetlands to agricultural uses and changes to the extent and composition of natural vegetation has had significant impacts on the provision of ecosystem services. Water quality and quantity may be greatly affected by overuse and contamination, very often due to poorly informed behaviours and lack of knowledge by land users.

According to the literature on ecological infrastructure, degradation in natural resources is due to exploitative land uses which have contributed to the exhaustion of natural resources. As a result, key ecosystem services such as water quality and quantity, soil formation and pollution are undermined, which reduces their sustainability and undermines future economic development.

It is evident that land cover change is the key contributing factor in the changing landscape of the Goukou catchment. Different drivers have played different roles in altering land cover and land use change in this catchment, including natural disasters such as soil erosion due to heavy rainfall (CSIR, 2006), growing agricultural activities, and increasing disturbed surfaces. These can be quantified and revealed by analysis of BWAP and CIR images of 1940 to 2010, which are to be analysed using GIS and remote sensing in this thesis.

The impacts are likely to include adverse impacts on ecosystem services such as water quality and quantity. Due to significant shifts in hydrological function in this catchment, accelerated soil loss and wetland degradation are likely to lead to direct losses of many ecosystem services. At the same time, water extraction from the river has increased as a result of dam construction for the purposes of irrigation. Although the increase in agricultural practices at this catchment is in all likelihood playing some role in reducing poverty through providing work opportunities, its growth has affected many other natural resources.

CHAPTER 3: METHODS AND DATA

3.1 General overview of methods

In current strategies for monitoring environmental changes and managing natural resources, land use and land cover has become a central component. According to Meyer (1995), “land use and land cover are distinct yet closely linked characteristics of the Earth’s surface”. While remotely sensed satellite imagery is one of the most widely used primary sources of data for land use and land cover classification and change detection analysis, these data have become available only in the past two decades, post-dating significant changes in land cover during the 20th century.

Aerial photography was conducted long before multispectral satellite imagery was implemented via the deployment of satellites, and provides past land cover information with high spatial resolution. This is invaluable in land use and land cover change detection studies (Erener and Duzgun, 2009). The analysis of aerial photographs and remotely-sensed images with GIS has proved to be an effective method of displaying and analysing spatial and temporal data at different scales (Corrigan, 2009). A robust and consistent classification method is needed for monitoring land cover using digital Earth data that will allow the accurate mapping of land cover classes (Rodrriguez-Galiano and Chica-Rivas, 2014). This aim is complicated when different image types are involved.

In this study, GIS and remote sensing technologies were used in an attempt to reconstruct historical land use changes from 1940 to 2010, in the Goukou River catchment. Two distinct sets of images were used to gather information about land cover changes of the Goukou catchment, black and white aerial photographs (BWAP, for the period 1940 to 1991) and Colour Infrared Images (CIR, for the period 2006 to 2010). The images for 1940 to 1991 are high resolution images, which needed to be enhanced from their original state as they only provide a single spectral reflectance, and could not be used for automated land cover classification. As a result, textural analysis was employed to enhance information about land cover history in the Goukou catchment for this time period. Land cover types in CIR images of 2006 and 2010 were extracted more simply and directly by using on-screen digitising process.

Specific steps that were performed on BWAP were: image projection, geo-referencing, image cropping, mosaic, image subset, textural analysis. The analysis of the images was done by running filters over kernel windows (Filters: variance, skewness, kurtosis and mean euclidean distance (MED), and kernel window sizes used were: 3x3, 5x5, 7x7, 13x13 and 25x25), followed by layer stacking, principal component analysis (PCA), unsupervised classification, image smoothing and manual editing. The resulting classified images were used to create land use and land cover maps, and were further interrogated to examine land cover changes for each time period. Furthermore, change detection and projection was conducted using Markov chain analysis and stochastic analysis, using IDRISI Kilimanjaro v18.05.

Textural analysis is a process that enhances the texture of an image. The arrangement of the brightness differences or grey-scale levels within an image or area is referred to as an image texture (Tuttle et al., 2006). The dimensionality of the original one band image was increased by creating layer stacked images using textural analysis to improve classification results. Then, PCA was used to eliminate redundancy within bands and to create a pseudo colour image which has different spectral signatures. Layer stacking was used to combine selected PCA image results.

3.2 Description of data

This study used two sets of data which, namely Colour Infrared Images (CIR), which were projected to Transverse Mercator, Datum WGS84 Ellipsoid WGS84, and black and white aerial photographs (BWAP), which were in jpeg format and had no spatial reference. The black and white aerial photographs and colour infrared images were selected for years where there was clear imagery available that covered as par as possible the whole study area, to represent a decadal land use and land cover information. Filling data gaps was achieved by using a majority rule, for example gaps in 1940 were filled with available data in 1954 as long as the data covered the major part of the study area in that decade (see appendix B). It is important to note that the interval between selected years is not equal, but there is a year representing each decade from 1940 to 2010. The following table (Table 2) provides information regarding the datasets used in this study.

Years Scale of photography Camera Lens Format Principal Date Acquired distance (mm)

1940 1:34 000 F24 5ꞌꞌ 5ꞌ x 5ꞌ

1954 1:30 000 0s 196 9ꞌꞌ x 9ꞌꞌ March- April 1954

1967 1:30 000 ZEISS RMK A 21171 WILD 9ꞌꞌ x 9ꞌꞌ December 1963- 15/23 15AG39 February 1964

1974 1:150 000 to 1:60 000 WILD RC 10 8.8 1340 88.18 Sag11 2021

1985 1:250 000 ZEISS A 8.5/23 132014 85.54 Jul-85

1991 1:50 000 ZEISS RMK A 127808 152.7 Dec-91 15/23 127758

2006 1:30 000 WILD RC 30 15/4 13370 153.45 2006-07-24 to 2006-08- UAG-S 2006/05/12 26 2004/04/21

2010 0.50m GSD DMC 01-135 2010-04-02 to 2010-04- 26

PAN1- 120mm- 00117324 RED- 80mm- 00116839

PAN2- 120mm- 00117304 GRN- 80mm-00115869

PAN3- 120mm- 00117325 BLU- 80mm- 00115872

PAN4- 120mm- 00117327 NIR- 80mm- 00116858

Reference datasets such as fine scale land cover maps, National Freshwater Water Priority Areas (NFEPA), National Wetlands, Google Earth maps, road maps, field trip photographs and other basic datasets such as geology, were considered in this study. The study area (Goukou River catchment) was delineated from DWAF quaternary catchment shape file to

31 create an accurate planning domain of the Goukou catchment. The rainfall database was drawn from the South African Weather Service (SAWS) for the years 1940, 1954, 1967, 1974, 1985, 1991, 2006 and 2010.

3.3 Software utilised

These data were prepared by using GIS and remote sensing software ArcMap 10.1, ENVI 4.4, Erdas Imagine 2013 and IDRISI v18.05, which is now called TerrSet. These software packages played different roles in ensuring accurate elaboration of the land cover of the study area. ArcMap 10.1 was used to geo-reference images in order to establish the accurate coordinates of each image, and GIS analysis was used to derive basic statistics about the data, via data cleaning, on-screen digitising, post-classification and accuracy assessment of land cover classes. The images required stitching by using the Mosaic tool in Erdas Imagine. Spatial land cover classification and textural analysis was also performed in Erdas. The remote sensing software ENVI 4.4 was used to perform Principal Component Analysis (PCA) and layer stacking of non-correlated images. Finally, change detection, Markov chain analysis and stochastic analysis were all done using TerrSet.

3.4 Data pre-processing

Image pre-processing, also called image restoration (Anon, 2015), involves the rectification of distortions, degradation, and noise introduced through the imaging process (Jensen, 1996). This method creates a corrected image that is as close as possible, both geometrically and radiometrically, to the radiant energy characteristics of the original scene (Civco, 1993). Radiometric and geometric are the most common types of errors encountered in remotely sensed imagery (Liew and Lee, 2013). Pre-processing normally involves the processing of digital images to enhance the fidelity of the brightness value magnitudes (Vibhute et al., 2013). The main purpose of pre-processing is to reduce the effect of errors or irregularities in image brightness values that may cap one’s capabilities to interpret or quantitatively process and analyse digital remotely sensed images (Openshaw and Openshaw, 1997).

In this study, it was impractical to investigate methods of eliminating cloud cover which resulted in shadows especially in mountainous areas. This highly labour intensive exercise was not conducted at the stage of pre-processing. Future analysis may investigate methods of removing the effect of clouds and shadows in an image to improve the quality of the output.

3.4.1 BWAP pre-processing

Image pre-processing was done to extract significant information from BWAP so that they may become more amenable to analysis. To prepare BWAP, a number of pre-processing steps were performed. These steps were projection, geo-referencing, subset, mosaic, textural analysis, PCA, layer stacking, classification, post-classification, smoothen and generalisation. Projection is a mathematical transformation that takes spherical coordinates (latitude and longitude) and transforms them to an XY (planar) coordinate system (Qihe et al., 2000). This allows the creation of a map that correctly shows distances, areas and directions (ArcGIS 10, 2010). Detailed information on how pre-processing techniques assisted in preparing BWAP for the creation of final land cover and land use of the Goukou catchment is explained below.

3.4.1.1 Geometric corrections

The first pre-processing step conducted was to project the images. The images were projected to Albers Equal Area. After projection, images were then geo-referenced. Geo-referencing is simply registering the images to their true geographical locations. A minimum of ten control points was selected, a RMSE of ±10 m, was achieved, with the transformation type of 1st Order Polynomial (affine). The 2006 CIR image was used as reference layer during this process. The following figure (Figure 3) below provides an example of a geo-referenced tile of 1940.

Figure 3: Geo-referenced BWAP single tile of 1940 by using 2006 as a reference layer.

When all images had been registered to their geographical locations, black borders were then removed using a subset tool in Erdas 2013. The following figure (Figure 4) provides an example of subset tile of 1940.

Figure 4: Subset BWAP single tile of 1940.

The georeferenced images were then mosaicked, and the area of interest was masked. Mosaicking means merging together two or more images of the same area to create one single image, and masking means selecting or extracting the area of interest from the image. Masking aids in reducing the area to be processed to only the area of interest. The following figure (Figure 5) shows the mosaicked and subsetted image for 1974.

Figure 5: Subsetted and mosaicked BWAP for 1974.

3.4.1.2 Textural analysis

Texture is one of the most important characteristics used in identifying objects in an image, whether the image is a photomicrograph, an aerial photograph or a satellite image (Robert et al., 1973). Texture signifies important tonal information about the physical arrangement of objects in the image (Awwad, 2003). Texture determines the overall visual smoothness or coarseness of the image. The role of the texture is to generate brightness values for each pixel in the image (Hsu and Shin-Yi, 1978).

Texture is statistically measured by means of moving kernel window sizes throughout the image using statistical operators for assessment (Lillesand and Kiefer, 2000). Statistical operators such as skewness, kurtosis, variance and mean Euclidean distance are used for the textural analysis in image processing to characterise the distribution of grey levels of an

35 image (Kushwaha et al., 1994). The use of textural analysis creates images with more enhanced edges (Jensen, 1996). The application of textural analysis on this study improved the tonal information of the images and it was found to be a useful tool for the development of automated feature extraction.

Textural analysis involved running or moving filters over kernel window sizes. The method created new values for a raster file by using surrounding pixel values within a designated area to calculate the new value of the central pixels (Tuttle et al., 2006). When the calculation for the central pixel is completed, the window then moves over one unit (that is column or a row) and repeats the process for all the pixels over the entire area of interest (Tuttle et al., 2006).

Expert practitioners have used larger window sizes to assimilate large extents of the image, followed by smaller window sizes. This method has proven to be successful in image processing because the analysis allows the determination of the most appropriate window sizes to be used for the final image that will be classified (Carr, 1999). Different window sizes ranging from small (3x3) to large (25x25) were used in this study in order to test the effectiveness of this analysis and to avoid possible spurious results from an analysis conducted at one particular scale (Chen, 1998). The window sizes were then chosen based on the spectral homogeneity of the images and filters were selected based on the range of filters available on Erdas 2013.

The images had the similar spectral reflectances, and this required different window sizes to separate and improve their textural information. Four filters were used: skewness, kurtosis, variance and mean Euclidean distance, together with five kernel window sizes, 3x3, 5x5, 7x7, 13x13 and 25x25. Kurtosis was used to measure the relative degree of image sharpness of the peak on a frequency distribution curve (Brown, 2014). Kurtosis is a statistical operator used to describe the distribution of the data around the mean. It is derived from the field of statistics (Salimon et al., 2009).

From an image processing textural analysis point of view, large kurtosis values show that many pixel values are close to the mean within a local neighbourhood and few pixels are further away from the mean, that is, towards the tails of distribution curve (Bartels et al., 2006). Most of the pixels in a local area are related to one another but some pixels may be very different. The results of kurtosis image can assist in the determination of some land cover types, or support to the segmentation of the image into areas with different texture patterns (Kim and White, 2004).

Mean euclidean distance was used to measure the ordinary distance between two points in euclidean space (Himmelstein et al., 2010). This distance metric is normally used to measure similarities in image retrieval due to its efficiency and effectiveness. It measures the distance between two images by computing the square root of the sum of the squared absolute differences (Malik and Baharudin, 2013).

Variance measures were used to measure the distribution of variability in the image (Jensen, 1996). The more the pixels vary from the mean pixel standards of a particular window size, the greater the variance of that specific pixel related to its neighbours (Jeter, 2014). The mean parameter is the average pixel value of neighbouring pixels combined to the middle pixel of a particular window size (Jeter, 2014). The correlation parameter is articulated by a coefficient between two random pixels within a co-occurrence matrix (Baraldi and Parmiggiani, 1995). Variance can be expressed as the grey tone linear dependencies within an image. High correlation values mean that there is a strong linear association subsists among greyness levels of pixel pairs (Bharati et al., 2004).

Skewness was used to measure the asymmetry of the distribution of the intensity in the image (Jensen, 1996). Skewness describes the asymmetrical attributes of a group of pixel values in a digital image. If the skewness is equal to zero, it means that the data are more symmetrical and further from zero the more asymmetrical (Xi et al., 2008). As a significant statistical measure, skewness is extensively used in finance, engineering and atmospheric research. Skewness is becoming a useful tool in the field of remote sensing, though not currently widely used. Skewness can be applied in order to detect and enhanced the edges in the image, and also used as a tool to detect small targets in the image (Chiang, 2001; Irons, 1981). Skewness can reveal the symmetry of diverse land cover types; it is an effective and simple feature to differentiate between land cover types (Xi et al., 2008).

The textural analysis resulted in twenty band textured images, which were then stacked together with the original grey scale image to form a twenty-one band image. This was then used to run Principal Component Analysis (PCA). The following two figures (Figure 6 and 7) shows examples of filters and kernel window sizes used during textural analysis and figure 7 show the stacked results of filters and kernel windows of 1940 BWAP.

3.5 Principal Component Analysis (PCA)

Principal Component Analysis (PCA) is a method used to emphasise differences and bring out strong patterns in a dataset (Richardson, 2009). PCA was used to decrease the dimensionality of the multiple layer texture images. PCA transforms the original data into a set of uncorrelated variables, where the first components signify the majority of variance from the original data sets and the subsequent orthogonal components accounts for less variance and a higher proportion of noise (Eastman and Fulk, 1993).

PCA was performed by using standardised covariance matrix in ENVI 4.4 with all 21 bands as input (see Figure 7). The 21 band image produced from the textual analysis was used to run the PCA analysis, which gave the statistics of the bands that qualified to be used as a PCA band (see table 3). Furthermore, the PCA bands were identified based on the variance explained by Eigenvalues where the first four bands with high Eigenvalues were selected and then stacked together to produce the final PCA band. The final PCA band was then used for land cover classification. The remainder of the 17 bands which were not selected in the PCA results were bands that contained most of the noise. The following figure (Figure 8) shows the variance of more than 98% from the final PCA band of 1967.

Figure 8: The output results of four PCA images of 1967. This is pseudo colour image obtained by running PCA on the twenty textured images and one original image of 1967.

Basic Stats Eigen Values Percentages Sum (%)

41525.004 0.866096867 86.6097

3630.765082 0.075727729 7.57277

1256.011853 0.026196937 2.61969

773.370249 0.016130367 1.61304 98.4152

393.212543 0.008201327 0.82013

153.428171 0.003200088 0.32001

84.717855 0.00176698 0.1767

57.045792 0.001189818 0.11898

27.453296 0.0005726 0.05726

13.740874 0.000286597 0.02866

12.290621 0.000256348 0.02563

5.700364 0.000118894 0.01189

3.509274 7.31938E-05 0.00732

2.983802 6.22339E-05 0.00622

2.889969 6.02768E-05 0.00603

1.240376 2.58708E-05 0.00259

0.837301 1.74638E-05 0.00175

0.317647 6.62524E-06 0.00066

0.273134 5.69682E-06 0.00057

Basic Stats Eigen Values Percentages Sum (%)

0.195896 4.08585E-06 0.00041

0 0 0

47944.9881 1 100

3.5 Classification

Characteristics, activities, and status of specific areas on the Earth’s surface is the fundamental information that can be provided by the classification and mapping of the land cover (Friedl et al., 2010). Ecologists and resource managers depend on the quality of the final classification in order to make effective decisions for developing a landscape level and understanding of an ecosystem service delivery, so the quality of final classification is critical in providing accurate information (Enderle et al., 2005).

There are two main techniques of image classification (Tadesse et al., 2003), supervised and unsupervised classification (Perumal and Bhaskaran, 2010). In the supervised classification, the analyst identifies training areas; these areas of known pixels are then used in a classification algorithm to classify the pixels of unknown identity using spectral characteristics (Campbell, 2002). Unsupervised classification is an automated classification algorithm that involves the grouping of image pixels together based upon similar spectral characteristics and the resultant assignment of those groupings to informational classes by the analyst (Enderle et al., 2005). There are advantages and disadvantages associated with the use of these classification methods (Auria and Moro, 2008).

Several advantages associated with supervised classification include:

1. The allocation of final information classes is controlled by the analysist. This permits for easier assessment with other classifications by means of individual classes for both.

2. The final classification is tied to a precise area on the image of known identity through the process of choosing training sites,

3. Matching of spectral categories to information classes is addressed during the selection of training sites, so the analyst does not faced with this issue.

4. One way of discovering serious errors or issues in the classification method is by equating training data with the final classification (Enderle et al., 2005; Campbell et al., 2005; Enderle et al., 2005; Campbell, 2002).

According to Archana and Elangovan (2014), “there are also disadvantages and limitations to the use of supervised classification”; which are:

1. The training areas selected by the analyst may not necessary be present in the data, as the analyst is imposing a classification structure upon the data by selecting training areas and of specific information classes.

2. The use of spectral properties may lead to overlap and ambiguity during the classification process as the spectral properties are not the primary characteristics used in identifying training areas.

3. The analyst must have extensive knowledge about the area and invest a lot time and resources, which are required for unsupervised classification.

4. Analyst may overlook some of unique classes that are present in the image during selection of training areas and selection of classes (Jensen, 1996; Campbell, 2002; Horning, 2004).

On the other hand, there are three primary advantages of using unsupervised classification (Serra et al., 2003).

1. Initial separation of image pixels does not require extensive knowledge of the area being classified.

2. The analyst is not required to make many decisions during classification, whereas there is less opportunity for human error.

3. Supervised classification may overlook unique classes in the data that will be recognised by unsupervised classification (Jensen, 1996; Serra et al., 2003; Horning, 2004; Holden and Kelley, 2010; Richards, 1999).

The disadvantages of using unsupervised classification method ranges from:

1. The natural groupings recognised by the classification method have the similar spectral band, which might not necessarily correspond with the information classes of importance.

2. The association amongst the natural groupings of spectral categories and that of the desired information classes are not always directly correlated as the results there is a limited control over the classes selected by the classification method for an analyst (Enderle et al., 2005; Jensen, 1996).

3.5.1 Unsupervised classification

Unsupervised classification was adopted for this study due to its advantages and the nature of the data being used. Historical aerial photographs are predominately black and white images, which means there is only one spectral band; hence the textural information was improved by using textural analysis. Unsupervised classification technique was chosen over supervised classification because the imagery data used in this study was homogeneous, which means that this data had similar spectral band, which could lead to more misclassified pixels in the supervised classification (Wu et al., 2013). Unsupervised classification identifies spectrally homogeneous classes within the data (McAffrey and Franklin, 1993). Furthermore, spectral properties are not the primary characteristics used in identifying land use types (Jensen, 1996).

Unsupervised classification can classify a much large number of spectrally distinct classes than an analyst might consider when using supervised classification (Peacock, 2014). The much larger amount of spectral categories in unsupervised classification permits additional detailed and distinct approaches to assign land cover classes to smaller groups of pixels (Peacock, 2014). Moreover, methods used in this study and time frame were also a contributing factor for eliminating supervised classification. Future studies could compare both classification schemes to compare their accuracy assessment scores. Also, it is recommended to increase the number of initial classes to a larger number than used in this study (25 initial classes), as this might also improve the accuracy assessment results.

The unsupervised classification was performed by using ISODATA in Erdas 2013. According to Jensen (1996) “ISODATA is a clustering algorithm that uses an iterative process to separate image pixels into spectrally similar clusters based upon their position in

43 nth dimensional spectral space”. Firstly, the algorithm assigns a random initial cluster vector. Each pixel is then classified to the neighbouring cluster, in the second step. New mean cluster vectors are considered in the third step, based on all the pixels in one cluster (Jensen, 1996; Zhou et al., 2009; Wu et al., 2013). The change between the iteration should be small otherwise step two and three are repeated (Anderson et al., 1976; Jensen, 1996, Busch, 2000). The ISODATA algorithm has some additional enhancements that operate by splitting and integrating clusters (Jensen, 1996; Halder et al., 2011). If the centres of two clusters are closer than a certain threshold or the number of pixels in a cluster is less than a certain threshold, then clusters are fused together. If the cluster’s standard deviation surpasses the predefined value and the number of pixels is double the threshold for the least number of members, then clusters are divided into two (Jensen, 1996).

The parameters for unsupervised classification were set to 25 initial classes with maximum iterations of hundred. These classes were further reclassified into eleven classes, because there were a lot of misclassified pixels due to different effects such as the time of the year and the time of the day that the image was taken. Classification results were then further refined for misclassified pixels via a process called image post-classification smoothing and generalisation which is explained in the section below.

3.6 Image smoothing and generalisation

In the classified image output, some misclassified isolated pixels or small regions of pixels may exist (Mylonas et al., 2015). This gives the output a "salt and pepper" or speckled appearance (Vasuki, 2012). In such situations, it is important to smooth the classified output to show only the dominant classification (Lillesand et al., 2008). In that regard, post classification smoothing and generalisation was applied to the classified images so as to eliminate isolated pixels and generate a less noisy image (Al-Ahmadi and Hames, 2008).

This process was performed using ArcMap 10.1, where three processing steps were used. First: filtering the classified image output, to remove the isolated pixels or noise from the classified image output by using the majority filter. Secondly, smoothing class boundaries and clumping classified output, this process smoothens the untidy class boundaries and clumps the classes, by using boundary clean tool and, thirdly, generalising classified output by removing small isolated regions. Image smoothing and generalisation simple reclassifies

44 small isolated regions to the nearest classes by using three sets of tools; region group, set null and nibble tools (ArcGIS 10.1 manual help, 2013).

3.7 Post-classification

Even though the images have after the above processing been generalised and smoothed, there remain misclassified pixels when overlaying the classified image with the original classified image. To address this, the classified image is converted to vector polygons for further post-classification refinements. To improve the image, each class on the image is extracted and shape files created using Catalog in ArcMap, to explore which polygons have been misclassified and need to be reclassified.

The select tool, and in some instances the cut polygon tools, are used to select and split the polygons. The cut tool was mostly used to extract roads, which were classified often with bare surfaces, and was also used to extract dams, which were sometimes misclassified as vegetation and bare surfaces. The reason for misclassification of these classes may be the seasonal variations and the similar spectral properties of land cover classes.

Due to unavailability of topographical maps (1:50000) of the Goukou River catchment from NGI, land cover and land use referencing was done by using the BGIS S.A shape files, Google Earth, 2006 CIR images, 2010 CIR images, visual interpretation and 1map, as the references. It was impossible to map built-up areas due to spectral reflectance that these areas have. 1map is online GIS software which was launched in mid-2010, which offers basic data for the whole of South Africa, including erven cadastre, road centre lines, Comprehensive Street addresses, Aerial Photography, dams and agricultural lands through the Internet.

They have similar spectral reflectance as disturbed areas (degraded areas) as a result these were classified as such. After all the polygons were examined, and assessed if they fall within their designated classes, they were then merged and dissolved together. From the cleaned polygons, accuracy assessment was performed, to determine how well the classification process accomplished the task (Strahler et al., 2006).

3.8 Accuracy assessment

In any study of this nature where uncertainty in the resulting output may be high, it is vital to perform accuracy assessment of the output in order to check that the approach has performed

45 to an acceptable level. The quality of the information derived from remote sensed data can be determined by performing accuracy assessment (Congalton and Green, 1999). Accuracy assessments are performed on classified images to determine how well the classification process accomplished the task (Strahler et al., 2006), therefore; it should be an important part of any classification and increases confidence in using information for major decision making such as policy formulation (Foody, 2002). The most common form of expressing classification accuracy is the error matrix or confusion matrix (Congalton and Green, 1999). An error matrix approach takes into consideration both the corresponding results of the classification procedure and the link amongst known reference data (ground truth), and compares them on a class by class basis (Congalton and Biging, 1992).

3.8.1 Accuracy assessment for BWAP

Because of land cover change, it is best to collect the ground reference as close as the date of remote sensing data acquisition as possible (Congalton and Green, 1999). A comparison between the classified image and reference data that is believed to be accurate reflections of the true land cover of area being classified can be used to assess the accuracy of the classification (Congalton, 1991). Ground truth, higher resolution satellite images, and maps derived from aerial photo interpretation are some of the reference data that can used during accuracy assessment (Zhuang et al., 1991).

Accuracy assessment was not performed on all the years of this study, due to time constraints, the size of the study area, availability of data, and the method used to perform accuracy assessment. It was thus only performed on the years 1940, 1974 and 1991. In order to determine accuracy, a digital reference layer was created by digitising the original year being classified. Digitising is assumed to be more accurate since the extraction of classes is solely dependent on the visual interpretation of land use categories by the analyst.

Samples of 100 random points were created using the digitally-created reference layer, then classified them using visual interpretation. It is recommended that a practical minimum number of samples should be ten times the number of classes obtained (Swain and Davis, 1978; Congalton, 1991). A confusion matrix (Congalton, 1991) was then built and Overall accuracy (OA), User’s accuracy (UA), Producer’s accuracy (PA), Omision error (OE), Commission error (CE) and Cohen’s Kappa Coefficient (CKC) were computed.

Gu et al., (2015) define Overall Accuracy as, “the sum of pixels correctly classified (diagonal elements) divided by the overall number of samples in the comparison”. According to Plourde and Congalton (2003) the user’s accuracy is the reliability of the classification or the probability that a pixel on a map actually represents that category on the ground. OE is the measure of how well the algorithm has conducted the classification, and is calculated as the number of pixels of the reference layer that were correctly classified divided by the total number of reference samples, whereas CE indicates the probability that the algorithm has properly allocated an image pixel (Petropoulos et al., 2012) (Appendices D). All of these errors and accuracies are expressed in percentages.

A significant component of the accuracy assessment is Cohen’s Kappa Coefficient (Congalton, 2001). Kappa tells us how well the classification process performed as compared to just randomly assigning values; did we do better than randomly assigned samples? The Cohen’s kappa coefficient, unlike the overall accuracy, includes errors of omission and commission (Congalton, 1991). The following figure (Figure 9) shows the formula for Kappa as defined by Congalton (1991).

(1)

3.9 CIR images

3.9.1 Preliminary pre-processing

Preliminary pre-processing for CIR images was conducted via the same process performed on BWAP as described above. The first process conducted was to project CIR images to Albers

Equal Area. The individual projected images were mosaicked to obtain a single layer of the study area for each decade. The output of mosaicked image was then delineated using subset tool in Erdas, to reduce the digitising and analysis process to the only area of interest.

3.9.2 Manual digitising

The colour infrared images of 2006 and 2010 were prepared by manual digitisation using an on-screen digitising method. Manual digitizing is the method by which coordinates from a map, image, or any other sources of data are converted into a digital format by tracing them on the screen using a mouse (Eitsch, 2002). High accuracy can be achieved through manual digitizing (Paul et al., 2013), there is usually no loss of accuracy compared to the source map and usually no editing or post-processing needed when to compare to automated classification process used to classify BWAP.

Manual digitising comprises the creation of blank shape files to represent the classes within the image, using Catalog in ArcMap. The process involves uploading of these blank shape files on ArcGIS, and activation of Editor Tool and start the digitising process. The scale of digitising varied between 1: 500 for small features like dams and 1: 1250 for bigger features like roads and cultivated fields. The accuracy of each feature extracted from the images was based on visual interpretation. Field visits were conducted and coordinates of randomly selected areas in the catchment were taken and were verified with the CIR images. This process is called ground-truthing. The following two figures below (Figures 10 and 11) shows examples of some sites visited and their coordinates during study area visits, and the corresponding land cover types that were extracted from CIR image of 2006 using digitising process.

Figure 10: Examples of sites visited during study area visits on September 2013, their land cover types, and their coordinates.

Figure 11: Examples of 2006 CIR digitised land use classes.

Thirty three land use types were extracted from 2006 and 2010 CIR images. These land use classes were identified with the assistance of the 2009 National Land Cover map, visual interpretation, and field visits, and they were defined as priority classes for future hydrological analysis to be conducted in the same area. The following table (Table 4) shows the initial 33 land use classes that were digitised from CIR images of 2006 and 2010.

Table 4: Initial 33 land use types as obtained by digitising 2006 and 2010 CIR images.

Land Use Land Use

 Channel dams  Fallow fields  Coastal bare sand  Golf courses  Cultivated fields dryland (banks)  Main roads  Cultivated fields dryland (no banks)  Major dam  Cultivated irrigated non-pivot  Mines  Cultivated irrigated pivots  National roads  Disturbed veld I  Natural bare rock  Disturbed veld II  Natural low shrub  Disturbed veld III  Natural tall shrub pure  Non-coastal bare sand  Off-channel dams  Old-fields  Parks  Plantation and woodlots  Plantation clear-felled  Railway line  Recent fire scars  Rivers  Secondary roads  Tall aliens mix  Tall natural shrub  Tall trees and tall alien shrubs  Water natural  Wetlands

After completing the digitising process, the data were then cleaned for errors that may have occurred during the process. These errors are called topological errors (Maras et al., 2010), which include dangling nodes, overshoot, undershoot and slivers. After cleaning and validating of the errors, the polygons were further reclassified into 11 classes, which were same as in black and white images, to standardise both CIR images and BWAP land use classes. Land cover maps were then created from the standardised land use classes.

To merge the initial 33 classes, the following procedure was taken as demonstrated in the following table:

Table 5: The procedure taken to merge the 33 initial classes from CIR images.

Natural vegetation Cleared plantation Cultivated fields Wetlands

 Old-fields  Plantation clear-felled  Cultivated fields dryland  Wetlands  Fallow fields (banks)  Golf courses  Cultivated fields dryland  Natural low shrub (no banks)  Natural tall shrub  Cultivated irrigated non- pure pivot  Tall natural shrub  Cultivated irrigated pivots

Disturbed surfaces Plantation Roads Bare surfaces

 Disturbed veld I  Plantation and  Railway line  Coastal bare sand  Disturbed veld II woodlots  Main roads  Non-coastal bare  Disturbed veld III  National roads sand  Mines  Secondary roads  Natural bare rock  Parks  Recent fire scars

River Dams Aliens

 Rivers  Channel dams  Tall aliens mix  Water natural  Major dam  Tall trees and tall aliens  Off-channel dams shrubs

Reclassification was done because not all these classes identified from CIR images were able to be identified on BWAP, so to standardise CIR images with the BWAP, reclassification was necessary. The following table (Table 6) below show resulting land use and land cover classes after reclassification was done.

Table 6: The Goukou River catchment final land use classes as obtained by digitising 2006 and 2010 CIR images.

Land use Description 1 Natural vegetation Natural growing vegetation 2 Cleared plantation Cut down forestry 3 Cultivated fields Commercial and non-commercial agricultural fields 4 Wetlands Areas that are saturated with water 5 Disturbed surfaces Areas that have loss some basal cover, moderate or severe basal cover loss. 6 Plantation An area of land usually privately or government owned which cultivate a single commercial crop 7 Roads Transportation system 8 Bare surfaces Surfaces with no vegetation cover or non-vegetated surfaces. These include bare rocks on the mountain and coastal areas, sand (coastal and non-coastal sand). 9 River Natural stream of water flowing in a channel to the sea or lake 10 Dams Receive water natural, either from the rain, river or canal 11 Aliens Plant species that are non-native to an ecosystem

3.10 Change detection analysis

Change detection is the process of identifying differences in the state of an object by observing it at different times (Singh, 1989). The change detection analysis used in this study was performed using the Land Use Change Model (LCM) in IDRISI. Land use change models focus on (a) the amount of change amongst two or more land cover classes (b) the location of change to one or more classes and (c) the rate and location of change between two or more classes. The IDRISI land use change models are calibrated using past time periods or hypothesis of driving factors of change like any other models. Predicted land use at one or more specified dates or the likelihood of each cell converting to a given class (at an unspecified time) may be the output.

This study focused on the rate of change between two or more classes only. The rate of change between two classes provides an understanding of what classes resulted in the reduction or increase of the other class. These reductions and increases are called losses and

52 gains, which can be computed in percentages or hectares. This study did not focus on (b) and (c) above, because the objective of this study was to determine historical changes in land cover and their rates.

The data used to perform LCM were the images prepared above by classifying and digitising the images of 1940, 1954, 1967, 1974, 1985, 1991, 2006 and 2010. The images (1940, 1954, 1967, 1974, 1985 and 1991) were all prepared using unsupervised classification and the CIR images of (2006 and 2010) were prepared using digitising process. The ArcGIS shape files were first converted to IDRISI vector file format, and were uploaded on the IDRISI software in order to convert them to the IDRISI raster compatible format, which is the acceptable format file to perform LCM. The two images were uploaded to LCM; the first year was uploaded as the “Early land cover map” and the second year as “Late land cover map” for example 1940 and 1954, 1940 would be uploaded as “early land cover map” and 1954 would be uploaded as “late land cover map”. This process was repeated for all other images 1954, 1967 and so on. The LCM results are graphs that represent gains, losses and contributors of each land cover change.

3.11 Markov chain analysis for future land use change

Projections are used in different number of study fields, including the science of land cover and land use change, to better understand the dynamics of systems, to develop theories that can be scientifically verified, and to create projections that can be used to evaluate scenarios for use in assessment activities (Goodchild et al., 1992). There are many existing land use change models that could be used for this purpose, such as the FOREcasting SCEnarios of Land-use Change model (FORE_SCE), but for the purpose of this exploratory “proof of concept” study a simple modelling technique to give a basic insight into how the land cover could change in the near future was selected, which is Markov chain analysis. When changes and processes in the landscape are complex Markov chain analysis is the most appropriate technique to use for modelling such land use changes. Markov chain analysis presents an exploratory approach for projecting future land use change, based on historical trends. Markov chain analysis describes a sequence of likelihood events in which the probability of each event depends only on the state attained in the previous event, and it also known as a stochastic model that (Norris, 1996).

There are two main limitations when using Markov chain analysis. These are:

1. Markov chain analysis does not take into consideration the drivers of land use and land cover change. It simply assumes that those drivers will continue be the same in the future.

2. Markov chain analysis does not provide a sense of geography. Although the transition probabilities may be accurate for a particular class as a whole, there is no spatial element to the modelling process.

This can be overcome by integrating cellular automata (CA) with Markov chain analysis (Eastman, 2003). However, in a study conducted by Memarian et al., 2012 in the Langat Basin, Malaysia, it was found that CA-Markov calibration is only applicable based on a single period of time, which renders difficulty in simulating land cover dynamics on a temporal scale. Furthermore, it was found that the use of CA-Markov did not accurately simulate land use change dynamics, and this resulted in the high values of quantity and allocation disagreements and low values of figures of merit computed using a three- dimensional approach.

There are also advantages of using Markov chain analysis. These include:

Markov chain analysis has an ability to simulate multiple land covers and complex patterns. It is a very powerful and effective tool which is able to provide estimates and predictions of future land use change (Yang et al., 2012).

Markov chain analysis (Araya and Cabral, 2010) was therefore used to project future land cover change in the Goukou catchment. A transition probability matrix, transition area matrix, and a set of conditional probability images were all produced as the results of performing Markov chain Analysis. Transition area probability matrices contain information about the probability that each land use type would be found in a different class at each pixel in a subsequent timer period, and transition areas probability matrices present the number of pixels which are expected to change from one land use class to another land use type over the specified time period (Eastman, 2012). The transition probabilities matrix calculated was used to project the future probable land use change over a specified time period, namely to 2025.

The data used to perform Markov chain analysis were the digitised images of 2006 and 2010, which were prepared using the same procedure as in the change detection analysis of converting them to IDRISI raster format. After mapping the current state of land use and land

54 cover of 2006 and 2010, Markov chain analysis in IDRISI was used to project the probability of land cover change from one class to another in the next fifteen years.

The two images were uploaded to the IDRISI software, the first year was uploaded as “Early land cover map” and the second year as “Current land cover map” and the period was set to fifteen years; that is from 2010 to 2025. The results are two tables and an image, as explained above. The image is the projected land cover and land use map of the year 2025. The model results were not validated using the Kappa Cohen’s coefficient, because for this exploratory study it was not the intention to exploit the full methodology, but rather to indicate for researchers the potential for this approach, and to obtain a “first-cut” view of a potential future trajectory.

CHAPTER 4: RESULTS, DISCUSION AND RESEARCH LIMITATIONS

4.1 Summary

A series of six black and white images was classified in this study using unsupervised classification (1940, 1954, 1967, 1974, 1985 and 1991), and the other two colour infrared images were prepared by on-screen digitising (2006 and 2010) and eleven classes were identified and extracted from the images. These classes were: natural vegetation, cleared plantation, cultivated fields, wetlands, disturbed surfaces, plantation, roads, bare surfaces, river, dams and aliens. The average Overall accuracy of 60% and Kappa Cohen’s coefficient of 55% was achieved for the years 1940, 1967 and 1991.

4.2 Accuracy assessment results

The following table (Table 7) contains detailed information on the confusion matrix errors and accuracies computed for land cover and land use classes of 1940. Detailed information on the confusion matrix errors of 1974 and 1991 can be found in appendices E and F. Accuracy metrics presented include: Overall accuracy (OA), User’s accuracy (UA) and Producer’s accuracy (PA) and errors metrics presented are: Errors of Omission (OE) and errors of Comission (CE).

LU N Cp Cf W Ds P Ro Bs R D A RT

N 81 13 9 13 52 11 19 38 10 6 27 279 Cp 0 79 0 0 0 4 0 0 0 0 0 83 Cf 1 0 83 2 13 0 4 0 2 4 0 109 W 0 0 0 70 1 0 1 0 40 1 2 115 Ds 14 0 5 9 27 5 47 20 17 6 23 173 P 0 0 0 0 0 74 0 0 0 0 0 74 Ro 0 0 0 1 0 0 26 0 0 0 0 27 Bs 4 5 3 0 7 6 3 42 2 0 6 78 R 0 0 0 1 0 0 0 0 17 0 0 18 D 0 3 0 0 0 0 0 0 0 83 0 86 A 0 0 0 4 0 0 0 0 12 0 42 58 CT 100 100 100 100 100 100 100 100 100 100 100 1100 K 0.52 OA 0.57 PA 81 79 83 70 27 74 26 42 17 83 42 UA 29 95 76 61 16 100 96 54 94 97 72 CE 71 5 24 39 84 0 4 46 6 3 28 OE 19 21 17 30 73 26 74 58 83 17 58

The following tables (Tables 8 and 9) shows the Producer’s accuracy, User’s accuracy, errors of comision, errors of omision, overall accuracy and Cohen’s Kappa Coefficient in percentages for 1974 and 1991, respectively.

Accuracies N Cp Cf W Ds P Ro Bs R D A PA 83 68 82 57 17 94 40 33 51 87 48

UA 29 99 89 84 11 98 100 49 88 100 61

CE 71 1 11 16 89 2 0 51 12 0 39

OE 17 32 18 43 83 6 60 67 49 13 52

K = 0.56 OA = 0.60

Accuracies N Cp Cf W Ds P Ro Bs R D A PA 59 87 25 19 83 81 53 75 88 44 59

UA 27 84 78 10 89 100 70 99 100 50 100

CE 73 16 22 90 11 0 30 1 0 50 0

OE 41 13 75 81 17 19 47 25 12 56 41

K = 0.57 OA = 0.61

One hundred reference points were randomly created for each land use class; summing to 1100 random points for all the eleven classes that were identified. Of this total, 624, 660 and 673 points were correctly classified by the algorithm for 1940, 1974 and 1991, respectively. An overall accuracy of 57%, 60% and 61% was achieved for 1940, 1974 and 1991,

58 respectively. However, user’s accuracy and producer’s accuracy for disturbed surfaces are relatively low, being less than 30% in all time periods. User’s accuracy for the natural vegetation in both years is also relatively low, being less than 30%.

Error of omission for disturbed surfaces, bare surfaces and aliens were all above 50% in all time periods except for rivers and roads, while the error of comission for natural vegetation and disturbed surfaces are above 50%. A Cohen’s Kappa Coefficient of 52%, 56% and 57% for 1940, 1974 and 1991 was obtained. Cohen’s Kappa Coefficient is the indication of agreement between the algorithm and reality. These values show moderate agreement between land cover classification and reality, except for 1991 where substantial agreement is indicated.

- < 0: Less than chance agreement

- 0.01- 0.20: Slight agreement

- 0.21- 0.40: Fair agreement

- 0.41- 0.60: Moderate agreement

- 0.61- 0.80: Substantial agreement

- 0.81- 0.99: Almost perfect agreement (Pontius, 2000).

The following maps (Figure 12) shows the Goukou catchment final land cover and land use classes which were obtained from digitised and classified images of 1940, 1954, 1967, 1974, 1985, 1991, 2006 and 2010.

4.3 Classification results

The following figures (Figures 13 and 14) and tables (Tables 10 and 11) show the shift of the Goukou catchment land use classes from 1940 to 2010. These land use classes are divided in terms of the area they occupied, major classes being land use types that occupied larger areas, while minor classes are land use types that occupied relatively smaller area in relation to the catchment as a whole. Table 10 illustrates the shift of land use classes that occupied larger area in the catchment, while Table 11 illustrates shift of land use classes that occupied relatively smaller area in the catchment. The area occupied by natural vegetation in 1940 was 53.5% and that by disturbed surfaces was 19.5%, while bare surfaces and cultivated fields occupied about 13% of the total area of the catchment. Wetlands, rivers, dams and aliens occupied 1.2%, less than 0.5%, less than 0.5% and 0.6%, respectively. Plantations, cleared plantations and roads covered less than 1% of the total catchment area. In 1954, the cover of natural vegetation and cultivated fields increased to 62.9%, and 15.5%, respectively, while dams remain less than 0.5%, plantations increased from 0.1% to 0.5% and roads remain

62 roughly constant at 0.2%. There was a reduction in cover of wetlands, disturbed surfaces, bare surfaces, river and aliens of less than 0.5%, 9%, 3.5%, less than 0.5% and 0.5%, respectively.

120000

100000

1940 80000 1954 1967 60000 1974 1985 40000 1991 2006 20000 2010

0 Natural Cultivated Disturbed Bare surfaces Wetlands vegetation fields surfaces

4000

3500

3000 1940 1954 2500 1967 2000 1974

1500 1985 1991 1000 2006

500 2010

0 Aliens Plantation Roads Cleared Dams River plantation

Major LC Classes 1940 1954 1967 1974 1985 1991 2006 2010

Natural vegetation 86685 102135 84508 91373 59843 72025 66459 55681

Cultivated fields 20314 25014 34200 35117 30942 42247 37248 39394

Disturbed surfaces 31335 16672 26265 18506 40986 29299 45813 51930

Bare surfaces 20300 14802 12105 12789 22654 11219 4483 6848

Wetlands 1960 1877 1523 1480 1479 1381 1200 1083

Minor LC Classes 1940 1954 1967 1974 1985 1991 2006 2010

Aliens 981 408 1863 959 3300 3373 2968 3557

Plantation 232 800 969 1058 932 1528 1741 1143

Roads 324 354 474 508 982 607 381 403

Cleared plantation 73 72 83 185 655 949

Dams 45 65 136 203 185 236 435 388

River 159 153 192 161 203 128 204 208

In 1967 and 1974, the cover of natural vegetation decreased from 53.5% to 52.5% and increased again to 56.3% in 1974. There was minor change in the cover of cultivated fields between these years, cultivated fields cover increased from 21.1% to 21.6% which is an increase of 0.5%. A more sizable increase in cultivated fields occurred between 1954 and 1967, when cultivated fields increased from 15.4% to 21.1%. The cover of alien vegetation

65 decreased from 1.2% to 0.6% by 1974, while the surface area occupied by dams remains roughly constant at 0.1% between these years. There was also a slight increase and decrease in plantations, wetlands and rivers between these years. Bare surfaces decreased from 9.1% to 7.5% in 1967 and increase again to 7.9% in 1974, also the cover of disturbed surfaces, increased between 1954 and 1967 from 10.3% to 16.2% and decrease again in 1974 to 11.4%.

Between 1985 and 1991, alien vegetation cover increased from 2.0% to 2.1%. Disturbed surfaces and bare surfaces also increased to 25.4% and 14.0% respectively, while these values decreased in 1991 to 18.1% and 6.9%, respectively. There was slight increase in cover by plantations from 0.6% to 0.9% while there was not much change in the cover of rivers; rivers remained roughly constant at 0.1%. The cover of natural vegetation decreased to 37.0% in 1985 and increased to 44.0% in 1991. Cultivated fields decreased in 1985 to 19.2% and increased to 26.0% in 1991, whereas the surface area occupied by dams increased from 0.1% to 0.2% in 1991. There is also a decline in the cover of wetlands between these years, the cover of wetlands decreased from 0.9% to 0.7% in 2010, while the surface area occupied by the rivers remain roughly constant at 0.1% between these years.

The area covered by disturbed surfaces increased from 18.0% to 28.4% in 2006 and to 32.1% in 2010, however, there was a slight decrease in the cover of cultivated fields from 26.0% to 23.1% and 24.4% in 2006 and 2010, respectively. The area covered by bare surfaces decreased from 6.9% to 2.8% and 4.2% in 2006 and 2010, respectively. The cover of natural vegetation also decreased from 44.4% to 41.1% and 34.5% in 2006 and 2010, while there were increases and decreases in the cover of alien plants, the cover of alien plants decreased from 2.1% to 1.8% in 2006 and increased to 2.2% in 2010. There is also a sizable increase in the surface area occupied by dams between these years, though in 2010 the area occupied by dams decreased from 0.3% to 0.2%.

4.1.4 Trend analysis results

A trend is a gradual change in a state, output, or process, or an average or overall propensity of a sequence of data points to move in a certain direction over time, which can be represented as a graph (Gregory and Hansen, 2009). The computed trend change for this study was from 1940 to 2010 and the results are presented graphically. It was computed between two successive decades. The total average change of seventy years (1940 to 2010)

66 was also computed to be 14555 (ha) per year. In order to assess land cover trends during the study period, the rate of change was calculated by using the following formula.

∆퐴푟푒푎 The rate of change = (2) ∆푡푖푚푒

The results showed that the average trend for natural vegetation cover from 1940 to 2010 was 50%, while cultivated fields, disturbed surfaces, bare surfaces and dams occupied about 19.5%, 18.3%, 8.8% and 1.2%, respectively, and all other land use classes were less than 1%. This might be because of the size they occupy relative to the size of the catchment. Most of natural vegetation cover has been converted into other land cover types. Furthermore; an average trend change was also computed both in hectares and percentages to be 14555 ha and 9.1%. The following table (Table 12) contains percentage details of trend analysis for each land cover type and their averages from 1940 to 2010. The two figures (Figures 15 and 16) show the average trend and individual land use and land cover trends.

LCLU 40-54 54-67 67-74 74-85 85-91 91-06 06-10 Average%

Natural vegetation 52.6 63.8 51.4 58.3 35.6 44.6 43.4 50.0 Cleared plantation 1.2 1.2 0.9 0.9 0.9 0.7 0.8 1.0 Cultivated fields 12.3 15.0 21.0 22.0 17.8 26.3 22.6 19.5 Wetlands 20.0 9.8 17.0 10.0 26.8 17.3 27.1 18.3 Disturbed surfaces 0.1 0.5 0.6 0.7 0.5 1.0 1.2 0.6 Plantation 0.2 0.2 0.3 0.3 0.7 0.4 0.2 0.3 Roads 12.8 9.3 7.4 7.3 15.5 7.2 2.3 8.8 Bare surfaces 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 River 0.0 0.0 0.0 0.1 0.1 0.1 0.3 0.1 Dams 0.6 0.2 1.2 0.5 2.0 2.1 1.7 1.2 Aliens 0.0 0.0 -0.0 0.0 0.0 0.19 0.3 0.1

Figure 15: Average rate of land cover change in (ha/year) from 1940 to 2010. As calculated by using average change for each combination of years.

16000 14000 12000

10000 8000

Area(ha) 6000 4000 2000 0 1940 to 1954 to 1967 to 1974 to 1985 to 1991 to 2006 to 1954 1967 1974 1985 1991 2006 2010 Change in Years

Figure 16: Graphical representations of the Goukou catchment land cover trends from 1940 to 2010 in (ha/year).

4.1.5 Rainfall results

The following table (Table 13) and figure (Figure 17) seasonal rainfall in (mm) for four seasons of the year from 1940 to 2010, for the Goukou catchment, which was obtained by using a “Moving Average” of three intervals. The data were obtain from South African Weather Service, for the station in Heidelberg town (-[0009815x] - Heidelberg C/K- POL). The four seasons are defined as summer (December to February), autumn (March to May), winter (June to August) and spring (September to November). These results are based only on the years that were used on this study. The highest annual rainfall observed was in 2006, at 556 mm, followed by 1985 at 554.8 mm and the lowest annual total rainfall observed was in 2010, which is 285.5 mm.

Year JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Annual Totals 1940 54.9 78 35.5 38.1 6.8 21.2 11.7 2.5 33.1 16.7 66.1 2 366.6 1941 23.9 9.7 7.9 75.6 29.2 41.4 30.8 40.4 37.1 45 27.1 26.4 394.5 1942 25.2 15.3 44.9 17 40.9 21.3 21.2 17.5 39 38.9 10.9 58.4 350.5 1943 116.5 44.4 35.6 20.1 9.9 3 6.4 26.9 79.5 17.8 125.3 21.6 507 1944 0 8.4 53.9 21.3 72.4 32.5 36.6 44 60 36.8 8.4 5.1 379.4 1945 0 6.6 5 22.7 87.2 50.1 26.7 47.2 23.6 58.1 14.7 2 343.9 1946 3.5 10.2 96.5 7.6 4.8 17 29.9 32.2 24.1 24.6 18.3 0 268.7 1947 5.1 15.5 118.2 16 37 27 50 7.9 50.6 19.3 28.5 7.4 382.5 1948 45.7 11.7 42 49.6 15.7 24.1 31 0 39.6 82.6 15.8 12.7 370.5 1949 32.8 7.9 5.3 46.3 46.2 4.1 28.9 13.2 16.3 28 127.2 0 356.2 1950 22.6 9.7 10.9 46.3 41.2 2 44.9 46.5 26.8 69.6 73.6 7.4 401.5 1951 107.6 2.6 30.9 21.8 42.2 49.1 81.8 24.3 80.1 12.7 2.3 8.5 463.9 1952 31.4 16.8 5.6 35.2 14.4 20.3 10.9 48.4 107 33.5 64.5 35.1 423.1 1953 18.3 19.5 5.5 31 4.5 49 96.5 23.5 69.1 63.6 43.9 4 428.4 1954 9.5 17 17.5 101 74 24.5 32.5 87.5 12.2 12 46.1 6 439.8 1955 27 120.2 0 14 13.6 28.5 45 40.5 31 46.5 39.2 4.5 410 1956 18.5 19.6 36 29 95 23.5 28 36.5 33.5 87 0 70.5 477.1 1957 8 66.1 35 14 61.1 85 25.5 58.5 46.5 28.5 0 0 428.2 1958 22 14.5 47 36.5 108 32 3.5 52 16 18.5 19 2 371 1959 49 67.5 62 98.5 58 9.5 91 62.5 20.5 73 0 0 591.5 1960 24 16 40.5 29.5 29 57.5 21.5 17 33.5 14.5 66 58 407

Year JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Annual Totals 1961 29 23.5 15.5 15.5 41.5 12.5 30.5 30 25 30.5 23 10 286.5 1962 30 55 54 26 15 24 16.5 142 0 33.5 68 0 464 1963 31.5 0 52.5 14 65.5 10 39.5 10 0 36 28.5 31.5 319 1964 0 15 59.5 16.5 20 59.5 0 62 57 19 41 9.5 359 1965 11 26 37.6 41.5 41 0 27.5 31.5 8 101.8 44.5 40.5 410.9 1966 20.5 0 44.5 33 56.5 18 19.5 80 49.8 0 0 0 321.8 1967 38 20 61 149.5 56 48 33 25 48 9.8 38.5 12.5 539.3 1968 0 6 26.5 49 47.5 52.5 0 50.5 31.6 21 45 7 336.6 1969 6 12 18 48.5 6 59.1 22.7 32.5 20.3 21.5 29.5 0 276.1 1970 45 21.5 3.5 7.5 14.1 22.5 21.5 63.5 10 37 13 34.5 293.6 1971 35 53 69 55 54 24.5 73 66.5 12.5 26 58.5 6.5 533.5 1972 22.1 37.5 30 57 30.5 0 26.5 53.5 47 0 29.5 16.5 350.1 1973 18 6.5 14.5 35.5 13 39 23.7 30.5 16.5 8.4 18 45 268.6 1974 34 51.5 52.5 0 90.3 21.5 6.5 75 27 21.5 28.5 0 408.3 1975 27 16.5 17 18.5 44 49.4 36.5 63.5 48.5 5 55.6 19.5 401 1976 23 45.8 40.5 33 41.5 71.3 32.7 19 31.5 100.9 46.2 75.5 560.9 1977 7.5 27.5 27.5 55.2 100.5 45 17 10 28 26 35.6 18.4 398.2 1978 9 20 45.9 45 14 27 24.9 49 13 23.7 24 37.3 332.8 1979 26.5 69.5 4.8 7.5 38.5 20.5 68 55.4 40.7 44.8 9.5 31 416.7 1980 26.5 15.5 4 21 15.5 32 7.5 31.7 33.7 30 99 0 316.4 1981 121.5 146 86.8 111 42.9 12.5 53.5 104.5 11 14 34.7 26.5 764.9 1982 7 38.5 27.5 202.4 0 41 38 27 52.5 33.5 14.7 24.9 507 1983 4 34.1 5.2 28 42 38.7 45 9 57.5 23.9 45.5 12 344.9 1984 35 27.1 33 22.3 12.6 17 30.2 3 10.5 44.7 13.8 20 269.2 1985 90.4 46.3 19.7 53.5 9.5 39 74 13.3 9.5 98.5 57.4 43.7 554.8 1986 15.3 25.6 28.3 27.1 6.6 12.6 18.4 168 13.6 41.6 11 4 372.1 1987 6 4.4 15.2 112 26.5 34.6 20.5 46.2 49 10 0.2 25.6 350.2 1988 3.5 5.2 40.8 62.8 19.5 46.1 16.4 22.9 30.9 38 15.5 35.2 336.8 1989 26.5 19.5 40.3 82 8.3 19.9 16 36.5 8.2 96 37 0.5 390.7 1990 10.5 33.5 11.5 114.2 50.5 53.2 4 18.5 20.8 22.4 25.8 29.4 394.3 1991 35.5 46.5 12.6 9.8 23.4 21.8 18.2 2.7 17.7 132.7 6.5 24.1 351.5 1992 20.1 36.5 39.6 30.2 15.3 55.7 36.3 17.5 20.9 96.6 76.5 9.5 454.7 1993 25.7 20.5 11.1 125.2 33.2 22 19.5 21.5 39 8 14 68.5 408.2 1994 11.5 37.5 24.5 39 22.7 26.6 22.7 48.7 24.8 43 5.5 79.4 385.9 1995 14.4 34.6 51.9 49.5 68.8 12.4 29.6 23.5 35.5 23.8 92.9 92.7 529.6 1996 22.4 11 39.8 15 10.6 8.5 43.4 9.5 35.5 87.8 125.1 19 427.6 1997 12.2 27.5 69.5 32 58.7 29.8 63.9 31.3 8 17.2 30 8 388.1 1998 25 15.1 67.1 56.3 37.8 10.8 23.7 23.1 21.1 13 35.3 39 367.3 1999 51.5 55.5 30.7 30.3 20.2 11.2 24.4 19.2 35 44.3 5.5 18.5 346.3 2000 51 19.5 105 12 58.7 6 16.9 16.6 16 30 48.5 37.7 417.9 2001 24.5 7.2 26 97.6 14 6.6 14.7 51.1 27 19.4 41.5 12.1 341.7 2002 21.4 34.4 2.5 46.5 38.3 23.8 33.4 58.6 38.7 4 28.8 28 358.4

Year JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Annual Totals 2003 24.5 20 177 25.6 79.6 21.5 15.8 41.9 13.8 65 6 16 506.7 2004 12 23 13.8 78.3 17.1 32.4 37.5 12.8 25.3 133.5 4.5 189.8 580 2005 41 3.3 47.5 80.7 56.2 49.1 5 29.2 7.5 10 12.7 0.5 342.7 2006 17.5 15.5 17.5 86.5 49.9 36.5 89.7 118 24.4 54 23 23.5 556 2007 13 15 20.5 64.5 62 34.8 55 19.5 6 24.5 269.9 58.5 643.2 2008 11.5 40.5 17 15.8 12 36.5 9.5 50.5 30.3 37 126 1.5 388.1 2009 9.5 22.5 2.1 46 18 62.7 24.5 6 24 63 12 10 300.3 2010 3 5 14.5 14 14.5 56.5 27 16 0 82 26 27 285.5

Average 405.904

300.0

250.0

200.0

Summer 150.0 Autumn Winter Spring 100.0

50.0

0.0 1920 1940 1960 1980 2000 2020

The average rainfall per season (Table 14) from 1940 to 2010 for each year was computed. The highest average rainfall was in autumn at 118.4 mm, followed by spring at 109.2 mm, while in winter average rainfall was 100.3 mm and the lowest average rainfall is in summer at 77.9 mm.

Year Summer Autumn Winter Spring 1940 134.9 80.4 35.4 115.9 1941 60 112.7 112.6 109.2 1942 98.9 102.8 60 88.8 1943 182.5 65.6 36.3 222.6 1944 13.5 147.6 113.1 105.2 1945 8.6 114.9 124 96.4 1946 13.7 108.9 79.1 67 1947 28 171.2 84.9 98.4 1948 70.1 107.3 55.1 138 1949 40.7 97.8 46.2 171.5 1950 39.7 98.4 93.4 170 1951 118.7 94.9 155.2 95.1 1952 83.3 55.2 79.6 205 1953 41.8 41 169 176.6 1954 32.5 192.5 144.5 70.3 1955 151.7 27.6 114 116.7 1956 108.6 160 88 120.5 1957 74.1 110.1 169 75 1958 38.5 191.5 87.5 53.5 1959 116.5 218.5 163 93.5 1960 98 99 96 114 1961 62.5 72.5 73 78.5 1962 85 95 182.5 101.5

Year Summer Autumn Winter Spring 1963 63 132 59.5 64.5 1964 24.5 96 121.5 117 1965 77.5 120.1 59 154.3 1966 20.5 134 117.5 49.8 1967 70.5 266.5 106 96.3 1968 13 123 103 97.6 1969 18 72.5 114.3 71.3 1970 101 25.1 107.5 60 1971 94.5 178 164 97 1972 76.1 117.5 80 76.5 1973 69.5 63 93.2 42.9 1974 85.5 142.8 103 77 1975 63 79.5 149.4 109.1 1976 144.3 115 123 178.6 1977 53.4 183.2 72 89.6 1978 66.3 104.9 100.9 60.7 1979 127 50.8 143.9 95 1980 42 40.5 71.2 162.7 1981 294 240.7 170.5 59.7 1982 70.4 229.9 106 100.7 1983 50.1 75.2 92.7 126.9 1984 82.1 67.9 50.2 69 1985 180.4 82.7 126.3 165.4 1986 44.9 62 199 66.2 1987 36 153.7 101.3 59.2 1988 43.9 123.1 85.4 84.4 1989 46.5 130.6 72.4 141.2 1990 73.4 176.2 75.7 69 1991 106.1 45.8 42.7 156.9 1992 66.1 85.1 109.5 194 1993 114.7 169.5 63 61 1994 128.4 86.2 98 73.3

Year Summer Autumn Winter Spring 1995 141.7 170.2 65.5 152.2 1996 52.4 65.4 61.4 248.4 1997 47.7 160.2 125 55.2 1998 79.1 161.2 57.6 69.4 1999 125.5 81.2 54.8 84.8 2000 108.2 175.7 39.5 94.5 2001 43.8 137.6 72.4 87.9 2002 83.8 87.3 115.8 71.5 2003 60.5 282.2 79.2 84.8 2004 224.8 109.2 82.7 163.3 2005 44.8 184.4 83.3 30.2 2006 56.5 153.9 244.2 101.4 2007 86.5 147 109.3 300.4 2008 53.5 44.8 96.5 193.3 2009 42 66.1 93.2 99 2010 35 43 99.5 108

Average 77.9 118.4 100.3 109.2

4.1.6 Change detection analysis

The process of identifying differences in the state of an object by observing it at different times is known as change detection (Singh, 1989; Ramachandra and Kumar, 2004; Xu et al., 2009). There are many techniques for change detection analysis (Alqurashi and Kumar, 2013). Reclassification and overlay processes (Eric and Forkuo, 2012) were used in this study to analyse change between land use classes of the Goukou catchment. The reclassification process involves classifying one image according to new land use classes to match the legend of both years being analysed. An overlay involves combining information from different layers. By using these methods amount of gain and losses, net change, and contributors to change from each land use were computed (Johnson, 2009). Changes in each land use class were observed for each decade.

The following tables (Table 15 and 16) (graphs can be found as appendix J) shows the area gains and losses in hectares and percentages from 1940 to 2010. Change detection analysis results indicated that, between 1940 and 1954 natural vegetation gained about 9.6% of the total catchment area and lost about 5.6%. Cultivated fields gain about 3% of the total catchment area and lost about 1.3%, with disturbed surfaces lost about 6.4% and gained about 2.6% of the total catchment area. Bare surfaces gained about 2.7% of the total catchment area and lost about 4.2% between these years, and losses and gains in wetland cover were roughly the same at 0.2%. Between 1954 and 1967, natural vegetation lost about -9% and gained 4.9% of the total catchment area, whereas cultivated fields gains and losses remain roughly constant at 3% and 1.3% respectively, and disturbed surfaces gained about 6% of the total catchment area and lost about 3%. Bare surfaces gained about 2.4% of the total catchment area and lost about 3% and aliens gained about 0.4% of the total catchment area and lost about 0.1% between these years. The highest percentage gain of natural vegetation cover was between 1967 and 1974 where it gained about 70% of the total catchment area and only lost about 5.2%. Cultivated fields and wetlands gains and losses between these years are roughly constant at 2% and 0.2% respectively. Disturbed surfaces gained about 3% of the total catchment area and lost about 5.5%, while alien vegetation gained about 0.33% of the total catchment area between these years.

Between 1974 and 1985, natural vegetation lost about 13% and only gained about 4.7% of the total catchment area, while cultivated fields gained about 2.5% of the total catchment area and lost about 36%. Disturbed surfaces gained about 10% of the total catchment area and lost about 3% and aliens gained about 0.8% of the total catchment area and lost about 0.2%. Between 1985 and 1991, natural vegetation gained about 9% of the total area and lost about 6%. Cultivated fields gained about 4.5% of the total catchment area and only lost about 1.9%. Disturbed surfaces gained about 5.1% of the total catchment area and lost about 8.2%, while bare surfaces gained about 2.1% of the total catchment area and lost about 5.1% between these years. Between 2006 and 2010, natural vegetation gained about 3.2% of the total catchment area and lost about 6%, while cultivated fields gains and losses are roughly constant at 1.6%. Disturbed surfaces gained about 5.4% of the total catchment area and lost about 3.8%, while aliens gained about 0.5% of the total catchment area and there were no losses for alien vegetation between these years.

Table 15: Net gains and losses in (ha) and in (%) from 1940 to 2010. These are as calculated by overlaying images for two consecutive decades in IDRISI.

Year Class Gains (ha) Losses (ha) % (Gains) % (Losses) 1940-1954 Natural vegetation 36805 -21319 9.6 -5.6 Cultivated fields 9862 -5154 2.6 -1.3 Wetlands 681 -764 0.2 -0.2 Disturbed surfaces 1003 -24593 2.6 -6.4 Plantation 588 -19 0.2 -0.01 Roads 277 -246 0.1 -0.1 Bare surfaces 10468 -15958 2.7 -4.2 River 77 -83 0.02 -0.02 Dams 52 -32 0.01 -0.01 Aliens 251 -824 0.1 -0.33 Cleared plantation 0 -72 0 -0.02

1954-1967 Natural vegetation 18772 -36361 4.9 -90.5 Cultivated fields 14167 -4988 3.7 -1.3 Wetlands 467 -820 0.1 -0.2 Disturbed surfaces 21044 -11436 5.5 -3 Plantation 310 -141 0.1 -0.04 Roads 412 -292 0.1 -0.1 Bare surfaces 9205 -11904 2.4 -3.1 River 123 -84 0.03 -0.02 Dams 119 -48 0.03 -0.01 Aliens 1673 -217 0.4 -0.1 Cleared plantation 0 0 0 0

1967-1974 Natural vegetation 26888 -20031 70.1 -5.2 Cultivated fields 8754 -7831 2.3 -2.04 Wetlands 739 -781 0.2 -0.2 Disturbed surfaces 13136 -20876 3.4 -5.5 Plantation 244 -154 0.1 -0.04 Roads 433 -400 0.1 -0.1

Year Class Gains (ha) Losses (ha) % (Gains) % (Losses) Bare surfaces 9761 -9006 2.5 -2.4 River 0 -154 0 -0.04 Dams 111 -44 0.03 -0.01 Aliens 731 -1635 0.2 -0.4 Cleared plantation 0 0 0 0

1974-1985 Natural vegetation 18100 -48990 4.7 -12.8 Cultivated fields 9514 -13679 2.5 -35.6 Wetlands 815 -818 0.2 -0.2 Disturbed surfaces 35175 -12703 9.2 -3.3 Plantation 436 -652 0.1 -0.2 Roads 937 -463 0.2 -0.1 Bare surfaces 19273 -9406 5.02 -2.5 River 162 -121 0.04 -0.03 Dams 171 -135 0.03 -0.04 Aliens 3078 -739 0.8 -0.2 Cleared plantation 0 -72 0 -0.02

1985-1991 Natural vegetation 36095 -24519 9.4 -6.4 Cultivated fields 18452 -7166 4.8 -1.9 Wetlands 791 -888 0.2 -0.2 Disturbed surfaces 19587 -31274 5.1 -8.2 Plantation 977 -380 0.3 -0.1 Roads 522 -898 0.1 -23 Bare surfaces 7995 -19442 2.1 -5.1 River 94 -196 0.02 -0.4 Dams 170 -118 0.04 -0.3 Aliens 2349 -2278 0.6 -0.6 Cleared plantation 183 -81 0.1 -0.2

2006-2010 Natural vegetation 12341 -23122 3.2 -6.01

Cultivated fields 6237 -4090 1.6 -1.1 Wetlands 0 -200 0.02 -0.1

Year Class Gains (ha) Losses (ha) % (Gains) % (Losses) Disturbed surfaces 20787 -14667 5.4 -3.8 Plantation 25 -624 0.01 -0.25 Roads 0 -1 0 -0.4 Bare surfaces 3751 -1384 1 -0.01 River 0 -25 0 -0.04 Dams 0 -160 0 -0.4 Aliens 2132 -1542 0.6 -0.43 Cleared plantation 325 -31 0 0

4.1.7 Markov chain analysis results

The following results were obtained by using 2006 and 2010 images to create a 15 year land cover projection of the Goukou catchment. These images were the latest images used in this study, which made them most appropriate to project future changes of land cover and land use in the Goukou catchment. The following two tables (Tables 16 and 17) shows the results of transition areas probability matrices and transition probability matrices obtained after performing Markov chain analysis for a period of fifteen years (2010-2025). The number of pixels will be persisted during the simulation are presented by the diagonal matrices while the number of pixels that are projected to change during the simulation period are presented by the off-diagonals of the matrices.

These results shows how much of each land use and land cover are projected to transform to another land cover over this specified time period. By analysing the results in Table 17, it is observed that only 38.8% of natural vegetation cover is projected to remain constant by 2025. Natural vegetation is projected to transform into cultivated fields (12.3%), wetlands (0.1%), disturbed surfaces (37.5%), plantations (less than 0.5%), roads (0.1%), bare surfaces (about 8%), river (0.1%), dams (0.1%) aliens (3.1%), and cleared plantations (less than 0.5%), by the year 2025. Nonetheless, certain proportions of each of these land cover classes are in turn projected to revert to natural vegetation, indicating substantial turnover in land cover class, possibly related to fire and fire recovery (prominently in the form of bare surface and disturbed surface areas).

Nv Cf W Ds P Ro Bs R D A Cp Nv 2367873 747138 7401 2288383 2235 3927 486845 2570 4323 187309 798

Cf 429623 2955205 10128 872367 611 848 29482 686 2766 13218 138

W 9202 13140 60193 20892 13 20 1676 1969 11104 493 3

Ds 1737437 1350527 5803 2308887 3639 3674 180029 4682 5587 86210 1937

P 15020 7968 36 28556 23647 22 1027 22 62 695 48169

Ro 97 7 0 48 74 43757 7 3 0 4 24

Bs 244670 49724 369 230723 168 464 209395 172 289 14190 40

R 1874 909 1350 4443 3 2 112 13889 144 79 1

D 7881 10173 461 15633 15 10 504 19 7632 256 4 A 160654 32093 490 135881 113 206 26693 170 209 33165 31

Cp 3760 2049 38 11228 115 5 171 7 12 92 86464

Table 17: Conditional probability of changing land cover and land use classes. Transition probability matrices in (% cover and area per pixel) used for projections of land use and land cover in 2025 (2006-2010). This reports the probability that each land use and land cover type would be found at each pixel after the specified time period (Eastman, 2012). Rows represent initial state and columns represent future state. Natural vegetation is represented by Nv, Cultivated fields by Cf, Wetlands by W, Disturbed surfaces by Ds, Plantations by P, Roads by Ro, Bare surfaces by Bs, River by R, Dams by D, Aliens by A and Cleared plantations by Cp.

Area Nv Cf W Ds P Ro Bs R D A Cp (p.p)

Nv 2367873 0.3883 0.1225 0.0012 0.3752 0.0004 0.0006 0.0798 0.0004 0.0007 0.0307 0.0001

Cf 429623 0.0996 0.6849 0.0023 0.2022 0.0001 0.0002 0.0068 0.0002 0.0006 0.0031 0.0000

W 9202 0.0775 0.1107 0.5071 0.1760 0.0001 0.0002 0.0141 0.0166 0.0935 0.0042 0.0000

Ds 1737437 0.3054 0.2374 0.0010 0.4059 0.0006 0.0006 0.0316 0.0008 0.0010 0.0152 0.0003

P 15020 0.1199 0.0636 0.0003 0.2280 0.1888 0.0002 0.0082 0.0002 0.0005 0.0055 0.3847

Ro 97 0.0022 0.0002 0.0000 0.0011 0.0017 0.9940 0.0002 0.0001 0.0000 0.0001 0.0006

Bs 244670 0.3261 0.0663 0.0005 0.3075 0.0002 0.0006 0.2791 0.0002 0.0004 0.0189 0.0001

R 1874 0.0822 0.0399 0.0592 0.1948 0.0001 0.0001 0.0049 0.6090 0.0063 0.0035 0.0000

D 7881 0.1850 0.2389 0.0108 0.3671 0.0003 0.0002 0.0118 0.0005 0.1792 0.0060 0.0001

A 160654 0.4122 0.0824 0.0013 0.3487 0.0003 0.0005 0.0685 0.0004 0.0005 0.0851 0.0001

Cp 3760 0.0362 0.0197 0.0004 0.1080 0.0011 0.0000 0.0016 0.0001 0.0001 0.0009 0.8319

About 68.5% of cultivated fields are projected to remain persistent over the 15 year time interval used in this study. Transition probability matrices show that about 10.0% of cultivated fields is projected to transform into natural vegetation cover, 0.2% (wetlands), 20% (disturbed surfaces), less than 0.5% (plantation), less than 0.5% (roads), 0.7% (bare surfaces), less than 0.5% (river), 0.1% (dams) and 0.3% (aliens).

0.5% (plantations), less than 0.5% (roads), 1.4% (bare surfaces), 1.7% (river), 9.4% (dams), 0.4% (aliens) by the year 2025.

About 40.6% of disturbed surfaces are projected to be persistent by the projected time interval of 15 years used in this study. About 30.5% is projected to transform into natural vegetation, 23.7% (natural vegetation), 0.1% (cultivated fields), 0.1% (wetlands), 0.1% (plantations), 0.8% (roads), less than 0.5% (bare surfaces), 0.1% (river), 0.6% (dams) and 38.5% (aliens) is projected to transform into these respective land use classes by the year 2025, respectively.

Only about 18.9% of plantations are projected to remain persistent over the time interval of 15 years used in this study. About 12.0% is projected to transform into natural vegetation, 6.4% (natural vegetation), less than 0.5% (cultivated fields), 22.8% (wetlands), less than 0.5% (disturbed surfaces), 0.8% (roads), less than 0.5% (bare surfaces), 0.1% (river) and 0.1% (dams).

Only about 27.9% of bare surfaces are projected to remain persistent over the 15 year time interval used in this study. About 32.6% is projected to transform into natural vegetation, 6.6% (cultivated fields), 0.1% (wetlands), 30.8% (disturbed surfaces), less than 0.5% (plantations), 0.1% (roads), less than 0.5% (river), less than 0.5% (dams), 1.9% (aliens) and less than 0.5% (cleared plantations).

About 18.0% of dam will remain persistent over the 15 year time interval used in this study. About 18.5% is project to transform into natural vegetation, while 23.9% (cultivated fields), 1.1% (wetlands), 36.7% (disturbed surfaces), less than 0.5% (plantation), less than 0.5% (roads), 1.2% (bare surfaces), 0.1% (river), 0.6% (aliens) and less than 0.5% (cleared plantations) is projected to transform into these respective land use classes by the year 2025, respectively.

About 8.5% of aliens are projected to remain persistent by the year 2025. About 41.2% of alien vegetation cover is projected to transform into natural vegetation, 8.2% (cultivated fields), 0.1% (wetlands), 34.9% (disturbed surfaces), less than 0.5% (plantation), 0.1% (roads), 6.9% (bare surfaces), less than 0.5% (river), 0.1% (dams) and less than 0.5% (cleared plantations).

4.2 Discussion

In this research, 1940 was used as an initial year to document and quantitatively analyse the rate of impacts of land cover and land use change in the Goukou catchment. Using advanced GIS and remote sensing techniques to process and analyse BWAP and CIR images, overall accuracies of 57%, 60% and 61% for 1940, 1974 and 1991, respectively, were achieved. These results were found to have a moderate fit (Pontius, 2000). In addition to the quantification of historical change using these images, field visits provided for targeted ground-truthing and key information required to explore the possible socioeconomic dimensions of land cover change drivers. Remotely sensed data have a long history of useful service for deriving land cover maps, and grey-scale aerial photographs provide a far deeper time reach to allow capture of changes to land cover that occurred well before the launch of the first space-based multi-spectral sensors on the Landsat platform in 1972 (AFPO, 2005; Bell, 1995). This section will discuss trends, drivers, change detection and future projections which were examined from BWAP and CIR images by using advanced GIS and remote sensing.

4.2.1 Trends

Land use trends revealed that the Goukou catchment land use has undergone a significant change with differing magnitude and rates. Different factors played major roles in accelerating and transforming land use in the Goukou catchment, such as conversion of natural habitat to agricultural fields, and invasion of alien species. The adverse impacts of heavy rainfall that flooded the area over the years may have been exacerbated by smaller but critical trends in the conversion of wetlands to cultivation fields, and the invasion of wetland and riparian zones by alien vegetation. When wetlands are modified by alien invasions or are replaced by agricultural activities, natural capital is lost and the system is no longer capable of sufficiently providing the same quality of hydrological ecosystem services (Rebelo, 2012).

The invasion of alien species has been predicted to cause a reduction in river flow over time and results in the vulnerability in land surface area where alien vegetation was removed (Rebelo, 2012). After alien vegetation has been removed, it makes the area susceptible to erosion especially because alien trees are concentrated near river banks in this catchment. When there is a high rainfall and the river floods, the area where the alien vegetation was removed may become easily eroded, because the area was left bare or un-vegetated. The area where the alien vegetation was removed should be replaced as fast as possible with another

82 vegetation type, preferable indigenous vegetation to protect the area from floods and erosion (Western Cape Environmental Affairs and Development Planning/Western Cape Department Agriculture, 2015).

During the seventy years of this study over which land cover change could be reconstructed from aerial photography in the Goukou catchment, an increase in disturbed surfaces, alien vegetation, construction of dams and cultivated fields can be observed, whereas there is a decline in river, natural vegetation, bare surfaces and wetlands. This is the general trend of land cover change pattern for the Goukou catchment from 1940 to 2010. These trends would very likely have negative impacts on the delivery of ecosystem services such as water quality and the regulation of streamflow (Power, 2010). Deterioration in ecosystem service delivery may create additional vulnerabilities to projected climate change (Dallas et al., 2014). The main driver of these trends appeared to be human and natural activities. Reduction in natural vegetation and transformation in wetlands were mostly due to unsustainable agricultural practices and alien invasion (Rebelo, 2012).

4.2.2 Drivers

One of the key goals of global change research is to understand the drivers and implications of land cover change (Lambin et al., 2006; Rindfuss et al., 2004). The main purpose for studying land cover and land use drivers is that by understanding the past and current land cover and land use drivers we can understand future drivers and be able to address their impacts on natural resources to prevent future degradation of natural habitats such as wetlands.

Land cover change observations and monitoring rely largely on remotely sensed data coupled with field observations, the latter of which can add an important detail that describes social, economic and physical dimensions of land cover change (Raumann and Soulard, 2007). Understanding the causes and the significances of land cover change requires robust quantification, documentation of the nature and the rate of land cover change, the elaboration of the socioeconomic, and biophysical context in which that change is occurring (Sherbinin, 2002) and the understanding of both how people make use of each piece of land and how social and environmental factors interact to influence the use of that land (Olson, et al., 2014).

There were two types of land use drivers that were identified as likely to be important in the Goukou catchment as inferred from processing and analysing of BWAP and CIR images, being environmental and socioeconomic drivers. Socioeconomic drivers are those that include agricultural practices, water quality and quantity. Socioeconomic drivers lead to changes in the state of ecosystem services and other natural resources while degrading the environment. Changes in the functioning of natural ecosystem have impacts on the provision of ecosystem service delivery which directly or indirectly benefits human well-being.

Environmental drivers include floods, soil erosion and land degradation (Druga and Faltan, 2014). The term “environmental drivers” is synonymous with “direct drivers” as used in the Millennium Ecosystem Assessment (2005). Environmental drivers affect or influence natural habitats directly (Alkemade et al., 2009). There is evidence in this catchment of wetland degradation, decline in natural vegetation, and increase in disturbed surfaces. These are caused by both man and natural activities such as floods. There is a recorded event of heavy rainfall that flooded this catchment in 2006 (CSIR, 2006) and some rehabilitation projects have been initiated to rehabilitate the areas affected by the rainfall floods.

The development of agricultural practice has highly influenced changes in land use of Goukou catchment, such as construction of dams, conversion of wetlands into agricultural land. The on-going efforts to maintain economic growth and development are the consequences of intensification in agriculture and other ways of using landscapes (Rebelo et al., 2012). This consequently, affects ecosystem resilience to degradation (Scheffer et al., 2001).

4.2.3 Rainfall

In the 20th century, rainfall regime globally became more extreme and characterised by fewer and large rainfall occasions and such variations are anticipated to remain through the current century (Ross et al., 2012). For this study, annual total rainfall for each year was computed using rainfall data collected from South African Weather Service.

A recorded flooding event was in 2006, heavy rains flooded the area and destroyed infrastructure (CSIR, 2006). This can be the result of an increase in disturbed surfaces in 2010 from 28.2% to 32.1%, which is the highest percentage of disturbed surfaces recorded from 1940 to 2010 in this area, also dams increased from 0.2% to 0.3% by 2006. Cultivated

84 fields also increased in 2010 to 24.4% from 23.1% of 2006, this might show that there was still a lot of irrigation water stored from 2006 heavy rainfall.

There is also an observed reduction in wetlands, slight increase in cultivated fields, and decrease in disturbed surfaces and river cover between 1967 and 1974; this could be due to the lowest average rainfall in this year. The increase in agriculture and decrease in wetlands can be due to three factors; wetlands are being converted to agricultural fields, wetland degradation and low rainfall. Rainfall impacts on this catchment can be small related to other land use changes drivers such as agricultural activities but have contributed to the increase of other land use types such as increase of dams, disturbed surfaces and increase of agricultural activities.

4.2.4 Change detection

Land cover change can be the results of many factors, in order to investigate these factors; we need to investigate land use and land cover variations (Jara et al., 2012). Transitions from one land use to another can be observed by using a change detection analysis method which can help to understand the interaction between them (Brown et al., 2000). Understanding biophysical and human causes of land use and land cover change, and the land use and land cover patterns and dynamics that affect the structure and functionality of the earth systems are the main objectives of environmental change science (Rindfuss et al., 2004).

This study revealed that there were losses and gains in each land use and land cover class over the period of study, with the major changes in bare surfaces, disturbed surfaces, cultivated fields, natural vegetation, dams and alien vegetation. The graphs and tables for contributors in each class gains and losses can be found as appendix G. and also as overlay map results as appendix H.

Different trends in each land use type played different roles in land class conversions. A positive contribution means that one class contributed to the increase of another class, whereas negative contribution means that, one class resulted in the reduction of the other class. For example, between 1940 and 1954 (Appendix G), disturbed and bare surfaces contributed positively to natural vegetation, whereas cultivated fields contributed negative to natural vegetation.

Agricultural expansion is the major contributor in loss of many land use types, especially with regard to natural vegetation and wetlands in this catchment. Agricultural expansion also affects river water by introducing fertilisers into the river. This has its own negative consequences on the quality and quantity of water (FAO, 2012).

4.2.5 Future projections

The modelling and projection of land use change are important for the assessment of the effects of environmental impacts (Li et al., 2004). A Markov chain model is able to generate predictive understanding of land cover changes from one period to another and thus use this as a basis to project future changes (Estmen, 1995).

The transition areas matrix records the number of pixels that are expected to transform from each land cover type to other land cover type over the specified number of time units (Mandal, 2014). The land use and land cover images of 2006 and 2010 with a period of 15 years were used as variables to project land cover and land use of the Goukou catchment. Land cover change is spatially and temporally persistent over ten to fifteen year intervals, hence a fifteen year interval used in this study was deemed appropriate and within that range. The 10 to 15 year interval available provides fairly regular snapshots in time of the major land cover change. But it should be, however, noted that certain events can change trends significantly and rapid.

The Markov chain analysis results predicted a further increase in agricultural fields, disturbed surfaces, dams and alien vegetation and subsequently a decrease in natural vegetation, wetlands and river, with the probability that cultivated fields will persist in about 69.0% of the total surface area of the catchment by the year 2025 (2.955 million ha), and could show net increases gained from natural vegetation (318 000 ha), but could see net degradation of 737 000 ha lost to the degraded surface land cover class. The highest probability transformations observed were natural vegetation into disturbed surfaces, natural vegetation into alien vegetation and natural vegetation into agricultural fields, at about 39%, 41% and 13%, respectively. If current land use trends are not managed, such a scenario will likely have further adverse effects on ecosystem functionality. The decline of natural vegetation, degradation of wetlands and increase of disturbed surfaces has been observed to accelerate degradation of ecosystem service delivery (Kakembo, 2009).

A decline in natural vegetation and wetlands, and increase in degraded surfaces, has severe implications for future wetland resilience, and will impact negatively on rural livelihoods. In this analysis, agricultural expansion seems to be acting as a catalyst that transforms other land use types such as a decline in natural vegetation, conversion of wetlands thereby degrading the catchment. This is a useful prediction that could serve and be used as an early indicator of possible future scenarios for development and planning. Although this could be prevented if effective environmental strategies to control degradation of natural resources could be put in action. This study provides an important contribution to environmental planning in a catchment level that can be applied even on a global scale.

It is clear that increased agricultural activity resulted in the land cover and land use change in the Goukou catchment. The trend of shifting other natural or semi-natural land cover classes such as natural vegetation and wetlands, into agricultural activities, has increased over the study period. This is because other economic activities such as fishing and tourism might not be sufficient to provide for economic growth and developments in this catchment. Land cover and land use change aspects are the result of complex interactions between several biophysical and socio-economic conditions (Morie, 2007). The effects of human activities on the environment are immediate and often radical while the natural effects take a relatively longer period of time (Chase et al., 2000).

4.3 Research limitations

Research studies that depend upon remote sensing analysis are prone to several limitations. Despite techniques used in this study to try to reduce them, it is important to point these out.

- Historical data are derived from black and white images which provide only one band of spectral reflectance, complicating the interpretation and analysis of the data (Siljander, 2010). In this regard, the textural analysis performed successfully enhanced these images.

- Data availability. Aerial photographs that were available from NGI were not consistent across all the years. In some years, there were not enough images to cover the whole area of the Goukou catchment, and as a result gaps were patched with data available from other years within the same decade or the following decade.

- Shadows in mountainous areas in BWAP covered the different categories of land use in these areas, which were often confused with water bodies. These areas were classified as natural vegetation to avoid highly misclassification. To overcome this drawback, the appropriate image enhancement techniques and methods that will able the detection and discrimination of these land use types need to be studied.

- Seasonal changes or time of the year which the image was taken contributed to the uncertainty of classification of different land use types. In some seasons or years, for an example, agricultural fields appeared as bare surfaces.

CHAPTER 5: CONCLUSIONS AND PRELIMINARY RECOMMENDATIONS

5.1 Introduction

This study was conducted in the Goukou catchment which is known to have undergone a significant change in land cover over the past seven decades, a period for which aerial photographs and colour infrared images exist. These changes would have had likely implications for ecosystem services in this catchment (Falkenmark et al., 2007). However, the rate and extent of change, and the possible drivers have not been formally analysed. Analysis was conducted to provide useful insights into how land cover in this catchment might be managed into the future to achieve a sustainable future flow of services especially as climate continues to change (Maina et al., 2013).

This study was conducted by using advanced GIS and remote sensing tools. Two sets of data were processed and analysed using textural analysis and on-screen manual digitising. BWAP of 1940, 1954, 1967, 1974, 1985 and 1991 were processed and analysed using textural analysis and were classified using unsupervised classification method, and CIR images of 2006 and 2010 were processed and analysed using on-screen manual digitising process. These images were classified into eleven land use classes which were: Natural vegetation, Cleared plantation, Cultivated fields, Wetlands, Disturbed surfaces, Plantation, Roads, Bare surfaces, River, Dams and Aliens.

The main objectives of this study were (i) to determine the main types of land cover, their geographical distributions, and historical changes in land cover and their rates (ii) to develop a Markov model of the observed land cover change, and (iii) to project the implications of continued land cover transformation assuming that historical trends continue. To achieve these objectives the following questions were asked and investigated.

1. What are the historical geographical patterns of land cover change in this catchment?

2. What are the main changes between land cover types and what are their rates?

3. What are the potential implications of these changes?

4. What are the potential future patterns of land cover change assuming historical trends continue?

5. What recommendations could have value in reducing the impacts of land use and land cover change in this catchment?

Results from classified and digitised images indicated that, as expected, land cover in the Goukou catchment has undergone considerable change over the study period (1940 to 2010), revealing a significant change of land use practices. Results show that there is a rapid increase in the area of cultivated fields over the study period. There were also noticeable increases in disturbed areas, alien vegetation and dam construction over the entire time period, while natural vegetation, wetlands and bare surfaces are diminishing. Overall, this study revealed that; there was a decline in the cover of natural vegetation and wetlands which are overwhelmingly due to the conversion of natural habitat by agriculture. Conversion via invasive alien species encroachment and the results of natural hazards such as floods and soil erosion appear to be minor components of land cover change, but may have effects that are out of proportion with their surface area due to their position in the landscape relative to water courses.

The Overall accuracies of 57%, 60% and 61% for 1940, 1974 and 1991, and Kappa Cohen’s coefficient of 52%, 56% and 57% was achieved, respectively. These accuracies are fairly low; according to Pontius (2000),

- < 0: Less than chance agreement

- 0.01- 0.20: Slight agreement

- 0.21- 0.40: Fair agreement

- 0.41- 0.60: Moderate agreement

- 0.61- 0.80: Substantial agreement

- 0.81- 0.99: Almost perfect agreement

Trend analysis results showed that the average trend for natural vegetation cover from 1940 to 2010 was about 50.0%, for cultivated fields (19.5%), disturbed surfaces (18.3%), dams (8.8%) and bare surfaces (1.2%) was found to be, respectively. Most of natural vegetation cover has been transformed into other land use and land cover types. Contributors of these

90 transformations are explained by change detection analysis. Furthermore; average trend change was also computed both in hectares and percentages and it was found to be 14555 ha and 9.1%.

Change detection analysis results indicated that, between 1940 and 1954 natural vegetation gained about 9.6% of the total catchment area and lost about -5.6%. Cultivated fields gain about 3.0% of the total catchment area and lost -1.3%, with disturbed surfaces lost about 6.4% and gained about 2.6% of the total catchment area. Bare surfaces gained about 2.7% of the total catchment area and lost about 4.2% between these years, and losses and gains in wetlands were roughly constant at 0.2%. Between 1954 and 1967, natural vegetation has lost about -9.0% and gained 4.9% of the total catchment area, whereas cultivated fields gains and losses remain roughly constant at 3% and 1.3% respectively, and disturbed surfaces gained about 6.0% of the total catchment area and lost about 3.0%. Bare surfaces gained about 2.4% of the total catchment area and lost about 3.0%, and aliens gained about 0.4% of the total catchment area and lost about 0.1% between these years. The highest percentage gain of natural vegetation cover was between 1967 and 1974 where it gained about 70.0% of the total catchment area and only lost about 5.2%. Cultivated fields and wetlands gains and losses between these years are equal at 2.0% and 0.2% respectively. Disturbed surfaces gained about 3.0% of the total catchment area and lost about 5.5%, while alien vegetation gained about 0.2% of the total catchment area between these years.

Between 1974 and 1985, natural vegetation lost about 13% and gained only 4.7% of the total catchment area, while cultivated fields gained about 2.5% of the total catchment area and lost about 3.6%. Disturbed surfaces gained about 10.0% of the total catchment area and lost about 3%, and aliens gained about 0.8% of the total catchment area and lost about 0.2%. Between 1985 and 1991, natural vegetation gained about 9.0% of the total catchment area and lost about 6.0%. Cultivated fields gained about 4.5% of the total catchment area and lost only 1.9%. Disturbed surfaces gained about 5.1% of the total catchment area and lost about 8.2% while, bare surfaces gained about 2.1% of the total catchment area and lost about 5.1% between these years. Between 2006 and 2010, natural vegetation gained about 3.2% of the total catchment area and lost about 6.0%, while cultivated fields gains and losses are roughly constant at 1.6%. Disturbed surfaces gained about 5.4% of the total catchment area and lost about 3.8%, while aliens gained about 0.5% of the total catchment area and there were no losses for alien vegetation between these years.

Markov chain analysis results indicated about 68.8% of cultivated fields are projected to remain persistent over the 15 year time interval used in this study. Transition probability matrices shows that about 10.0% of cultivated fields is projected to transform in natural vegetation cover, while 0.2% (wetlands), 20.2% (disturbed surfaces), less than 0.5% (plantation), less than 0.5% (roads), 0.7% (bare surfaces), less than 0.5% (river), 0.1% (dams) and 0.3% (aliens) of cultivated fields is projected to transform to respective land use classes and no projection observed that will transform into cleared plantations by the year 2025.

Only about 50.7% of wetland cover is projected to remain persistent by the year 2025, a 15 year time interval that was used in this study. About 7.8% is projected to transform into natural vegetation cover, while 11.1% (cultivated fields), 17.6% (disturbed surfaces), less than 0.5% (plantations), less than 0.5% (roads), 1.4% (bare surfaces), 1.7% (river), 9.4% (dams), 0.4% (aliens) is projected to transform into respective land use classes, and no projections observed that will transform into cleared plantations by the year 2025.

About 40.6% of disturbed surfaces are projected to be persistent by the projected time interval of 15 years used in this study. About 30.5% is projected to transform into natural vegetation, while 23.7% (natural vegetation), 0.1% (cultivated fields), 0.1% (wetlands), 0.1% (plantations), 0.8% (roads), less than 0.5% (bare surfaces), 0.1% (river), 0.6% (dams) and 38.5% (aliens) is projected to transform into these respective land use classes by the year 2025, respectively. Disturbed surfaces are characterised by areas that have some basal cover loss, moderate and severe basal cover loss.

Only about 18.9% of plantations are projected to remain persistent over the time interval of 15 years used in this study. About 12.0% is projected to transform into natural vegetation, while 6.4% (natural vegetation), less than 0.5% (cultivated fields), 22.8% (wetlands), less than 0.5% (disturbed surfaces), 0.8% (roads), less than 0.5% (bare surfaces), 0.1% (river) and 0.1% (dams) is projected to transform into these land use classes by the year 2025, respectively.

Only about 27.9% of bare surfaces are projected to remain persistent over the 15 year time interval used in this study. About 32.6% is projected to transform into natural vegetation, while 6.6% (cultivated fields), 0.1% (wetlands), 30.8% (disturbed surfaces), less than 0.5% (plantations), 0.1% (roads), less than 0.5% (river), less than 0.5% (dams), 1.9% (aliens) and less than 0.5% (cleared plantations) is projected to transform into these respective land use classes by the year 2025, respectively.

About 8.5% of aliens are projected to remain persistent by the year 2025. About 41.2% of alien vegetation cover is projected to transform into natural vegetation, while 8.2% (cultivated fields), 0.1% (wetlands), 34.9% (disturbed surfaces), less than 0.5% (plantation), 0.1% (roads), 6.9% (bare surfaces), less than 0.5% (river), 0.1% (dams) and less than 0.5% (cleared plantations) is projected to transform into these respective land use classes by the year 2025, respectively. The transformation of all other land use and land cover classes is mostly dependent on various aspects, for example, for plantation and cleared plantation, these two land uses depend on when the plantation plantations are harvested and how long if and when they will grow.

5.2 Evaluation of the research questions and the objectives

5.2.1. What are the historical geographical patterns of land cover change in this catchment? - The Goukou catchment is divided into three sections; the low (coastal), the middle and the upper (mountain) sections. By the year 1940, the low reaches of the Goukou catchment were mostly characterised by bare surfaces which were sand dunes, coastal and non-coastal sand. There were few scattered agricultural activities on the far south-east of the catchment. The remaining area was mostly natural vegetation and some few small quantities of other land cover classes identified in this research. The middle reaches were characterised mostly by agricultural fields, wetlands and few dams. This section was also highly vegetated and there were some degraded areas, bare surfaces and relatively few alien invasive species. The upper reach is a mountainous area which is covered by rocks; there were few agricultural activities, dams, disturbed and bare surfaces on this section. The upper reaches remain fairly covered by rocks and natural vegetation, throughout the study period.

- In 1954, the lower reaches were mostly characterised by natural vegetation, fewer agricultural activities and bare surfaces. The stabilisation of sand dunes was introduced around this decade which resulted in the increase of the natural cover and decrease in bare surfaces. On the middle reaches, agricultural fields started to increase as the results dam construction increased, natural vegetation cover decreased, and wetlands degradation also started to increase and plantations were introduced in this decade.

- By the year 1967, most of the natural vegetation in this catchment was mostly distributed between the lower and the upper reaches, the middle parts of this catchment were highly characterised by agricultural fields, dams and wetlands. The main dam, Korente-Vet dam was also constructed in this decade. Plantation and alien vegetation also continue to increase. Alien vegetation is mostly concentrated on the river course which has impacts on the river water.

- Overall, general geographical patterns of the Goukou catchment land cover are; the upper reaches of the catchment are characterised by rocks and natural vegetation, while the middle reaches are heavily infested by alien vegetation, characterised by agricultural activities, dams, wetlands and disturbed surfaces. The lower reaches are characterised by coastal and non-coastal sand, natural vegetation and few alien vegetation. The river runs from the mountain to the sea at the bottom of the catchment.

5.2.2. What are the main changes between land cover types and what are their rates?

The main changes observed in this study for this period of seventy years are:

(i) Increase in agricultural fields, dams and disturbed surfaces

(ii) Decline in natural vegetation, bare surfaces and wetland

5.2.3. What are the potential implications of these changes?

- The increase in agriculture is the key strategy for economic development, poverty alleviation and sustainable development. Although agricultural activities in this catchment have negative implications to other land use types

such as wetlands. Wetlands are mostly degraded by agricultural activities which results in the deterioration of ecosystem services that are essentially provided by wetlands. Fertilisers from agricultural fields near the river, pollute river water, as the results the quality and the quantity of available water deteriorate rapidly. This compromises the lives of the people living in this catchment. The increase of dam construction has also added more pressure on the quality and the quantity of water as this water is usually abstracted from the river. Decline in natural vegetation and wetlands, and increase in degraded surfaces have severe implications for future ecosystem service delivery and impacts negative on rural livelihoods.

5.2.4. What are the potential future patterns of land cover change assuming historical trends continue?

- Markov chain analysis which was used in this study to predict future land cover and land use change of the Goukou catchment revealed different quantities and amounts of land cover change by the defined time period of fifteen years (from 2010 to 2015). The highest future probability of change for natural vegetation is in disturbed surfaces, which is 37.5% by the year 2025; while the highest future probability of change for cultivated fields and wetlands is in disturbed surfaces, which is 20.2% and 17.6%, respectively. Furthermore; the future probability of change for disturbed surfaces is in natural vegetation, which is 30.5%, for plantation is in cleared plantations (38.5%), for bare surfaces is in natural vegetation (32.6%), for river and dams is in disturbed surfaces which is 19.5% and 36.7%, while for aliens and cleared plantations is in natural vegetation and disturbed surfaces and is 41.2% and 10.8%, respectively.

5.2.5. What recommendations could have value in reducing the impacts of land use and land cover change in this catchment?

- Regulation and monitoring of water abstraction from the river, irrigation, and dam construction and expansion.

- Monitoring and maintenance of water canals.

- Community awareness about wetlands and water related issues within the catchment and South Africa as a country, and

- Removal of alien vegetation within the catchment.

5.4 Preliminary recommendations

There are negative impacts on the outcomes of the land use decisions that are intended to maximise a single land use activity such as agricultural production or timber, are likely to produce an accompanying deterioration in the delivery of other services (Polasky et al., 2010). Major challenges are related with understanding and mitigating the negative impacts of land use and land cover change on natural resources. This can be simplified by seeing the potential impacts before and after land use change occurs. Several recommendations that could have value in reducing the impacts of land use and land cover change in the Goukou catchment are discussed below.

5.4.1 Water abstraction from the river

The analysis reveals a substantial intensification in the number of dams and the area of dams in this catchment over the period of observation. These dams are used for irrigation purposes and mostly get their water through water abstraction from the river. The abstraction of this water can have effects on streamflow, water quality and quantity due to return flows from irrigation. Given the current national concerns with water quality and quality; which promotes continuing development and implementation of water resource protection measures, water abstraction and storage should be registered and authorised. Moreover; water abstraction can reduce the ability of a river water to dilute pollutants such as agricultural return flows. Over-water abstraction can also result to river drying up.

5.4.2 Monitoring and maintenance of water canals

Water canals have also increased substantially in this catchment according to this analysis. These canals are used as a waterway for mainly irrigation purposes. This can suffer losses due to agricultural expansion and increasing alien vegetation. Canals should be constantly monitored and maintained to avoid blockages and vegetation growth, which can lead to

96 reduced water flow on the canal system, water loss as results of floods and little or no water reaching the farmers.

5.4.3 Alien vegetation

This analysis revealed that alien plant cover increased over the period of study. However, the cover was overall very low, and this could be due to misclassification errors. In some years, it is possible that alien species cover has been misclassified as natural vegetation. The Goukou catchment has been identified as one of the worst areas affected by alien vegetation invasion in the Western Cape (Tasneem and Earl, 2013). The removal of alien invasive plants could limit future adverse impacts (Western Cape IWRM Action Plan, 2011), and “there is an obligation under the Conservation of Agricultural Resources Act, 1983, to manage invasive plants” (Resource Act 43 of 1983, 1996). However, it would be valuable to be able to quantify their effects using hydrological models to make the economic case for further efforts in this regard. A well-developed hydrological modelling approach that uses the data compiled in this thesis could assess how much alien vegetation affects the proper functionality of ecosystem services thereby degrading wetlands and riparian zones. Moreover, responsibility needs to be given on a civic and municipal level regarding strategic awareness (Chamier et al., 2012).

5.4.4 Awareness about wetlands and water

The wetland ecosystem systems have been severely reduced in cover according to this analysis, due to both human activities such as agricultural expansion, and exacerbated by resulting floods and soil erosion. It would therefore be useful to explore how palmiet vegetation in wetlands could be rehabilitated and restored. Palmiet communities appear to act as a “super glue” of the wetland and thereby ensure the provision of water supply, water quality, biodiversity, and flood protection (Rebelo, 2012). Farmers and communities could be more directly educated about the importance of wetlands and their role in ecosystem service delivery.

5.4.6 Flood structures

Large rainfall events have compromised or destroyed flood and erosion protection structures in this catchment over many years (CSIR, 2006). Properly planned flood and erosion

97 protection structures, such as gabions, levee, reservoirs and weirs can reduce the effects of floods, soil erosion and thereby lessen land degradation. Preventing the source of floods can result to a reduction of its impacts to natural habitats. It would be valuable to assess the potential for rehabilitation in the river course to reduce the impacts of such events.

5.4.7 Dam construction and expansion

This analysis revealed that dam construction has increased over this study period, but not in a simple fashion, rather over particular periods, indicating shifting economic incentives or technology development and adaptation by land users in the catchment. It is unclear if all dams in this area are registered on the Water Authorisation and Use (WARMS) database as suggested by Reconciliation Strategy for Stillbaai (2010). Nevertheless, dam construction should be monitored and regulated by government institutes such as Duiwenhoks Water User Association. Dam water is usually abstracted from the river; if this abstraction is not regulated and monitored this can affect water availability and water quality from the river for downstream users, and also affect ecosystem health. Moreover; the quality of water in these dams should be inspected at the same time that dam safety inspection is carried out, which should be done every five years, as suggested by dam safety regulations.

More dams have been built over the past years in this catchment to enhance water storage for irrigation, this requires an increase on the amount of water used for irrigation and also irrigation strategies such as; farmers should irrigate during cool conditions to avoid evaporation, also should avoid over-irrigation because that can cause problems such as diseases, especial for the fields closer to the river, should be applied. Also, irrigation schemes should be monitored and maintained to ensure efficient water use and rainwater harvesting should be used in order to supplement water supply for irrigation. Moreover, community awareness about water quality and quantity in South Africa and locally should be initiated.

5.4.8 Summary

Overall, this analysis revealed that there has been a rapid increase in agricultural activities in this area over the past six decades, and that expansions occurred at different rates over the time period, probably reflecting economic incentives and changing agricultural technology and its adoption. This agricultural expansion probably reflects that the main source of income in this area is through agricultural practice. While the increase in agriculture maintains

98 economic growth and development, it also results in the conversion of some land cover types with a higher long term environmental value such wetlands. The fact that this type of conversion is still on the increase suggests that further emphasis needs to be placed on the effective management of land use activities, such as agricultural expansion (Kotze et al., 2010).

Furthermore; if the government and other stakeholders can get involve and encourage farmers and livestock keepers to exercise sustainable and conservative agriculture, the impacts of land use change can be reduced, especial those associated with agricultural practices. Moreover; more robust approaches are also desired to ensure that these people earn income from other activities which are non-farming activities. Also, the government should reinforce institutions dealing with land and conservation issues to integrate and address environmental issues that are related to land use change and other factors.

5.5 Conclusions

While human socioeconomic development has been a significant driver of substantial land cover change globally, the relationship between societal development and regional to local land cover change is complex (Haines-Young, 2009). Many factors play potentially powerful roles in accelerating or slowing the rate of land cover change as society develops, including biophysical, cultural, political and economic drivers (Outlook, 2005). This study shows that in the Goukou catchment, socioeconomic development in the form of changing agricultural practice is the main driving factor, with contributing or interacting factors including alien plant spread, especially along river banks. There are periods during which development accelerates and land cover changes increase, and periods during which land cover is relatively stable, probably reflecting economic drivers. Thus there appears to be no single on-going effort to maintain economic growth and development, rather there are periodic shift and changes.

The Goukou catchment faces a number of significant challenges in relation to its environmental sustainability. These include apparent degradation of especially the riparian ecosystem, with loss of natural habitats, and associated loss of water quality and quantity likely. It seems important to be able to project the ecosystem service implications of these impacts, and through such tools to assess potential recovery and rehabilitation options.

To a greater extent, these challenges are motivated by human activities in terms of consumption, production processes and agricultural profit incentives. Rapid agricultural development stimulated the land use change on the Goukou catchment during the period of 1940 to 2010. The positive aspect of agricultural intensification is that it provides work opportunities to communities and thereby reduces poverty (Cervantes-Godoy et al., 2010), and also contributes to regional and national food security. Agricultural intensification resulted in the increase in the number of dams constructed during the study period of seventy years. These dams are built for irrigation purposes and they obtain water through abstraction from the main river courses, likely affecting the flow of river water. It will be valuable to quantify the effects of this practice in order to determine long term sustainability of this agricultural production model.

There are many potential adverse impacts due to the intensification of agricultural production models (Rodriguez et al., 2004). Fertilisers in run-off from agricultural fields closer to the river can pollute river waters, thus affecting water available for household use (EPA, 2002). Furthermore, the infestation of alien vegetation along river banks and throughout the catchment is likely to be reducing water flows (Chamier et al., 2012). The most common alien vegetation in this area is black wattle (Ivers and Ealth, 2006), and the riparian zones and wetlands in the Goukou catchment are mostly infested by black wattle. These aliens reduce the healthy hydrological function of these areas and thereby reduce river flows (River health programme, 2006). Wetlands are important for flood attenuation, erosion control, improving water security and providing essential hydrological ecosystem goods and services to mankind (Mooney et al., 2005). Thus the land cover conversions observed are incipient drivers behind land degradation and associated reduction in some ecosystem service delivery in this catchment.

Alien invasive plants are known to use increasingly significant volumes of water in correlation to the plants biomass (Chamier et al., 2012). Thus control and removal of alien vegetation especially from the riparian zone and wetlands areas, and rehabilitation of cleared areas with indigenous plants (IPW, 2006) would likely provide towards an increasing hydrological reserve (DAFF, 2011) and thus increase ecosystem service resilience in the face of potential future climate shocks. In addition to benefiting biodiversity in the catchment, and the country as a whole, and benefiting the water resources, removal of alien vegetation is having other site-specific benefits, such as job creation (Driver et al., 2004).

100

Poorly designed flood control structures, have resulted to the conversion of wetlands to cultivated fields by farmers, which have reduced and likely increased the probability of flooding events in this area. Water quality and quantity is being compromised due to the return flow from irrigation which increasingly contains fertiliser. These changes in land cover and land use reflect the dynamics of human impacts on this catchment (Zhang et al., 2014). In order to grow and sustains the country’s economy, water is needed, but also to sustain people and livelihoods (DWA, 2012). National Department of Water Affairs is responsible for responding to expected developments and initiatives in order to provide guidance on water resources and supply. However, the delivery of these resources is the responsibility of the local municipalities.

Results from GIS and remote sensing for land use and land cover change in this catchment shows that there has been expansion, development and intensification in agricultural practices, dams and disturbed surfaces, whereas there is a decline in river, natural vegetation, bare surfaces and wetlands. For this reason and based on the analysis of these results, agriculture has been identified as the major driver of land use and land cover change in the Goukou catchment, and the resulting landscape will interact with likely future increases in extreme climate events. It can be concluded therefore, that land cover change in this catchment has mainly been characterised by expansion of land use types with higher immediate economic benefit, such as agriculture, and the shrinking of some land cover types with higher long term ecosystem service value such as wetlands.

GIS and remote sensing has been used as one of the essential tools in assessing land cover and land use change and their impacts in an area (Muzein, 2010). The use of GIS and remote sensing in this study has successfully achieved its objective of assessing the rate of impacts of individual land use type, document and quantify them. This study may contribute to the policy making for the balanced land use and ecosystem protection on the Goukou catchment.

101

REFERENCES

Aerial Photography Field Office (AFPO). (2005). Historical Imagery Holdings for the United States Department of Agriculture (USDA).

Agarwal, C., Green, G. M., Grove, J. M., Evans, T. P., & Schweik, C. M. (2002). A review and assessment of land-use change models: Dynamics of space, time, and human choice. General Technical Report NE- 297. Newton Square, Pennsylvania, U.S. Department of Agriculture, Forest Service, North-eastern Research Station. 61 pp.

Alhaddad, B., Roca, J., & Burns, M. (2009). Monitoring urban sprawl from historical aerial photographs and satellite imagery using texture analysis and mathematical morphology approaches. In a European Congress of the Regional Science Association International. "Territorial Cohesion of Europe & Integrative Planning: 49th European Congress of the Regional Science Association International". Belgium, European Regional Science Association, 1-9.

Alkemade, R., Van Oorschot, M., Miles, L., Nellemann, C., Bakkenes, M., & Ten Brink, B. (2009). GLOBIO3: A framework to investigate options for reducing global terrestrial biodiversity loss. Ecosystems, 12(3), 374–390. doi:10.1007/s10021- 009-9229-5

Alqurashi, A. F., & Kumar, L. (2013). Investigating the Use of Remote Sensing and GIS Techniques to Detect Land Use and Land Cover Change. Advance Remote Sensing, 2013 (2), 193–204.

Anderson, J. R., Hardy, E. E., Roach, J. T., & Witmer, R. E. (1976). A land use and land cover classification system for use with remote sensor data. Geological Survey Paper, 964, 2001.

Araya, Y. H., & Cabral, P. (2010). Analysis and Modeling of Urban Land Cover Change in Setúbal and Sesimbra, Portugal. Remote Sensing, 2(6), 1549–1563. doi:10.3390/rs2061549

102

ArcGIS 10.1 help manual. (2013). Available: http://resources.arcgis.com/en/help/main/10.1/index.html [Accessed 25 May 2013].

ArcGIS. (2010). what are map projections ? Page 1 of 155 what are map projections? Available: http://training.esri.com/gateway/index.cfm?fa=catalog.webCourseDetail&coursei d=2151 [Accessed 18 May 2014].

Archana, S., & Elangovan, K. (2014). Survey of Classification Techniques in Data Mining. International Journal of computer Science and mobile Applications, 2(2), 65-71.

Asia Development Bank. (2012). Infrastructure for Supporting Inclusive Growth and Poverty Reduction in Asia.

Auria, L., & Moro, R. A. (2008). Support Vector Machines (SVM) as a technique for Solvency Analysis.

Awwad, W. A. (2003). Land cover mapping: A comparison between manual digitising and automated classification of black and white historical aerial photography. MSc thesis, University of Florida.

Ayivor, J. S., & Gordon, C. (2013). Impact of Land Use on River Systems in Ghana. Institute for environmental and sanitation studies.

Badger, A. M., & Dirmeyer, P. A. (2015). Climate response to Amazon forest replacement by heterogeneous crop cover. Hydrology and Earth System Sciences Discussions, 12(1), 879–910.

Bagley, J. E. (2011). Impacts of land cover change. Energy regulation, Breadbasket Production, and Precipitation.

Bajocco, S., De Angelis, A., Perini, L., Ferrara, A., & Salvati, L. (2012). The impact of land use/land cover changes on land degradation dynamics: A Mediterranean case study. Environ Manage, 49(5):980-9. doi: 10.1007/s00267-012-9831-8. Epub 2012 Mar 15.

103

Baraldi, A., & Parmiggiani, F. (1995). An Investigation of the Textural Characteristics Associated with Gray Level Co-occurrence Matrix Statistical Parameters. IEEE Transaction ON Geoscience and Remote Sensing, 33(2), 293-304.

Bartels, M., Wei, H., & Mason, D. (2006). DTM Generation from LIDAR.

Bell, T. E. (1995). Remote sensing. IEEE Spectrum, 32(3), 3–28. doi:10.1109/6.367967

Belward, A. (1996). The IGBP-DIS global 1 km land cover data set “DISCover” – Proposal and implementation plans. Report of the Land Cover Working Group of the IGBP-DIS. IGBP-DIS Working Paper, No. 13. Stockholm.

BenDor, T., Shoemaker, D. A, Thill, J.-C., Dorning, M. A, & Meentemeyer, R. K. (2014). A mixed-methods analysis of social-ecological feedbacks between urbanisation and forest persistence. Ecology and Society, 19(3).

Berry, M. W., Flamm, R. O., & Maclntyre, R. L. (2011). The Land-Use Change Analysis System (LUCAS) for evaluating landscape management decisions.

Bharati, M. H., Liu, J. J., & MacGregor, J. F. (2004). Image texture analysis: methods and comparisons. Chemometrics and Intelligent Laboratory Systems, 72(1), 57-71.

Biediger, J. M. (2012). The use of digital image processing to facilitate digital land covers zones from grey level aerial photos. MSc thesis, Northwest Missouri State University Maryville, Missouri.

Biggs, R., & Scholes, R. J. (2002). Land cover changes in South Africa 1911-1993. South African Journal for Science, 98.

Bowman, K., Brown, D., Comim, F., Kouwenhoven, P., Manders, T., Milimo, P., Mohamed- Katerere, J., & De Oliveira, T. (2009). Chapter 1: Environmenta for development.

Boysen, L. R., Brovkin, V., Arora, V. K., Cadule, P., de Noblet-Ducoudré, N., Kato, E., & Gayler. (2014). Global and regional effects of land-use change on climate in 21st century simulations with interactive carbon cycle. Earth System Dynamics Discussions, 5(1), 443–472.

Bristow, L. K., Marchant, M. S., Deurer, M., & Clothier, E. B. (2010). Enhancing the ecological infrastructure of soils. World, 13–16.

104

Brown, D. G., & Duh, J.-D. (2004). Spatial simulation for translating from land use to land cover. International Journal of Geographical Information Science, 18(1), 35–60.

Brown, D. G., Page, S., Riolo, R., Zellner, M., & Rand, W. (2005). Path dependence and the validation of agent-based spatial models of land use. International Journal of Geographical Information Science, 19, 153–174.

Brown, D. G., Pijanowski, B. C., & Duh, J. D. (2000). Modeling the relationships between land use and land cover on private lands in the Upper Midwest, USA. Journal of Environmental Management, 59(4), 247–263.

Brown, D. G., Polasky, C., Bolstad, P., Brody, S. D., Hulse, D., Kroh, R., & Thomson, A. (2014). Chapter 13: Land use and land cover change. Climate Change Impacts in the United States: The Third National Climate Assessment, 318-332.

Brown, O. (2014). Measures of Shape : Skewness and Kurtosis. Journal of Gerontology, 200, 1–10.

Busch, G. (2000). Remote Sensing and Land Use Classification : Supervised vs. Unsupervised Classification. MSc thesis, Utah State University.

Büttner, G., Feranec, G., and Jaffrain, G. (2002). Corine Land Cover Update 2000, Technical Guidelines. EEA Technical report No 89. Available at: http://reports.eea.europa.eu/technical_report_2002_89/en [Accessed 07 July 2014].

C.A.P.E Estuary Programme. (2007). Goukou Estuary. Situation Assessment Report Version ii.

Campbell, D., Gichohi, H., Reid, R., Mwangi, A., Chege, L., & Sawin, T. (2003). Interactions between People and Wildlife in Southeast Kajiado District, Kenya. LUCID Working paper 18. Int. Livestock Res. Institute. Nairobi.

Campbell, J. B. (2002). Introduction to Remote Sensing, 3rd ed. Guilford Press, New York. 621p.

Carr, J. R. (1999). Classification of digital image texture using variograms. Advances in remote sensing and GIS analysis (ed. by P.M. Atkinson and N.J. Tate), pp. 135– 146. Wiley, West Sussex, UK.

105

Centre for Remote Sensing (CRS). (2005). Spatial scaling of reflectance and surface albedo over a mixed-use, temperate forest landscape during snow-covered periods. Boston University, Boston, MA 02115.

Ceraudo, G. (2013). Good Practice in Archaeological Diagnostics. Natural Science in Archaeology, 11(31).

Cervantes-Godoy, D., & Dewbre, J. (2010). Economic Importance of Agriculture for Poverty Reduction. Development, (23), 27. doi:10.1787/5kmmv9s20944-en.

Chamier, J., Schachtschneider, K., Le Maitre, D. C. Ashton, P. J. & Van Wilgen, B. W. (2012). Impacts of invasive alien plants on water quality, with particular emphasis on South Africa. Water SA, 38(2).

Charles, O, John, O, & Onisimo, M. (2014). Land use and land cover change in the fringe of eThekwini Municipality: Implications for urban green space using remote sensing. South African Journal of Geomatics, 3(2).

Charted Institute of Management Accounts (CIMA). (2006). Accounting for climate change. How management accounts can help organisation mitigations and adapt to climate change.

Chase, T. N., Pielke, R. A., Kittel, T. G. F., Nemani, R. R., & Running, S. W. (2000). Simulation impacts of historical land cover changes on global climate in northern winter. Climate Dynamics, 16, 93–105.

Chen, J., Sun, B. M., Chen, D., Wu, X., Guo, L. Z., & Wang, G. (2014). Land use changes and their effects on the value of ecosystem services in the small Sanjiang Plain in China. The Scientific World Journal, 2014(2014).

Chen, K. (1998). Correlations between census dwelling data and remotely sensed data. Presented at the 10th Annual Colloquium of the Spatial Information Research Centre. University of Otago, Dunedin, New Zealand.

Chen, M. Q., Zhao, B. P., & Z. J. Luo. (2013). “The ecosystem services value change in the upper reaches of Ganjiang River based on RS and GIS”. Acta Ecologica Sinica, 33(9), 2761–2767.

106

Chiang, S. (2001). Unsupervised target detection in hyperspectral images using projection pursuit. IEEE Transactions on Geoscience and Remote Sensing, 39(7), pp.1380- 1391.

Chin D. A. (2006). Water quality engineering in natural systems. John Wiley & Sons, Inc. 610 pp.

Chu, H-J., Liu, C-Y & Wang, C-K. (2013). Identifying the Relationships between Water Quality and Land Cover Changes in the Tseng-Wen Reservoir Watershed of Taiwan. Int. J. Environ. Res Public Health, 2013(10), 478-489.

Civco, D. L. (1993). Artificial neutral network for land classification and mapping. International Journal of Geographical Information System, 7, 173-187.

Clawson, M. S., & Charles, L. (1965). Land use information. A critical survey of U.S. statistics including possibilities for greater uniformity. Baltimore, Md., the Johns Hopkins Press for Resources for the Future, Inc., 402 p.

Cohen, L., Manion, L & Morrison, K. (2000). Research Methods in Education Routledge. Falmer, London.

Cohen, S. J. (1986). Impacts of CO2- induced climatic change on water resources in the Great Lakes Basin. Climatic Change, 8, 135-53.

Cohen, W. B. (2003). Human Population Growth and Trade-offs in Land Use: Human as Agent to Ecological Change. USA, American Institute of Biological Sciences.

Cohen, W. B., Hasen, J. A., & Rasker, R. (2002). Ecological Causes and Consequences of Demographic Change in the New West. American Institute of Biological Sciences. Bioscience, 52(2), 151-162. 2002.

Cohen, W. B., Kennedy, R. E., Townsend, P. A., Gross, J. (2009). Remote sensing change detection tools and natural resource managers: Understanding concepts and trade- offs in the design of landscape monitoring projects. American Institute of Biological Sciences. Remote Sensing of Environment, 113(7), 1382-1396.

Congalton, R. G. & Green, K. (1999). Assessing the accuracy of remotely sensed data: Principles and Practices. Lewis Publisher, Boca Raton.

107

Congalton, R. G. (1991). A review of assessing the accuracy of classifications of remotely sensed data. Remote Sensing of Environment, 37(1), 35-46.

Congalton, R. G. (2001). Of Wildland Fire and other spatial information. International Journal of Wildland Fire, 10, 321–328.

Congalton, R. G., & Biging, G. S. (1992). A pilot study evaluating ground reference data collection efforts for use in forest inventory. Photogrammetric Engineering and Remote Sensing, 58, 1669-1671.

Conservation of Agricultural Resources Act 43 of 1983. (1996). Permission to Burn Veld and to Graze Burned Veld in the Magisterial Districts of the, (April 1983), 1–24.

Corrigan, B. M. (2009). Historical changes in the extent, composition and structure of the forest patches on the KwaNibela Peninsula. MSc Thesis, University of Johannesburg.

Cots-Folch, R., Aitkenhead, M. J., & Martinez-Casasnovas, J. A. (2007). Mapping land cover from detailed aerial photography data using textural and neural network analysis. International Journal of Remote Sensing, 28(7), 1625–1642.

Crews, K. A. & Walsh, S. J. (2009). Remote sensing and the social sciences. In T. Warner, M. D. Nellis, and G. M. Foody (Eds.), The Sage Handbook of Remote Sensing (pp. 437–445). London: Sage.

CSIR. (2006). State of Western Cape rivers report, 2006. Available: www.csir.co.za/rhp/events/06_day1_sum_report_draft.pdf [Accessed 18 May 2013].

CSIR. (2011). Natural resource and the environment research report 2011. Available: www.csir.co.za/nre/docs/Final_IMPACT_STRATEGY.pdf [Accessed 18 May 2013].

DAFF. (2011). A profile of South African Market Value Chain. Department of Agriculture, Forestry and Fisheries. Pretoria. South Africa.

Dale, V. H. (1997). The relationship between land-use change and climate change. Ecological Applications, 7(3), 753-769.

108

Dale, V. H., & Hacuber, R. A. (2001). Applying ecological principles to land management: ISBN 0-387-95100-8.

Dallas, H. F., Elizabeth, P., Africa, S., & Dallas, H. (2014). Ecological consequences of global climate change for freshwater ecosystems in South Africa, 110(5), 1–11.

Daniel, T. B. (2008). Assessing the effects of land-use / cover change on ecosystem services in Ejisu-Juaben District, Ghana : The case study of carbon sequestration Ejisu- Juaben District, Ghana : The case study of carbon sequestration.

Darwin, R., Tsigas, M., Lewandrowski, J. & Raneses, A. (1996). Land use and cover in ecological economics. Ecological Economics, 17, 157–181.

Davis, E. L., & de Noblet-Docoundre, N. (2010). Climate impacts of global-scale deforestation-radioactive versus non-radioactive processes. Journal of climate, 23(1), 97-112.

De Sherbinin, A., (2002). A CIESIN Thematic Guide to Land-Use and Land-Cover Change (LUCC). Centre for International Earth Science Information Network (CIESIN) Columbia University.

De Sherbinin, A., K. Kline, & K. Raustiala. (2002). Remote Sensing Data: Valuable Support for Environmental Treaties. Environment, 44(1), 20-31.

Deng, X., Zhao, C., & Yan, H. (2013). Systematic Modeling of Impacts of Land Use and Land Cover Changes on Regional Climate: A Review. Advances in Meteorology, 2013.

Deng, X., Zhao, C., Lin, Y., Zhang, T., Qu, Y., Zhang, F., Wu, F. (2014). Down scaling the Impacts of Large-Scale LUCC on Surface Temperature along with IPCC RCPs: A Global Perspective, 2720–2739.

Dent, M. C. (2001). Installed water resource modelling systems for catchment management agencies. Durban. University of Natal, Computing Centre for Water Research.

Department of Economic and Social Affairs. (2013). World Economic and Social Survey 2013: Sustainable Development Challenges.

109

Department of Environmental Affairs. National Biodiversity Framework. (2009). Natl. Environ. Manag. Biodivers, Act 2004.

Department of Land Affairs. (2010). White Paper on South African Land Policy April 1997 Department of Land Affairs. Pretoria.

Di Gregorio, A. & Jansen, L. J. M. (1997). A new concept for a land cover classification system. Proceedings of the Earth Observation and Environmental Information 1997 Conference. Alexandria, Egypt, 13–16 October 1997.

Di Gregorio, A. (1991). Technical report on the land cover mapping of Lebanon. FAO Project NECP/LEB/001/SAU.

Di Gregorio, A. (2005). Land cover classification system: classification concepts and user manual: LCCS (No. 8). Food & Agriculture Org.

Di Gregorio, A., & Jansen, L. J. M. (1996c). The Africover Land Cover Classification System: A Dichotomous, Modular-Hierarchical Approach. Working Paper with the Proposal for the International Working Group Meeting. Dakar, 29–31 July 1996. FAO, Rome.

Di Gregorio, A., Jansen, L. J., & Louisa, J. M. (1997). A new concept for a land cover classification system. Proceedings of the Earth Observation and Environmental Information 1997 Conference, 2(1), 1–9.

Dirmeyer, P. A., Niyogi, D., De Noblet-Ducoudré, N., Dickinson, R. E., & Snyder, P. K. (2010). Impacts of land use change on climate. International Journal of Climatology, 30(13), 1905–1907.

Dixon, T. H. (1991). An introduction to the Global Positioning System and some geological applications. Reviews of geophysics, 29(2), 249-276).

Driver, A., Maze, K., Rouget, M., Lambard, A. T., Nel, J., Turpie, J. K., Cowling, R. M., Desmet, P., Goodman, P., Harris, J., Jonas, Z., Reyers, B., Sink, K., & Strauss, T. (2004). National Spatial Biodiversity Assessment 2004: Priorities for Biodiversity conservation in South Africa. Strelitzia 17. South African National Biodiversity Institute. Pretoria. South Africa.

110

Druga, M., Faltan, V. (2014). Influences of environmental drives on land cover structure and its long term changes: A case study of the villages of Malachov and Podkonice in Slovalia. Moravian Geographical Reports, 22(3), 2014

Duhamel, C. (1995). Programme télédétection et statistique. Cadre de travail statistique utilisation des sols. Draft. Eurostat/CESD-Communautaire, Luxembourg.

DWA. (2009). Strategic planning for water resources in South Africa. A situation analysis. Report No: P RSA 000/00/7809. Department of Water Affairs. Pretoria. South Africa.

Eastman, J. R. & M. Fulk. (1993) “Long Sequence Time Series Evaluation using Standardized Principal Components”. Photogrammetric Engineering and Remote Sensing, 59(8), 1307–1312.

Eastman, J. R. (2012). IDRISI Selva Manual. Worcester, MA: Clark Labs, Clark University. 322p. provided as a PDF with the IDRISI Selva software package.

Eastman, J.R. (2003). “IDRISI Kilimanjaro, Guide to GIS and Image Processing,” Clark Lab, Clark University, Worcester.

Eiten, G. (1968). Vegetation forms. A classification of stands of vegetation based on structure, growth form of the components, and vegetative periodicity. Boletim do Instituto de Botanica (San Paulo), 4.

Eitsch, S. L. N. (2002). ArcView Interface for Swat2000 User’s Guide.

Ejaro, S. P., & Abdullahi, U. (2013). Spatiotemporal Analyses of Land Use and Land Cover Changes in Suleja Local Government Area, Niger State, Nigeria. Journal of Environment and Earth Science, 3(9), 72–84.

Enderle, D., & Jr, R. W. (2005). Integrating supervised and unsupervised classification methods to develop a more accurate land cover classification. Journal of the Arkansas Academy of Science, 59, 65–73.

EPA. (2002). Drinking Water From Household Wells, EPA 816-K-02-003, 19.

111

Erener, A., & Duzgun, H. S. (2009). Supporting user-orientated analysis for multi-view domain-specific visual languages. Information and Software Technology.

Eric, K., Forkuo, A. F. (2012). Analysis of Forest Cover Change Detection. International Journal of Remote Sensing Applications, 2(4), 82–92.

Estuaries of the Cape (1990): Report 34 - Goukou CSW24 – CSIR.

Falkenmark, M., Finlayson, C. M., & Gordon, L. J. (2007). Agriculture, water, and ecosystems: avoiding the costs of going too far. Water for Food, Water for Life- A Comprehensive Assessment of Water Management in Agriculture, 233–277. doi:10.4324/9781849773799.

Fall, S., Diffenbaugh, N. S., Niyogi, D., Pielke, R. A., & Sr., Rochon, G. (2010). Temperature and equivalent temperature over the United States (1979–2005). International Journal of Climatology, 30, 2045-2054.

FAO. (2000). Agriculture towards 2015-30. Technical Interim Report, Rome: FAO.

FAO. (2000). Land Cover Classification System (LCCS). Classification Concepts and User Manual for software version 1.0. By A. Di Gregorio and L. J. M. Jansen. Rome.

FAO. (2012). Coping with water scarcity: An action framework for agriculture and food security. doi:http://www.fao. org/docrep/016/i3015e/i3015e.pdf

Fensham, R. J. & Fairfax, R. J. (2002). Aerial photography for assessing vegetation change: A review of applications and the relevance of findings for Australian vegetation history. Australian journal of Botany, 50(4), 415-429.

Fisher, J. A., Patenaude, G., Giri, K., Lewis, K., Meir, P., Pinho, P., & Williams, M. (2014). Understanding the relationships between ecosystem services and poverty alleviation: A conceptual framework. Ecosystem Services, 7, 34-45.

Foley, J. A., Monfreda, C., Patz, J. A., & Ramankutty, N. (2009). Agriculture, Land Use and the Transformation of Planet Earth. In. The Princeton Guide to Ecology, 724-730.

Foley, J. A., Ramankutty, K. A. N., Brauman, E. S., Cassidy, J. S., Gerber, M., Johnston, N. D., Mueller, C., O'Connell, D. K., Ray, P. C., West, C., Balzer, E. M., Bennett, S. R., Carpenter, J., Hill, C., Monfreda, S., Polasky, J., Rockstrom, J., Sheehan, S.

112

Siebert, D., Tilman, & Zaks, D. P. M. (2011). Solutions for a cultivated planet. Nature, 478(7369), 337-342.

Food and Agriculture Organisation of the United Nations (FAO). (2011). Save and grow: A policymaker’s guide to the sustainable intensification of smallholder crop production. Rome.

Food and Agriculture Organisation of United States (FAO). (1995). Planning for sustainable use of land resources towards a new approach. Land and Water bulletin 2. United Nations Conference on Environment and Development (UNCED).

Foody, G. M. (2002). Status of land cover classification accuracy assessment. Remote Sensing of Environment, 80, 185–201.

Foody, G. M., McCulloch, M. B., & Yates, .B. (1995). Classification of remotely sensed data by artificial neural network. Issues related to training data characteristics. Photogrammetric Engineering & Remote Sensing, 61,391-401.

Freibauer, A., Mathijs, E., & Brunori, G. (2011). Sustainable food consumption and production in a resource-constrained world, 0–147.

Friedl, M. A., Sulla-Menashe, D., Tan, B., Schneider, A., Ramankutty, N., Sibley, A., & Huang, X. (2010). MODIS Collection 5 global land cover: Algorithm refinements and characterization of new datasets. Remote Sensing of Environment, 114(1), 168–182.

Ganoulis, J. (2009). Risk Analysis of Water Pollution. WILEY-VCH, Weinheim achtnebel HP (1994). Multicriterion Decision Making Methods with Ordinal and Cardinal Scales. In: Bogardi J, Nachtnebel HP (Eds) ELECTRE I-III. Multicriterion Decision Analysis in Water Resources Management, UNESCO, Pari.

GEO4/UNEP. (2007). Environment for development. Available: www.unep.org/geo/geo4/report/geo-4report_full_en.pdf

Gerard, F., Petit, S., Smith, G., Thomson, A., Brown, N., Manchester, S., Wadsworth, R., Bugar, G., Halada, L., Bezak, P., Boltiziar, M., De Badts, E., Halabuk, A., Mojses, M., Petrovic, F., Gregor, M., Hazeu, G., Mucher, C. A, Wachowicz, M., Huitu, H., Tuominen, S., Koehler, R., Olschofsky, K., Ziese, H., Kolar, J.,

113

Sustera, J., Luque, S., Pino, J., Pons, X., Roda, F., Roscher, M., & Feranec, J. (2010). Land cover change in Europe between 1950 and 2000 determined employing aerial photography. Progress in Physical Geography, 34(2), 183–205.

Gibbs, J. P. (2000). Wetland loss and biodiversity conservation. Conservation Biology 14, 314-317.

Gleik, P. H., Ajami, N., Christian-Smith, J., Cooley, H., Donnelly, K., Fulton, J., Ha, M,-L., Heberger, M., Moore, E., Morrison, J., Orr, S., Schulte, P., & Srinivasan, V. (2014). The World’s Water Volume 8: The Biennial report on Freshwater Resources. Island Press, Washington, USA.

Global Environment Outlook (GEO). (2000). United Nations Environment Programme. United Nations Environment Programme (UNEP).

Godfray, H. C. J., Beddington, J. R., Crute, I. R., Haddad, L., Lawrence, D., Muir, J. F., Pretty, J., Robinson, S., Thomas, S. M., & Toulmin, C. (2010) . Food security: The challenge of feeding 9 billion people. Science, 327(2010), 812–818.

Gorlinski, G. (2012). The history of agriculture. New York: Britannica Education Pub. in association with Rosen Educational Services.

Gregory, A. W, Hansen B. E. (2009). PRACTITIONERS CORNER: Tests for Co-integration in Models with Regime and Trend Shifts. Oxford Bulletin of Economics and Statistics, 58(3), 555-560.

Groppo J. D., Moraes J. M. D., Beduschi C.E., Genovez A.M. & Martinelli L. A. (2008). Trend analysis of water quality in some rivers with different degrees of development within the Sao Paulo State, Brazil. River. Res. Applic, 24, 1056– 1067.

Gu, J., Congalton, R. G., & Pan, Y. (2015). The Impact of Positional Errors on Soft Classification Accuracy Assessment: A Simulation Analysis. Remote Sensing, 579–599.

Gulbaz, S. Kazezyilmaz-Alhan, C. M. (2012). Impact of Land Use/Cover Changes on Water Quality and Quantity in a Calibrated Hydrodynamic Model. 10th International

114

Congress on Advances in Civil Engineering, 17-19 October 2012 Middle East Technical University, Ankara, Turkey.

Haines-Young, R. (2009). Land use and biodiversity relationships. Land Use Policy, 26(SUPPL. 1), 178–186. doi:10.1016/j.landusepol.2009.08.009.

Halder, A., Ghosh, A., & Ghosh, S. (2011). Supervised and unsupervised landuse map generation from remotely sensed images using ant based systems. Applied Soft Computing, 11(8), 5770–5781. doi:10.1016/j.asoc.2011.02.030.

Hancock, J. F. (2012). Origins of Agriculture. Plant Evolution and the Origin of Crop Species, 99–113.

Harris, J. M. (2013). Green Keynesianism: Beyond Standard Growth Paradigms. Global Development and Environmental Institute: Working Paper No. 13-02. Tufts University, Medford MA 02155, USA.

Harrison, S., Rice, K., & Maron, J. (2001). Habitat patchiness promotes invasions by alien grasses on serpentine soil. Biological Conservation 100:45–53

Himmelstein, D. S., Bi, C., Clark, B. S., Bai, B., & Kohtz, J. D. (2010). Chapter 4 Measures of distance between samples : Euclidean. BMC Developmental Biology, 10(1), 1– 11.

Hoare, R. E. (1999). Determinants of human-elephant conflict in a landuse mosaic. J. Applied Ecol, 36, 689–700.

Hobbs, R. J. (2000). Land-use changes and invasions. Invasive species in a changing world. Island Press, Washington, D.C., USA. 55–64.

Hoffmann, W. H. (2010). Farm Modelling for Interactive Multidisciplinary Planning of Small Grain Production Systems in South Africa. PhD thesis, University of Stellenbosch. Stellenbosch.

Holden, J. E., & Kelley, K. (2010). The Effects of Initially Misclassified Data on the Effectiveness of Discriminant Function Analysis and Finite Mixture Modeling. Educational and Psychological Measurement, 70(1), 36–55.

115

Horning, N. (2004). Land cover classification methods, Version 1.0. American Museum of Natural History, Centre for Biodiversity and Conservation.

Hsu, Shin-Yi. (1978). “Texture-Tone Analysis for Automated Land-Use Mapping.” Photogrammetric Engineering and Remote Sensing, 44, 1393-1404.

Hu, Q., Guo, H., & Jiang, T. (2007). Annual and seasonal stream-flow responses to climate and land-cover changes in the Poyang Lake basin, China. Institute of Geographical Sciences and Natural Resources Research. Chinese Academy of Sciences, Beijing, Chine, 355, 106–122.

Hughes, C. J., Schulze, R. E., Mander, M., & Marais, C. (2012). Altered Land Use and Climate Change Effects on Water Yield and Downstream Ecosystem Services. Golder Associates, Johannesburg, South Africa; University of KwaZulu-Natal, Pietermaritzburg, South Africa; Future Works, Kloof, South Africa, Department of Environment Affairs, Cape Town, South Africa.

Hussain, S., & Miller, D. (2014). The Economics of Ecosystems and Biodiversity (Teeb) for Agriculture & Food Concept Note, 1–20.

Inland Water Chapter. (2013). State of Environment Outlook Report for the Western Cape Province.

IPCC. (2007). Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, 2007. Climate Change 2007. Working Group II. Impacts, Adaptation and Vulnerability.

IPCC. (2012). Glossary of Terms, 555–564. Available: https://www.ipcc.ch/pdf/special- reports/srex/SREX-Annex_Glossary.pdf [Accessed 23 September 2015].

IPCC. (2014): Summary for policymakers. In: Climate Change 2014: Impacts,Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Field, C.B., V.R. Barros, D.J. Dokken, K.J. Mach, M.D. Mastrandrea, T.E. Bilir, M. Chatterjee, K.L. Ebi, Y.O. Estrada, R.C. Genova, B. Girma, E.S. Kissel, A.N. Levy, S. MacCracken, P.R. Mastrandrea, and L.L.White (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 1-32.

116

IPWA Farms guidelines chapter 2. (2006). Conservation and Improvement of the Farm and Vineyard Environment (referred to as “Biodiversity Guidelines”).

Irons, J. (1981). Texture transforms of remote sensing data. Remote Sensing of Environment, 11, pp.359-370.

Ivers, D. U. R., & Ealth, R. I. H. (2006). Ecological State of the Goukou and Duiwenhoks Rivers, (March), 2006.

James, R. A., Ernest, E. H., John T. R., & Richard R. W. (2002). A land use and land cover classification system for use with remote sensing data.

Jara, F. J., Juli, C., Jos, M. A., & Guti, A. (2012). Land Cover and Land Use Change in the Central Spanish-Portuguese Border, 369–398.

Jensen, J. R., (1996). Introductory Digital Image Processing: A Remote Sensing Perspective. Prentice Hall. Upper Saddle River, New Jersey.

Jepson, A. (2011). Image Projection. Introduce the basic concepts and mathematics for image projection. Available: www.cs.toronto.edu/~jepson/csc420/notes/image.pdf.

Jeter, W. G. (2014). A Vegetation Analysis on Horn Island, Mississippi, CA. 1940 Using Characteristic Dimensions Derived from Historical Aerial Photography. MSc thesis, University of Mississippi.

Johnson, S. J. (2009). An evaluation of Land Change Modeler for ARCGIS for the ecological analysis of landscape composition. PhD dissertation, Southern Illinois University.

Joint Federal State Land Use Planning Commission for Alaska. (1973). Major ecosystems of Alaska Anchorage Joint Federal State Land Use Planning Comm for Alaska map scale 1 2,500,000 incl text.

Kadmon, R. & Harari-Kremer, R. (1999). Studying long-term vegetation dynamics using digital processing of historical aerial photographs. Remote Sensing of Environment, 68, 164-176.

Kaihura, F., & Stoking, M. (2003). Agricultural biodiversity in smallholder farmers of East Africa. United Nations University Press.

117

Kakembo, V. (2001). Trends in vegetation degradation in relation to land tenure, rainfall and population changes in Peddie District, Eastern Cape, South Africa. Environmental Management, 28, 39-46.

Kaphengst, T. (2014). Towards a definition of global sustainable land use? A discussion on theory, concepts and implications for governance. GLOBALANDS Discussion Paper AP 3.1, (January).

Karl, T. R., Gleason, B. E., Menne, M. J., McMahon, J. R., Heim, J. R. R., Brewer, M. J., Kunkel, K. E., Arndt, D. S., Privette, J. L., Bates, J. J., Groisman, P. Y., & Easterling, D. R. (2012). U.S. temperature and drought. Anomalies of spring and summer 2011-12 and trends. EOS Transactions, 473.

Kim, T. H., & White, H. (2004). On more robust estimation of skewness and kurtosis. Financ Res Lett, (1), 56-74.

Kotze, I., Beukes, H., Van den Berge, E., & Newby, T. (2010). The impacts of invading alien plants on surface water resources in South Africa: A preliminary assessment. Water SA, 26, 397-408.

Kratky, V. (1989). Rigorous photogrammetric processing of SPOT images at CCM Canada. ISPRS Journal of Photogrammetry and Remote Sensing, 44(2), 53-71.

Kushwaha, S. P. S., Kuntz, S., & Oesten, G. (1994). “Application of Image Texture in Forest Classification.” International Journal of Remote Sensing, 15(11), 2273-2284.

Lambin, E. F., Geist, H. J., & Lepers, E. (2003). Dynamics of land use and land cover change in tropical regions. Annual Rev. Environ. Resources, 28, 206-241.

Lambin, E. F., Geist, H., & Rindfuss, R. R. (2006). Introduction: local processes with global impacts. Pages 1–8 in E. F. Lambin and H. Geist, editors. Land-use and land- cover change: local processes and global impacts. Springer, Berlin, Germany.

Lambin, E. F., Meyfroidt, P. (2011). Global land use change, economic globalization, and the looming land scarcity. Proceedings of the National Academy of Sciences of the United States of America, 108(2011), 3465–3472.

118

Lawler, J. J., Lewis, D. J., Nelson, E., Plantinga, A. J, Polasky, S., Withey, J. C & Radeloff, V. C (2014). Projected land-use change impacts on ecosystem services in the United States. Current Issue, 111(20), 7492-7497.

Leadley, P., Pereira, H. M., Alkemade, R., Fernandez-Manjarres, J. F., Proenca, V., Scharlemann, J. P. W., & Walpole, M. J. (2010) Biodiversity Scenarios: Projections of 21st century change in biodiversity and associated ecosystem services. Secretariat of the Convention on Biological Diversity, Montreal. Technical Series No. 50, 132 pages.

Li, W., Zhang, C., Burt, J. E., Zhu, A. X., & Feyen, J. (2004). “Two-Dimensional Markov Chain Simulation of Soil Type Spatial Distribution”. Soil Science Society of America Journal, 68(5), 1479-1490.

Liew, L. H., & Lee, Y. (2013). Aerial Images Rectification Using Non-parametric Approach, Journal of Convergence, 4(2), 15–22.

Lillesand, T. M., & Kiefer, R.W. (2000). Remote sensing and image interpretation 4th Edition. John Wiley and Sons Inc. New York.

Lohani, B. N., Degoma, S. E., Polprasert, C. (2007). Integrated Planning for Rural Water Supply and Sanitation. doi:10.1111/j.1752-1688.1980.tb03890.x.

Loveland, T. R., Zhu, Z., Ohlem, D. O., Brown, J. F. (1999). An analysis of the IGBP global land-cover characterisation process. Photogrammetric Engineering and Remote Sensing, 65(9), 1021-1032.

Magaba, S. (2012). An Assessment of land use and land cover change due to land degradation in the Alice and surrounding areas, Eastern Cape Province. MSc thesis, Nelson Mandela Metropolitan University, Department of Geoscience.

Mahmood, R., Loveland, T., Patel-Weynand, T., Karstensen, K., Beckendorf, K., Bliss, N., & Carleton, A. (2012). National Climate Assessment Technical Report on the Impacts of Climate and Land Use and Land Cover Change. USGS. Science for the changing world.

119

Maina, J., de Moel, H., Zinke, J., Madin, J., McClanahan, T., & Vermaat, J. E. (2013). Human deforestation outweighs future climate change impacts of sedimentation on coral reefs. Nature Communications, 4(05), 1986. doi:10.1038/ncomms2986.

Maitima, J., Reid, R. S., Gachimbi, L. N., Majule, A., Lyaruu, H., Pomeroy, D., Mugatha, S., Mathai, S., & Mugisha, S. (2004). The linkages between land-use change, land degradation and biodiversity across East Africa. LUCID Working Paper Series 42- 63. Available: www.lucideastafrica.org [Accessed 15 March 2015].

Malik, F., & Baharudin, B. (2013). Analysis of distance metrics in content-based image retrieval using statistical quantized histogram texture features in the DCT domain. Journal of King Saud University-Computer and Information Sciences, 25(2), 207–218. doi:10.1016/j.jksuci.2012.11.004

Mandal, U. K. (2014). Geo-information Based Spatio-temporal Modeling of Urban Land Use and Land Cover Change in Butwal Municipality, Nepal. ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. XL-8(December), 809–815. doi:10.5194/isprsarchives-XL-8-809-2014.

Maras, S. S., Maras, H. H., Aktug, B., Maras, E. E., & Yildiz, F. (2010). Topological error correction of GIS vector data. International Journal of the Physical Sciences, 5(05), 476–483.

Mateo-sagasta, J. (2008). Solow background thematic report- TR08. Agriculture and water quality interactions: A global overview. Javier Mateo-Sagasta.

Maze, K., & Joubert, L. (2014). Restoring water catchment areas leaves us flush: Spending on catchment areas provides solutions to the problems of water supply and demand. SANBI, Pretoria.

McAffrey, T. M., & Frankin, S. F. (1993). Automated Training Site Selection for Large-Area Remote Sensing Image Analysis. Computer and Geoscience, 10, 1413-1428.

McCarthy, T. (2007). Compact Airborne Image Mapping System (CAIMS). The International Archives. 19–22.

120

McWilliams, R. N., Teeuw, M. W. & P. Z. (2005). Field Techniques Manual: GIS, GPS and Remote Sensing. Field Techniques Manual: GIS, GPS and Remote Sensing, 68– 76.

Meiyappan, P., & Jain, A. K. (2012). Historical land-cover changes and land-use conversions global data. Frontiers of Earth Science, 6.

Memarian, H., Balasundram, S. K., Talib, J., Bin, Teh, C., Sung, B., Sood, A. M., & Abbaspour, K. (2012). Validation of CA-Markov for Simulation of Land Use and Cover Change in the Langat Basin, Malaysia, 2012(December), 542–554.

Mendoza-González, G., Martínez, M., Lithgow, D., Pérez-Maqueo, O., & Simonin, P. (2012). Land use change and its effects on the value of ecosystem services along the coast of the Gulf of Mexico. Ecological Economics, 82, 23–32.

Metzger, M. J., Rounsevell, M. D. A., Acosta-Michlik, L., Leemans, R., & Schroter, D. (2006). The vulnerability of ecosystem services to land use change. Agriculture Ecosystems and Environment, 114(2006), 69–85.

Meyer, W. B. (1995). Past and Present Land-use and Land-cover in the U.S.A. Consequences. P.24-33.

Midgley, G. F., Wakeling, J. L. (2012). Exploring the significance of land cover change in South Africa. S Africa J Sci. 2012, i108 (5/6), Art# 1247, 3 pages.

Millennium Ecosystem Assessment. (2005). Ecosystems and Human Well-being: Desertification Synthesis: World Resource Institute, Washington, DC.

Miller, C., & Cotter, J. (2013). An impending storm. Impacts of deforestation on weather patterns and agriculture. Greenpeace Research Laboratories: Technical Report (Review) 04-2013. Greenpeace International, Ottho Heldringstraat, Amsterdan, The Netherlands.

Mirkatouli, J., Hosseini, A., & Neshat, A. (2015). Analysis of land use and land cover spatial pattern based on Markov chains modelling. City, Territory and Architecture, 2(1). doi:10.1186/s40410-015-0023-8

Mojtaba, Z., Amir, S., & Arash, Z. G. (2012). Assessing land-cover/land-use change and its impacts on surface water quality in the Ziarat Catchment, Golestan Province-

121

Iran. MSc thesis, Gorgan University of Agricultural Sciences and Natural Resources, Department of Environment.

Mooney, H. A, Cropper, A., Leemans, R., Arico, S., Bridgewater, P., Peterson, G., & Peter, A. W. (2005). Ecosystems Wetlands. doi:10.1007/BF02987493.

Moran, E. F. (2001). Progress in the Last Ten Years in the Study of Land Use/Cover Change and the Outlook for the Next Decade. Human Dimensions of Global Change. Cambridge, Mass. MIT Press.

Mukhopadhyay, A., Mondal, A., Mukherjee, S., & Khatua, D. (2014). Forest cover change prediction using hybrid methodology of Geoinformatics and Markov chain model : A case study on sub-Himalayan town Gangtok, India. 6, 1349–1360.

Mukhopadhyay, A., Mukherjee, S., Garg, R. D., & Ghosh, T. (2013). Spatio-temporal analysis of land use - land cover changes in Delhi using remote sensing and GIS techniques. Computational Ecology and Software, 4(1), 213–223.

Muzein, B. (2010). Remote Sensing & GIS for Land Cover, Land Use Change Detection and Analysis in the Semi-natural Ecosystems and Agriculture Landscapes of the Central. PhD thesis, Technische University Dresden, Fakultät Forst-Geo-unde Hydrowissenschaften, Institut für Photogrammetrie und Fernerkundung, Germany.

Mylonas, S., Stavrakoudis, D., Theocharis, J., & Mastorocostas, P. (2015). A Region-Based Genesis Segmentation Algorithm for the Classification of Remotely Sensed Images. Remote Sensing, 7(3), 2474–2508.

Ndlovu, N. B. (2013). Quantifying indigenous forest change in Dukuduku from 1960 to 2008 using GIS and remote sensing techniques to support sustainable forest management planning. MSc thesis, Stellenbosch University.

Norris, J. R. (1996). Cambridge series in statistical and probabilistic mathematics. Markov cahins. University of Cambridge, United Kingdom.

O’Farrell, P. J., Reyers, B., Cowling, R. M., Egoh, B. N., Le Maitre, David C., & Vlok, J. H. J. (2009). Ecosystem Services, Land-Cover Change, and Stakeholders. Finding a

122

Sustainable Foothold for a Semiarid Biodiversity Hotspot. Ecology and Society, 14(1), 38.

O’Keeffe, J. H. (1998). Ecological research on South African rivers- a preliminary synthesis. South African National Scientific Programmes Report No. 121.

Obata, O. O., (2010). Effects of land use on water quantity and quality: A case study of Benin City, Nigeria. University of Abertay: School of contemporary Sciences; Environmental biotechnology.

Olson, J. M Misana, S., Campbell, D. J., Mbonile, M., & Mugisha, S. (2004). A Research Framework to identify the root cause of land use change leading to land use change impacts and dynamics (LUCID), project working Paper No. 48. International Livestock research Institute, Nairobi, Kenya.

Ometto, J. P., Sampaio, G., Marengo, J., Assis, T., & Aguiar, A. P. (2013). Climate Change and Land Use Change in Amazonia.

Openshaw, S., & Openshaw, C. (1997). Artificial intelligence in Geography. Chichester: John Wiley & Sons.

Organisation for Economic Co-operation and Development (OECD). (2008). Land Use Restrictions as Barriers to Entry. Policy Roundtables.

Otari, G. V. (2012). A Review of Application of Data Mining in Earthquake Prediction, 3(2), 3570–3574.

Palmer, M. A., Moglen, G. E., Bockstael, N. E., Brooks, S., Pizzuto, V. E., Wiegad, C., & VanNess, K. (2010). The ecological consequences of changing land use for running waters, with a case study of urbanizing watersheds in Maryland.

Paul, F., Barrand, N. E., Baumann, S., Berthier, E., Bolch, T., Casey, K., & Winsvold, S. (2013). On the accuracy of glacier outlines derived from remote-sensing data. Annals of Glaciology, 54(63), 171–182.

Peacock, R. (2014). Accuracy assessment of supervised and unsupervised classification using LandSat imagery of little rock, Arkansas. MSc thesis, Northwest Missouri State University, Maryville Missouri.

123

Perumal, K., & Bhaskaran, R. (2010). Supervised Classification Performance of Multispectral Images. Journal of computing, 2(2), 124.

Petropoulos, G. P., Kalaitzidis, C., & Prasad Vadrevu, K. (2012). Support vector machines and object-based classification for obtaining land-use/cover cartography from Hyperion hyperspectral imagery. Computers & Geosciences, 41, 99-107.

Pezanowski, S., Tomaszewski, B., MacEachren, A., Li, J., Zlatanova, S., & Fabbri, A. G. (2007). Geomatics Solutions for Disaster Management. doi:10.1007/978-3-540- 72108-6.

Pielke, R. (2005). Land use and Climate Change. Science, 310(5754), 1625-1626.

Pielke, R. A. & R. L. Wilby. (2012). Regional climate downscaling – what’s the point? EOS: Climate Science.

Plourde, L., & Congalton, R. G (2003,). Sampling method and sample placement: How do they affect the accuracy of remotely sensed maps? Photogramm. Eng. Remote Sens, 69, 289–297.

Polasky, S., Nelson, E., Pennington, D., & Johnson, K. A. (2010). The Impact of Land-Use Change on Ecosystem Services, Biodiversity and Returns to Landowners: A Case Study in the State of Minnesota. Environ Resource Econ, 2011(48), 219–242.

Pollack, H. N. (2004). Global Change and the Earth System. Eos. Transactions American Geophysical Union, 85. doi:10.1029/2004EO350005.

Pongratz, J., Reick, C. H., Houghton, R. A., & House, J. I. (2014). Terminology as a key uncertainty in net land use and land cover change carbon flux estimates. Earth System Dynamics, 5(1), 177–195.

Pontius, R., & Chen, H. (2006). Land change modeling with GEOMOD. Idrisi Kilimanjaro, 2006.

Pontius, R.G. (2000). Quantifying error versus location error in comparison of categorical maps. Photogrammetric Engineering and Remote Sensing, 66(8), 1011-1016.

124

Power, A. G. (2010). Ecosystem services and agriculture: Trade-offs and synergies. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 365(1554), 2959–2971. doi:10.1098/rstb.2010.0143

Press, W. H., Flannery, B. P., Teukolsky, S. A, & Vetterline, W. T. (1990). Moments of a Distribution : Mean Variance, Skewness, and So Forth. Numerical Recipes: The Art of Scientific Computing, 1, 610–615.

Proenc, V. M., Pereira, H. M., & De Lisboa, U. (2011). Ecosystem Changes, Biodiversity Loss and Human Well-Being.

Qihe, H. Y., Snyder, J. P., & Tobler, W. R. (2000). Map projection transformation principles and applications, Tylor and Francis.

Ramachandra, T. V, & Kumar, U. (2004). Geographic Resources Decision Support System for land use, land cover dynamics analysis, 12–14. Centre for Ecological Sciences + Centre for Sustainable Technologies Energy & Wetlands Research Group, Centre for Ecological Sciences, Indian Institute of Science, Bangalore – 560 012.

Raumann, C. G. & Soulard, C. E. (2007). Land cover trends in the Sierra Nevada ecoregion, 1973-2000: U.S. Geological Survey Scientific Investigations Report 2007-5011, 29 p.

Rebelo, A. J., Esler, P. K., & Cowling, P. R. (2012). An Ecological and Hydrological Evaluation of the Effects of Restoration on Ecosystem Services in the Kromme River System, South Africa.

Reed, C. (2014). Fresh water impacts assessment for the construction of proposed river bank rehabilitation structures in the Kruis River, Riversdale.

Reyers, B., O’Farrell, P. J., Cowling, R. M., Egoh, B. N., Le Maitre, D. C., & Vlok, Jan H. J. (2009). Ecosystem Services, Land-Cover Change, and Stakeholders. Finding a Sustainable Foothold for a Semiarid Biodiversity Hotspot. Ecology and Society 14(1), 38.

Richards, J. A. (1993). Remote sensing digital image analysis: an introduction (second edition). Springer-Verlag Berlin Heidelberg

Richardson, M. (2009). Principal Component Analysis.

125

Rindfuss, R. R., S. J. Walsh, B. L. Turner, J. Fox, & V. Mishra. (2004). Developing a science of land change: challenges and methodological issues. Proceedings of the National Academy of Sciences of the United States of America, 101, 13976– 13981.

River Health Programme. (2006). State-of-Rivers Report, the Mokolo River System. Department of Environmental Affairs and Tourism, Pretoria.

Robert, M. H., Shanmugam, K., & Dinstein, I. (1973). Textural Features for Image Classification. IEEE, SMC, 3(6), 610-621.

Rodriguez, E., Sultan, R., & Hilliker, A. (2004). Negative Effects of Agriculture on Our Environment. The Traprock, 3, 28 – 32.

Rodrriguez-Galiano, F. V., & Chica-Rivas, M. (2014). Evaluation of different machine learning methods for land cover mapping of a Mediterranean are using multi- seasonal Landsat images and Digital Terrain Models. International journal of Digital Earth, 7(6), 492-509.

Ross, I., Misson, L., Rambal, S., Arneth, A., Scott, R. L., Carrara, A., Genesio, L. (2012). How do variations in the temporal distribution of rainfall events affect ecosystem fluxes in seasonally water-limited Northern Hemisphere shrublands and forests? Biogeosciences, 9(3), 1007–1024.

SAEON. (2011). Observations on environmental change in South Africa: Section 2. SUN MeDIA Stellenbosch, SUN PRESS imprint.

Salimon, C. I., Guilherme, P., Wadt, S., & Alves, S. D. S. (2009). Decrease in carbon stocks in an oxisol due to land use and cover change in southwestern Amazon Diminuição nos estoques de carbono em um latossolo devido à mudança de cobertura e uso da terra no sudoeste da Amazônia. Revista Ambiente & Agua-An Interdisciplinary. Journal of Applied Science, 4(2), 57–65. doi:10.4136/1980- 993X.

SANBI (South African National Biodiversity Institute). (1999). Updating National Land Cover. Pretoria.

SANBI. (2014). A Framework for Investing in Ecological Infrastructure in South Africa.

126

Scanlon, B. R., Jolly, I., Sophocleous, M., & Zhang, L. (2007). Global impacts of conversions from natural to agricultural ecosystems on water resources. Quantity versus quality, Water Resource. Res, 43, W03437.

Schaffartzik, A., Haberl, H., Kastner, T., Wiedenhofer, D., Eisenmenger, N., & Erb, K-H. (2015). Trading Land: A Review of Approaches to Accounting for Upstream Land Requirements of Traded Products. Journal of Industrial Ecology.

Scheffer, M., Carpenter, S., Foley, J. A, Folke, C., & Walker, B. (2001). Catastrophic shifts in ecosystems. Nature, 413(6856), 591–596.

Schutte, S. (2012). Linkages between selected hydrological system services and land use change. A case study on the Mpushini/Mkhondeni sub-catchments of the Umngeni catchment, South Africa. MSc Proposal, School of Agricultural, Earth and Environmental Sciences, University of KwaZulu-Natal.

Sulle, E. (2014). MAPS, Provocateur Briefing Report: Forum on Development and Mitigation. Institute for Poverty, Land and Agrarian Studies, University of Western Cape.

Serra, P., Pons, X., & Saurí, D. (2003). Post-classification change detection with data from different sensors: some accuracy considerations. International Journal of Remote Sensing, 24(23), 4975–4976.

Seymour, F., & Busch, J. (2014). Why Forests? Why now?

Shiklomanov, I. A. (2000). Appraisal and assessment of world water resources. Water Int, 25, 11 –32.

Siljander, M. (2010). Geospatial environmental data modelling applications using remote sensing, GIS and spatial statistics.

Sims, D. (1995). Background note on ongoing activities relating to land cover and land use

Singh, A. (1989). Review Article Digital change detection techniques using remotely-sensed data. International journal of remote sensing, 10(6), 989-1003.

127

Singh, R. B. & Shi, C. (2014). “Advances in observation and estimation of land use impacts on climate changes: Improved data, upgraded models, and case studies”. Advances in Meteorology, 2014, Article ID 748169, 7 pages.

Solomon, C. (2009). The Need to Consider Human Forcings besides Greenhouse Gases. Climate Change. Department of Terrestrial Magnetism, Carnegie Institution of Washington. Eos, 90(45), 10- 2009.

State of rivers report. (2007). Rivers of the Gouritz water management area. Conservation Planning for River and Estuarine Biodiversity in the Fish to Tsitsikamma Water Management a Rea. Water, (June).

Steffen, W. A. Sanderson, P. D., Tyson, J., Jager, P. M., Matson, B., Moore, F., Oldfield, K., Richardson, H. J., Schnellnhuber, B. L., Turner, II, & R. J. Wasson. (2004). Global change and the Earth system: a planet under pressure. Springer-Verlag, New York, New York, USA. 336 pp., hardcover, ISBN 3540408002.

Steffen, W., Rockstrom, J., Cornell, S., Fertzer, I., Biggs, O., Folke, C., & Reyers, B. (2015). Planetary boundary: Guiding human development on changing planet. Science, doi:10.1126/science.1259855.

Stipinovich, A. (2005). Change in the land cover and water abstraction. Modelling runoff effects in the Bot river catchment. MSc thesis, University of Stellenbosch.

Stone, B. (2009). Land use as climate mitigation: Environmental Science and Technology. Georgia Institute of Technology, Atlanta. Georgia. Environ. Sci. Technol, 43, 9052-9056.

Strahler, A. H., Boschetti, L., Foody, G. M., Friedl, M. A., Hansen, M. C., Herold, M., & Woodcock, C. E. (2006). Global Land Cover Validation : Recommendations for Evaluation and Accuracy Assessment of Global Land Cover Maps. Luxemburg: Office for Official Publications of the European Communities.

Swain, P. H., & Davis, S. M. (1978). Remote Sensing. The Quantitative Approach, McGraw- Hill, New York.

Tadesse, W., Coleman, T. L., & Tsegaye, T. D. (2003). Improvement of land use and land cover classification of an urban area using image segmentation from Landsat

128

ETM+ data. 30th International Symposium Remote Sensing of the Environment, 3–6.

Teo, Y. M., Low, S. C., Tay, S. C., & Gozali, J. P. (2003). Distributed geo-rectification of satellite images using Grid computing. Proceedings International Parallel and Distributed Processing Symposium, 2003, 1–13.

The Environmental Literacy Council. (2013). Land Use Changes & Climate.

Townshend, J. R. G. (1992). Improved global data for land applications. IGBP Report No. 20. IGBP Secretariat/Royal Swedish Academy of Sciences, Stockholm.

Trail, M., A. P. Tsimpidi, P. Liu, K. Tsigaridis, Y. Hu, A. Nenes, B. Stone, & A. G. Russell. (2013). Potential impact of land use change on future regional climate in the South-eastern U.S.: Reforestation and crop land conversion. J. Geophys. Res. Atmos, 118, (20), 11577-11588.

Tsegaye, L. (2014). Analysis of Land Use and Land Cover Change and Its Drivers Using GIS and Remote Sensing: The Case of West Guna Mountain , Ethiopia, 3(3), 53–63.

Turner, B. L., Skole, D., Sanderson, S., Fischer, G., Fresco, L., & Leemans, R. (1995). Land- Use and Land-cover change Science/Research Plan. Joint publication of the International Geosphere Biosphere Programme (Report No. 35) and the Human Dimensions of Global Environmental Change Programme (Report No. 7). Royal Swedish Academy of Sciences, Stockholm, Sweden.

Turner, M. G., Romme, W. H., Gardner, R. H., & Hagrove, W. W. (1997). Effects of patch size and fire pattern on early post-fire succession on the Yellowstone Plateau. Ecological Monographs, 67, 411-433.

Turner, R. K., Fisher, B., & Morling, P. (2008). Defining and classifying ecosystem services for decision making. Ecological Economics, 68(3), 15- 2009.

Tuttle, E. M., Jensen, R. R., Formica, V. A., & Gonser, R. A. (2006). Using remote sensing image texture to study habitat use patterns: A case study using the polymorphic white-throated sparrow (Zonotrichia albicollis). Global Ecology and Biogeography, 15(4), 349–357. doi:10.1111/j.1466-822X.2006.00232.x

129

Tyler, M. E., Quinn, M., Barchard, E., Cecchetto, J., Humber, L., Selivig, C., & Woodhouse, N. (2008). Ecological infrastructure in the Calgary Region. “What we know now”. Final Report- January 2008: The Calgary Regional partnership. University of Calgary. Available: http://www.rockies.ca/files/reports/Ecological%20Infrastructure%20in%20the%2 0Calgary%20Region%20-%20What%20we%20know%20now.pdf [Accessed 11 June 2015].

U.S Natural Resources Conservation Service (NRCS). (2006). National Conservation Practice Standard: Riparian Forest Buffer. Code 391. Available: www.nrcs.usda.gov/Internet/FSE_DOCUMENTS/nrcs143_026098.pdf [Accessed 07 June 2015].

UNEP. (2012). The Fifth Global Environmental Outlook Report. Chapter 3: Land. Managing Increasing Pressures on Land; based on GEO-5 Land chapter 1. Available: www.unccd.int/Lists/SiteDocumentLibrary/.../GEO5%20land.pdf [Accessed 25 July 2015].

UNEP/FAO. (2005). Land cover Classification System (LCCS): Classification concepts and user manual. Software version 2. Environmental and Natural resources services series, No.8-FAO, Rome, 2005. Available: www.glcn.org/downs/pub/docs/manuals/lccs/LCCS2-manual_en.pdf [Accessed 23August 2015].

USGS. (2014). How does land-use change affect climate change? Available: http://www.usgs.gov/faq/node/56021 [Accessed 17 September 2015].

Van Den Hurk, B., Seneviratne, S. I & Batlle-Bayer L. (2012). Land cover change - To what degree do human land cover dynamics affect climate change? Earth System Dynamics Discussions, 3(1), 201–234. doi:10.5194/esdd-3-201-2012.

Vasuki, P. (2012). Man-made Object Feature Extraction in SAR Images using Gabor Wavelet. IJCST, 3 (8491), 111–113.

Verburg, P. H., Erb, K. H., Mertz, O., & Espindola, G. (2013). Land System Science: Between global challenges and local realities. Current Opinion in Environmental Sustainability, 5(5), 433–437.

130

Vibhute, A. D., Nagne, A. D., Gawali, B. W., & Mehrotra, S. C. (2013). “Comparative Analysis of Different Supervised Classification Techniques for Spatial Land Use / Land Cover Pattern Mapping Using RS and GIS” International Journal of Scientific & Engineering Research, 4(7), 2013- 1938.

Vitousek, P., Mooney, H., & Lubchenco, J. (1997). Human domination on earth’s ecosystems: Science, 277, 494-499.

Vorosmarty, C. J., Green, P., Salisbury, J., & Lammers, R. B. (2000), Global water resources: Vulnerability from climate change and population growth. Science, 289, 284– 288.

Vorosmarty, C. J., McIntyre, P. B., Gessner, M. O., Dudgeon, D. (2010). Global threats to human water security and river biodiversity. Nature, 467, 555–561.

Wang, X., Zheng, D., & Shen, Y. (2008). Land use change and its driving forces on the Tibetan Plateau during 1990-2000. Catena, 72(1), 56–66.

Western Cape Environmental Affairs and Development Planning/Department of Agriculture. (2015). A Status Quo Review of Climate Change and the Agriculture Sector of the Western Cape Province Chapter 7 Overview of the natural resource base for public comment. Available: http://acdi.uct.ac.za/sites/default/files/filesattached/SmartAgri%20Review%20Ch 7%20Natural%20resource%20base.pdf [Accessed 13 September 2015].

Western Cape IWRM Action Plan. (2011). Executive Summary: Status Quo Report. Available: www.docsunrise.com/d/Western-Cape-Integrated-Water- Resources.pdf [Accessed 28 August 2015]

Wetzstein, G., Ihrke, I., Lanman, D., & Heidrich, W. (2011). Computational Plenoptic Imaging. Computer Graphics forum, 30(8), 2397–2426.

World Energy Outlook. (2005). Middle East North Africa Insights Middle East North Africa. Middle East, 634. Available: http://www.iea.org/weo/docs/weo2005/WEO2005.pdf [Accessed 18 April 2015].

World Health Organisation. (2003). Guidelines for safe recreational water environments. Volume 1, Coastal and Fresh waters.

131

World Meteorological Organisation. (2013). Planning of Water Quality Monitoring Systems. Available: https://drive.google.com/file/d/0BwdvoC9AeWjUSFB1eWxJTXBaWFU/view?u sp=sharing [Accessed 23 August 2015].

Wu, J., & Bauer, M. E. (2013). Evaluating the effects of shadow detection on quickbird image classification and spectroradiometric restoration. Remote Sensing, 5(9), 4450–4469.

Xi, L., Zequn, G., & Tiyan, S. (2008). Extract profile of urbanized area in Beijing assisted by skewness. Archives, 124, 463–466.

Xu, L. X. L., Zhang, S. Z. S., He, Z. H. Z., & Guo, Y. G. Y. (2009). The comparative study of three methods of remote sensing image change detection. 2009 17th International Conference on Geoinformatics, 1595–1598.

Yang X., Zheng X-Q. (2012). A spatiotemporal model of land use change based on ant colony optimization, Markov chain and cellular automata. Ecol Model 233:11– 19. doi:10.1016/j.ecolmodel.2012. 03.011

Zhang, T., Zhang, X., Xia, D., & Liu, Y. (2014). An Analysis of Land Use Change Dynamics and Its Impacts on Hydrological Processes in the Jialing River Basin. Water, 3758–3782. doi:10.3390/w6123758.

Zhou, W., Huang, G., Troy, A., & Cadenasso, M. L. (2009). Object-based land cover classification of shaded areas in high spatial resolution imagery of urban areas: A comparison study. Remote Sensing of Environment, 113(8), 1769–1777.

Zhuang, X. B. A. Engel, M. F. Baumgardner, & P. H. Swain. (1991). Improving classification of crop residues using digital land ownership data and Landsat TM imagery. Photogrammetric Engineering & Remote Sensing, 57(11), 1487-1492.

132

APPENDICES

Appendix A. Missing data as represented by DRID-reference

GRID 3221CD 3421AA 3421AB 3421BA 3421AC 3421AD 3421BC 3421BD REF

YEAR/JOB NUMBER

1940_08 NOT x x √ √ x √ √ √ (partial) COVER

8_1942 NOT X √ X √ √ X √ √ (partial) COVER

171_1942 NOT X X X √ X X √ √ COVER

344_1954 NOT √ √ √ √ √ √ x x COVER

492_1964 NOT X X X √ X X √ √ COVER

564_1967 NOT √ √ √ √ √ √ √ X COVER

647_1969 NOT X X X X X √ √ √ COVER

735_1974 COVER √ √ √ √ √ √ √ √

180_1983 NOT x √ x x √ x x x COVER

211_1984 NOT x x √ √ x √ √ √ COVER

889_1985 COVER √ √ √ √ √ √ √ √

959_1991 COVER √ √ √ √ √ √ √ √

1026_1999 COVER √ √ √ √ √ √ √ √

2006 COVER √ √ √ √ √ √ √ √

2010 COVER √ √ √ √ √ √ √ √

133

: A tick (√) represents a present grid reference in each year and a cross (x) represent a missing grid reference in each year

: Grid 3421AA from Job8_1942, 3421BD from Job 171_1942 and 3221CD from Job 344_1954; cover the missing grids of 1940.

: Grid 3421BD from Job 647_1969 was used to cover the missing grid of 1967.

Appendix B. Accuracy assessment Errors and Accuracies, and formulas Omision = [(incorrectly classified pixels in the column)/ total number of the column] x 100

Comision = [(incorrectly classified pixels in the row)/ total number of pixels in the row] x 100

Producer’s Accuracy = [(correctly classified pixels)/ reference total points] x 100

User’s Accuracy = [(correctly classified pixels)/ total points in a class)] x 100

LU N Cp Cf W Ds P Ro Bs R D A RT

N 83 11 6 15 70 3 19 53 8 2 20 290

Cp 1 68 0 0 0 0 0 0 0 0 0 69

Cf 0 0 82 0 5 0 4 1 0 0 0 92

W 0 0 0 57 0 0 0 0 5 0 6 68

Ds 8 12 8 21 17 3 34 13 11 10 16 153

P 0 0 0 0 0 94 0 0 0 0 2 96

Ro 0 0 0 0 0 0 40 0 0 0 0 40

Bs 8 9 4 1 8 0 3 33 0 1 1 68

134

LU N Cp Cf W Ds P Ro Bs R D A RT

R 0 0 0 0 0 0 0 0 51 0 7 58

D 0 0 0 0 0 0 0 0 0 87 0 87

A 0 0 0 6 0 0 0 0 25 0 48 79

CT 100 100 100 100 100 100 100 100 100 100 100 1100

K 0.56

OA 0.60

PA 83 68 82 57 17 94 40 33 51 87 48

UA 29 99 89 84 11 98 100 49 88 100 61

CE 71 1 11 16 89 2 0 51 12 0 39

OE 17 32 18 43 83 6 60 67 49 13 52

LU N Cf W Ds P Ro Bs R D A Cp RT

N 59 3 15 67 5 3 26 1 3 21 18 221

Cf 6 87 1 6 0 2 0 0 1 1 0 104

W 1 0 25 0 0 0 0 2 0 4 0 32

Ds 23 9 28 19 8 12 20 19 6 30 8 182

P 0 0 0 0 83 0 0 0 0 0 10 93

135

LU N Cf W Ds P Ro Bs R D A Cp RT

Ro 0 0 0 0 0 81 0 0 0 0 0 81

Bs 7 1 0 6 0 2 53 0 2 0 5 76

R 0 0 0 0 0 0 1 75 0 0 0 76

D 0 0 0 0 0 0 0 0 88 0 0 88

A 4 0 31 2 4 0 0 3 0 44 0 88

Cp 0 0 0 0 0 0 0 0 0 0 59 59

CT 100 100 100 100 100 100 100 100 100 100 100 1100

K 0.57

OA 0.61

PA 59 87 25 19 83 81 53 75 88 44 59

UA 27 84 78 10 89 100 70 99 100 50 100

CE 73 16 22 90 11 0 30 1 0 50 0

OE 41 13 75 81 17 19 47 25 12 56 41

Appendix E. Land cover trends from 1940 to 2010 in (ha)

Land 40-54 54-67 67-74 74-85 85-91 91-06 06-10 40-10 Use

Nv 79390 95634 71455 85933 47839 67594 52539 85890

Cf 18527 22383 29183 32304 23901 39764 27400 19751

W 1826 1760 1312 1346 1249 1301 929 1945

136

Land 40-54 54-67 67-74 74-85 85-91 91-06 06-10 40-10 Use

Ds 30144 14652 23621 14780 36103 26245 32831 30593

P 175 725 818 973 677 1412 1455 216

Ro 299 318 401 419 881 582 280 318

Bs 19243 13871 10278 10730 20784 10920 2771 20202

R 148 138 169 143 182 114 152 156

D 40 55 107 186 146 207 338 39

A 952 265 1726 659 2738 3175 2079 930

Cp 73 0 -10 64 52 141 418 59

Average 13711 13618 12642 13412 12232 13769 11017 14555

Appendix F. Landscape changes over years (1: wetlands converted to agriculture, 2: natural hazards and 3: aliens species)

137

Appendix G. Contributors to land cover change of the Goukou catchment from 1940 to 2010

138

1985-1991

139

2006-2010

Appendix H. Overlay maps of 1940 to 2010

140

Appendix I. Percentage of variance explained by Eigen values of Principal Component Analysis

Sum Basic Stats Eigen Values Percentages (%)

1954

30987.38604 0.835550683 83.5551

3369.275168 0.090849876 9.08499

1426.843509 0.038473722 3.84737

686.144317 0.018501346 1.85013 98.3376

361.852656 0.009757074 0.97571

112.552134 0.00303488 0.30349

62.0427 0.001672933 0.16729

32.342583 0.000872093 0.08721

18.751179 0.000505611 0.05056

8.842226 0.000238424 0.02384

6.874738 0.000185372 0.01854

3.838304 0.000103497 0.01035

2.69932 7.27851E-05 0.00728

1.793276 4.83543E-05 0.00484

1.466579 3.95452E-05 0.00395

1.436134 3.87242E-05 0.00387

141

Sum Basic Stats Eigen Values Percentages (%)

0.8289 2.23506E-05 0.00224

0.55648 1.5005E-05 0.0015

0.301987 8.14284E-06 0.00081

0.21271 5.73556E-06 0.00057

0.142685 3.84739E-06 0.00038 sum 37086.18363 1 100

1967

42746.7778 0.832152345 83.2152

4953.836965 0.096436439 9.64364

1867.882069 0.036362096 3.63621

892.37903 0.01737196 1.7372 98.2323

488.477938 0.009509209 0.95092

192.132402 0.003740245 0.37402

86.497194 0.001683843 0.16838

60.477035 0.001177308 0.11773

28.487973 0.000554576 0.05546

14.481167 0.000281905 0.02819

12.134294 0.000236219 0.02362

7.581661 0.000147592 0.01476

142

Sum Basic Stats Eigen Values Percentages (%)

5.215061 0.000101522 0.01015

3.552013 6.91471E-05 0.00691

2.897001 5.6396E-05 0.00564

2.853746 5.55539E-05 0.00556

1.22245 2.37975E-05 0.00238

1.063254 2.06984E-05 0.00207

0.469818 9.14596E-06 0.00091

0.27053 5.26641E-06 0.00053

0.24326 4.73555E-06 0.00047 sum 51368.93266 1 100

1974

32983.48774 0.817349192 81.7349

4390.510909 0.108799305 10.8799

1671.331891 0.041416535 4.14165

704.135113 0.01744886 1.74489 98.5014

381.463496 0.009452878 0.94529

106.468705 0.002638354 0.26384

47.871451 0.001186281 0.11863

29.049169 0.000719855 0.07199

143

Sum Basic Stats Eigen Values Percentages (%)

15.326213 0.000379792 0.03798

7.511823 0.000186147 0.01861

5.933877 0.000147045 0.0147

3.175198 7.86832E-05 0.00787

2.221571 5.50518E-05 0.00551

1.643046 4.07156E-05 0.00407

1.346398 3.33645E-05 0.00334

1.020035 2.5277E-05 0.00253

0.695673 1.72392E-05 0.00172

0.477145 1.18239E-05 0.00118

0.268767 6.6602E-06 0.00067

0.161788 4.0092E-06 0.0004

0.118314 2.93189E-06 0.00029 sum 40354.21832 1 100

1985

33208.86139 0.83362872 83.3629

4034.663439 0.101280537 10.1281

1272.529794 0.031943805 3.19438

686.884413 0.017242584 1.72426 98.4096

144

Sum Basic Stats Eigen Values Percentages (%)

384.553448 0.009653291 0.96533

104.24489 0.002616818 0.26168

65.184423 0.001636298 0.16363

34.921584 0.000876623 0.08766

21.150755 0.000530939 0.05309

9.061045 0.000227456 0.02275

6.097113 0.000153053 0.01531

2.060018 5.17118E-05 0.00517

1.595911 4.00615E-05 0.00401

1.503129 3.77324E-05 0.00377

0.900939 2.26159E-05 0.00226

0.857871 2.15348E-05 0.00215

0.544139 1.36593E-05 0.00137

0.432512 1.08572E-05 0.00109

0.215359 5.40607E-06 0.00054

0.144982 3.63942E-06 0.00036

0.105928 2.65907E-06 0.00027 sum 39836.51308 1 198.41

1991

145

Sum Basic Stats Eigen Values Percentages (%)

43219.94505 0.853997487 85.3997

4303.703601 0.085038333 8.50383

1358.692592 0.026846866 2.68469

836.094607 0.016520676 1.65207 98.2403

464.980132 0.009187699 0.91877

176.815887 0.003493765 0.34938

95.468143 0.001886387 0.18864

67.016129 0.001324194 0.13242

32.775576 0.000647624 0.06476

18.977119 0.000374975 0.0375

14.885678 0.000294131 0.02941

5.469606 0.000108076 0.01081

3.684868 7.28106E-05 0.00728

3.448422 6.81385E-05 0.00681

2.834683 5.60115E-05 0.0056

1.638997 3.23855E-05 0.00324

1.254419 2.47865E-05 0.00248

0.676595 1.33691E-05 0.00134

0.361224 7.13755E-06 0.00071

0.260527 5.14784E-06 0.00051

146

Sum Basic Stats Eigen Values Percentages (%)

0 0 0 sum 50608.98385 1 100

Appendix J. Graphs for Gains and Losses from 1940 to 2010 as obtained by using change detection analysis in Markov

147