Water quality assessment of the Sabie River catchment with respect to adjacent land use

M Stolz orcid.org/0000-0003-3243-3946 Previous qualification (not compulsory)

Dissertation submitted in fulfilment of the requirements for the Masters degree in Environmental Science at the North-West University

Supervisor: Prof S Barnard Co-supervisor: Dr TL Morgenthal

Graduation May 2018 23567597

PREFACE This research study was embarked with background knowledge in geo-spatial sciences. This included undergraduate knowledge in geography and geology and postgraduate knowledge in geohydrology. It has been an intense learning process that I am very thankful for.

Acknowledgements

First and foremost,

Soli Deo Gloria,

without which we are none but slaves to work.

For the financial and logistical support during this study I would like to thank the following contributing supporters: Rand Water Analytical Services, Water Resource Commission SANParks and North-West University.

Thank you to the following people for their support and contribution to this dissertation:

The financial and immense emotional support given to me by my parents, Magda and Willie Stolz not only over the past two years, but during my entire university studies. Without you I would not have been able to come this far.

Xander Rochér for your love, support and especially your patience throughout the whole process.

My supervisor, Prof. Sandra Barnard for your guidance, time and patience. Not only with regards to science, but always making time to include personal support as well. Thank you for your assistance with data processing and data analysis, especially with regards to the water quality section. I have learned a lot over the past two years from you.

My co-supervisor, Dr. Theunis Morgenthal for your guidance, thoughts and inputs regarding the spatial aspects pertaining to this dissertation. Thank you for the thorough guidance and input in the background information of this study area pertaining to the vegetation description.

Thank you to the various NWU lecturers for initial thoughts and supplementary data regarding the spatial methodology: Dr. D. Cilliers, Mr. D. van Schalkwyk and Dr. G. Mahed.

i ABSTRACT Situated in the north-eastern region of South Africa is the Sabie-Sand River catchment. Initially seen as the separate Sabie catchment draining the Sabie River, and the Sand catchment draining the Sand River, these drainage basins are collectively classified by The Department of Water Affairs and Sanitation (DWS) as the X3 secondary drainage region. This catchment lies to the north of the greater Inkomati-Usuthu Water Management Area (IUWMA). The primary Sabie River flows from its headwaters originating east of the Great Escarpment through various towns, such as Sabie, Hazyview and Skukuza, which is situated in the Kruger National Park (KNP) to its cross- boundary end in the Corumana Dam, Mozambique. Two of its tributaries were considered important within this study (Sand and Marite rivers) as well as the Inyaka Dam (also called Injaka Dam) situated within the Marite River.

Forestry, commercial plantations, cultivated commercial lands and urban land uses were suspected of having increasing water demand on the water resources of the catchment. The densely populated, rural or semi-urban developments within the catchment have raised increasing concerns over the past two decades due to the substantial increase in population. Initially it was pressures from the increasing water supply (lifted by the Integrated Water Supply Scheme from the Inyaka Dam) that impaired the livelihoods of the still expanding Bushbuckridge community. Now, water quality has raised increasing concerns with regards to both Bushbuckridge and the urban township Thulamahashe. As responsibility towards the conservation practices of the KNP; the international water agreement between South Africa, Swaziland and Mozambique; and the protection of catchment water resources, this study was initiated with regards to the influence of land use on water quality.

It is very important to note that certain parts of South Africa had suffered a climatological drought from 2013 to 2016, so this study assessed the river water quality experienced within a dryer rainfall period and should be interpreted as such. This relates to an international issue of concern, namely dwindling potable water supply. Intense human activities conducted by using earth’s surfaces, influence water quality. It has become highly important to assess water issues with regards to land use as indicator of human activities. The main aims of this study were to do an in-depth determination of the different land uses surrounding the Sabie-Sand River Catchment; and to evaluate the influence thereof on the water quality of the Sabie River and its two major tributaries including the Inyaka Dam. It also focused much attention on the possible effects that urbanization had on water resource quality. Therefore, it was necessary to conduct in situ field investigations and GIS spatial analysis towards answering the research problem.

Every three months, surface water samples were collected form 12 sites in the Sabie-Sand catchment. This was done over a period of a year, as to include results from all four climate seasons. Major chemical, biological and physical water quality parameters were measured; ii analysed with accredited methods from the South African National Accreditation System (SANAS) by Rand Water Analytical Services laboratory, and interpreted by use of Spearman Rank correlation and Kruskal-Wallis ANOVA statistical methods. Spatial analysis conducted through ArcGIS version 10.4 (Esri, 2015), made use of the 3D analyst, spatial analyst and network analyst extensions in ArcGIS for land use interpretation. The main spatial datasets used for interpretation were the 2013/2014, 72 Class South African National Land Cover Dataset (GeoTerraimage, 2015) and the SRTM90 Digital Elevation Model (DEM) (Jarvis et al., 2008).

Assessing spatial influences on water quality, a multivariate redundancy analysis (RDA) was conducted to compare both datasets. The results indicated that there was no significant relationship between land use and water quality for the Sabie-Sand catchment, but it was identified that land use still played a major role in the water quality of the catchment. Since the state of the rivers report conducted by Roux and Selepe (2011), the land uses in terms of forestry and the upper part of the Sabie River have not changed as much as the urban expansion of peri- urban areas, urban villages and urban townships have. The reason for saying this is that forestry, commercial plantations and cultivated commercial lands showed lesser adverse influence on water quality and that the major negative influences were urban in nature. The urban township Thulamahashe had the most negative impact on the water quality of the Sand River, tributary of the Sabie River.

This was related back to effluent from the wastewater treatment plant on the Sand River. Discussed in Chapter 4 of this study are the water quality parameters that identify organic pollution from anthropogenic activities. These parameters were either high or in exceedance of drinking water standards. This is seen in exceedingly high Escherichia coli (E. coli) and Total Coliforms (coli) concentrations that were associated with high Chlorophyll-a (Chl-a), Chemical Oxygen Demand (COD), dissolved inorganic nitrogen (DIN), Dissolved Organic Carbon (DOC) and geosmin. With low concentrations of DO for this site, accompanied with the high coli, DOC, Total Organic Carbon (TOC) and COD this might indicate rapid bacterial growth.

Support should be provided to this urban settlement in the form of adequate water supply and sanitation facilities, primarily starting with the known point pollution wastewater treatment plant. Adequate maintenance and monitoring should be insisted on at the plant to ensure effluent that conform to national water quality guidelines.

Key terms: Land use, water quality, correlation analysis, redundancy analysis, Sabie-Sand River Catchment

iii TABLE OF CONTENTS

PREFACE ...... I

ABSTRACT ...... II

LIST OF TABLES ...... VII

LIST OF FIGURES ...... IX

LIST OF ABBREVIATIONS ...... XIV

TABLE OF DEFINISIONS ...... XV

CHAPTER 1 INTRODUCTION ...... 1

Background ...... 1

Aims and Objectives ...... 4

CHAPTER 2 OVERVIEW AND GENERAL STUDY AREA ...... 5

Location and Boundaries ...... 5

Topography and Climate ...... 8

Vegetation and Geology ...... 10

2.1.1 Vegetation ...... 10

2.1.2 Geology ...... 12

CHAPTER 3 LITERATURE REVIEW ...... 16

Land use effect on river quality (brief South African context) ...... 16

Importance of river health and remedial assistance given through policies (Ecological Reserve and Target Water Quality Study) ...... 19

Climate change effects on river heath in general and the South African context ...... 21

Population increases and socio-economic drivers ...... 25

GIS approaches and statistical analysis methods to assess land use influence on water quality...... 26

iv CHAPTER 4 WATER QUALITY ...... 35

Methodology ...... 35

4.1.1 Sampling ...... 35

Ancillary Data Acquisition ...... 38

4.1.2 Rainfall data ...... 38

Statistical Analysis ...... 38

Results ...... 39

4.1.3 Rainfall Results ...... 39

4.1.4 Water Quality Analysis ...... 39

4.1.4.1 Major Ions ...... 48

4.1.4.2 Nutrients ...... 49

4.1.4.3 Heavy metals ...... 51

4.1.4.4 Biological Indicators ...... 52

4.1.5 Principal Component Analysis (PCA) of the variables influencing water quality ...... 54

Discussion ...... 59

CHAPTER 5 LAND USE ...... 61

Methodology ...... 61

5.1.1 Data Collection and Preparation ...... 62

5.1.1.1 GIS Database ...... 62

5.1.1.2 Digital Surface Model Data ...... 64

5.1.2 Data Interpretation (Spatial) ...... 64

5.1.2.1 Site sectioning ...... 64

5.1.2.2 Digital Elevation Model ...... 67 v 5.1.2.3 Land use (diffuse sources)...... 70

5.1.2.4 Land use (point sources) ...... 73

5.1.2.5 Ancillary data ...... 78

Results and Discussion ...... 78

5.1.3 Land use and tenure ...... 78

5.1.4 Soil, Geology and Slope ...... 84

CHAPTER 6 LAND USE INFLUENCE ON WATER QUALITY ...... 89

Methodology ...... 89

6.1.1 Statistical Data Interpretation ...... 89

Results ...... 91

6.1.2 NMDS (Non-metric multidimensional scaling) ...... 91

6.1.3 RDA (Redundancy Analysis) and Cluster ...... 92

6.1.3.1 Land use influence on water quality ...... 92

6.1.3.2 Soil and slope combination influence on water quality ...... 97

Discussion ...... 100

Conclusion ...... 102

References ...... 103

METADATA ...... 122

APPENDIX A ...... 129

APPENDIX B ...... 134

APPENDIX C ...... 139

ANNEXURE ...... 141

vi LIST OF TABLES

Table 2-1: A description of the water quality assessment sites of the Sabie River catchment sampled from February 2016 to October 2016...... 7

Table 3-1: Literature examples of implementing GIS to detect land use influence on water quality. Selected relevant studies of land use and land cover influence on surface water quality with respect to the use of mainly multivariate statistical analysis...... 28

Table 4-1: Physical-chemical parameters measured in situ with a YSI 556 handheld field multimeter at each sampling site...... 36

Table 4-2: Summary of the physical-chemical variables measured by Rand Water’s Analytical Services as well as the method number listed by Rand Water, the unit and reporting limit...... 37

Table 4-3: The average values and range (minimum and maximum) of the water quality parameters measured at sampling localities 1-12 from February 2016 to October 2016. The results that have been shaded indicate that the SANS 241: 2015 limits have been exceeded indicating aesthetic  and health risks . These values were also compared to RQOs and TWQR (DWAF, 1996a,b&c)...... 41

Table 4-4: Eigenvalues and related percentages on the main contributing principal components...... 57

Table 4-5: Component variable correlations for the three extracted components...... 58

Table 5-1: Indicating SRTM data product specifications (USGS, 2015)...... 64

Table 5-2: Symbology legend for the combined 18 class land use interpretation extracted from 57 class clipped 2014 National Land Cover layer...... 72

Table 5-3: Reclassification values of land uses during weighted overlay analysis...... 75

Table 5-4: Land use results for watersheds 1 – 12 as the extracted 18 class land use categories. The table depicts the watersheds’ first and second most dominant land use as percentages in mustard and faded yellows respectively...... 86

Table 5-5: Land tenure table used within the multivariate analysis...... 86

vii Table 5-6: The distance in meters to the nearest waste water treatment plant above the sampling locality...... 86

Table 5-7: Soils summary statistics of average soil depth, percentage clay and percentage occurrence of broad land types (Land Type Survey Staff, 2002) within the Sabie-Sand catchment. (Symbology description can be seen on the following page)...... 87

Table 5-8: Percentage occurrence of main lithological forms within the Sabie-Sand catchment, representing the geology spatial characteristic used for multivariate analysis...... 87

Table 5-9: The mean slope for the Sabie-Sand catchment is shown as derived from the SRTM90 DEM. The mean elevation of the watersheds in meters was obtained. Also shown are the percentage rise in slope for the river and the percentage rise in slope for the watersheds (Jarvis et al., 2008)...... 88

Table 5-10: Land type descriptions (Land Type Survey Staff, 2002) within the Sabie- Sand catchment...... 88

Table 8-1: Spearman rank correlation test was used to determine whether there was significant (p < 0.05) differences between water quality indicators for the 12 sites...... 134

Table 8-2: Spearman rank correlation test was used to determine whether there was significant (p < 0.05) differences between water quality indicators for the 12 sites (continued)...... 135

Table 8-3: Kruskal-Wallis ANOVA and multiple test results...... 136

Table 8-4: Component-variable correlations (factor correlations), based on correlations of water quality variable concentrations. (Component heading abbreviated to C)...... 137

Table 8-5: Hardness of water is classified as follows by Kunin (cited by DWAF, 1996a)...... 138

viii LIST OF FIGURES

Figure 2-1: Locality map. Study area with regards to Mpumalanga, South Africa and various neighbouring country boundaries...... 6

Figure 2-2: Sampling sites locality map. Study sites within Sabie-Sand catchment, South Africa...... 8

Figure 2-3: Digital Elevation Model (DEM) representing topographical differences in mamsl (meters above mean sea level) of the Sabie-Sand catchment...... 9

Figure 2-4: Rainfall column chart (from January 2002–December 2008). Representing rainfall of the Sabie-Sand catchment from the west (operational rainfall station at Inyaka Dam) to the east (operational rainfall station in Kruger National Park in Skukuza) (DWS, 2017)...... 10

Figure 2-5: The general lithology of the Sabie-Sand catchment describing the properties of the surface rocks in the catchment...... 14

Figure 2-6: The first two lithological layers of the Sabie-Sand catchment represents the various rock types found at the surface of the catchment. These rocks will predominantly weather to form the catchment soils that contribute to surface water characteristics...... 15

Figure 4-1: Sampling sites locality map listing this study sites within Sabie-Sand catchment, used during this study...... 36

Figure 4-2: Indicating the monthly rainfall from October 2015 to October 2016 at Inyaka Dam monitoring station (DWAF, 2017)...... 39

Figure 4-3: Box and whisker plots illustrating the differences in a) SPC, b) TDS and c) Turbidity observed between the sites 1-9 (Sabie River), 10 (Marite River), 12, 7 (Sand River) and 11 (Inyaka Dam) during 2016. n = 4; ±SE (Standard Error) and ±SD (Standard Deviation)...... 47

Figure 4-4: Box and whisker plots illustrating the differences in a) Na+, and b) Cl- concentrations observed between the sites 1-9 (Sabie River), 10 (Marite River), 12, 7 (Sand River) and 11 (Inyaka Dam) during 2016. n = 4; ±SE (Standard Error) and ±SD (Standard Deviation)...... 49

3- Figure 4-5: Box and whisker plots illustrating the differences in a) DIN, b) PO4 and - c) NH3 observed between the sites 1-9 (Sabie River), 10 (Marite River),

ix 12, 7 (Sand River) and 11 (Inyaka Dam) during 2016. n = 4; ±SE (Standard Error) and ±SD (Standard Deviation)...... 50

+ 2+/3+ Figure 4-6: Box and whisker plots illustrating the differences in a) Al3 , and b) Fe concentrations observed between the 1-9 (Sabie River), 10 (Marite River), 12, 7 (Sand River) and 11 (Inyaka Dam) during 2016. n = 4; ±SE (Standard Error) and ±SD (Standard Deviation)...... 52

Figure 4-7: Box and whisker plots illustrating the differences in a) Chl-a, and b) E. coli concentrations observed between the sites 1-9 (Sabie River), 10 (Marite River), 12, 7 (Sand River) and 11 (Inyaka Dam) during 2016. n = 4; ±SE (Standard Error) and ±SD (Standard Deviation)...... 53

Figure 4-8: Scree plot indicating the eigenvalues of the correlation matrix...... 55

Figure 5-1: Digital Elevation Model (DEM) with Shreve ordered river network and 12 delineated watersheds used for spatial method of analysis in this study outlined in red...... 65

Figure 5-2: Comparison of multi-buffered segment 1 (above node 1) and segment 5 (above node 5) of Sabie River, converted to polygons for further analysis such as zonal statistics for each distance (e.g. 0.01km, 0.5km, 1km, 1.5km, 2km, 2.5km in watershed 1)...... 67

Figure 5-3: a) Represents the individual slopes (in % rise) of the 12 watersheds and b) indicates the slope (in % rise) for the whole Sabie-Sand catchment. The colour scheme indicates low lying areas as darker and areas increasing in percentage slope rise as lighter colours (Jarvis et al., 2008)...... 70

Figure 5-4: Combined 18 class land uses for this study area as seen in Table 5-2. Extracted from the 57 class, 2014 NLC layer (GeoTerraimage, 2015)...... 73

Figure 5-5: Major Possible Point Source Pollution Layer for Sabie-Sand catchment...... 74

Figure 5-6: Weighted overlay (b), cost analysis (c, d) and slope (f) of the western part of the Sabie-Sand catchment (watershed 1) used in combination with the land cover point and river point layers (a and e) to obtain a final cost path (g) for this watershed...... 77

Figure 5-7: Pie charts representing land use results in percentages of watershed areas 1-4. Land use abbreviation meanings: CuCom - Cultivated

x Commercial; CuOrc – Cultivated Orchards; CuSub – Cultivated Subsistence; EBare – Erosion-Bare; GrasL – Grassland-Low Shrubland; IndFo – Indigenous Forest; PlanW – Plantations-Woodlots; UrBuU – Urban Built-Up; UrRes – Urban Residential; UrVil – Urban Village; Water – Water; WdlnO – Woodland; Wetld – Wetlands...... 81

Figure 5-8: Pie charts representing land use results in percentages of watershed areas 5-8. Land use abbreviation meanings: CuOrc – Cultivated Orchards; CuSub – Cultivated Subsistence; EBare – Erosion-Bare; GrasL – Grassland-Low Shrubland; UrBuU – Urban Built-Up; UrRes – Urban Residential; UrVil – Urban Village; WdlnO – Woodland; Wetld – Wetlands...... 82

Figure 5-9: Pie charts representing land use results in percentages of watershed areas 9-12. Land use abbreviation meanings: CuCom - Cultivated Commercial; CuOrc – Cultivated Orchards; CuSub – Cultivated Subsistence; GrasL – Grassland-Low Shrubland; IndFo – Indigenous Forest; PlanW – Plantations-Woodlots; UrTsh – Urban Townships; UrVil – Urban Village; Water – Water; WdlnO – Woodland; Wetld – Wetlands...... 83

Figure 6-1: Non-metric multidimensional scaling (NMDS) ordination plot of the percentage land use surrounding the river with regards to the sampling localities. The Bray Curtis 2D stress value was 0.02 and the Euclidian distance 2D stress 0.01, indicating a good fit. Sites shown are: 1 – Sabie River Headwaters; 2 – Sabie River After Waste Water Treatment; 3 – Sabie River Before Hazyview; 4 – Sabie River After Hoxane Water Treatment Works; 5 – Sabie River at Kruger National Gate; 6 – Sabie River Skukuza; 7 – Sand River Kruger National Park; 8 – Lower Sabie River after Sabie - Sand Confluence; 9 – Sabie River Bordering Mozambique; 10 – Marite River After Inyaka Dam; 11 – Inyaka Dam Outlet; 12 – Sand River Thulamahashe After Waste Water Treatment Plant...... 92

Figure 6-2: Redundancy Analysis (RDA) triplot of land use percentages (predictor) effect on mean measured water quality parameters (response), at the 12 different sampling sites (orange dots) for the Sabie-Sand catchment...... 95

Figure 6-3: Dendrogram showing the cluster classification of the watersheds of the 12 sampling sites in the Sabie-Sand catchment, according to their land-use

xi influences on water quality parameters. Break was inserted in the axis between 20 000 and 2 500 000. Inset: The smaller inset is the full Euclidean distances between the clusters before axis breakage. Interpretation of the large axis was difficult and therefore the above- mentioned break was introduced...... 96

Figure 6-4: Redundancy Analysis (RDA) triplot of soil percentages (predictor) effect on mean measured water quality parameters (response), at the 12 different sampling sites (black dots) with regards to a co-variable namely slope for the Sabie-Sand catchment. Watershed slopes in percentage rise represented as the co-variables were divided into classes: Class 1 (0-5%) (not shown in RDA); class 2 (>5-15%); class 3 (>15-25%); class 4 (>25%)...... 99

Figure 7-1: Site 1. Sabie River Headwaters. Left photo captured by Michaela Stolz during October 2016 and right captured by Dr. Annelie Swanepoel during July 2016 sampling...... 129

Figure 7-2: Site 2. Sabie River below Wastewater treatment plant. July, (A. Swanepoel)...... 129

Figure 7-3: Site 3. Sabie River Below Waste Water Treatment. July 2016, (A. Swanepoel)...... 130

Figure 7-4: Site 4. Sabie River Below Hoxane Water Treatment Works. July 2016, (A. Swanepoel)...... 130

Figure 7-5: Site 5. Sabie River KNP Gate. July 2016, (A. Swanepoel)...... 131

Figure 7-6: Site 6. Sabie River Skukuza KNP. July 2016, (A. Swanepoel)...... 131

Figure 7-7: Site 7. Sand River KNP. July 2016, (A. Swanepoel)...... 131

Figure 7-8: Site 8. Sabie River Lower. July 2016, (A. Swanepoel)...... 132

Figure 7-9: Site 9. Sabie River close to Mozambique. July 2016, (A. Swanepoel)...... 132

Figure 7-10: Site 10. Marite River after Inyaka Dam Outlet. July 2016, (A. Swanepoel)...... 132

Figure 7-11: Site 11. Inyaka Dam Outlet. July 2016, (A. Swanepoel)...... 133

xii Figure 7-12: Site 12. Sand River Thulamahashe Below Waste Water Treatment. July 2016, (A. Swanepoel)...... 133

Figure 9-1: Map indicating the three main land tenure classes for the Sabie-Sand catchment (as seen in Table 5-5). To the west is a dominant commercial tenure, centre of the catchment is predominantly communal as indicated with the red and light blue polygons and to the east, the dominant land tenure class is conservational as represented by the SAPAD colour legend...... 139

Figure 9-2: Map indication and description of the soil types that occur in the Sabie- Sand catchment as seen in Tables 7 and 12 (Land Type Survey Staff, 2002)...... 140

Figure 10-1: South Africa: PDF files of South African landcover from the CSIR ARC national 1: 250 000 land cover data set segmented by secondary drainage region X3 (Resource Quality Services, 2003)...... 141

xiii LIST OF ABBREVIATIONS CMA Catchment Management Area DD Decimal Degrees DWA Department of Water Affairs DWAF Department of Water Affairs and Forestation DWS Department of Water and Sanitation DRDLR Department of Rural Development and Land Reform EMC Ecological management class EWR Environmental Water Requirements IUCMA Inkomati-Usuthu Catchment Management Agency IWMA/ IUWMA Inkomati / Inkomati-Usuthu Water Management Area LULC Land Use and Land Cover mamsl Meters above mean sea level MAP Mean Annual Precipitation mbgl Meters below ground level NAEHMP National Aquatic Ecosystems Health Monitoring Programme NGA National Groundwater Archive NTU Nephelometric Turbidity Units NWA National Water Act NWRS National Water Resource Strategy PES Present Ecological State PPS/ PSP Point Pollution Source/ Point Source Pollution RC Resource Class RDA Redundancy Analysis REMP (formerly River Eco-status Monitoring Programme (formerly known as The River known as RHP) Health Program) RQO’s Resource Quality Objectives SAWQG South African Water Quality Guidelines SRTM Shuttle Radar Topography Mission SWMM Strom Water Management Model TWQR Target Water Quality Ranges WMA Water Management Agency WTP Water Treatment Plant WWT Wastewater Treatment WWTP Wastewater Treatment Plant

xiv TABLE OF DEFINISIONS Catena “In soil science a catena is a sequence or series of soils derived from similar parent material, under similar climatic conditions at approximately the same time, but varies in characteristics due to differences in relief and drainage” (Goudie, 2004). Cation Soil’s ability to hold positively charged ions is called the cation exchange exchange capacity (CEC) and as EC is conventionally expressed in meq/100g this is capacity (CEC) equal to and expressed in centimoles of charge per kilogram of exchanger (cmol(+)/kg) (Soil Quality, 2009). Detailed method The step-by-step use of ArcMap to complete the described method (work flow). Earth Explorer “Earth Explorer user interface, developed by the United States Geological Survey (USGS), is an online tool used for the query, search, discovery, and ordering of satellite images, aircraft and other remote sensing inventories. This is done by interactive and textual-based query capabilities of the remote sensing resources from several databases. Users can identify search areas, datasets, and display metadata, browse and integrate visual services within the interface” (LP DAAC, 2014 and USGS, 2018a). Ecological A classification system for South African rivers serve as different levels of management water resource protection. Three ecological management classes exist. class Natural; moderately used/ impacted; and heavily used or impacted each representing a different level of protection (NWRS, 2004). Environmental Water required for aquatic ecosystem protection of the water resource – Water where operational limitations and stakeholder considerations are also Requirements taken into account (Pollard et al., 2011). Initial method Within the present study initial method carries the meaning of a method, tools or approaches initially used or attempted before a decision was made for the final method and outcome. Land Tenure FAO, 2002 have an elaboration on what is land tenure. Broadly defined, land tenure is the rules how property rights to land are to be allocated within societies; how access to use, control, and transfer land, and its associated responsibilities are granted; and it is the relationship among individual people or groups with respect to land. Land Use and As described by NOS (2015) land cover describes the type of natural Land Cover landscape that covers a region e.g. agriculture, forests, wetlands etc. and land use is the method in which these land covers are used by people, e.g. for conservation, development, recreation etcetera. Landsat Landsat is a succession of 8 satellites launched between 1972 and 2013 that provides a continuous collection of space-based, moderate-resolution land remote sensing data” (e.g. multispectral images of the Earth) (USGS, 2018b). The latest Landsat-8 satellite caries two instruments namely the Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS). “The OLI and TIRS images consists of nine spectral bands with a spatial resolution of 30 meters for bands 1 to 7 and 9. Band 8 has a 15 meter (panchromatic) resolution. Thermal bands 10 and 11 are collected at 100 meters and provide more accurate surface temperatures” (Barsi, 2014 and USGS, 2018b).

xv Natural The quantifiable capacity of a river to remediate non-hazardous waste assimilative discharges without violating its predetermined ecological management capacity class (NWRS, 2004). Every water resource’s assimilative capacity will differ. Present A classification of water resources according to ecological status or health Ecological State compared to natural conditions. Classes include: A – near natural, B – largely natural C – moderately modified D – largely modified E – seriously modified F - critically modified. Redundancy “Method to extract and summarise variation in a set of response variables Analysis (RDA) explained by a set of explanatory variables. It is a direct gradient analysis technique which summarises linear relationships between components of response variables that are “redundant” with (or “explained” by) a set of explanatory variables. To achieve this, RDA extends multiple linear regression by allowing regression of multiple response variables on multiple explanatory variables” (Buttigieg and Ramette, 2014). Resource In order to ensure resource protection, RQOs are a descriptive statement Quality of the quality conditions which should be met by the water resource Objectives (NWRS, 2004). SPOT “Satellite Pour l’Observation de la Terre (SPOT), is a high-resolution, optical imaging system operating from space” (CNES, 2009). SPOT-6 optical satellite carries a number of cameras, imaging (recording) the Earth with a 1.5 meter panchromatic and 6-meter multispectral (green, red, blue, near-IR) resolution (Kressler, et al., 2003 and Satellite Imaging Corporation, 2017). Water A “Water Management Area” is an area established as a management unit Management in the NWRS within which a catchment management agency will conduct Area the protection, use, development, conservation, management and control of water resources (DWAF, 1999).

xvi CHAPTER 1 INTRODUCTION

Background

River health assessment is of utmost importance in South Africa; whether to indicate the quantities of clean water required for domestic use and the sustainable ecological flow of rivers, or to indicate river qualities for these uses. Healthy ecosystems are dependent on the services provided by healthy rivers. Brauman et al. (2007) emphasised the importance of maximizing a river’s health to improve water quality. This is an uplifting thought as studies such as “Entering an Era of Water Scarcity: The Challenges Ahead” conducted in 2000 was especially concerned with the finite freshwater resources that South Africa has at its disposal (Postel, 2000). Also drawn from the previously mentioned study, is that at the time, South Africa was the only country in which the definite water laws, such as the National Water Act (Act No. 36 of 1998) (see section 3.2) focused on reserving water to enhance ecosystem health (Postel, 2000). The NWA contributed to efforts from other countries and continents such as the United States and Europe, to augment their existing legislation to a more advanced status – hence the introduction of legislation such as the EU Water Framework Directive. The developed countries not only conducted monitoring and water health assessments, but since 2000 also made it a national policy to protect their water resources (Hallett et al., 2016).

Various management guidelines have been implemented to support the protection of South Africa’s water resources (such as the eight volumes of the South African Water Quality Guidelines) (NWRS, 2004). Adhering to these standards researchers have come a long way in initiating awareness and conducting studies aimed at enhancing the existing knowledge of South African water resources, its associated influences, and affected health. The river’s influences are experienced either through determining a regions’ demography, socio-economic structures and physical manmade landscapes due to the type, quality and amount of flow present in the river; or the other side of the spectrum where the river is being influenced by the various anthropogenic activities once an areas’ niche is established (Kusangaya et al., 2013).

River water quality research generally aims to be of such value that it may be incorporated in future and on-going legislations; management strategies (see section 3.2) set-up by governing bodies in order to protect these valuable resources; and decision making by water service providers. To ensure effective water resource management to take place as required by the National Water Resource Strategy (NWRS) (see section 3.2), South Africa’s water resources have been divided into 9 Water Management Areas (WMA). The WMA division takes into account the catchment and aquifer boundaries; stakeholder participation; financial viability and equity considerations (DWAF, 1999).

1 Required under the NWRS (see section 3.2) these areas are managed by respective catchment management agencies (CMA). This study area for the present study falls within the Inkomati- Usuthu Water Management Area (IUWMA). The study will hopefully provide decision support information by analysing the effect of current land use on the water quality of the well-known Sabie River. The Sabie River and its main tributaries – the Sand and Marite rivers are the main water courses formed from the Sabie-Sand drainage region, better known as the Sabie-Sand catchment. This catchment forms part of the Inkomati-Usuthu Water Management Area (IUWMA). As stated by Mallory and Beater in 2009 the region previously referred to as the IWMA is a water stressed Management Area, meaning the supply cannot hold up against the demand for freshwater resources.

The Sabie-Sand may soon be a water-stressed catchment because of conflicting demands for potable water and the requirements of the ecological Reserve (DWA, 2013 as cited by Mallory et al., 2013). If full implementation of the Inyaka Integrated Water Supply Scheme takes place, treating and supplying water from Inyaka Dam to the fast expanding Bushbuckridge semi-urban area); while other parts of the catchment are in conservation areas. During a water requirement and availability study by Beumer and Mallory (2014) which analysed the water amounts of the Sabie River already allocated for use, it also concluded that the water available from the Sabie River is fully allocated. Therefore, even though a surplus of water supply exists after implementing the Reserve, in the event where additional allocations are required, the risk exists that the Ecological Management Class (EMC, see definitions) of the Sabie River will then have to degrade (Beumer and Mallory, 2014).

Regarding water quality, this type of degradation will negatively influence the Sabie River’s status. Degrading of its Present Ecological State (PES, see definitions) will take place when lowering the desired Reserve from the existing Class A/B (Beumer and Mallory, 2014). The EMC is closely related, or a representative value of the Sabie River’s health and available water quality. As briefly described above with the implementation of the Integrated Water Supply Scheme, water quantities and qualities vary greatly when considering the inflow and abstraction of different natural and anthropogenic activities. This relates to the importance of land use as a driver of river ecosystem health (Vrebos et al., 2017). Below is a brief description of the typical land use and associated activities found in the area. Further investigation of land use and anthropogenic influences on water quality form part of the literature review (see Section 3) and will also be described in an in-depth analysis in Chapter 5. Mallory et al. (2013) mentions the typical land use with associated activities found in this study area that can possibly alter the physical, biological and chemical characteristics of the river water or that can pose water quality problems include:

 Economic drivers such as commercial forestry, irrigated agriculture and industries causing surface alterations and major water abstractions from water resources; 2  Conservational practices causing remediating effects on the one hand, but also influences water resources on the other hand though natural wildlife intervention and possible overgrazing in riparian zones – resulting in runoff erosion and sedimentation;  Social and demographic drivers: where 1) urban infrastructure, semi-urban and rural township expansion contribute as being point source- (e.g. non-compliant Waste Water Treatment Plant (WWTP) and non-point source (e.g. stormwater runoff) polluters; as well as 2) an increase in water users;  Impoundments contributing to bulk water supply for irrigation and increasing domestic purposes, dams also impact temperature, oxygen levels and sediment transport.

As mentioned above, previous studies such as O'Keeffe et al. (1996) and Mallory et al. (2013) assessed water demand upstream of the Kruger National Park (KNP) as well as ecological river flow requirements of the Sabie River Catchment (Pollard et al., 2011). These studies provided information on water supply schemes to cater for future developing land use activities and an increasing population. The positive and negative outcomes that the construction of the Inyaka Dam had on the Marite River, a tributary of the Sabie River, are also mentioned (see section 3.1). Unfortunately, a full water quality assessment in terms of land use after construction of the dam was not conducted. The future increases in urbanization and especially rural settlement development (such as the Bushbuckridge area) have started to raise concerns in terms of water quality of this major South African river and its tributaries (Mallory et al., 2013 and Tlou, 2011). A study by Tlou (2011) focuses on the demographics of the Sabie-Sand River Catchment and the Inyaka Dam Water Supply Scheme. It indicated that, although the water quality is relatively good, supply problems and quality deterioration may be experienced in the future, especially in the case if water sanitation facilities are not properly managed. These problems relate to population and rural land use increases or they may even – in drought periods – be enhanced by a shifting climate. Recent climatic situations in this study area include that 2015 was a very dry year with minimal precipitation. Minor flash floods were only experienced in March 2016 and higher amounts of rainfall in October 2016. A nationwide drought was experienced from 2013–2016 where effective or significant rainfall was only experienced in the summer of 2016/7. It is suspected that the land uses and rural urbanization within the area of the Sabie River have a negative effect on the water quality of the Sabie River. Therefore, the purpose of this study was not only to conduct water quality assessments in terms of physical, chemical and land utilization but also to achieve a river quality analysis within such dry or low flow circumstances. Hereby the study aims to contribute to this field of study in terms of river health and usage.

3 Aims and Objectives

The main aims of this study are to do an in-depth determination of the different land uses surrounding the Sabie River Catchment and to evaluate the influence thereof on the water quality of the Sabie River and two of its major tributaries. These aims will be accomplished with the completion of the following objectives:  Undertake an assessment of the land use surrounding the Sabie, Sand and Marite rivers.  Analyse the water quality of the Sabie, Sand and Marite rivers by measuring their physical, chemical and biological factors.  Compare water quality parameters to Target Water Quality Ranges (TWQR) set for the Sabie River.  Assess whether or not rural urbanization and surrounding land uses have taken a negative toll on the water quality.

This study will therefore not only focus on the possible effects that urbanization and land uses have on the areas’ water quality, it will also assess the river water quality that will be experienced during a dryer rainfall period. This will entail various limitations or constraints such as: the crystallisation of certain minerals in the streambeds and (sediment interstices) in soil; crystallised minerals sporadically dissolving into water when occasional rainfall is present; the limited amount of sampling water taken to be analysed by a laboratory; and higher rates of evapotranspiration caused by heat, therefore increasing vegetation loss and exposed soils. These are only a few points taken into consideration when viewing some of the influences for field analysis and chemical data collection.

4 CHAPTER 2 OVERVIEW AND GENERAL STUDY AREA

Location and Boundaries

The Sabie-Sand River Catchment is one of three sub-catchments that fall under the Inkomati- Usuthu Water Management Area (IUWMA). As seen in Figure 2-1, the Sabie-Sand Catchment lies in the north of the IUWMA, where the Sand River joins with the Sabie River, as one of its main tributaries, within the Kruger National Park (Pollard and Du Toit, 2011). The Marite River, the other main tributary of the Sabie flows from the Inyaka Dam, forming part of the Inyaka Dam Water Supply Scheme (Tlou, 2011). The main urban and semi-urban areas in the Sabie-Sand sub-catchments (from here on referred to as the catchment) include towns such as Sabie, Hazyview, and Skukuza that are centrally located near the Sabie River. Bushbuckridge between the Marite and Sand rivers, is a representative town of densely populated, rural or semi-urban development within the catchment. Hills formed by the North-eastern part of the Great Escarpment form the western border of the catchment. The headwaters of the Sabie River within these hills mark the start of the catchment’s most western side and stretches through to the Kruger National Park, bordering Mozambique to the East of the catchment. This results in the Sabie River flowing through Mozambique, into the Corumana Dam and into the Incomati River near the town of Moamba (Pollard and Du Toit, 2011). These rivers, forming part of the IUWMA, contribute to an international water course, namely the Incomati basin. This is a shared basin with a transboundary nature Reserve between South Africa, Swaziland and Mozambique (Tlou, 2011). It requires certain international obligations from South Africa in terms of cross-border flow, especially with regards to planners, managers and everyday decision-makers (Ashton et al., 2006). Not only is the Tripartite Interim Agreement applicable to this study area, but according to Pollard and Du Toit (2011) a large portion of this study area also falls within the Great Limpopo Transfrontier Conservation Area (GLTFCA), which was promulgated in 2002. It would therefore be very important for South Africa to adhere to these freshwater agreements and responsibilities pertaining to water quality and quantity. In order for various reaches within the Sabie, Sand and Marite rivers to be examined. Table 2-1 and Figure 2-2 briefly describe the habitat and locality information of the 12 sites for sampling during this study.

5

Figure 2-1: Locality map. Study area with regards to Mpumalanga, South Africa and various neighbouring country boundaries.

6 Table 2-1: A description of the water quality assessment sites of the Sabie River catchment sampled from February 2016 to October 2016. Site name and abbreviation Coordinates Description (in DD) (Site photo Appendix A) 1 - Sabie River Sabie HW -25.14747, This site is situated in the mountainous Sabie River Headwaters 30.66872 headwaters above Sabie town. Site 1 serves as a convenient reference point to compare land use impacts further down the Sabie River. As the only major influence is forestation, most water quality parameters still indicate a stream of pristine conditions (Figure 7-1). 2 - Sabie River Sabie WWT -25.07386, Still situated at a high relief, between forestry practices Waste Water 30.85080 (and at the time of sampling deforestation) is site 2 in Treatment the Sabie River. It lies ~6.7km below Sabie town’s WWTP situated at (-25.0912441, 30.7942149) Decimal Degrees (Figure 7-2). 3 - Sabie River Sabie BH -25.03008, Site 3 is situated before Hazyview town, right before the Hazyview 31.02530 Sabie River’s confluence with the Mac-Mac River. The main adjacent land use influences are forestry and plantation practices (Figure 7-3). 4 - Sabie River Sabie Hoxane -25.01933, Site 4 within the Sabie River is situated in a lower Hoxane Water 31.21733 moderate relief (flatter region) ~1.2km below Hoxane Treatment Water Treatment Works situated at (-25.015117, Works 31.208857) Decimal Degrees (Figure 7-4). 5 - Sabie River Sabie KNP -24.97972, Sampling conducted from a high bridge above the Sabie Kruger Gate Gate 31.48275 River at the Kruger Gate of the KNP. Site 5 has a prevailing low relief with sparse riparian vegetation and progressive sediment accumulation (Figure 7-5). 6 - Sabie River Sabie Skukuza -24.99088, Sabie River site at Skukuza town in the KNP just after Skukuza 31.60175 the confluence with the intermittent Nwaswitshaka stream (a tributary that was dry throughout the sampling timeframe) (Figure 7-6). 7 - Sand River Sand KNP -24.96777, Site 7 pinpoints the Sand River near Skukuza within the Kruger National 31.62561 KNP before its confluence with the Sabie River. This site Park clearly distinguishes where the river name has its origin, flowing/ transecting through thick sandy beds (Figure 7- 7). 8 - Lower Sabie Sabie Lower -24.97527, This point within the Sabie River has a very broad, River after 31.76805 dendritic channel and is situated below the confluence Sabie - Sand of the Sabie and Sand rivers at Lower Sabie in the KNP. confluence Site 8 is located opposite to the exclosure sites in the KNP next to the Nkuhlu picnic spot (Figure 7-8). 9 - Sabie River Sabie Moz -25.16041, Site 9 is located in very low relief with a savanna type bordering 31.99875 terrain intersecting a few geological outcrops. Situated Mozambique in the Sabie River at Lower Sabie ~5km before crossing the border to Mozambique in the KNP (Figure 7-9). 10 - Marite River Marite -24.96081, Nearly ~18.2km from where the Marite River flows from 31.10850 the Inyaka Dam, lies site 10 within the Marite River, above Hazyview town. In this reach of the river it traverses through natural bushy vegetation, over quite a few bare rock outcrops (Figure 7-10). 11 - Inyaka Dam Inyaka Dam -24.88538, Site 11 situated in the stagnant water, at the closest at the dam wall Wall 31.08469 point of access to the Inyaka Dam Wall, before the water leaves through the dam sluices/ outlet (Figure 7-11). 12 - Sand River Sand -24.72186, Situated underneath a high-water train bridge, within the Thulamahashe Thulamahashe 31.23716 Sand River is site 12. The site is ~0.5km below the Waste Water Thulamahashe WWTP situated at (-24.720727, Treatment Plant 31.231072) Decimal Degrees (Figure 7-12).

7

Figure 2-2: Sampling sites locality map. Study sites within Sabie-Sand catchment, South Africa.

Topography and Climate

As this study catchment stretches from the Drakensberg Escarpment (Middleveld) on the west at an altitude of ~2200m it gradually flattens towards the (Lowveld) KNP and Mozambique on the east at an altitude of ~150m (Figure 2-3). The Lowveld zone of the catchment with its gradually sloping topography is classified as a pediplain with a gentle eastward slope. A pediplain being an extensive plain formed by coalescence of numerous pediments – which are gently inclined slopes (usually 0.5°-7°) of erosion and/or transportation that truncates rock formations and connects eroding slopes or scarps to lower levels which then form areas of sediment deposition – (bedrock surfaces) (Oberlander, 1989 as cited by White, 2004). This topographical change also brings about a change in catchment climate. Smits et al. (2004) mentions that the catchment has a higher rainfall in the western sub-humid (sub-tropical) part compared to the rainfall experienced in the eastern semi-arid part of the catchment (Figure 2-4). With a semi-arid climate; warm to hot conditions and according to WR (2012) an average summer rainfall of ~700mm per year within the catchment, winters are mild and generally frost-free (DWAF, 2004; Rountree et al., 2000; Venter et al., 2003 and Woodhouse, 1995). Figure 2-4 indicates monthly rainfall for the years 8 2002–2008, represented by Inyaka Dam (western side) and the Skukuza Station in the KNP (eastern side). This histogram only gives an overall comparison to visually represent the western high to eastern low monthly rainfall differences.

To the north of the catchment lies the sand-region, which according to Pollard and Walker (2000) is a semi-arid; majority Lowveld region; with erratic rainfall. This variability in rainfall seasons result in droughts every three to four years (Cousins et al., n.d). Therefore, the region has a vulnerable water security and a suspected high-risk quality – judging by the amount of increase in rural expansion of the catchment (Pollard and Walker, 2000). The southern-part of the catchment drained by the Sabie River, do not vary drastically from the northern part. It also testifies to a higher rainfall closer to the escarpment (west) as Highveld conditions prevail; and lower rainfall occurs to the east where Lowveld conditions are experienced. This west to eastward trend is also true for the evaporation rate of the catchment where, according to DWAF (2004) the rate is at 1 400mm/a in the west and 1 700mm/a in the east (Jewitt et al., 1997 as cited by DWAF, 2004).

Figure 2-3: Digital Elevation Model (DEM) representing topographical differences in mamsl (meters above mean sea level) of the Sabie-Sand catchment.

9 According to Middleton and Bailey (2005) the catchment receives varying amounts of rainfall contributing to surface water. The average annual rainfall decreasing along the topographical gradient from the mountainous areas through the foothills and towards the lower, flatter reaches of the Sabie River range between 1500–900mm/a to the west, and 600-348mm/a to the east of the catchment. This result in the catchment’s surface runoff MAR (mean annual runoff) originating predominantly in the wetter, western parts of the catchment (Woodhouse, 1995). Average temperature during midday in summer is ~28 ᵒC and in winter, average midday temperature is ~21.5 ᵒC (WR, 2012). During sampling times, site specific river water varied on average with 2ᵒC according to ambient temperatures, as well as with differences in topography.

Average Rainfall for Sabie-Sand Catchment (2002-2008) 280 260 240 220 200 180 160 140 120 100

80 Monthly Averages (mm) Monthly 60 40 20 0

Month

Inyaka Dam Rainfall Station Kruger National Park Skukuza Rainfall Station

Figure 2-4: Rainfall column chart (from January 2002–December 2008). Representing rainfall of the Sabie-Sand catchment from the west (operational rainfall station at Inyaka Dam) to the east (operational rainfall station in Kruger National Park in Skukuza) (DWS, 2017).

Vegetation and Geology

2.1.1 Vegetation

The vegetation of the Sabie-Sand catchment is strongly related to topography and geological formations. Mucina and Rutherford (2006) classify the lowland pediplains and escarpment 10 foothills as Savanna Biome with only the upper mountain plateau as Grassland Biome. Forests occur as zonal and intrazonal patches along sheltered valleys and steep slopes of the escarpment (Coetzer et al., 2010 and WRC, 2001).

The Grassland Biomes are divided by Mucina et al. (2006) into the Lydenburg Montane Grassland, the Northern Escarpment Dolomite Grassland and the Northern Escarpment Quartzite Sourveld along the topographic catena. The vegetation of the Lydenberg Montane grassland is described as forb rich short grassland with small forest and thicket patches along drainage lines. The Lydenburg Montane Grassland transcend into the Northern Escarpment Dolomite Grassland characterised as highly variable shrub rich grassland. Because of the dolomitic geology the soils have a high pH and are rich in calcium and magnesium but low in phosphorous. The Northern Escarpment Quartzite Sourveld occurs on quartzites of the Black Reef Group at the base of the Transvaal Sequence. Due to the rocky relief and acidic soils grass composition is characteristically sour. Northern Mistbelt Forest occurs as isolated patches within the Northern Escarpment Quartzite Sourveld extending along river valleys into the Savanna Biome below (Matthews et al., 1991). Land use is mainly plantations with some limited cultivation. A detailed phytosociological analysis of the communities of the Escarpment grassland was done by Matthews (1991) and parts of the vegetation of the Lydenburg Montane Grassland was described by Burgoyne (1995).

Further down the catena climate becomes drier and warmer and Grassland grades into Savanna of the Lowveld Bioregion. Rutherford et al. (2006) recognised seven Savanna vegetation types on the lower plains of the Sabie Catchment. The Legogote Sour Bushveld occurs on the foothills of the Escarpment. Vegetation structure is characteristically open sparse woodland. Dominant trees are Pterocarpus angolensis, Sclerocarya birrea subsp. caffra, Vachellia sieberiana var. woodii and Vachellia davyi. The central part of the Sabie catchment is described by Rutherford et al. (2006) as Granite Lowveld. Vegetation structure is a tall shrubland/low woodland dominated by the trees Senegalia nigrescens, Sclerocarya birrea subsp caffra, Terminalia sericea, Combretum zeyheri and Combretum apiculatum.

Within the Granite Lowveld sparse woodland, the herbaceous layer is dominated by Themeda triandra is present on vertic soils derived from the Timbavati gabbro. Rutherford et al. (2006) refers to this as the Gabbro Grassy Bushveld. Related to the Legogote Sour Bushveld is the Pretoriuskop Sour Bushveld on granite derived soil relating to upland topography. To the far east of the catchment the Sabie River transects the Delagoa Lowveld, Tshokwane-Hlane Basalt Lowveld and Northern Lebombo Bushveld (Rutherford et al., 2006). The Delagoa Lowveld is associated with sodium rich duplex soils derived from the Karoo Supergroup shale (Gertenbach, 1983). The vegetation structure is an open savanna dominated by Senegalia senegal var rostrata and Senegalia welwitschii subsp. delagoensis. To the east of the Delagoa Lowveld the 11 Tshokwane-Hlane Basalt Lowveld occurs on Letaba Formation Basalts of the Karoo Supergroup. The vegetation is an open tree savanna dominated by the trees Sclerocarya birrea and Senegalia nigrescens. The Northern Lebombo Bushveld occurs on the Lebombo Mountains. The savanna vegetation type is an open Combretaceae dominated woodland on clayey shallow soils.

Siebert, 2003 (as cited by Ayres, 2012) mentions four major riparian zones along the elevation gradient of the Sabie River. From the higher reaches a dry riparian zone (or dry woodland) zone has predominant Acacia spp, Combretum imberbe, Gymnosporia senegalensis and Philenoptera violacea species moving towards a more gradual topography with dominant Ficus spp., Diospyros mespiliformis and Trichilia emetica species grouped to form the wet riparian forest zone. Closer or within the KNP a shrubby reed scrub is the major riparian zone predominated by Flueggea virosa, Gymnosporia spp., Grewia spp., Phragmites mauritianus and Trema orientalis species and ending the final stages of the Sabie River with a reed scrub zone with predominant Combretum spp., Phragmites mauritianus and Ficus species within this zone (Siebert, 2003 as cited by Ayres, 2012).

2.1.2 Geology

The area comprises various ranges of bedrock lithologies including: metamorphic rocks such as quartzite; intrusive and extrusive igneous rocks such as granite; and sedimentary rocks such as shale. Other main geological rock types (Figure 2-6) in the area include basalts, conglomerates, granites, andesites, irons and gneiss (Middleton and Bailey, 2005). These all form part of the three main litho-stratigraphic units that underlie the catchment. To the west the Transvaal Sequence (2600-2200Ma old) (blue, yellow and light green in Figure 2-5) overlies the older Basement granite-gneiss complex (3500 - 2650Ma old) (light pink in Figure 2-5) which is followed to the east by the youngest litho-stratigraphic unit of the area, the Karoo sequence (approx. 250– 183Ma old) (dark pink, green and purple in Figure 2-5) (Norman and Whitfield, 2006).

Norman and Whitfield (2006) mentioned that to the topographically higher eastern part (Transvaal highlands), the catchment’s Transvaal Sequence consists of various (mainly sedimentary) rock formations. Typically for this catchment, the Transvaal Sequence is represented by quartzite (of the Black Reef Formation and the Wolkberg Group) (Figure 2-6 Lit1) (Norman and Whitfield, 2006), comparatively soft shale (of the Wolkberg Group) (Figure 2-6 Lit1 and 2) (Norman and Whitfield, 2006), dolomite (of the Chuniespoort Group – Malmani Subgroup) (Figure 2-6 Lit1) (Norman and Whitfield, 2006) breccia, conglomerate (Figure 2-6 Lit2), lava (Figure 2-6), tuff (Figure 2-6), diamictite, basalt (Figure 2-6) and chert (Figure 2-6 Lit2) (Norman and Whitfield, 2006 and Chunnett et al., 1990 as cited by Wells, 1992). Chunnett et al. 1990 (as cited by Wells, 1992) mentions that the frequent gold deposits that occur within the Graskop and Sabie areas led to a small amount of mining activity in this part of the catchment.

12 According to Norman and Whitfield (2006) the catchment’s largest lithological unit is the crystalline granite-gneiss Basement Complex (of the Nelspruit Suite) which forms the Lowveld (light pink Figure 2-5 and Figure 2-6). It contains granodiorite (Figure 2-6) and massive, grey, coarse- grained, crystalline granite (Figure 2-6) (Norman and Whitfield, 2006) as well as intrusions of gabbro (Figure 2-6) and diabase (also called dolerite) in the south west, with a large tonalite intrusion (Figure 2-6) in the centre of this Complex (Chunnett et al., 1990 as cited by Wells, 1992). As the quartzites within the Transvaal Supergroup are more resistant to weathering over time than the Basement granite-gneiss, it resulted in a plateau (Lowveld) in the centre of the catchment (described above as pediplanation). The KNP is underlain by several varieties of Basement granite-gneiss; several younger gabbroic and syenite intrusions and minor greenstone remnants (Norman and Whitfield, 2006). The underlying geology is an important factor in the catchment’s topography and strongly controls its landscape.

According to Norman and Whitfield (2006) the eastern side of the catchment (especially the entire eastern strip of the KNP) mainly consist of sediments and volcanics of the Karoo Supergroup such as: river-channel sandstone (arenite in Figure 2-6), grey-green to reddish mudstone (Figure 2-6), dark shale (Figure 2-6) and coal seams (of the Beaufort and Ecca Groups) (Figure 2-6); yellowish to pinkish fine-grained sandstone and siltstone (dark pink Figure 2-5) (of the Clarence formation) (Norman and Whitfield, 2006); lower basaltic lavas (from the Lebombo Group – Letaba Formation) (Figure 2-6) and upper rhyolitic lavas (from the Lebombo Group – Jozini Formation) (Figure 2-6) (Norman and Whitfield, 2006); as well as basalts (from the Drakensberg Group) (Figure 2-6) (Norman and Whitfield, 2006) and some granophyre (Figure 2-6) and Karoo dolerite intrusions (Chunnett et al., 1990 as cited by Wells, 1992).

Chunnett et al. 1990 (as cited by Wells, 1992) is of opinion that the soils in this catchment, compared to other areas in southern Africa, have a low erosion risk (therefore relatively high erosion resistance). The mountain grasslands and Afromontane forests of the upland areas and lower slopes respectively, are underlain by shallow lithosols (beneath grasslands) and well- developed, sometimes leached, mature soils (beneath forest floor). According to O’Keeffe (1985) the lower catchment (outside of the KNP) is represented by ferrallitic clays and arenosols, whereas they KNP’s higher areas are representative of shallow sandy soils and its lower lying areas of sodic duplex soils. In the KNP’s western part, black and red clays overlie the gabbro that outlines the Sabie River’s channel (O’Keeffe, 1985). The east of the catchment, accompanied by lower topographies, is represented by river channels typical to floodplains (e.g. anastomosing channels) (Nanson and Gibling, 2004). This, in drought periods results in an accumulation and deposition of sediment with subsequent thicker vegetative establishment and stabilisation within the stream compared to the density found on the stream banks (Heritage and Van Niekerk, 1995 and Rountree et al., 2000). Rountree et al. (2000) mentions that with low flow and sediment

13 accumulation, terrestrial vegetation would rather establish than water dependent riparian vegetation. This observation was confirmed during fieldwork for this study. Also, Norman and Whitfield (2006) mentions that to the far eastern side where one finds the Karoo Sequence, the Letaba Lava weathers to form a characteristic dark, clayey soil and the successive Jozini rhyolite, yield lithosols (O’Keeffe, 1985).

Figure 2-5: The general lithology of the Sabie-Sand catchment describing the properties of the surface rocks in the catchment.

14

Figure 2-6: The first two lithological layers of the Sabie-Sand catchment represents the various rock types found at the surface of the catchment. These rocks will predominantly weather to form the catchment soils that contribute to surface water characteristics.

15 CHAPTER 3 LITERATURE REVIEW Although land use and land cover (LULC) are two different concepts (as seen in definitions table) often the interconnected nature between the two entities compel studies and researchers to use these two words interchangeably. The FAO (Di Gregorio and Jansen, 2000) defines Land cover as “the observed (bio)physical cover on the earth's surface” whereas land use refers to “the arrangements, activities and inputs people undertake in a certain land cover type to produce, change or maintain it”. Although land use is greatly determined by the type of land cover, in my opinion the latter is also dependent on land use. As land use is such a dynamic entity, it is difficult in a study to refer only to land use as predictor variable for example describing water quality influence and is therefore mostly based on broad land cover classes. Since the early 90's a greater consideration was given to development and sustainable use of the natural environment, therefore a continuous deliberation and review is and should be applied to land use practices.

Land use effect on river quality (brief South African context)

Studies that have examined land cover and influence on water quality in the area are somewhat directed to a vegetation viewpoint. They are not as strongly associated with the land use effect on water quality as would be desired for this dissertation (for example Bredenkamp, 1991). A meagre amount of published South African studies that analyse the influence of land use on water quality exist in comparison to the abundance of international publications on the subject. It also seems as if the South African studies on the subject have only in recent years started being published in readily accessible journals and resource spaces as they generally range from 2013 until this year. Three published examples of water quality analysis with land use influence on South African catchments are considered.

Changes in land use activities in the Olifants River catchment was analysed by Dabrowski and De Klerk (2013) to identify and link nutrient and metal concentrations to spatial trends in the catchment. Their approach to analysing the land use was firstly to section out the monitoring sites using ArcGIS and thereafter they quantified the upstream land use activities for each point using the National Land Cover dataset for South Africa. Apart from routine water sampling, they also conducted once-off sampling downstream of abandoned and current mining; agriculture and wastewater treatment works to assess those possible effects on water quality. Their multivariate analysis assessed the associations between the median measurements of water quality variables and the different sites that were representatives of the different land use activities. By conducting a Principal Component Analysis (PCA), they ordinated their water quality variables with CANOCO software. The reason for using a PCA approach was that a linear relationship between sites and water quality existed when the length of the gradient in the data set was less than 4 standard deviations. Some of their findings were high or significant nutrient loads such as ortho-phosphate from WWTP and sulphate increases or AMD from mining. 16 Trends in land cover change and the influence thereof on the water quality of the Blesbok Spruit catchment for 1994-2009 was analysed by Du Plessis, Harmse and Ahmed (2014). The National Land Cover data set was used and reclassified into 5 broad land use classes. Within this study, the catchment's land cover data for 2000, 2005 and 2009 were used in the Partial Least Squares (PLS) correlation and regression analysis and that of 1994 was used as reference conditions. This study found that a change in land cover resulted in various negative influences on water quality parameters measured. The that the conclusion of this study, was that the most recent negative effects of land cover change on water quality characteristics of the catchment is associated with an increase in economic growth. Increasing urbanization, industrial and agricultural growth accompanied by mining and sewage effluent resulted in the highest negative impact on the catchment's water resources.

Physical chemical and microbiological parameters of the Umhlatuzana, Umbilo and Amanzimnyama River catchments were analysed in relation to land use change and spatiotemporal characterization by Moodley, Pillay and Pather (2015). They selected sampling sites to lie at the interface of each land use type along the three river systems. These land uses were identified via topographical maps, aerial photographs and ground truthing from site visits. Although their results concluded that pollution associated with catchment activities was the main factor governing water quality, they did not see apparent trends in water quality based on specific land use patterns which linked sites across different catchments. Data analysis was conducted from a spatiotemporal dissimilarity viewpoint though PCA with varimax rotation using XLSTAT 2014 software. This study concluded that the intensification of anthropogenic activities such as industries, WWTW and urbanization over the past few decades have caused a deterioration of microbiological, heavy metal and nutrient water quality across all land use types on the water quality variables that were analysed.

It can be expected that river water quality will decline as land accessibility improves. River health is influenced by the fact that in low lying areas, urban boundaries expand, and land is more readily accessible for development through agriculture, plantations forestry and pastoral development (Larned et al., 2004). River health in a study by Larned et al. (2004) followed a downward trend within developed –compared to undeveloped– catchments. This study especially assessed how river health is negatively impacted in these developed catchments with low-elevations and they confirmed the link between lower elevations, increasing development and as a result degradation of river health. A study by Yu et al. (2016) also touches this concept but with slightly different results. They analysed land use influence on water quality within a seasonal as well as a topographic context. Their main objective was to establish if water quality can be improved with sufficient land use management which was also confirmed by Ding et al. (2016). Yu et al. (2016) confirmed that it can. They concluded that in the wet season the land use – water quality

17 correlation had a high significance. In the dry season the land use – water quality – and quantity correlation, had an even higher significant correlation across a broader range of land uses. Furthermore, they concluded that average slope and proximity did have an impact and therefore mention that steeper land generally had a stronger influence on water quality (Yu et al., 2016).

A definite statement can therefore be made that land use and geographical land characteristics (e.g. slope) play a very important role in general river ecosystem health. Water quantities and qualities vary greatly when considering the inflow and abstraction of different natural and anthropogenic activities. Studies conducted by Mallory et al. (2013); Pollard et al. (2011) and O'Keeffe et al. (1996) provided information on water supply schemes to cater for future developing land use activities and an increasing population. They also mention the possible outcomes that the construction of the Inyaka Dam (also called Injaka Dam) had on the Sabie-Sand catchment. Unfortunately, there has not been a full water quality assessment of the Sabie and Marite rivers after construction of the dam. Only the DWAF (2004) study that states that there is considerable assimilative capacity available for the Sabie River to maintain its current good state.

Taking the initial flow assessment of the pre-impoundment study for Inyaka Dam as an example: The study conducted by O'Keeffe et al. (1996) assessed water demand upstream of the Kruger National Park (KNP) as well as ecological river flow requirements of the Sabie River Catchment (Pollard et al., 2011). In a summary of the environmental effects of the Inyaka Dam, they concluded that the dam would at times have flow consequences for the Marite River. They also mentioned that the KNP lies far enough downstream within the Sabie River after the dam to enable full recovery of flow conditions. This statement was supported by including that the dam was large enough to support the KNP with compensation flows as the dam is in a tributary of the system it will not intercept high flows (O'Keeffe et al., 1996). After construction of the Inyaka Dam 2001 and flow assessment by Mallory et al. (2013) it was clear that water resource management within the catchment was majorly influenced by the KNP and surrounding privately owned game parks. True to the initial assessment by O'Keeffe et al. in 1996 the Inyaka Dam is managing to ensure adequate river flows to the KNP as well as water supply for domestic use in the Sabie- Sand catchments (DWAF, 2004). This is occurring even though the detailed operating rules set by the then Department of Water Affairs and Forestation to ensure the ecological Reserve as well as other water requirements within the catchment would be met in order for the catchment not to become stressed was never implemented (Mallory et al., 2013). Mallory et al. (2013) suggested that this might have been reasoned that at the time the Sabie catchment (exclusively) was not unduly stressed.

18 Importance of river health and remedial assistance given through policies (Ecological Reserve and Target Water Quality Study)

Water has distinct purposes from a human perspective. The most important purpose includes it being a prerequisite for life; and the other carries economic importance (Roux, 1999 and Postel, 2000). The trouble we have as water users, is to balance economic and social growth in terms of agriculture, industries, dam and borehole constructions, channelization etc. with ecological well- being (Belcher, 2004 and Postel, 2000). Catering for this difficult task is the manner in which the National Water Act (36 of 1998) is structured to “ensure that South Africa’s water resources are protected, used, developed, conserved, managed and controlled in a sustainable and equitable manner, for the benefit of all persons” (South Africa, 2016). In other words, ensuring that the environment is protected without excessively curtailing the country’s economic and social development (South Africa, 2016). This reminds us, as noted earlier, that healthy rivers provide services to healthy ecosystems, (Brauman et al., 2007). In order for ecosystems to be healthy, the rivers or water resources sustaining them should be too. Even the NWA considers ecosystems essential to the hydrological cycle – incorporating and driving water resources.

What is then defined as a good and poor-quality water resource? A study conducted by Boulton (1999) intensely analysed and debated this question. He mentioned that the value or “health” of a river to humans depend on the use they have for that water resource. When some people view it as a means to fish or as recreational and aesthetic places, or even use it as conduits for pollutants they will assign different values from those viewing rivers as a source of clean water for drinking and washing, or for agricultural, industrial and Reserve purposes (Boulton, 1999). This means that a river at a certain level of quality or “health” can satisfy some of the mentioned needs, and a river at another level of “health” (for instance one with poor quality water) can only satisfy another set of needs. This concept with reference to an aquatic viewpoint is described by Dalas and Day (2004) as something that needs to be approached with extreme caution, as improvement of water quality from a human’s perspective, might not resemble improvement from an ecological perspective. Therefore, improving the water quality for human use risks making the water quality less tolerable for natural inhabitants (Dalas and Day, 2004).

As an example of international water policies that influence water quality is the United States Clean Water Act. The Clean Water Act (Karr, 1999), amended from the 1972 US Water Pollution Control Act Amendments, describes river health. In its section 101(a) the statement resulted in a set of achievements or standards that has to be reached in order for rivers to reside in a state of good quality in the United States. Karr (1999) cited the Act, and its river health objectives resulted in: “a restoration and maintenance of the physical, chemical and biological integrity of the Nation’s waters. This integrity spoken of, implies a condition of an ecosystem in which the natural structure and function of the ecosystem is maintained”. Karr (1999) further mentions that baseline 19 ecological integrity is dependent on geographical changes of the river, as the river’s biota change with regards to regional and local constraints. This change in geographical extent (e.g. changes in adjacent landscapes) alters biota and should always be taken into account when ecological integrity is measured as a result of human interaction; as well as when remediation is implemented to counteract this anthropogenic interaction (Karr, 1999). This concept will usually alter the answer to the question asking if a certain level of change is acceptable or if a river is healthy. River health associated with water quality is assessed by using certain components or water parameters described as: turbidity and suspended solids, toxic substances, temperature, heavy metals, dissolved oxygen, organic and inorganic chemicals, conductivity, total dissolved solids, major ions, acidity (pH), alkalinity and salinity (Karr and Chu, 1997 and Dalas and Day, 2004). Ensuring that these parameters remain within appropriate limits can safeguard a type of ecosystem protection.

The National Water Resource Strategy (NWRS) assists the South African Department of Water and Sanitation (DWS) in decisions on the health of water resources and usable regulations pertaining to this preferred health. The NWRS was drawn up by incorporating various documents pertaining to the protection of South African water resources. These documents included: policies, legislations, guidelines and frameworks on the topics of water resources planning, resource protection, water use, water quality management, water conservation and water demand management, water pricing, institutional arrangements, monitoring and information, environmental management and lastly international water resources management (NWRS, 2004). Below, some of the concepts in these documents will be discussed where they are considered more applicable to this dissertation. An important resource protection concept that the NWRS uses to regulate river health, when complying with the NWA, is the Reserve (NWRS, 2004).

The general consensus is drawn to a point that the river should be healthy enough to sustain the Reserve. In the NWA, there are two main parts when referring to this Reserve:

Basic Human Needs Reserve (BHNR) and Ecological Reserve (ER). The Reserve refers to the quantity and quality of water in the resource required to:

“(a) satisfy basic human needs – potable water, food preparation and personal hygiene;

(b) protect aquatic ecosystems in order to secure ecologically sustainable development and use of the relevant water resource

The ecological Reserve is a function of the natural flow and it refers to the modified EWR (Environmental Water Requirements) – water required for aquatic ecosystem protection of the water resource – where operational limitations and stakeholder considerations are also taken into 20 account (Pollard et al., 2011). It also depends on the resource class and once a water resource’s Reserve is determined, it is just as binding as the class and the resource quality objectives (RQO’s) (see definitions)” (South Africa, 2016). As a means to water quality management, South Africa has a set of eight water quality guidelines that act as criteria for specific water constituents. These guidelines also include background information about the effect of these constituents on the user (e.g. domestic, agriculture etc.) in certain concentration ranges and the user can thus assess whether the water is suitable for use (DWAF, 1996a). One document is a field guide, while the remaining 7 guidelines have been drawn up for the following uses: Domestic, Recreational, Industrial, Aquatic Ecosystems, Mariculture, land Agricultural: Irrigation, Livestock Watering and Aquaculture. The results obtained through sampling will be compared to these water quality guidelines (see chapter 4). The River Eco-status Monitoring Programme (REHP) or the National Aquatic Ecosystems Health Monitoring Programme is a very inclusive program considering all aspects of water resources and the ecological state of important and frequently overused South African river ecosystems by the continuous assessment and monitoring thereof (Strydom et al., 2006). According to DWAF (2006b) the REMP measures the water resource’s quality through assessing its health or also known as the condition of the river. The motivation behind the REMP is “to insure the ecologically sound management of the country’s rivers”. The guiding documents of this program perfectly outlines the need to assess water resource quality and the value of South Africa’s rivers.

Climate change effects on river heath in general and the South African context

It is getting very hard to escape the reality of climate change. For decades the shift in climate have been affecting our everyday lives as the slower paradigm shift of humans have had an increasing effect on earth’s once over-abundant natural resources. The consensus drawn up by Wilbanks et al. (2007) is that the effect of climate change will not only be experienced as the rise in sea levels but will especially be felt as increases in weather extremes such as floods and droughts. The availability and demand of water resources are predicted to take the greatest strain as a result of climate change (Matondo et al., 2005 and Wetherald and Manabe, 2002). Wetherald and Manabe (2002) is of opinion that a warmer climate as experienced in recent years will increase the risk of floods and droughts at various degrees. Floods occurring as flash floods, river floods, sewer floods and urban floods are dependent on the volume, intensity and timing of precipitation; preceding conditions of rivers and drainage basins (Kundzewicz et al., 2007). Different types of droughts can also be experienced: when precipitation is well below average it is known as meteorological drought; if soil moisture is low, it is known as agricultural drought; when surface and groundwater levels are low, it is known as hydrological drought (Kundzewicz et al., 2007). A combination of all these is an ecological (environmental) drought. If South Africa

21 experience this type of drought in the past few years form 2013 up until the 2016/7 rainy season is unknown as of yet and it would take a very intensive study to determine this (Savage, 2016 as cited by Africa Check, 2016).

A review study by Kusangaya et al. (2013) explained the various approaches to climate change and the modelling of climate change effects on South African environments, focusing on temperature and rainfall. From the study, they concluded that climate change in the long-term will likely affect various factors contributing to human welfare, aspects such as: municipal and industrial water supply; flood control; health; agriculture; energy use; ecosystems and biodiversity such as wildlife management especially in areas of scarce water resources (Kusangaya et al., 2013). A heterogeneous experience will be evident for different parts of South Africa as a result of either a wetter or dryer climate. Mid-and high latitudes are reportedly going to receive higher precipitation intensities and subtropical and mid-latitude continental interiors, an increase in droughts especially in summer time (Meehl et al., 2007). As southern Africa is partly low to mid - latitude at 22° to 19°S the effects will be varying. General climate consensus predicts the east of Southern Africa to experience higher average precipitation events compared to the western parts. This will result in more frequent flooding events, but as the eastern part of this study area is considered as a drought prone area, a dryer climate will be evident in these parts of the catchment and even wetter conditions will still be experienced in the western parts (towards Sabie town), as this is closely related to the topographically downward gradient changes experienced from west to east. The water quality, erosion and sediment transport, are some of the broader influences analysed by Kundzewicz et al. (2007) that will be drastically altered from now into the future as a result of climate change. Human health, ecosystems and water available for use is at risk of deterioration as a result of water quality degradation due to higher temperatures and changes in runoff patterns (Hurd et al., 2004; O’Reilly et al., 2003 and Patz, 2001). Considering the importance of water quality to this study, as summarized by Kundzewic et al. (2007) the following will only be some of the influences from climate change on water quality due to floods and droughts:

 Higher rainfall and intense rainfall events will increase suspended solids (turbidity) and pollutants (pesticides, heavy metals, organic matter, etc.) washed from soils into reservoirs and lakes as a result of fluvial erosion (Boorman, 2003; Bouraoui et al., 2004; Fisher, 2000; Harker et al., 2004 and Neff et al., 2000).  Water quality will be negatively altered and possibly rendered unusable unless treated properly states Hamilton et al. (2001) (as cited by Kundzewicz et al., 2007) and Harker et al. (2004) when an increase in nutrients and sediments from runoff is experienced during lower water levels.

22  Soil and Water Conservation Society (2003) (as cited by Kundzewicz et al., 2007) is of opinion that when an increase in runoff takes place during fertiliser and pesticide application times in lower vegetative periods, these agricultural substances will be mobilised into water bodies.  Other pollution increases that Hall et al. (2002); Rose et al. (2001) and Rose et al. (2000) is concerned with during these extreme rainfall events, are the expected rise in water- borne diseases.  Lipp et al. (2001) also mentions the possibility of overwhelmed wastewater treatment plants during wet weather conditions that may cause higher faecal loads and pathogens in waterways as a result from direct surface runoff and saturated soils and therefore decrease in septic system drainfield capability.  Ferrier and Edwards (2002) mentions three major drivers that can be identified with analyses of changes in emissions and their connected hydro-chemical responses. They express concern towards the increase in acidification of lakes and rivers through atmospheric deposition, global change (e.g. altered temperatures) and land uses (Ferrier and Edwards, 2002 and Soulsby et al., 2002).

If dryer and hotter conditions occur, then surface water temperatures will rise. This will in time promote algal blooms (even with enhanced phosphorous removal in WWTPs), increased fungal and bacterial content and enhance transference of volatile and semi-volatile compounds to the atmosphere from surface water bodies such as pesticides, dioxins, ammonia, mercury etc. (Chambers et al., 2001; Harding and Paxton, 2001; Kumagai et al., 2002; Schindler, 2001; Van Ginkel et al., 2000 and Wade et al., 2002). Robarts et al. (2005) (as cited by Kundzewicz et al., 2007) as well as Moulton and Cuthbert (2000) are of opinion that this increase in fungi and bacteria can possibly lead to bad tastes and odours and a resultant increase in chemicals for purification, in chlorinated drinking water. The deterioration in water quality will, in drought prone areas increase the frequency to water-related diseases for example diarrhoea (Patz, 2001 and Harker et al., 2004).

Jiménez (2003); Lipp et al. (2001); and Maya et al. (2003) is of opinion that the lack of sanitation and proper potabilisation methods as well as poor health conditions in developing countries will also impair the biological quality of river water (WHO, 2004). Atkinson et al. (1999) is of opinion that lower river- and lake water levels might re-suspend bottom sediments and liberate compounds that will contaminate water supplies. Coupled with a risk of increased salinization in estuaries and inland reaches as a result of decreased streamflow (Beare and Heaney, 2002). Not only does the lower river water levels create a salinization risk but disturbance of humans to the natural salt cycle can cause an additional secondary salinization of water resources. Haron and Dragovich (2010) mentions that this occurrence of secondary salinization (also referred to as

23 dryland salinity) accounts especially for practices such as increased irrigations, impoundments and water diversions that humans use to relieve drought and an increasingly dry climate. Closer to sea-level and coastal areas, a possibility exists, according to Bobba et al. (2000) and Chen et al. (2004) that groundwater salinization will increase due to sea water intrusion in arid and semi- arid areas especially those with low surface runoff.

The Southern African context of these effects take a more drastic toll on the livelihoods of its citizens compared to those from developed nations. Being a developing country, it possesses a lower adaptive capacity and risks to climate change are intensified due to this fact, as well as widespread poverty and low technological uptake compared to developed countries (Callaway, 2004). Even if developing countries such as in Southern Africa are more prepared in the future, interaction between land use and climate is still inevitable.

Not only does climate change effect different land-use practices, but studies have shown socio- economic changes due to droughts arising from this interaction of human and natural conditions. These changes can emerge in the form of land use and change in land cover. This in turn can influence certain climate and air quality factors, as well as increase water use and demands (Kundzewicz et al., 2007). As Foley et al. (2005) mentions, the net radiation of an area can be altered via surface clearing thus affecting albedo; as well as dividing precipitation into different transportation pathways, forming part of soil water, evapotranspiration and runoff. As mentioned excessive water withdrawals intensifies the impact of droughts states Kundzewicz et al. (2007). Urban heat islands can store heat; lower the cooling effect brought about by evaporation and warm surface air as a result of impervious surfaces; reduced vegetative cover and the morphology of buildings in the city landscapes (Bonan, 2002). Biomass burning, vehicle emissions and other pollution sources are some of the land-use practices causing a direct effect –through altering emissions– on air quality.

The important part of conducting studies such as this one is being able to use this study’s outcomes to assist in the preparation and set-up of management guidelines for the next possible extreme drought and associated risks within this study area. Therefore, one gap that was identified was a lack of knowledge on the Sabie River system’s response to drought and the resulting effects it has on the water users and ecosystems dependent and influenced by the river system as well as overland and groundwater flow. The importance was realised in analysing this study area with regards of a dryer rainfall year, as this sets the stage to 1) future stakeholders; 2) mitigation management strategies and policies; and 3) water users to know what to expect and remediate future droughts and the effects of such periods of rainfall uncertainty. If the demand from the Bushbuckridge treatment plant increases and the upstream conservation demands that hold threat if the ecological Reserve is not met is implemented, the Sabie - Sand river catchment will soon be regarded as a water stressed catchment especially to the east where the climate is 24 semi-arid (Mallory et al., 2013). Current climate situations regarding this study area includes that 2015 was a very dry year with minimal precipitation. Minor floods were only experienced in March and higher rainfall in October 2016. A nationwide drought was experienced from 2013 and effective rainfall was only experienced in the summer of 2016/7. By conducting water health assessments in terms of physical, chemical and land utilization, river qualities can be determined within such dry or low flow circumstances also called baseflow, therefore greatly contributing to this field of study in terms of river health and usage.

Population increases and socio-economic drivers

When considering the direct and indirect benefits of a river and its associated ecosystem one starts considering functions such as: water for potable and domestic use; industrial cooling via water; industries and production; plants having medicinal properties; fishing and firewood all being considered as direct or primary uses. Whereas typical indirect uses of ecosystems can for example be through groundwater recharge; waste disposal; eco-tourism and cultural rituals to name a few (Boulton, 1999). Strydom et al. (2006) as well as Boulton (1999) mentions that with decline in river health, these benefits obtained from rivers decline as well, rendering water resources and associated ecosystems useless for social and economic benefits.

Strydom et al. (2006) made a clear understanding of the reasons why we should protect aquatic resources. As mentioned above it is a prerequisite for humans and animals and carries economic benefit (Postel, 2000; Roux, 1999 and Strydom et al., 2006). The fact that the volume of our water resources is limited motivates South Africa to incorporate management and efficient legislation in protection of this commodity. The natural assimilative capacity (see definitions) is the ability of the water resource to “cleanse” itself of waste and toxins. Without protection against anthropogenic interactions, this natural assimilative attribute will decline with incline in negative changes in river systems (Stydom et al., 2006).

It is therefore very important to assess and assist in mitigating these changes as far as possible, before prolonged effect alters river systems completely, rendering them useless for human and ecosystem functions. Protection strategies enhances knowledge and vice versa. The importance thus lies in understanding the interactions and effect thereof as well as the efficient management between water resources; land uses and aquatic ecosystem health. Strydom et al. (2006) mentions that we therefore measure each aquatic ecosystem component to assess overall health of these resources and rely on this information for important policy and decision making, such as estimating RQO’s. Supporting freshwater ecosystems and their functions, whilst simultaneously catering to human water demands is still an important 21st century challenge that was predicted by Postel in 2000.

25 GIS approaches and statistical analysis methods to assess land use influence on water quality

Various software packages exist that assist in surface water runoff modelling especially in South Africa where water related studies such as hydrological flow responses to land use change have been conducted (Baishya, 2006; Schütte and Schulze, 2017 and Warburton et al., 2012). The use of these hydrological models such as the Storm Water Management Model (SWMM) was considered. Hydraulic modelling was conducted in the catchment before by Heritage et al. (2004) which analysed the morphological influence of extreme flood events in the Sabie River, using the Agricultural Catchments Research Unit (ACRU) hydrologic model (Heritage et al., 2004). Although an increasing need for an update on this type of surface runoff analysis and a hydrological model analysis of the inflows and outflows of the catchment is needed, it was decided that they were beyond the scope and time-frame of this study (Li et al., 2016).

Geographic Information Systems (GIS) are of the most widely used computer-based tools amongst others, for the analysis – capturing, manipulating, managing, displaying, assembling and storing – of geographic and related information that have been geographically referenced (Dempsey, 2017 and UW-Madison, 2018). This includes information such as: topography, changes in surface characteristics, hydrology etc. ArcGIS is a geospatial processing program created by Environmental Systems Research Institute (Esri) for the processing of GIS data (Esri, 2015). As explained in section 3.1.1 many international studies exists that analysed LULC, the change in LULC and the change’s influence on water quality (Baker, 2003 and Vrebos, 2017). Many land use – water quality analysis studies referred to during this study, made use of GIS and satellite derived data as part of the methodology and tools to conduct a spatial analysis. Frequently these countries had readily accessible LULC datasets, SPOT as well as Landsat imagery available for use (if not obtained from ancillary data viewer such as Earth Explorer) retrievable from their National Environmental Department (Table 3-1). Not only is this an effective way of obtaining LULC data but majority of these national datasets were recently updated and only needed small enhancements before use. Considering some of the most recent documents, the following was written just as a brief summary of the different methods used in some cases and their final conclusions. For a more descriptive summary of these studies and the methods used, refer to Table 3-1.

A study analysing land use impact on water quality was conducted in the Kleine Nete catchment in Belgium (Vrebos et al., 2017). Using ArcGIS9.3 for any GIS calculations the 49 class Land use vector maps were converted to a 1 m-raster and the categories aggregated to 8 different classes namely: greenhouses, water, buildings, woodlands, pasture, paved area, cropland and others. They also used digital soil maps that contributed to their spatial predictors of land use influences. Areas upstream a sampling point were used to classify a sub-catchment and were delineated 26 from a 1:5000 DEM expressed as a 5 m-raster using a D8-runoff model. These sub-catchments were the basis for extracting upstream land use, soil texture and soil drainage acreages. They obtained maps that indicated which buildings in the urban areas were connected to a WWTP and reconnected them to the sub-catchments which received their treated wastewater (Vrebos et al., 2017).

Studies that investigated land use influence on water quality with regards to topographical changes or spatial characteristics have confirmed the response of water degradation to an increase in land use surface surrounding the river as well as proximity of land use to the river. Zhao et al. (2015) analysed the land use effect on water quality regarding a scale factor in a reticular river network in Shanghai, China. They identified that industrial land had a greater negative effect on water quality on a small scale than on a large scale and that the influence of urban land use on water quality was more intense on a larger spatial scale. This study used ArcGIS9.3 software to interpret 2.5m resolution pan-sharpened natural colour SPOT-5 imagery acquired during 2006. Using buffer zones as hydrological units they drew 100m, 200m, 400m, 800m and 1500m buffers around a sampling point to analyse the scale effect on water quality. Zhao et al. (2015) is of opinion that the riparian zone has the most direct impact on river water quality at 100m scale. After conducting a Pearson's correlation and redundancy analysis this study concluded that the larger the spatial scale of land use influence, the greater the negative effect on the water quality and more diverse in character the water quality became. This was also true for a study conducted by Yu et al. (2016) as mentioned above, where a 30m x 30m resolution DEM was used to obtain 44 sub-basins (catchments) as well as four different categories of slope: I (0-5ᵒ), II (5-15ᵒ), III (15-30ᵒ) and IV (30-68ᵒ). These slopes as well as land use distance to stream were correlated with water quality and found to have a definite influence thereon.

When considering the literature examples of studies that deployed GIS in combination with water quality influences, there are many international examples each with their own variations of methodologies. Seven examples are summarised in Table 3-1, but numerous other studies exist within international literature. These examples gave the most recent insight during the decision making of similar methods to use in this study. Table 3-1 highlights the type of data that they had at their disposal for use in predominantly GIS based approaches and what type of methods they used to be able to compare it in a multivariate analysis. To view a similar comparison table of older studies (2006-2015), please refer to Yira et al. (2016) which analysed different methods of studies but within tropical regions. After considering these approaches, it was decided within this specific study area to use a watershed approach, delineated from an SRTM90 DEM in correspondence to each sampling point locality. Considering water quality as response (species) values, land use, slope, soil, geology, land tenure and distance from wastewater treatment plant was used as predictor (environmental) variables during multivariate analysis.

27 Table 3-1: Literature examples of implementing GIS to detect land use influence on water quality. Selected relevant studies of land use and land cover influence on surface water quality with respect to the use of mainly multivariate statistical analysis. Reference Study Scale/ Land Use Analysis Additional Spatial data Type of Statistical Analysis Key Results Study Location

1 – Vrebos Kleine Nete Analysed scaled dependent land Texture and drainage soil 25 WQ parameters were statistically WQ parameters for winter differed largely (2017). catchment, use impact on water quality (WQ). properties were compared by using Tukey HSD (honest from the spring, summer and autumn Belgium. Area ~ 49 category, high accuracy land calculated from a 1: 20 significant difference) between the 4 results (highly variable between and 780km2. use maps were converted to raster 000 soil raster map. seasons, for 73 sample localities. within periods). WQ significantly layers and aggregated to 8 influenced by land use both summer Thereafter, RDA was used to determine different land use classes: (combined) and winter. significant relationships between land uses 73 sub catchment buildings, cropland, greenhouses, and WQ parameters of 2 seasons: winter delineation from pasture, paved area, water, and combined summer, spring and autumn 1:5 000 DEM woodland and others. parameters. corresponding to each sampling Further, spatial scales of the resulting site. predicted pattern, built from geographic coordinates, were determined by means of weighted distance matrix. Significance of these relationships were tested using a permutation test of 9999 iterations.

28 Reference Study Scale/ Land Use Analysis Additional Spatial data Type of Statistical Analysis Key Results Study Location

2 – Kändler et Nisa River in the Difference in WQ due to land use GIS layers created for 25 WQ parameters from 29 weekly sample 6 groups of sampling sites could be al. (2017). Czech-German- in upper transboundary catchment transboundary localities. Measured concentration values differentiated when cluster analyses Polish triangle. evaluation. Land use classification catchment: DEM, land below limit-of-detection assumed to be considered all measured parameters. Area ~ 694 km2. done according to biotope-types use, soil, stream network. zero. Values between limit-of-detection and These groups of sampling sites had (BTLNK Saxony 2012); analogical limit-of-quantification assumed to be mean similar chemical water composition in

land-use data mapping of these limits. each group and reflected the land use, 29 sub (ZABAGED); mapping of arable regardless of sub-catchment size. This Sub catchments were grouped into classes catchments land through Land Parcel indicated that land-use classes were according to land use and water chemistry. delineated from Identifications System (LIPIS) and good indicators of physio-chemical Hierarchical agglomerative cluster 10m resolution land cover data Coordination of parameters of the river water. analyses performed on these normalized DEM information on the environment data sets by means of the Ward method – RDA analysis of land-use classes and corresponding to (CORINE). Land-use categories using Euclidean distances as a measure of WQ groups confirmed that the portions of each sampling characterized by different weight similarity (threshold of 87.5% similarity). land use were significant predictors of site. of human impact: settlement- individual physio-chemical parameters (P affected, mixed land use, forested, Effects of land use on log-transformed, = 0.001). This was also true when WQ grassland-dominated and mainly centred and standardized physio-chemical groups were used as predictor. arable land. parameters were revealed by conducting RDA, followed by Monte Carlo permutation One-way ANOVA resulted in significant Land-cover categories: arable test (999 permutations). differences between the WQ groups. land, settlement or urban areas, and forests. One-way ANOVA: Tukey HSD test used for post hoc comparison to identify significant differences between groups of different hydro-chemical signatures.

29 3 – Ding et al. Dongjiang River in Diversity of land use patterns and The stream network and K–S test showed that WQ parameters were Important RDA outputs: (2016). Guangdong scale effects developed through Strahler derived stream not normally distributed. Therefore, 1) canonical axes explained the Province, China. empirical models. Multi-scale order, were based on nonparametric Kruskal–Wallis tests were %proportion of total variance of the WQ Catchment area approach to land use patterns that 30m DEM and 1:250 000 used to test differences of WQ between parameters; ~35 340 km2. were related to WQ of low-order topographic maps. geomorphic regions. Scale: reach, streams at different geomorphic 2) ordination diagrams (biplots), explicitly To avoid the effects of Multiple linear regression (MLR) modeling riparian and whole regions. reveal relationships between WQ land use on WQ masked and RDA were used to examine land use catchment. parameters and land use variables. Land use patterns were quantified by heterogeneous metrics WQ parameter relationships. MLR in terms of 3 main land use geomorphic conditions, modelling – to determine the relationships Land use metrics varied from reach, metrics through FRAGSTATS: low-order catchments between predictors (i.e., land use metrics) riparian and catchment scale. 56 sampling composition (%landscape); were divided into two and response (i.e., single WQ parameter). localities in a The different WQ parameters varied in configuration (patch density, groups of relatively variety of low- Only the significant predictors that were their response to changes in scale. landscape shape index, largest similar geomorphology: order streams. chosen from the final MLR models were Observed obvious spatial differences in patch index and aggregation group1 – 24 mountain Catchments used as the explanatory variables in the WQ among the low-order streams as index); and hydrological distance catchments (average upstream of sites RDA analyses. well as at different (flow length in m) of the land use slope and elevation were comparable types. greater than 500m and The RDA allowed for the simultaneous geomorphic regions (mountain vs plain). in size, with mean 16°); group 2 – 32 plain examination of the influences of the area = 100 km2 Land uses categorized into 7 catchments (average multiple land use variables on all of the WQ groups: cropland; forest including elevation and slope were parameters. A Monte Carlo permutation wooded areas and mixed forest less than 220 m and test (499 permutations) was used to areas; grassland; orchard; urban 12°). determine the statistical validity of the areas including residential, RDA. commercial and industrial lands; water bodies including rivers, reservoirs and ponds; and other lands, including barren- and other unused lands.

30 Reference Study Scale/ Land Use Analysis Additional Spatial data Type of Statistical Analysis Key Results Study Location

4 – Chen et Wen-Rui Tang Identified the impact of land use River flow direction 5 WQ variables of 52 sampling localities Spatial patterns such as urbanization al. (2016). River watershed, and population density (pollution derived from 5m were analysed. Kolmogorov-Smirnov test intensity within the watershed, is being China. Area ~ 740 sources) on surface WQ. 96 land- resolution DEM data. determined that the 5 WQ variables were clearly reflected by the WQ indicators in km2. use categories were retrieved via not normally distributed and had to be the different clusters. Population density 0.5m resolution aerial imagery and logarithmic transformed. obtained from census Significant seasonal (temporal) variations aggregated to 7 broader land-use data. Each location’s Ordinary least squares (OLS) and in WQ variables are experienced in the 7 drainage basins categories: agriculture; population density was geographically weighted regression (GWR) river due to land use. were derived from commercial (commercial, calculated: each models were developed to explore DEM. 201 sub- administrative, cultural The largest contributors that influenced administrative sub district relationship between WQ indicators and catchments all entertainment, municipal utility WQ parameters varied with time and received a centroid to land use. larger than 100 lands); industrial and mining; space. which corresponding 000 m2 resulted residential; transportation; Standard regression coefficient, coupled population data was Multicollinearity effects, were resolved after catchments vegetated land (forest, grassland with cluster analysis were used to assigned; thereafter, through the use of manual variable smaller than100 and urban green belts) and water. determine which principal factor (variable) centroid was interpolated excluding-selecting method. 000 m2 were A few unused land categories e.g. had the greatest influence on WQ. Cluster into a raster (by using the merged. gravel, bare rock, waste lands analysis was applied to ensure comparison Kernel algorithm) were aggregated with residential of only similar structured land-use representing the category due to similar runoff and variables. population density; erosion characteristics. finally, the cells of the population density raster were extracted as grid values.

31 Reference Study Scale/ Land Use Analysis Additional Spatial data Type of Statistical Analysis Key Results Study Location

5 – Yu et al. Wei River basin, Determined correlation between DEM was used to obtain Mean and standard deviation for WQ Slopes were correlated with land use and (2016). China. Area ~ land use types and stream WQ at the sub-basins as well as parameters in dry and wet seasons were WQ and found to have a strong 134.3 × 104 km2. the sub-basin scale during dry and 4 different categories of calculated for each sub-basin. correlation. rainy seasons. slopes: I (0-5ᵒ); II (5-15ᵒ); The log-transformed WQ parameters were Spatially the slope coefficients at riparian III (15-30ᵒ) and IV (30- used for a one-way analysis of variance zones were weaker than at sub-basin 44 sub-basins 68ᵒ). (ANOVA) test, to test the degree of scale. were obtained 6 Land use categories were used significant differences between the two through a 30m x for analysis: agriculture land Temporal variations: WQ variables and seasons. 30m resolution (paddy fields and dry land), different land use relationships were DEM. forestland (shrub land and sparse PCA was used to determine whether there weaker in rainy –than in dry seasons. woodlot), grassland (different was relationship between land use types coverage types), waterbody and stream WQ at the sub-basin scale (p < (rivers, wetlands and sandy 0.05). beaches), urban land (industrial Normal distribution for all variables were and residential areas), barren land tested by K-S test. (gravel, bare ground and bare rocks). Mean values were used for each season in statistical analysis.

32 Reference Study Scale/ Land Use Analysis Additional Spatial data Type of Statistical Analysis Key Results Study Location

6 – Zhao et al. Shanghai, China Land use effect on WQ regarding After conducting a Pearson's correlation, a Clear WQ distribution along urban-to- (2015). a scale factor in a reticular river Monte Carlo permutation redundancy rural gradient was identified. Area ~ 6340.5 network. analysis between land use and km2, with a water Pearson correlations showed industrial hydrological variables and WQ indicators area of 569.6 km2. WQ monitoring stations were set and urban land correlation with all WQ were done with CANOCO 4.5. as geographical centres. 48 buffer indicators and correlations increased

zones as hydrological units 100m, with increase in buffer size. 48 buffer zones as 200m, 400m, 800m and 1500m Riparian zone had most direct impact on hydrological units buffers were drawn around river WQ at 100m scale. from 2.5m sampling points to analyse the resolution, 2006 scale effect on WQ. Land uses Study concluded that the larger the SPOT-5 satellite were classified into 5 classes spatial scale of land use influence, the data. namely: agricultural, forest, grass greater the negative effect on the WQ area, industrial and urban. and more diverse the WQ became

33 Reference Study Scale/ Land Use Analysis Additional Spatial data Type of Statistical Analysis Key Results Study Location

7 – Haidary et Hiroshima, Japan. Effect of changes in land use Soil types and geological All WQ parameter data was tested for There were significant positive al. (2013). Area ~ 635 km2. composition on wetland WQ. formations; geometric normality using the Shapiro–Wilk test with associations between % proportion of features such as: a p-value of less than 0.05. urban areas within the watersheds of the 24 wetland The wetlands were categorized catchment area, average wetlands and the EC, TDS, TN, DON, watershed into three main analysis classes Thereafter, Spearman rank correlation test catchment slope (%), NH4, NO2 values. Negative associations boundaries based on the extent of urban area: was applied to determine if any of the WQ average main channel were observed between DO and % defined by 30 m non-disturbed, moderately- variables were associated with changes in slope (%), drainage urban area in the watershed of the DEM were disturbed and highly-disturbed percentage of land use and the other density and Strahler wetlands. Possibly because of an examined. wetlands. mentioned spatial data. order within catchment of increase in nutrient concentrations. the wetlands. Accordingly, if the %proportion of urban area increased in the catchment, annual mean of the WQ parameters, except DO, increased in the wetlands.

This related to landscape degradation resulting from transformation of forest areas into urban areas and to soil losses in urban areas due to improper storm water management, which can cause degradation of WQ due to soil erosion and sediment transportation.

34 CHAPTER 4 WATER QUALITY River water quality is modified by the activities and processes taking place within the water as well as the surrounding landscape in its catchment area. These activities and processes include geomorphic processes, riparian and vegetation changes, climate variation, as well as anthropogenic land uses (Brierley, 2010). It is important to attain knowledge of the natural resources around us. Hydrology, aquatic science and other disciplines related to water and the use thereof, already have a tremendous amount of expanding and improving background knowledge. The challenge is to implement and improve this knowledge to ensure the continuous sustainable use of water. By monitoring the variations of the water quality and water quantity sustainable management of water resources and control of pollution are possible (Chen et al., 2012). Through conducting both a water quality assessment, which Meybeck et al. (1996) refers to as the chemical, physical and biological evaluation of water, and water quality monitoring, which is the collection of relevant information, we are a step closer to achieving this sustainable use. When conducting proper water quality monitoring, assessment of the different spatial and temporal variations in water quality is essential for pollution control (Chen et al., 2012). The successful managing of all anthropogenic and natural water users through an integrated approach includes the joint decisions and actions of stakeholder participation to protect the water resource (Bartram and Helmer, 1996). With increased participation, stakeholders will be able to understand, interpret and use this information to support various water management activities. It is thus essential that reliable water quality assessment in a watershed is conducted (Behemel et al., 2016). The aim of this chapter is to assess the water quality of the Sabie, Sand and Marite rivers by measuring their physical, chemical and biological factors and to comparing it to the TWQR and RQO.

Methodology

4.1.1 Sampling

Sampling of chemical and physical parameters occurred seasonally and commenced in February 2016 and repeated in April; July; October of 2016. The 12 samplings sites as described in Table 2-1 are indicated in Figure 4-1.

35

Figure 4-1: Sampling sites locality map listing this study sites within Sabie-Sand catchment, used during this study.

A surface water grab sample of 9l was taken at each sampling site. The samples were collected using a bucket fixed to a rope to allow the collection of water from bridges. The bucket was rinsed at each site to avoid cross contamination between the sites. The sample was then transferred to the different allocated containers. The following physical-chemical parameters were determined in situ:

Table 4-1: Physical-chemical parameters measured in situ with a YSI 556 handheld field multimeter at each sampling site. Quality Variable Symbol Unit Barometric pressure mmHg mmHg Dissolved oxygen DO mg/l Specific conductivity (related to EC) SPC µS/cm Percentage dissolved oxygen %DO % pH pH Water temperature Temp °C

All chemical and biological analyses were carried out by Rand Water’s Analytical Services. The APHA (2013) standard methods were used and the laboratory is accredited according to South

36 African National Accreditation System – affiliated at ILAC (SANAS) (Table 4-2). Table 4-2 shows a summary of the chemical parameters, microbiological parameters and other hydro-biological parameters determined. It also shows the symbol of these parameters that will be used in this study and the reporting limit used according to SANAS.

Table 4-2: Summary of the physical-chemical variables measured by Rand Water’s Analytical Services as well as the method number listed by Rand Water, the unit and reporting limit. Method Number Quality Variable Symbol Unit Reporting Limit 2.2.2.02.10* 2-Methyl isoborneol MIB ng/l <0.5 2.1.4.03.1 (Feb, Apr, Jul) Aluminium Al3+ µg/l <25 2.1.4.01.1 (Oct) - 2.1.8.04.2 Ammonia NH3 mg/l as N <0.2 (Feb, Apr) <0.05 (Jul, Oct) 2.1.4.03.1 (Feb, Apr, Jul) Calcium Ca2+ mg/l <0.90 2.1.4.01.1 (Oct) 2.1.3.03.1 Chemical Oxygen COD mg/l <10 Demand 1.1.2.01.1 Chlorophyll-a Chl-a µg/l <2 2.2.1.01.2 Dissolved Organic DOC mg/l as C <0.2 Carbon 1.2.2.09.1 Escherichia coli E. coli MPN/100ml 0 2.1.4.03.1 (Feb, Apr, Jul) Iron Fe2+/3+ µg/l <5 2.1.4.01.1 (Oct) 2.2.2.02.10* Geosmin Geo ng/l <0.5

2.1.4.03.1 (Feb, Apr, Jul) Hardness Hard mg/l CaCO3 <5 2.1.4.01.1 (Oct) 2.1.4.03.1 (Feb, Apr, Jul) Potassium K+ mg/l <1.5 2.1.4.01.1 (Oct) 2.1.3.01.2 Methyl orange M Alk mg/l CaCO3 <5 Alkalinity (total alkalinity) 2.1.4.03.1 (Feb, Apr, Jul) Magnesium Mg2+ mg/l <1.5 2.1.4.01.1 (Oct) 2.1.4.03.1 (Feb, Apr, Jul) Manganese Mn2+ µg/l <10 2.1.4.01.1 (Oct) 2.1.4.03.1 (Feb, Apr, Jul) Sodium Na+ mg/l <2.0 2.1.4.01.1 (Oct) 2- 2.1.7.01.1 Nitrate NO3 mg/l as N <0.1 - 2.1.8.01.2 Nitrite NO2 mg/l as N <0.03 3- 2.1.8.03.2 Ortho phosphate PO4 mg/l <0.1 (Apr <0.2 (Feb, Jul, Oct) 2.1.4.03.1 (Feb, Apr, Jul) Phosphorus P3- mg/l <0.5 2.1.4.01.1 (Oct) 2.1.4.03.1 (Feb, Apr, Jul) Silica Si mg/l <1 2.1.4.01.1 (Oct) 2.1.7.01.1 Sulphate Sulp or mg/l <1 2- SO4 1.2.2.09.1 Total Coliforms Coli MPN/100ml 0 2.1.8.02.2 Total Kjeldahl TKN mg/l as N <1 Nitrogen 2.2.3.02.1 Total Organic TOC mg/l as C <0.2 Carbon 3- 2.1.8.06.2* Total Phosphate T PO4 mg/l <0.036 37 Method Number Quality Variable Symbol Unit Reporting Limit 2.1.4.03.1 (Feb, Apr, Jul) Total Silica Tsi mg/l <0.15 2.1.4.01.1 (Oct) 2.1.4.03.1(Feb, Apr, Jul) Zinc Zn2+ µg/l (Feb, Apr, <0.015 (Feb, Apr, 2.1.4.01.1 (Oct) Jul) Jul) mg/l (Oct) <0.15 (Oct) 2.1.2.02.1 Turbidity NTU <0.25 2.1.2.04.1 Total dissolved TDS mg/l <15 solids 2.1.2.05.1 Suspended Solids SS mg/l <15 2.1.7.01.1 Chloride Cl- mg/l <0.5 1.1.2.01.1 Phaeophytin-a Phaeo µg/l <2

2.1.8.05.2 (Feb, Apr, Jul, Silicon dioxide SiO2 mg/l <0.5 Oct) 1.1.2.09.1 (Feb, Oct, Apr) Microcystin Mic_n µg/l <0.36 2.2.2.10.1*(Apr) Total Solids TS mg/l <15 * Method not Accredited by SANAS

Ancillary Data Acquisition

4.1.2 Rainfall data

Monthly and daily rainfall data was obtained from DWS (2017) and the monthly data was used for final analysis (Figure 4-2). Only one of the rainfall stations available in the catchment was constantly updated by The Department of Water and Sanitation (DWS) in the same time-frame as this study – the rainfall station (X3E005) at Inyaka Dam. Since the catchment’s overall climate changes from west to east, it would have been better to have a variety of real-time rainfall data of the catchment’s sampling localities including stations from Sabie, Hazyview, Skukuza (X3E001 – last entry 2008/09) and closer to Mozambique, but unfortunately the data is either not updated for working stations, or historic as the station is no longer active.

Statistical Analysis

The water quality dataset consisted of biological, chemical and physical parameters. Results that were below the limit of detection were assigned the value of half the detection value, to be included in data processing. Missing data were treated as gaps and 0 was used where the variable was measured as zero. All statistical analyses were carried out using Statistica version 13, Dell Inc. (2015). Initially, the Kolomogorov-Smirnov and Lilliefors tests for normality were conducted to determine if the data were distributed parametrically. The data did not meet the assumptions of normality in the distribution of all variables and thus non-parametric statistics were applied. The Kruskal-Wallis analysis (comparison of multiple groups) was used to compare multiple independent groups and the Spearman Rank correlation to determine any correlations between the different variables (Table 8-1 Appendix B). The significance of the results of a Kruskal-Wallis ANOVA can be determined as a z-value and/or a p-value (Table 8-3 Appendix B).

38 Principal Component Analysis (PCA) which elucidates the components that caused the most variance in the water quality data (Dabrowski and De Klerk 2013); as well as their positive and negative correlations with their component variables were determined with Statistica. The first three components on the scree plot (Figure 4-8) described the highest influence in the dataset and therefore indicates the highest variance amongst the water quality concentrations.

Results

4.1.3 Rainfall Results

As discussed under the topography and climate section in Chapter 2, this study area tends to receive summer rainfall with an average of ~700mm/a which varies from west to east. Prior to, and during sampling though, drought conditions were experienced over the majority of South Africa, resulting in low flows also referred to as baseflow, in the catchment. As seen in Figure 4- 2 the rainfall range varies between 0 and 219.2mm/month and the average monthly rainfall for 2016 was 50.92mm. In March 2016 a sudden increase due to minor (flash) floods experienced in the area. An increase in precipitation can be noticed in October 2016 during the start of the new rainfall season.

Monthly Rainfall (mm) from October 2015 to October 2016 at Inyaka Dam monitoring station (X3E005)

Rainfall (mm) 219,2 225 200 175 150 125 105,4 100,2 100 75,2 57,4 75

Rainfall (mm) 28 24,2 50 17,4 10 12,2 11 25 1,8 0 0 Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sept Oct (2015) (2016) Month

Figure 4-2: Indicating the monthly rainfall from October 2015 to October 2016 at Inyaka Dam monitoring station (DWAF, 2017).

4.1.4 Water Quality Analysis

Table 4-3 lists the values and ranges of the water quality variables determined during this study period compared to 1) South African National Standards (SANS) 241:2015 drinking water

39 standards; 2) target water quality ranges (TWQR) as well as 3) resource quality objectives (RQOs) for all the sampling sites. The average value for each water quality parameter is listed as the first value in each cell followed by the range (minimum and maximum) values at the bottom. The mean values that exceeded the SANS 241:2015 limits are shaded in colour. The parameters exceeding the SANS 241:2015 drinking water standards’ aesthetic values are shaded  and those exceeding health risks are shaded.. The RQOs for the specific resource unit within which the sampling site falls, are indicated in purple beneath the value and ranges (South Africa, 2016). The target water quality ranges (TWQR) from the South African Water Quality Guidelines (DWAF, 1996a&b) for each variable are indicated in italics in the column next to the variable symbol. Results of the Spearman rank correlations and Kruskal Wallis ANOVAs are in Appendix B.

The average in situ pH measurements in the Sabie River ranged from 6.18 to 7.40 with an average of 6.87 for all the Sabie River sampling sites. There was no significant difference in the pH between any of the sampling sites located within the Sabie River. Site 1 of the Sabie River experienced the lowest ranges in pH of 5.90 – 6.55 in comparison to the other sites. The lower pH observed at site 1 is possibly a result of the forestry practices and decomposition of organic materials surrounding site 1. These decomposed materials may either reach the water by increased runoff due to forestry usually occurring on steep slopes, or by leaching through the soil and thereafter into the water. The sites with pH most similar to site 1 were located on the Marite River, namely sites 10 and 11 (with pH values of 6.53 and 6.49 respectively) which are also still strongly influenced by forestry practices as the inflow into the Inyaka Dam (site 11) is also surrounded by forest plantations, while site 10 is downstream of the dam.

There was no significant difference for DO concentrations between the different sampling sites. However, a significant negative correlation between pH and DO was observed. Only site 12, located in the Sand River within the urban township of Thulamahashe, exhibited an average concentration of DO below the TWQR. Even though low concentrations of DO are usually associated with agriculture, forest harvesting, sewage treatment plants, pulp mills etc. (RIC, 2017), the low average DO concentration (3.7mg/l) observed at site 12 can most probably be ascribed to the anthropogenic influence of sewage treatment plant effluent (RIC, 2017). DO also showed significant negative correlation with other water quality variables, namely Chl-a, Total

3+ + 2+ - Coliforms (coli), COD, Al , K , Na , NH3 , DOC and TOC.

40 Table 4-3: The average values and range (minimum and maximum) of the water quality parameters measured at sampling localities 1- 12 from February 2016 to October 2016. The results that have been shaded indicate that the SANS 241: 2015 limits have been exceeded indicating aesthetic  and health risks. . These values were also compared to RQOs and TWQR (DWAF, 1996a,b&c). Variable TWQR Sampling localities (mean value above and range (minimum-maximum) below) RQO underneath both 1 2 3 4 5 6 7 8 9 10 11 12 1.00 1.00 1.28 1.00 1.00 1.00 2.30 2.38 1.38 1.00 1.28 51.50 Phaeo 1-1.2 1-3.6 1-6.5 1-2.5 1-2.1 20-120 1.25 1.50 2.75 1.30 1.60 3.55 4.58 5.15 2.63 1.950 4.25 113.80 T pigm 1-2 1-3 2.4-3 1-2.2 1-3.4 1-5.5 1-11 2.2-13 1-4.9 1-3.1 2.1-7.4 27-290 2.35 2.92 3.15 1.98 1.80 3.45 3.75 2.88 1.78 1.83 3.88 62.20 Chl-a 1-4.4 1-5.4 1-6.2 1-3.6 0.5-4.7 1-5.3 1-7.2 1-6.9 1-2.8 1-2.8 2.5-7.5 6.7-170 0 - 5 counts/ 2065 1720 4670 4962 2779 143594 507 2100 1230 2360 1196 326 coli 100 ml 1789- 1187- 1203- 1450- 2430- 15531- 31-1145 727-3790 727-1664 866-3990 435-1956 99-554 Domestic 2420 1986 6867 6867 2987 241960 use 26425 225 35 70 356 208 84 387 274 125 2 1986- 4-548 3-67 9-93 63-866 32-365 4-236 241-613 185-461 68-231 0-2 98040 Meet Meet Meet Meet Meet Meet Meet Meet Meet Meet 0-130 Meet 10 TWQR of TWQR of TWQR of TWQR of TWQR of TWQR of TWQR of TWQR of TWQR of TWQR of E,coli counts/ TWQR of 0-17 0-130 0-130 0-130 0-130 0-130 0-130 0-130 0-130 0-130 0-130 100 ml 0-130 counts counts counts counts counts counts counts counts counts counts counts per 100 per 100 per 100 per 100 per 100 per 100 per 100 per 100 per 100 per 100 per 100 ml ml ml ml ml ml ml ml ml ml ml <5>55 0.705 1.03 1.045 2.60 2.80 2.79 3.35 2.25 7.15 7.48 9.95 3.70 Turb NTU 0.45-1.1 0.6-1.6 0.72-1.6 1.4-4.5 1.8-3.5 0.85-6.2 2-5.9 1.6-3.9 2.6-15 1.5-21 1.7-34 1.5-6.9 0 – 450 33.00 90.00 51.00 48.00 64.50 96.25 159.25 110.25 86.50 67.75 98.75 161.00 TDS mg/l 16-67 63-150 17-71 40-63 49-82 52-170 82-225 33-245 53-170 32-165 43-220 69-285 <10% of the back 11.38 11.38 9.88 9.88 10.63 11.13 14.75 11.63 10.63 24.00 10.40 SS 7.50 ground 7.5-23 7.5-23 7.5-17 7.5-17 7.5-20 7.5-22 7.5-26 7.5-24 7.5-20 7.50-56 7.5-19 <100 mg 14.50 59.50 56.50 45.25 55.00 61.75 66.25 56.00 60.50 24.00 21.50 105.30 M Alk NA 8.2-11 54-64 51-61 39-50 45-66 45-81 55-83 44-66 49-72 21-28 20-24 68-150 10.05 58.50 54.75 39.00 33.38 49.00 48.50 47.00 50.25 13.75 11.50 62.00 Hard NA. 8.2-11 51-63 51-59 31-48 7.5-47 36-59 44-58 35-59 37-63 10-18 11-13 46-86 6.0 – 9.0 6.18 7.38 7.23 6.93 6.97 6.82 7.24 7.11 6.74 6.53 6.49 6.9 pH pH units 5.9-6.55 6.53-8.16 6.6-8.2 6.6-7.37 6.58-7.25 6.55-7.1 6.59-7.58 6.59-7.7 6.57-6.9 5.9-7.47 5.96-7.25 6.56-7.1 41 Variable TWQR Sampling localities (mean value above and range (minimum-maximum) below) RQO underneath both 1 2 3 4 5 6 7 8 9 10 11 12 82.23 83.65 88.15 85.40 82.93 85.40 86.00 91.90 76.33 79.43 75.43 40.20 %DO >80% 76.2-93.4 74.1-94.2 72.3- 66-105.3 77.8-85 63.8-98.1 76-95 87.4-98.8 62.8-83.3 70.2-95.6 62.4-89.4 21.1-69 102.2 8.11 7.75 8.38 7.70 7.16 7.00 6.75 7.78 6.58 7.05 6.45 3.7 5-9 DO 7.1-9.83 6.62-9.5 7.23- 5.38- 5.88-8.53 4.27-9.12 6.25-7.62 6.7-9.53 4.71-8.4 5.81-9.66 5.31-8.2 1.54-7.2 mg/l 10.76 10.45 46.78 143.38 130.68 124.63 126.90 155.93 292.63 148.63 161.88 76.93 62.05 389.20 41.4-49.4 122-157 115-142 119.2- 71.1-159 148.1- 243.5- 130.7- 133.9- 58.6- 53.3-71.1 276.2-602 95th 95th 95th 132-6 95th 170.1 329.4 158.3 177.2 100.4 95th 95th 0-70 percentile percentile percentile 95th percentile 95th 95th 95th 95th 95th percentile percentile mS/m or of the of the of the percentile of the percentile percentile percentile percentile percentile of the of the 0-700 SPC data must data must data must of the data must of the of the of the of the of the data must data must µS/cm be ≤30 be ≤30 be ≤30 data must be ≤30 data must must be data must data must data must be ≤30 be ≤42 (taste mS/m mS/m mS/m be ≤30 mS/m be ≤30 ≤42 mS/m be ≤30 be ≤30 be ≤30 mS/m mS/m threshold) therefore (≤300 (≤300 mS/m (≤300 mS/m (≤420 mS/m mS/m mS/m (≤300 (≤420 (≤300 µS/cm) µS/cm) (≤300 µS/cm) (≤300 µS/cm) (≤300 (≤300 (≤300 µS/cm) µS/cm) µS/cm) µS/cm) µS/cm) µS/cm) µS/cm) µS/cm) 5.00 6.75 5.00 5.00 8.50 7.25 5.00 5.00 5.00 5.00 5.00 19.00 COD NA 5-12 5-19 5-14 12-30 0 – 5 µg/l 12.50 17.38 15.88 83.00 41.38 61.38 158.75 54.88 118.25 132.75 34.13 31.30 Aquatic 12.5-32 12.5-26 66-110 12-83 12.5-105 12.5-460 12.5-100 45-190 56-185 12.5-61 12.5-60 systems Al3+ 0-150 µg/l Domestic use 0 – 32 1.98 12.75 12.25 8.60 6.09 9.73 9.90 8.95 9.58 3.23 2.60 14.80 Ca2+ mg/l as 1.2-2.5 11-14 11-13 6.8-11 0.45-9.4 6.4-12 7.9-13 5.8-12 5.5-13 2.1-4.4 2.3-3.3 10-21 Ca 0 - 100 20.13 77.25 125.00 338.75 153.63 172.50 288.75 142.50 318.75 545.00 540.00 161.0 µg/l 2.5-34 65-89 105-140 220-395 2.5-285 135-245 105-680 115-195 270-400 280-640 200-1440 24-300 Fe2+/3+ Domestic use 0 – 50 0.75 0.75 0.75 0.75 0.75 1.01 1.99 0.75 1.78 0.75 0.75 3.70 mg/l 0.75-1.8 0.75-3.4 1.5-2 1.6-6.7 K+ Domestic use 0 – 30 0.94 22.70 7.00 5.08 4.99 7.13 6.80 7.13 7.58 1.35 0.63 7.20 mg/l 0.75-1.5 7.7-67 6.2-7.6 4.1-6 0.75-6.9 5.8-8.5 5.9-7.3 6-8.4 6.7-8.7 0.75-2.1 0.25-0.75 5.7-9.6 Mg2+ Domestic use

42 Variable TWQR Sampling localities (mean value above and range (minimum-maximum) below) RQO underneath both 1 2 3 4 5 6 7 8 9 10 11 12 0 – 180 5.00 10.50 6.50 19.50 6.75 13.00 16.50 5.000 9.50 10.00 117.75 11.00 µg/l 5-20 5-11 5-35 5-12 5-32 5-51 5-15 5-14 5-435 5-29 Aquatic

2+ systems Mn 0 – 50 µg/l Domestic use 0 – 100 1.00 2.78 3.88 5.65 6.1 15.15 27.65 9.25 10.45 7.40 5.80 44.00 mg/l 2.4-3.7 3.1-5.6 5.2-6.4 1-8.5 7.9-35 8.6-38 8.9-10 9.8-11 6.2-8.4 5.5-6.3 31-67 Na+ Domestic use <0.03 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.70 P3+ mg/l 0.25-1.3 <0.002 0.01 0.01 0.01 0.01 0.33 0.01 0.01 0.01 0.01 4.76 4.78 7.12 mg/l 0.01-1.3 0.01-19 0.01-16 0.01-15 Aquatic Zn2+ systems 0 - 3 mg/l Domestic use 4.25 4.78 4.53 4.93 3.88 5.80 7.33 4.65 5.03 6.50 5.05 7.40 Si NA 2.9-4.8 3.4-5.4 3.4-5.1 3.7-5.8 0.5-5.9 4.1-8.4 4.5-12 3.2-6.7 3.4-8.2 4.1-8.8 2.9-5.9 5.1-10 9.00 10.33 9.58 10.48 8.60 12.35 15.65 9.73 10.80 13.95 10.80 15.80 Tsi NA 6.2-10 7.3-12 7.3-11 7.9-12 2.1-13 8.8-18 9.6-26 6.8-14 7.3-18 8.8-19 6.2-13 11-21

0 – 100 0.81 1.85 1.98 3.43 12.40 13.90 24.10 6.78 8.28 3.31 2.64 24.20 mg/l 0.69-0.91 1.8-1.9 1.8-2.2 3.3-3.7 5.3-33 5.1-38 6.4-35 6-7.9 7.4-10 0.25-4.7 0.25-3.8 1.6-40 Cl- Domestic use 0 – 6 mg/l 0.05 0.31 0.35 0.16 0.12 0.13 0.05 0.11 0.20 0.15 0.05 0.10 2- as N 0.05-0.45 0.21-0.51 0.05-0.28 0.05-0.19 0.05-0.22 0.05-0.27 0.05-0.64 0.05-0.34 0.05-0.2 NO3 Domestic use 0 – 6 mg/l 0.015 0.015 0.015 0.015 0.015 0.015 0.015 0.015 0.02 0.02 0.02 0.5 - as N 0.02-0.03 0.02-0.03 0.02-1.5 NO2 Domestic use

43 Variable TWQR Sampling localities (mean value above and range (minimum-maximum) below) RQO underneath both 1 2 3 4 5 6 7 8 9 10 11 12

2- 0.07 0.33 0.37 0.18 0.13 0.143 0.065 0.12 0.22 0.17 0.07 0.60 NO3 - 0.65-0.47 0.23-0.53 0.07-0.3 0.065- 0.07-0.24 0.07-0.29 0.07-0.66 0.07- 0.07-1.6 +NO2 0.21 0.370 0 – 1.0 0.07 0.06 0.06 0.06 0.06 0.06 0.06 0.06 0.06 0.06 0.20 5.0 mg/l N 0.03-0.1 0.02-0.1 0.03-0.1 0.03-0.1 0.03-0.1 0.3-01 0.03-0.1 0.03-0.1 0.03-0.1 0.03-0.1 1.3-6.8 0.1-11 Domestic Use - NH3 And 0.007 mg/l Aquatic systems 1.70 1.65 1.65 1.90 2.25 4.13 2.18 2.58 4.50 3.68 3.65 13.10 TKN 0.5-4.6 0.5-4.1 0.5-3.9 0.5-4.5 1.6-3.7 1.6-10 1.2-3.3 1.8-3.8 1.8-8.2 1.6-5 1.3-6.8 1.5-31

1.83 2.04 2.08 2.14 2.45 4.33 2.30 2.76 4.79 3.91 3.91 18.70 TN 0.59-4.7 0.59-4.59 0.93-4.32 0.67-4.82 1.87-3.93 1.94- 1.29-3.39 1.89-3.89 1.89-8.96 1.69-5.4 1.47-7.24 3.15-38.4 10.23 0.09 0.09 0.09 0.09 0.09 0.09 0.09 0.09 0.09 0.09 0.088 0.5 50th 0.05-0.1 0.05-0.1 0.05-0.1 0.05-0.1 0.05-0.1 0.05-0.1 0.05-0.1 0.05-0.1 0.05-0.1 0.05-0.1 0.05-1.1 percentile 50th 50th 50th 50th 50th 50th 50th 50th 50th 50th 50th of the percentile percentile percentile percentile percentile percentile percentile percentile percentile percentile percentile 3- PO4 data must of the of the of the of the of the of the of the of the of the of the of the be <0.015 data must data must data must data must data must data must data must data must data must data must data must mg/L be <0.015 be <0.015 be <0.015 be <0.015 be <0.015 be <0.125 be <0.015 be <0.015 be <0.015 be <0.015 be <0.125 PO4-P mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L PO4-P PO4-P PO4-P PO4-P PO4-P PO4-P PO4-P PO4-P PO4-P PO4-P PO4-P 0.07 0.06 0.06 0.07 0.07 0.06 0.05 0.05 0.05 0.05 0.054 0.10 TP 0.05-0.1 0.018-0.2 0.02-0.1 0.02-0.09 0.02-0.1 0.02-0.09 0.02-0.1 0.02-0.1 0.02-0.09 0.018- 0.02-0.09 0.02-0.1 0.09 0.15 0.15 0.15 0.15 0.15 0.15 0.14 0.14 0.14 0.14 0.14 0.50 3- T PO4 0.12-0.2 0.1-0.2 0.11-0.2 0.12-0.19 0.118-0.2 0.11-0.19 0.07-0.2 0.07-0.2 0.07-0.19 0.07-0.19 0.07-0.19 0.07-1.2 10.76 13.64 13.51 13.3 15.89 27.67 19.85 23.43 48.16 35.88 33.49 42.70 TN/TP 0 – 5 mg 0.96 0.90 1.22 1.53 1.85 1.62 2.58 1.85 2.45 2.05 2.20 6.20 C/l 0.5-1.6 0.52-1.6 0.72-2.4 1.3-1.9 1.4-2.5 0.88-2.8 1.7-4.2 1.3-2.3 1.9-2.8 1.5-2.7 1.6-2.9 4.2-9.6 DOC Domestic use

44 Variable TWQR Sampling localities (mean value above and range (minimum-maximum) below) RQO underneath both 1 2 3 4 5 6 7 8 9 10 11 12 0 – 200 0.50 5.48 4.25 3.15 4.28 3.93 5.93 3.65 4.10 0.56 0.50 7.30 mg/l as 4.6-6.8 3.3-5.5 2-3.7 3.6-5.5 2.5-5.7 3.5-9.2 2.3-4.4 2.5-5.7 0.05-1.2 2.1-13 Sulp SO4 Domestic use 0.25 0.25 0.25 0.25 0.25 0.25 0.250 0.250 0.25 0.25 0.25 5.70 MIB 1ng/l 0.25-22 0.77 1.45 2.80 4.10 4.89 2.67 1.26 3.32 7.85 3.85 2.13 6.80 Geos 1ng/l 0.25-1.76 0.91-2.4 2-3.8 2.9-5.3 0.25-15 0.52-8.8 0.25-3.5 0.99-7.1 1.7-17 1.8-7.3 1.5-3.4 4.1-8.2 <30 0.82 0.93 1.09 1.70 2.00 1.90 2.80 1.95 2.65 2.00 2.20 8.20 TOC mg/l 0.56-1.3 0.65-1.4 0.8-1.7 1.4-2.1 1.5-2.4 1.3-2.9 1.9-4.6 1.4-2.3 2-3.2 1.6-2.2 1.9-2.5 4-16

45 The different ways to measure dissolved material (inorganic salts and organic matter) in water are closely related and is often converted from one to the other by means of a mathematical conversion factor which varies with the type of water (DWAF, 1996b). Electrical conductivity (EC) is a measure of the total dissolved salts (TDS) and is ratio depended on the ions present. Therefore, the total dissolved solids (TDS); specific conductivity (SPC)/ EC and salinity are all closely correlated in certain types of water (Pieterse and Janse van Vuuren, 1997 and WHO, 2003). There was no significant variation in SPC levels within the Sabie River, Marite River and Inyaka Dam, with site 1 having the lowest average SPC (46.78µS/cm) (Figure 4-3a). SPC values in the Sand River at sites 7 and 12 were significantly higher than site 1 in the Sabie River as well as the Marite River (site10) and Inyaka Dam. SPC showed significant positive correlation with

2+ + 2+ + - 2- TDS as was expected. Besides its correlation with major ions (Ca , K , Mg , Na , Cl , SO4 ) and hardness, a significant positive correlation was also found between SPC and E. coli, as well as M alkalinity. The higher average SPC values measured in the Sand River at sites 7 (average concentration 292.63µS/cm) and 12 (average concentration 389.20µS/cm) is associated with the higher Na+ and Cl- concentrations (27.65mg/l and 44.00mg/l respectively) observed at these sites. It can be speculated that the high SPC from site 12 improved further downstream to site 7 on the Sand River due to a change in land use and positive influence from conservation practices. SPC increased downstream in the Sabie River from sites 6 to 9. Figure 4-3a shows that the Sand River did not significantly change the SPC values of the Sabie River as no significant increase in SPC was observed from site 6 to site 8 after the confluence of the Sabie and Sand rivers.

Even though TDS concentrations for the sampling localities on the Sabie River are quite similar; the average TDS values for the Sand River at sites 7 (159.25mg/l) and 12 (161.00mg/l) were significantly higher than that observed in both the Sabie and Marite rivers as well as the Inyaka Dam. This can be ascribed to the concentrations of the major ions, Ca2+, Mg2+, Na+, Cl-, being significantly different between these sites. The major ion measured in the Sand River was Na+ and the TDS value for site 12 (161.00mg/l) also positively correlated with the higher Na+ concentration of 44.00mg/l at this site. As the TDS properties in water are governed by dissolved salts, it is closely related to the total hardness of water; scaling as well as corrosion potential of water (DWAF, 1996b). An increase in hardness is seen for sites with higher TDS values because the major cations Ca2+ and Mg2+ are important contributors to water hardness.

In contrast to the higher SPC measured at site 7 in the Sand River, the higher TDS concentrations mentioned above for site 7 probably had an influence on the TDS concentrations seen at site 8 (which occurs after the Sabie-Sand confluence) as the concentrations increased from site 6 (with average TDS concentration of 96.25mg/l) to site 8 (with average TDS concentration of 110.25mg/l) in the Sabie River.

46 600 280

260

240 500 220

200

400 180

160

l) 140 300 120

TDS (mg/ TDS SPC(mS/m) 100

200 80

60

40 100 20

0

0 Mean -20 Mean 1 2 3 4 5 6 7 8 9 10 11 12 Mean±SE 1 2 3 4 5 6 7 8 9 10 11 12 Mean±SE Site Mean±SD (a) Site Mean±SD (b)

30

25

20

15

10

Turbidity (NTU) 5

0

-5

-10 Mean 1 2 3 4 5 6 7 8 9 10 11 12 Mean±SE Site Mean±SD (c) Figure 4-3: Box and whisker plots illustrating the differences in a) SPC, b) TDS and c) Turbidity observed between the sites 1-9 (Sabie River), 10 (Marite River), 12, 7 (Sand River) and 11 (Inyaka Dam) during 2016. n = 4; ±SE (Standard Error) and ±SD (Standard Deviation).

Another water quality indicator, which cannot be described without considering TDS concentrations, is turbidity. Turbidity relates to the suspended matter in water which restricts light from passing though the water. A higher turbidity (or lack of transparency) can therefore be a result of higher amounts of TDS (total dissolved solids) concentrations in the water. The Spearman rank correlation indicates a significantly positive correlation between turbidity and TDS. The highest average turbidity levels were observed at sites 9 (Sabie River), 10 (Marite River) and the Inyaka Dam (site 11). Inyaka Dam had a very high average turbidity (9.95 NTU) in comparison to the rest of the sites. This could be the result of the high Fe2+/3+ and Mn2+ concentrations of (540.00µg/l and 117.00µg/l) respectively found at this site (Figure 4-3c) (Khadse et al., 2015). Turbidity also showed a significant positive correlation with Fe2+/3+. Therefore, it is suspected that the high turbidity value downstream of the Inyaka Dam in the Marite River (site 10) of (7.48 NTU) can also be the result of the high average Fe2+/3+ concentrations (545.00µg/l) observed at site 10 (Figure 4-6b). Sites 9 (in the Sabie), 10 (in the Marite River) and 11 at Inyaka Dam exhibited

47 ranges of above 5 NTU which are objectionable to users and "have some chance of transmission of disease by micro-organisms associate with particulate matter" (DWAF, 1996a).

4.1.4.1 Major Ions

The combined concentrations (mg/l) for the cations in the Sabie, Sand and Marite rivers were higher than the anions concentrations. The dominant cation (highest average concentrations) in the Sabie River was Ca2+: Ca2+> Mg2+> Na+> K+. Ca2+ and Mg2+ which are the two cations that determine hardness as mg/l CaCO3 were significantly positively correlated to hardness water quality parameter. The highest average Mg2+ concentration of 22.70mg/l was observed at site 2. Even though both sites 7 and 12 show higher Ca2+ they do not seem to change the Ca2+ concentration of the Sabie River at site 8 after the Sabie-Sand confluence. The higher Mg2+ concentrations observed at sites 2 and 12 do not exceed the TWQR of 0 – 30mg/l. In the Sand River however, Na+ was the dominant cation: Na+> Ca2+> Mg2+> K+. The average Na+ concentrations measured at the sampling sites in the Sand River (27.65mg/l for site 7 and 44 mg/l for site 12) exceeded the TWQR of 0 – 100mg/l for Na+.

- The anion with highest average concentrations in all three rivers were HCO3 (M Alk), followed by Cl- in the Sand and Marite rivers. Ranging from higher to lower average concentration in all three

- - 2- 2- - rivers are: HCO3 > Cl > NO3 > SO4 . The highest Cl concentrations were observed in the Sand River at sites 7 and 12. It did however not influence Cl- concentrations in Sabie River which exhibited an increase in Cl- levels down the gradient of the river from site 1 up to site 6 after which there was a marked decrease in concentration to an average of 6.38mg/l at site 8. The TWQR of Cl- is 0 – 100.00mg/l and this level was never exceeded at any of the sampling sites (Figure 4- 4b). There was a significant difference in Cl- concentrations between sites 1, 7 and 9 (Table 8-1

2- Appendix B). There was no significant difference between the SO4 concentrations between the

2- different sampling sites. SO4 did however show significant positive correlations with all the other major ions analysed in this study. Sites 1, 10 and 11 in the Inyaka Dam exhibited the lowest average concentrations.

48 70 45

40 60

35 50 30

40 25

mg/l ) 30 mg/l) 20

(

(

-

+

Cl

Na 15 20

10 10 5

0 0

-10 Mean -5 Mean 1 2 3 4 5 6 7 8 9 10 11 12 Mean±SE 1 2 3 4 5 6 7 8 9 10 11 12 Mean±SE Site Mean±SD (a) Site Mean±SD (b) Figure 4-4: Box and whisker plots illustrating the differences in a) Na+, and b) Cl- concentrations observed between the sites 1-9 (Sabie River), 10 (Marite River), 12, 7 (Sand River) and 11 (Inyaka Dam) during 2016. n = 4; ±SE (Standard Error) and ±SD (Standard Deviation).

4.1.4.2 Nutrients

2- - Nitrate (NO3 ) and Nitrite (NO2 ) concentrations for the Sabie River sites were similar down the

2- gradient of the river with site 1 having the lowest average NO3 concentration of 0.05mg/l. The occurrence of high nitrogen (N) and phosphorous (P) compounds are highly unwanted in waterbodies and unfortunately natural nutrient enrichment can be accelerated by anthropogenic activities such as agriculture through the use of fertilizers. Site 2 and 3 of the Sabie River had

2- the highest NO3 concentrations, 0.31mg/l and 0.35mg/l respectively, which could possibly be a result of runoff from the fruit plantations surrounding these sites. Other sites with relatively low

2- - NO3 concentrations were Inyaka Dam (site 11) and site 7 in the Sand River. The combined NO2

2- and NO3 values were also low for both the Sabie and Sand rivers and significant positive

- 2- 2+ 2+ correlations were seen for NO2 + NO3 with coli, COD, Ca , SO4, Mg , Hardness and M

- - - alkalinity. The average dissolved inorganic nitrogen (DIN = NH3 -N+ NO3 -N+NO2 -N) concentrations at all the sites except site 12, were <0.5mg/l and are thus indicative of oligotrophic conditions (DWAF 1996a). Site 12 had an average DIN concentration of 5.58mg/l which indicates eutrophic conditions.

- 3- Ammonia (NH3 ) had significant positive correlation with Chl-a, coli, temp, COD, P and TOC and

- negative correlation with DO. The correlation of NH3 (Figure 4-5c) and Chl-a (Figure 4-7a) can especially be noticed at site 12 where both variables showed an increase in comparison to the

- other sites. The NH3 concentration at site 12 of 5mg/l N exceeded that of the TWQR of 1mg/l N.

3- Orthophosphate (PO4 ) concentrations detected at all the sites except site 12 in the Sand River were never above the reporting limit during 2016. This means that we cannot see how it differed

49 between the different sites. The level of orthophosphate for the whole catchment suggests an

3- oligotrophic system (DWAF, 1996a). There is no drinking limit standard mentioned for PO4 in

3- Table 4-3, because inorganic PO4 has a very low toxic potential therefore a drinking-water limit was not defined (Kempster and Smith, 1985). As seen in Table 4-3, the Sabie and Sand rivers

3- 3- th have different RQOs for PO4 . The Sabie River’s acceptable PO4 limit is “the 50 percentile of

3- th the data must be <0.015 mg/L PO4-P” and the Sand Rivers’ tolerable PO4 limit is “the 50 percentile of the data must be <0.125 mg/L PO4-P”. The difference in these tolerable and

3- acceptable PO4 limits may be because the target Ecological Class (EC) for the Sabie and Sand rivers differ. The target EC for the Sabie River is A/B where the Sand River’s is C.

12 1,0

10 0,8

8 0,6

l)

l) 6

(mg/ 0,4

3-

4

DIN (mg/DIN 4

PO

0,2 2

0,0 0

-2 Mean -0,2 Mean 1 2 3 4 5 6 7 8 9 10 11 12 Mean±SE 1 2 3 4 5 6 7 8 9 10 11 12 Mean±SE Site Mean±SD (a) Site Mean±SD (b)

12

10

8

6

(mg/l)

-

3 4

NH

2

0

-2 Mean 1 2 3 4 5 6 7 8 9 10 11 12 Mean±SE Site Mean±SD (c)

3- Figure 4-5: Box and whisker plots illustrating the differences in a) DIN, b) PO4 and

- c) NH3 observed between the sites 1-9 (Sabie River), 10 (Marite River), 12, 7 (Sand River) and 11 (Inyaka Dam) during 2016. n = 4; ±SE (Standard Error) and ±SD (Standard Deviation).

50 4.1.4.3 Heavy metals

No significant difference was determined between the sampling sites for Al3+ and Mn2+ concentrations. The Al3+ concentrations measured in the Sabie River showed an increase downstream from site 1 (Al3+ concentration of 12.50µg/l) to site 4 (Al3+ concentration of 83.00µg/l) site 6 and site 9 (118.28µg/l). The Sand River at site 7 had the highest average Al3+ concentration of 158.75µg/l (Table 4-3).

For most of the water quality parameters the Marite River had a possible dilution effect on the Sabie River, as these parameters decreased from site 3 to site 4 after the Marite-Sabie confluence. Water quality parameters that stood out were Al3+ and Fe2+/3+ as these concentrations increased after this confluence. On closer inspection it was seen that site 10 had a high Al3+ concentration with a mean concentration of 132.75µg/l. The Al3+ concentrations for site 3 and 4 were 15.88µg/l and 83.00µg/l respectively. The same increase can be seen for water quality parameter Fe2+/3+ which had a concentration of 545.00µg/l at site 10; 125.00µg/l at site 3 and 338.75µg/l at site 4. The Fe2+/3+ concentrations for site 10 and 4 were regarded as exceeding the SANS 241: 2015 aesthetic limits.

Fe2+/3+ concentrations were observed at high or concerning levels over the whole catchment (Table 4-3). Fe2+/3+ concentrations above 30µg/l can cause severe aesthetic effects. With values above 100µg/l long-term health effects gradually start to increase, combined with the aesthetic influences as was observed in the both the Sabie River at site 4 and the Sand River (sites 7 and 12). With concentrations above 300µg/l as was measured in the Marite River and the Inyaka Dam (sites 10 and 11) chronic health effects and acute toxicity may appear accompanied by the severe aesthetic effects (DWAF, 1996a). As the TWQR of Al3+ is exceeded at all the sites, those sites with accompanying high Fe2+/3+ concentrations are of concern as DWAF (1996a) states that severe aesthetic effects such as discoloration can occur due to Al3+ in the presence of Mn2+ and Fe2+/3+. This should be given attention as Mn2+ values at site 11 (close to the Inyaka Dam water purification plant) exceeds aesthetic values as indicated by the drinking water standards (SANS 214:2015 ranges).

In Table 4-3 it can be seen that Zn2+ was also measured at average concentrations that exceeded the TWQR of 0 – 3mg/l in the Marite River at site 10 (concentration of 4.78µg/l), as well as in the Inyaka Dam (site 11 concentration of 4.7mg/l). Only site 12 in the Sand River exceeded the TWQR with an average Zn2+ concentration of 7.12mg/l.

51 0,40 1,2

0,35 1,0 0,30

0,25 0,8

0,20 0,6

(µg/l)

(µg/l) 0,15

3+

/

3+

2+

Al 0,4 0,10 Fe

0,05 0,2

0,00 0,0 -0,05

-0,10 Mean -0,2 Mean 1 2 3 4 5 6 7 8 9 10 11 12 Mean±SE 1 2 3 4 5 6 7 8 9 10 11 12 Mean±SE Site Mean±SD (a) Site Mean±SD (b)

+ 2+/3+ Figure 4-6: Box and whisker plots illustrating the differences in a) Al3 , and b) Fe concentrations observed between the 1-9 (Sabie River), 10 (Marite River), 12, 7 (Sand River) and 11 (Inyaka Dam) during 2016. n = 4; ±SE (Standard Error) and ±SD (Standard Deviation).

4.1.4.4 Biological Indicators

According to DWAF (1996a) high Chl-a concentrations relate to the high total pigment concentrations. High Chl-a concentrations occur at site 12 (62.20µg/l) in the Sand River. The degradation product of Chl-a, phaeophytin, is also found at high concentrations at site 12. Site 12 also exhibited the highest abundance of the cyanobacteria (Erasmus, 2017). Even though the algal toxin microcystin (produced by genera from the cyanobacterial group) was measured during this study the concentrations are not shown as the levels were never above the limit of detection. Geosmin and 2-methylisoborneol (MIB) are taste and odour compounds produced by certain species of cyanobacteria as well as actinomycetes. All the Sabie River study sites (except site 1); as well as both sites within the Sand River (sites 7 and 12), the Marite River (site 10) and Inyaka Dam (site 11) experienced high concentrations of geosmin. Water collected from sites in the Sabie River therefore had a noticeable odour (geosmin 1-5.00ng/l) except for site 9 which had a geosmin concentration of 7.85ng/l representing a TWQR of a strong odour, likely to be objectionable to large sectors of the population (DWAF, 1996a). This range is also true for the Sand River at site 12 with an average geosmin concentration of 6.80ng/l and MIB concentration of 5.70ng/l. Although there was no significant correlation between geosmin and Chl-a, the high concentration of taste and odour components could be associated with the high concentrations of cyanobacteria at the site 12 (Erasmus, 2017). It is thus speculated that actinomycetes must be responsible for the production of geosmin at the other study sites.

3- Geosmin also showed significant positive correlations with PO4 , E. coli, M alkalinity, TOC and

3- DOC. This significant correlation is supported by the fact that PO4 and C serve as nutrients to

52 both cyanobacteria and or actinomycetes which can both produce geosmin and MIB. MIB, also showed a significant positive correlation with COD. The concentration of MIB measured at all the sites except for site 12 were otherwise low.

Total Coliforms (coli) and E. coli concentrations exceeded the TWQR as well as the drinking water standard (Table 4-3) and pose health risks at all sampling sites. As seen in Figure 4-7b, site 12 had an average E. coli concentration of about 26 000MPN/100ml higher than the average of all the other sites. Coli and E. coli concentration seem to increase from Sabie River sites 1 to 2 (10 and 225 E. coli MPN/100ml respectively) which occurs downstream of a wastewater treatment plant (WWTP). E. coli concentration increased from site 3 to site 4 (70 E. coli MPN/100ml) which also occurs downstream of a WWTP.

COD is a measure of the oxygen required for breakdown or oxidization of soluble and organic matter DWAF (1996a). This is probably the reason for the positive correlation between COD and TOC exhibited at site 12. There was also a significant negative correlation between DO and these parameters. TOC and DOC concentrations were significantly different between sites 1-3 in the Sabie River and site 12 in the Sand River located downstream of the WWTP in the urban township of Thulamahashe. An increase to 6.2mg/l C in DOC was noticed at site 12 which exceeded the TWQR of 0-5mg/l C. This, according to DWAF (1996a) can result in slight trihalomethane (THM) formation during the chlorination step of water purification.

160 80000

140 60000 120

100 40000

l)

80 µg/l) 20000

60 (MPN/100m

Chl-a ( Chl-a

E.coli 40 0

20 -20000 0

-20 Mean -40000 Mean 1 2 3 4 5 6 7 8 9 10 11 12 Mean±SE 1 2 3 4 5 6 7 8 9 10 11 12 Mean±SE Site Mean±SD (a) Site Mean±SD (b) Figure 4-7: Box and whisker plots illustrating the differences in a) Chl-a, and b) E. coli concentrations observed between the sites 1-9 (Sabie River), 10 (Marite River), 12, 7 (Sand River) and 11 (Inyaka Dam) during 2016. n = 4; ±SE (Standard Error) and ±SD (Standard Deviation).

53 4.1.5 Principal Component Analysis (PCA) of the variables influencing water quality

The PCA extracted 28 different principal components from the dataset that contains 29 variables (Table 8-4 Appendix B). To determine the number of components to extract for data interpretation, the components on the scree plot that showed substantial amount of common variance, (i.e., components before inflection point) were chosen (Hair et al., 2009). The following could also have been considered when choosing the final number of components according to Hair et al. (2009):

 “Components with eigenvalues >1.0 (Table 4-4 yellow colour)  “A predetermined number of variables based on prior research/ research objectives (3 to 4);  “Enough components to meet a specified percentage of variance explained, usually 60% or higher (Table 4-4, the first 3 variables cumulate to 61.28% of the variance and the first 4 variables to 69.23%);  “More components when heterogeneity is present among sample subgroups”.

As mentioned above, the scree plot’s first three components are of most importance (Figure 4-8). They were the components that explained the highest % variance in the dataset. Their cumulative variance described 61.28% of the variance in the water quality dataset. These components were inclusive of different variables. These variables ranged in degree of influence upon that specific component. The variables therefore contributed to the component’s higher or lower ranking (1 to 28) for describing % variance in the dataset. The degree of influence that these variables had on the component was measured in a value called component correlation. Variables with high component correlations were closer to 1 and those with weaker association to the component were lower than 1 (Table 4-5) (Hair et al., 2009). The closer the component correlation is to 1, the stronger is the influence of that variable on that component (Quinn and Keough, 2001). Table 8- 4 in Appendix B indicates the components describing highest variance in the water quality data. Table 4-5 lists the components and their associated variables with the highest component correlations indicated in red.

Figure 4-8 plots the eigenvalues (Table 4-4) for the 28 components of this study against the component number. Initially the plot slopes steeply down after starting with the first component and after the fifth component starts gradually sloping to end in a nearly horizontal line (Hair et al., 2009). The first point at which the line starts to “straighten out” (inflection point/ point 3) is the maximum number of components necessary to extract in the explanation of variance in the data.

54

Figure 4-8: Scree plot indicating the eigenvalues of the correlation matrix.

The first three components were extracted to indicate which variables in the dataset combine to explain the most variance in the dataset. Therefore, the components can be renamed as an indicator of influence. The first component with eigenvalue of 10.89 explained 37.55% of the variance and was highly negatively correlated with the variables Coli, TOC and K+. This means the influence of variance is the absence or low concentrations of Coli, TOC and K+ at most of the sites, compared to the high concentrations of these variables observed at site 12 in the Sand River. Other parameters that support this, were the negative correlations of SPC, E. coli, Na+ and

- - NH3 . E. coli can be closely linked to coli, and NH3 which is an indicator of sewage pollution.

The second component with eigenvalue of 3.69 explained 12.74% of the variance. The high negative variable correlations of Turb, Mn2+, Fe2+/3+ and TKN in component 2 leads to the influence of variance in the data again being the absence or low concentrations of these variables at most of the sites compared to the high concentrations of these variables observed at site 12 in the Sand River. Most of the sites did not have such a high turbidity or concentration of Mn2+, Fe2+/3+ and TKN as for example seen at Inyaka Dam or site 12 in the Sand River. This could therefore be an explanation for the negative correlation of these variables. Where there was high

55 turbidity, as seen at Inyaka Dam, there were also high concentrations of Fe2+/3+, Mn2+ and TKN. Fe2+/3+ and Mn2+ are parameters known for causing higher turbidity in water and this can be supported by the significant Spearman rank correlation between turbidity and Fe2+/3+. TKN may not be the highest at Inyaka Dam but as a measurement of organic materials in water can also link to cause these higher turbidity values in the water. As mentioned before Turbidity relates to suspended matter which can be increased with higher amounts of nutrient concentration in the water. TKN is organic N and therefore could represent particles that could be noted as turbidity. Another reason that may be apparent for the general absence of turbidity and its associated variables are drought or low flow conditions. As this study was conducted within a dry rainfall year, baseflow or low flow conditions dominated. With a lesser amount of soil and surface constituents being washed to the water by surface run-off this can substantiate these strong negative correlations. What baseflow conditions also indicate is a subdued stream flow amount and speed causing a settlement of total dissolved solids rather than the re-suspension usually associated with turbulent streams. These might all be explanations towards describing the influence of variance in the data.

The third component with eigenvalue of 3.186 explained 10.98% of the variance. The high component correlations with MIB, Fe2+/3+, Tsi concentrations leads to the influence of variance in data being nutrient enrichment by means of organic pollution. Especially with the negative component correlation of DO to this component, which might not be as high of a correlation, but might still indicate poor water quality. These values might be indicators of poor disposal of human (organic) waste. MIB is one of the parameters indicating tastes and odours in water. According to DWAF (1996a) these types of odours are predominantly caused by presence of organic substances of biological origin. High levels of MIB and its strongly associated variables might be the indicators of possible pollution from waste water treatment plants pollutions that were situated close to some of the sampling localities.

56 Table 4-4: Eigenvalues and related percentages on the main contributing principal components. Eigenvalues of correlation matrix. and related statistics (Water quality Value Eigenvaluedata) % Total Cumulative Eigenvalue Cumulative % 1number 10.88932 37.54938variance 10.88932 37.5494 2 3.69586 12.74433 14.58518 50.2937 3 3.18659 10.98826 17.77177 61.2820 4 2.30497 7.94816 20.07674 69.2301 5 1.75791 6.06174 21.83464 75.2919 6 1.47741 5.09451 23.31205 80.3864 7 1.15785 3.99259 24.46990 84.3790 8 0.94113 3.24528 25.41103 87.6243 9 0.77447 2.67060 26.18551 90.2948 10 0.68294 2.35498 26.86845 92.6498 11 0.55941 1.92898 27.42786 94.5788 12 0.38044 1.31187 27.80830 95.8907 13 0.26778 0.92339 28.07608 96.8141 14 0.25289 0.87205 28.32898 97.6861 15 0.17398 0.59995 28.50296 98.2861 16 0.17026 0.58712 28.67322 98.8732 17 0.09549 0.32928 28.76871 99.2025 18 0.07349 0.25343 28.84221 99.4559 19 0.05608 0.19339 28.89829 99.6493 20 0.04226 0.14573 28.94055 99.7950 21 0.02801 0.09659 28.96856 99.8916 22 0.01278 0.04407 28.98135 99.9357 23 0.00838 0.02891 28.98973 99.9646 24 0.00652 0.02248 28.99625 99.9871 25 0.00215 0.00741 28.99840 99.9945 26 0.00152 0.00524 28.99992 99.9997 27 0.00008 0.00026 28.99999 100.0000 28 0.00001 0.00003 29.00000 100.0000

57 Table 4-5: Component variable correlations for the three extracted components. Variables Component 1 Variables Component 2 Variables Component 3 Al3+ 0.050391 Al3+ 0.015884 Al3+ 0.474885 Ca2+ -0.566016 Ca2+ 0.324961 Ca2+ -0.335315 Chl-a -0.643298 Chl-a 0.179866 Chl-a 0.439117 Cl- -0.777944 Cl- 0.139501 Cl- 0.164476 COD -0.631142 COD 0.121820 COD -0.110138 Coli -0.918546 Coli -0.132505 Coli 0.035153 DO 0.300451 DO -0.378373 DO -0.569753 E. coli -0.811989 E. coli -0.370447 E. coli -0.375698 Fe2+/3+ 0.091917 Fe2+/3+ -0.653425 Fe2+/3+ 0.576661 Geos -0.488480 Geos -0.207281 Geos 0.236132 K+ -0.889344 K+ 0.057166 K+ 0.241939 M Alk -0.789522 M Alk 0.258458 M Alk -0.297429 Mg2+ -0.062994 Mg2+ 0.327335 Mg2+ -0.276885 MIB -0.421659 MIB 0.282190 MIB 0.582427 MMHg -0.349552 MMHg 0.221945 MMHg 0.151520 Mn2+ 0.050812 Mn2+ -0.665080 Mn2+ 0.313020 Na+ -0.866456 Na+ 0.133277 Na+ 0.243955 - - - NH3 -0.859645 NH3 -0.349597 NH3 -0.262140

NO3 +NO2 -0.147411 NO3 + NO2 0.349119 NO3 + NO2 0.098246 pH -0.062563 pH 0.476941 pH -0.084097 SPC -0.878569 SPC 0.271737 SPC -0.042154 Sulp -0.716171 Sulp 0.495666 Sulp 0.082864 TDS -0.204490 TDS 0.274771 TDS 0.129324 TKN -0.755785 TKN -0.509996 TKN -0.303954 TN -0.797951 TN -0.473634 TN -0.297632 TOC -0.892249 TOC -0.114905 TOC 0.247684 TP -0.789560 TP -0.411984 TP -0.377780 Tsi -0.338189 Tsi -0.110611 Tsi 0.575893 Turb 0.063589 Turb -0.665636 Turb 0.436459 According to Hair et al. (2013) (as cited by Swanepoel, 2015) variables indicating (±0.7 and higher) display the highest correlation to the component. Those equal to or greater than ±0.5 are considered significant and variables that display (±0.3 – ±0.4) may not be regarded as significant, but may aid in the identification of the component (Hair et al., 2009).

58 Discussion

Studies across the world have reported the decline in all aspects of water quality in rivers and catchments due to an increase in population, and as a result, an increase in urbanization, agriculture, industrial activities, and mining (Ding et al., 2016; Mei et al., 2014 and Van der Hoven et al., 2017; and references therein). The Sabie-Sand River catchment has also been experiencing an increase in population (Tlou, 2011). Furthermore, the impact of climate change will add to the deterioration of water quality (Munzhedzi, and Mgquba, 2013). Many regions in South Africa experienced a drought during 2016 and low baseflow conditions were evident in the Sabie, Sand and Marite rivers. Despite these negative impacts also evident in the Sabie-Sand catchment this study’s three contributing rivers as well as the Inyaka Dam, exhibit oligotrophic conditions. Only site 12 in the Sand River, which is located downstream of a WWTP in the urban township of Thulamahashe experienced eutrophic conditions.

During this study the sampling points were mostly compliant with the South African drinking water guidelines (SANS241:2015), the TWQR (DWAF, 1996a,b&c) and the RQO’s set for this catchment (South Africa, 2016). Only a few exceptions were observed especially pertaining to site 12 in the Sand River. The average SPC levels measured during this study were within the drinking water guidelines, TWQR and the RQOs at all sampling sites. Turbidity at site 1 was so low, it complied with the drinking water standard set at 0-1.00 NTU (SANS241:2015). The highest turbidity was observed at the Inyaka Dam which is the source for the Inyaka Water Supply Scheme. Water with high turbidity may not only be objectionable to water users but can also increase the treatment costs. There were also significant correlations observed between TOC, DOC, and E. coli concentrations and turbidity. Van der Hoven et al. (2017) observed a similar correlation between turbidity and E. coli in the Zandspruit and the North Ridingspruit in Gauteng. Both the total coliforms and E. coli exceeded not only the TWQR but also the drinking water guidelines at all of the sampling sites. This represents a health risk not only for drinking purposes but also recreational purposes and would again, add to the increased costs related to water purification. The water at all the sites (except site 1) also experienced taste and odour problems mainly due to geosmin present. Although geosmin does not pose a health risk, it is objectionable to consumers, who will perceive the drinking water as unsafe (Van Rensburg et al., 2016). The presence of geosmin would also add to increase the cost of water purification for drinking water purposes.

The concentrations of both Al3+ and Fe2+/3+ were a cause for concern as both exceeded the SAN241:2015 aesthetic limits especially in the Inyaka Dam. At a pH of above 6.5 Al3+ primarily exists in an insoluble form of mostly aluminium hydroxide. Only at pH <4 will it become soluble and toxic, so there is no health risk associated with these higher levels of Al3+ observed in the Sabie, Marite and Sand River as well as the Inyaka Dam.

59 Site 12 is located in the Sand River downstream of a WWTP at the edge of a densely populated and busy urban township. As was shown by the PCA analysis this sampling site could be associated with the worst water quality conditions observed during this study. It exhibited the highest levels for most of the variables determined. This sampling site is most probably adversely impacted on due to its location within a densely populated low-income settlement (with 1265 people per km² (Census, 2011). Such settlements are often negatively affected by neglected water resource management and failing sewage treatment infrastructures (Van der Hoven et al., 2017). The most concerning water quality variables are the high E. coli and coli concentrations observed which correlated with the high DIN, Chl-a, COD, TOC and DOC concentrations as well as the low DO concentration. The significant negative correlation of DO with coli, TOC and DOC can be a result of human waste pollution that favoured rapid growth of the bacteria leading to a depletion in DO as was found by Van der Hoven et al. (2017). This is furthermore supported by the high level of COD.

60 CHAPTER 5 LAND USE It is important to know about landscape dynamics for efficient land management. According to Foley et al. (2005) land management refers to the methods used for the managing of land development and land use, taking in consideration strategies of sustainable social, environmental and economic benefits. Land management forms the base upon which to conduct studies regarding land use change and land use influence on various other characteristics of the physical environment. Detecting urban growth or rate of expansion by monitoring land cover changes is one of the examples why monitoring land use is important. In short, land use as mentioned in Chapter 3 refers to the type of use assigned to the land cover (Di Gregorio and Jansen, 2000). GIS techniques are at the forefront of detecting surface changes and the causes of these changes, natural or anthropogenic in nature. The Spatial Planning and Land Use Management Act (Act 16 of 2013) enables a framework for the assessment of land use, land use influences and the management thereof in South Africa.

Methodology

There is no detailed and complete land use dataset available for the Sabie-Sand catchment area. Such a dataset is very dynamic and difficult to map, and land use is further a reflection of multiple factors, including anthropogenic activities and geomorphic land characteristics (Yu et al., 2016). This study used the land cover data as a surrogate for land use at a very high level. The land use dataset was the fundamental dataset used to answer the research question, using the ancillary datasets as a means to support the possible spatial influences. To accommodate for land use types that are likely to affect water quality, a point pollution dataset of mines, industries, waste water treatment plants etc. was created, evaluated and the most important points included in the analysis (section 5.1.2.4).

In order to answer the research question, it was important to group land cover classes derived from the 2013/14 national land cover (NLC) layer of the Sabie-Sand catchment into broad classes such as transformed – considered as urban, mine, agriculture and plantation and natural areas – such as grassland, wetland, woodland, thicket dense bush and indigenous forest (see Table 5-2 for this study’s land use legend). Surface runoff is influenced by various factors, two of which are land-use type and cover. As a result, the majority of land use and land cover (LULC) influences within the catchment, represents a non-point source pollution type. Therefore, the geographical information software, ArcGIS (Esri, 2015) was used to analyse the non-point and some possibly significant point source pollution in the catchment. The GIS database consisted of various map layers such as: land cover, conservation boundaries, geological layers, streams, communities and digital elevation model (DEM) maps. A DEM of this study area in 90m resolution was downloaded via Globalmapper from CGIAR-CSI (Jarvis et al., 2008). ArcGIS 3D Analyst tools were used to derive the upstream areas and watersheds corresponding to the 12 sampling

61 localities from the DEM. The watershed polygons were then used as zonal data to calculate statistics regarding the initial upstream land use, geology etc. mentioned above. Further data was derived from the DEM such as terrain differences e.g. slopes. It is suspected that the water quality varies according to the different land uses occurring upstream of the sampling localities (Kändler et al., 2017).

5.1.1 Data Collection and Preparation

5.1.1.1 GIS Database

GIS data for the Sabie-Sand catchment were sourced from existing databases. GIS has the capability to relate spatial data of a certain location (defined within a coordinate system) stored in multiple user-friendly layers that will eventually combine to form spatial information of that specific location on earth (Lee and White, 1992). The watershed boundaries, towns, conservation boundaries, geological lithologies, national land cover and Department of Rural Development and Land Reform community layers used for this study already existed in GIS format (see metadata).

National Land Cover Layer: The 72 Class GTI South African National 2013/2014 Land Cover Dataset derived by GeoTerraimage was used in this study as alternate for land-use information, because no detailed land use data is available for this study area for the 2015/2016 season. This layer was sourced from the South African Department of Environmental Affairs (DEA, 2016). It was last updated and previous errata corrected in June 2016. Derived from multi-seasonal Landsat 8 satellite imagery (April 2013 – March 2014) the NLC2013/14 dataset is based on 30x30m raster cells (30m resolution) and can be used for ~ 1: 60 000 – 1: 250 000 scale GIS- based mapping and modelling applications (GeoTerraimage, 2015). Based on the standard map projection for distributing Landsat 8 data, the original land-cover dataset was processed in Universal Transverse Mercator (UTM) 35 North, WGS84 datum format as provided by the USGS (GeoTerraimage, 2015). GeoTerraimage (GTI) Pty Ltd. developed repeatable and standardised semi-automated modelling procedures to generate this dataset. Desktop accuracy assessment has been done visually against high resolution imagery and photography of equivalent dates in Google Earth. Industry standard error (confusion) matrices were used to report accuracies namely Producer, User and Kappa values (GeoTerraimage, 2015). This raster dataset was re-projected by GeoTerraimage where necessary to a standardised UTM35(north), WGS84 map projection and afterwards received in a TIF format by DEA. The dataset was made available in both UTM35(north) and (south), WGS84 map projections and Geographic Coordinates, WGS84. Associated characteristics stored in an attribute data file (attribute Table) included the count, class name, class type and class description.

62 Watershed boundaries: The “X” primary drainage region (Inkomati catchment) and the “X3” secondary drainage region (Sabie-Sand catchment), both exported on 2007-07-17 by the Department of Water and Sanitation (DWS), was obtained from The Department’s Resource Quality Information Services (DWS, 2017). These exported KMZ/ KML files were then imported into ArcGIS10.4 to create catchment boundary shapefiles. The primary drainage region 1: 250 000 boundaries was only used for visual mapping as seen in Figure 2-1 in Chapter 2. It was decided not to work with already developed South African quaternary catchment boundaries, as new, site specific boundaries were rather delineated.

Conservation boundaries: Forming one of the three main land tenure classes is the conservation aspect of this study area (Table 5-5 and Figure 9-1 in Appendix C). The various South African National Parks (SANParks) conservation boundaries or game reserves were derived from the South African Protected Areas Database (DEA, 2016) layer obtained from DEA’s EGIS portal with the following link: https://egis.environment.gov.za/ (see metadata).

Urban boundaries: The Communities (DRDLR) shapefile (South Afrca, 2008) (communal also regarded as one of the three main land tenure classes) was used to indicate urban and semi- urban boundaries (Figure 9-1 Appendix C).

Geological boundaries: There were three different existing datasets available representing catchment geology. Firstly, the 1: 250 000 scale WR90 geological layer was used to obtain attribute data such as: area and one field of descriptive lithology (LITHOS) describing the characteristics of rocks found in the area (acidic, porous etc.). The second set of geological data layer at a minimum scale 1: 10 000 000 was used to obtain attribute data such as: old name (e.g., Timeball Hill and Rooihoogte), stratigraphic name (e.g., Timeball Hill), stratigraphic rank (e.g., formation), stratigraphic parent (e.g., Pretoria), chronological name (e.g., Vaalian), chronological rank (e.g., era), and 5 different lithologies (e.g., lithology 1 – shale) (GeoScience, 2003). At the end, the lithology 1 attribute from GeoScience (2003) was used when conducting multivariate analysis (Table 5-8).

Soils: A third layer was a 1: 250 000 scale soil and geological rock combination layer derived from the Agricultural Research Council (ARC) land types of South Africa (Land Type Survey Staff, 2002) (Figure 9-2 Appendix C). The soil data obtained from this layer (Table 5-7) was the main soil data used for the multivariate analysis.

Even though this layer contained soil and geology information for example Ab10 represents the soil properties: red-yellow apedal, freely drained soils; red, dystrophic and/or mesotrophic and geology: biotite granite and migmatite (Nelspruit granite), granodiorite (Hebron granite), mafic and ultramafic gneisses (Bandelierkop Complex), Cunning Moor tonalite (all of the Archaeozoic) and

63 diabase; a single lithology layer (lithology 1 from GeoScience (2003) was eventually used to represent the catchment geology as seen in Figure 2-6 Chapter 2. Soil descriptions of the land types occurring in the catchment are listed in Table 5-10 as a descriptive table.

5.1.1.2 Digital Surface Model Data

Digital Elevation Model Data: Digital Elevation Models (DEMs) are gridded elevation data of the earth’s surface. DEM data contain terrain morphological information that can be derived as secondary products, for example slope, aspect (slope direction), sub-basins (watersheds), river networks, curvature, etc. (Lee and White, 1992). The Shuttle Radar Topography Mission (SRTM) 90m DEM used for this study area was obtained through GlobalMapper (2016) in the form of six quadrangles (S25E030; S25E031; S25E032; S26E030; S26E031; S26E032) that were extracted from GlobalMapper on 04/06/2016 and re-projected using nearest neighbour resampling (Jarvie et al., 2008). The DEM was used to calculate land slope (Figure 5-3) and drainage patterns (Figure 5-1) for the Sabie-Sand catchment. Table 5-1 indicates the SRTM90 DEM product specifications.

Table 5-1: Indicating SRTM data product specifications (USGS, 2015). SRTM DEM Product Specifications used within this study Projection Geographic Horizontal Datum WGS84 Vertical Datum EGM96 (Earth Gravitational Model 1996) Vertical Units Meters Spatial Resolution 3 arc-seconds for global coverage (90 meters) Raster Size 1 degree x 1 degree tiles terrain grid

5.1.2 Data Interpretation (Spatial)

ArcGIS10.4, with the Spatial Analyst, Network Analyst and 3D Analyst Extensions was used for analysis and the presentation of spatial datasets. All GIS data was clipped according to the watershed boundaries presented by Sabie-Sand catchment and thereafter clipped to the watershed boundaries delineated through the DEM (see below).

5.1.2.1 Site sectioning

The sites were mainly chosen on the Sabie, Sand and Marite rivers as well as close to the outlet of the Inyaka Dam Wall. The Sabie River and its tributaries were sampled at 12 localities (in this study also referred to as nodes) (Figure 2-2, Chapter 2). Site locations were firstly based on existing survey points that corresponded with the sites monitored by the Inkomati-Usuthu Catchment Management Agency. Site accessibility was the second main concern for allocation of sampling localities. For analysis purposes, the river systems were divided into segments via the nodes and each individual drainage basin was delineated accordingly. The assumption was

64 that land use activities, climate, soil and terrain characteristics (physical parameters) associated with each upstream river segment will determine the water quality at the sampling point or node. The sampling point or node was also used as a pourpoint to determine the associated watershed (Figure 5-1). Therefore, it was decided to analyse the land use influence on the rivers via 12 watersheds that covered the entire extent of land uses influencing the river segment. These watersheds were used on the land cover dataset to calculate statistics per segment.

By considering the 12 nodes, 12 watersheds were delineated by using the DEM (Figure 5-1) and the 12 points as pourpoints or endpoints. These endpoints mark the lowest point of flow that a possible contaminant or river variable will pass before entering a new watershed. The river segments between the nodes transect different land uses. River quality and water features together with land uses were examined between each node considering the sampling results above that node within the river. The river segment together with its surrounding land uses formed its own watershed (Figure 5-1).

Figure 5-1: Digital Elevation Model (DEM) with Shreve ordered river network and 12 delineated watersheds used for spatial method of analysis in this study outlined in red.

65 Initial method: (Initial method meaning see definitions). Initially it was considered to create multiple buffers (5, 10, 15, 20km etc. depending on segment size) around each river segment upstream of the nodes (without separate watersheds). Figure 5-2 is a representation of what this looked like for river segment 1 above node 1 (with current watershed 1). As this is an irregular shaped secondary catchment, these buffer kilometres had to vary according to the segment length (Figure 5-2 watershed 5), which would result in node 1 (0.01km, 0.5km, 1km, 1.5km, 2km, 2.5km buffers) and node 5 (0.01km, 5km, 10km, 15km, 20km, 25km buffers) for example having a large difference in amount and distance of buffers surrounding the segment, making data abstraction and interpretation cumbersome. The two main purposes of the buffers were firstly to quantify the land use, by calculating land use statistics per segment for each buffer and secondly to see which land use has the most influence on the river segment at what distance from the segment. The latter was replaced by a point pollution layer that instead showed a possible point- pollution source at a certain distance from the nearest river or stream (section 5.1.2.4). Reconsidering the purpose of different buffers, it was decided to rather work with DEM delineated watersheds at sub-catchment scale (smaller than X3 secondary catchment). This would focus more on site specific values directly upstream of sample localities. Incorporating 1) land use as a whole; 2) slope and therefore the runoff possibility within a specific drainage basin. The watershed approach also eliminates buffers that possibly overlapped from adjacent river segments.

Once all the buffers for the segment were rasterized, statistics were calculated with the clipped land use (57 class) and specific buffer distance. These tables were joined per land use layer (thus for every distance). The land use statistics were calculated for each buffer (Figure 5-2). The 0.01km buffer is difficult to see at this scale, but is a representative of the riparian zone of the river.

66

Figure 5-2: Comparison of multi-buffered segment 1 (above node 1) and segment 5 (above node 5) of Sabie River, converted to polygons for further analysis such as zonal statistics for each distance (e.g. 0.01km, 0.5km, 1km, 1.5km, 2km, 2.5km in watershed 1).

5.1.2.2 Digital Elevation Model

The STRM90 DEM grid represents a 90m x 90m land surface area (Figure 5-1). Through use of data management in ArcGIS, the six individual DEM tiles were merged and clipped to secondary catchment boundary. The 90m DEM dataset was the basis for further hydrological and surface analyses.

Slope of catchment: The slope of a catchment in ArcGIS can be expressed in two ways, namely: degree or slope percentage (also called percentage rise). In this study slope percentage was used (Figure 5-3). Slope% ranges from 0 to infinity meaning that the more vertical a slope becomes, the larger the slope percentage becomes. Considering a slope with an angle of 45 degrees, the percentage rise or slope % is 100% (Esri, 2016a).

67 Detailed method: 1. spatial analyst 2. re-project DEM coordinate system to projected raster 3. spatial analyst 4. surface 5. slope (as percentage rise, Z Factor: 1) 6. layer properties change classes 7. reclassify 8. natural breaks 9. edit break values (e.g. 2, 8, 15, 30, 100) 10. extract zonal statistics as table.

Zonal Statistics of slope as a table: 11. input re-projected slope DEM from previous step 12. value 13. insert re-projected original raster DEM 14. ignore nodata 15. export table (Table 5-9). 16. clip slope DEM to 12 watersheds to extract zonal statistics for each watershed on land use layer.

River Slope: Apart from calculating the 12 watersheds’ overall slopes, the reach specific slope was also obtained. This enabled a closer analysis of the correlation between water quality parameters and slope.

Detailed method: 1. clip main river segments above sample localities 2. data management 3. features 4. points to line – include end points 5. extract values to points 6. select the interpolation option 7. extract table 8. calculate average values of Z (height) (Stack Exchange, 2011) (Table 5-9).

Drainage basin: The 12 watersheds used to interpret/ analyse data for this study were created with the hydrology tools of ArcMap10.4.

68 Detailed method: 1. using the clipped raster DEM 2. spatial analyst tool 3. hydrology 4. fill (which means to fill sinks) 5. flow direction 6. flow accumulation 7. create new shapefile – pourpoints using sample coordinates 8. watershed tool – input flow direction & pourpoints.

Convert watershed raster to polygon: 9. conversion tools 10. from raster 11. to polygon 12. input watershed layer 13. use output watersheds polygon layer to clip watersheds on DEM.

Stream network: Detailed method: After above mentioned: 1. spatial analyst tool 2. map algebra 3. raster calculator 4. “flowacc” 5. = 100.

Stream order: 6. 2 options to choose from is Strahler or Shreve 7. used Shreve 8. conversion tools 9. streams to features (shapefile) 10. convert stream raster to polyline (uncheck simplify lines) 11. clip stream networks according to watersheds boundaries.

69

Figure 5-3: a) Represents the individual slopes (in % rise) of the 12 watersheds and b) indicates the slope (in % rise) for the whole Sabie-Sand catchment. The colour scheme indicates low lying areas as darker and areas increasing in percentage slope rise as lighter colours (Jarvis et al., 2008).

5.1.2.3 Land use (diffuse sources)

National Land Cover data set: Visual analysis was done by overlaying and comparing two datasets. These were: the NLC data with 2015 SPOT 6/7 mosaic and multispectral data, obtained from the South African National Space Agency (SANSA, 2015), via the North-West University Geo- and spatial department mapping facility, which were captured on the 24th of March 2015 and the 30th of April 2015; and the combined satellite imagery of Google Earth Pro. An overall image classification of the SPOT6/7 data did not take place to adjust the land use dataset as there were only negligible differences between the three layers. The NLC layer (Figure 5-4) was clipped according to the watershed boundaries presented after the analysis of the DEM. It resulted in 12 watersheds each with their own land cover layer. These raster datasets were representative of the non-point pollution sources of each catchment.

70 Initially the 72 class NLC layer obtained from the DWA website was clipped to the Sabie-Sand catchment (X3 secondary catchment boundary). Only 57 lasses were applicable to this study area (Table 5-2 column 4). The clipped dataset was then used to merge similar land use types into broader categories. The broad summarised 6 – 8 class classification used by several studies as seen in chapter 3, Table 3-1 was considered. As the spatial scale of the 12 watersheds was relatively small compared to those in the literature study, it was decided to aggregate the 57 class land cover output data (Table 5-2 column 4), into an 18 class land use dataset (Table 5-2 column 2 and Figures 5-7, 5-8 and 5-9). This dataset was used within the final multivariate analysis. This data was obtained by zonal statistics through ArcMap and resulted in a count for every land use type in every watershed. These counts were used to obtain land use percentages for the watershed.

The data were divided into a binary class of transformed and natural (Figure 5-4). Transformed refers to all the anthropogenic land uses that have fully changed the landscape; and natural means, that although the landscape is used, it has not been changed in its entirety. These are the transformed and natural land uses based on the 57 class land cover layer: Transformed: ● Agriculture e.g. cultivated commercial fields; ● Barren land e.g. erosion donga; ● Forestry e.g. plantation; ● Mine e.g. bare mines. ● Urban land e.g. built-up; and Natural vegetation: grassland, indigenous forest, thicket, woodland; low shrubland; ● Water body e.g. water permanent. In addition, an indicative representation of land within this study is the way the land tenure (see definitions) classes divide the area into three broad classes of occupancy, namely; communal, commercial and conservation (Table 5-5 and Figure 9-1 Appendix C).

71 Table 5-2: Symbology legend for the combined 18 class land use interpretation extracted from 57 class clipped 2014 National Land Cover layer. Colour Land use class Land use Transformed/ 57 classes combined to form legend class Natural class of column 2 symbol Cultivated CuCom Transformed Cultivated commercial fields Commercial (high, low and medium classes); Cultivated commercial pivots (high and medium classes). Cultivated Orchards CuOrc Transformed Cultivated orchards (high, low and medium classes). Cultivated CuSub Transformed Cultivated subsistence (high, Subsistence low and medium classes). Erosion-Bare EBare Transformed Erosion donga; bare none vegetated. Grassland-Low GrasL Natural Grassland; low shrubland. Shrubland Indigenous Forest IndFo Natural Indigenous forest. Mines Mines Transformed Mines 1 bare; mines 2 semi- bare and mines water permanent. Plantations- PlanW Transformed Plantations / woodlots. Woodlots Urban Built-up UrBuU Transformed Urban built-up (bare, dense trees/ bush, low veg/ grass and open trees/ bush classes); Urban commercial and Urban industrial. Urban Informal UrInf Transformed Urban informal (dense trees/ bush, low veg/ grass, open trees/ bush classes). Urban Lawns UrLaw Transformed Urban school and sports ground; Urban sports and golf (bare, dense trees/ bush, low veg/ grass and open trees/ bush classes). Urban Residential UrRes Transformed Urban Residential (bare, dense trees/ bush, low veg/ grass and open trees/ bush classes). Urban Smallholding UrSma Transformed Urban Smallholding (bare, dense trees/ bush, low veg/ grass and open trees/ bush classes). Urban Townships UrTsh Transformed Urban Townships (bare, dense trees/ bush, low veg/ grass and open trees/ bush classes). Urban Village UrVil Transformed Urban Village (bare, dense trees/ bush, low veg/ grass and open trees/ bush classes). Water Water Natural Water permanent; water seasonal Woodland WdlnO Natural Woodland/ Open bush; and Thicket /Dense bush. Wetlands Wetld Natural Wetlands.

72

Figure 5-4: Combined 18 class land uses for this study area as seen in Table 5-2. Extracted from the 57 class, 2014 NLC layer (GeoTerraimage, 2015).

5.1.2.4 Land use (point sources)

Major point pollution: A major point source pollution layer (PPS) was created by creating a feature layer class consisting of identified pollution points near the rivers. It was suspected that their intensities and ranges of influence could have an influence on river water quality, therefore the main use for this layer was to serve as indicator of any outlier polluter that could not be explained through normal land use activities. This was accomplished with photo interpretation by overlapping the 2014 NLC dataset with 2015 SPOT-6/7 data (SANSA, 2015) using ArcMap. These two datasets were used in combination with ground truthing and extensive photo interpretation of mainly corresponding Google Earth satellite imagery using Google Earth Pro software, 2015 via the internet. This source layer (Figure 5-5) captures localities of most importantly wastewater treatment plants (green); large industries and factories such as saw mills, timber etc. (blue) and bare ground/ diggings from abandoned mines (red). Initially deforestation of timber plantation/ woodlots active during the sampling time was considered as point pollution due to increased probability of runoff, which was later revised as they were still regarded as non-

73 point sources. This layer can be consulted when changes in water quality reflect other pollution possibilities not explained by non-point pollution sources.

Figure 5-5: Major Possible Point Source Pollution Layer for Sabie-Sand catchment.

Point Pollution Source distance to sampling locality: The distance of the point pollution source to sampling locality was obtained using the Network Analyst extension on a Shreve river network (Figure 5-5). This meant that the route options of a point pollution source were determined by the river network (drainage line). Allowing it to follow a path created by water even if it was only following a smaller tributary stream into the primary river network. Within Network Analyst extension the shortest downstream flow path was determined using the new route option. The shortest flow path was determined for each point pollution source to the nearest downstream sampling point.

Initial method: It was initially considered to create a map that would describe the probability that a specific point source pollutant would reach a river within this study site. It would have been used to determine the contribution of that point (land use/ land cover) which was at a certain distance from a river or tributary. Therefore, it had to be known what type of water retention ability the land

74 cover had over which it would most probably flow. To manage this, a weighted overlay and cost analysis was done (see below). After obtaining the results of this, it was decided that the quality of the maps (Figure 5-6 b, c and d) were not sufficient to indicate the probability of a certain PPS reaching the river sampling locality. This unfortunately meant that only the distances of the point pollutions from the nearest tributary, stream or reach of river were calculated for the comparison analysis of water quality data (Figure 5-5). This calculation discarded the weighted overlay and cost analysis method and used the network analyst method instead.

Table 5-3: Reclassification values of land uses during weighted overlay analysis. Least resistance for Classes within this study Reference/ Source/ Reason surface flow and related surface types 9 Water (Rivers) Kansiime et al. (2007) (water usually follows the path of least resistance). Erosion (Donga) Nunes et al. (2011) and Wang et al. (2016). Bare ground and Mines Liu et al. (2016) and Nunes et al. (2011). Plantation / Woodlots Dye (1996) and Lin et al. (2015). Increase clearfelled Urban Abu-hashim et al. (2015) and Lin et al. (2015). in Low shrubland Nunes et al. (2011). resistance Plantation / Woodlots Nunes et al. (2011). young to Cropland and Farmland Abu-hashim et al. (2015) and Nunes et al. (2011). surface Cultivation, Orchids Wang et al. (2016). Wetlands Kansiime et al. (2007) flow Plantations / Woodlots Lin et al. (2015). mature / Afforested land Litter Liu et al. (2016) and Nunes et al. (2011). Woodland/ Open bush Nunes et al. (2011). Grass cover Liu et al. (2016), Nunes et al. (2011) and Yang 1 et al. (2014). Shrubland Nunes et al. (2011). Highest resistance for Indigenous Forest Lin et al. (2015). surface flow Thicket/ Dense bush Lin et al. (2015).

Weighted overlay and cost analysis: To obtain the route and distance of surface flow – in other words the “ease” with which the water flows over different types of surfaces – a weighted overlay and thereafter a cost path had to be calculated. This would give an indication of what the quickest route of flow would be for that land use/ land cover (point and non-point source) to reach the sampling point (least cost) (destination). The least resistance “easiest flowpath” would be for example where the existing streams are – they would acquire a factor of 9 meaning surface flow on that part of the land cover is very easy “most favourable” and factor of 1 “least favoured” with higher surface water retention ability. This type of basic factor of reclassification for each land use was given from 1 to 9 (Esri, 2016b). After the factors were allocated to each land use, a cost analysis classification took place to include the percentage influence of slope and land use (see below). The purpose of the cost path was to find out what the probability was of a specific point

75 pollution reaching the river. It was also used to investigate the contribution of a point that lies at a certain distance from the river.

Attempted method for cost analysis: By inserting a weighted overlay Figure 5-6b indicates an example of reclassified land use and slope created from a DEM (Figure 5-6f) using various land use influences (Table 5-3). The different factors allocated to the 2014 NLC raster are listed and referenced in Table 5-3 above. Firstly, the 2014 NLC raster was converted to a point layer (Figure 5-6a), thereafter the river layer was converted to a point layer (Figure 5-6e) and finally the point pollution layer was also used (Figure 5-5). With these layers, it was considered that during the cost analysis, the land use point will follow the shortest path to any closest river or tributary point. This would indicate if a certain land use affects a tributary or even a main stream, considering the influence experienced from surface roughness/ runoff ability and slope. This was also done to measure the most probable path or “least costly path” that a land use parameter or point source pollution would therefore have to take to reach the sample locality of that watershed. It was thought that each point would eventually combine to create a “gradient of travel” map toward a nearest stream and eventually we could create a gradient of travel cost path towards the chosen sampling locality, but instead the map did not execute the distance as intended during these trial and error attempts. For the cost path (Figure 5-6g), the closest land use point (source), to the sample location (destination) was used for calculation instead of having all the point sources travel to the destination. This type of trial and error method resulted in a polyline map such as Figure 5- 6g. This happened when both the point pollution and national land point cover layers were the source layers. Therefore, it did not work as expected to have various start and end points within a cost analysis.

When allocating the influence of the land uses, ArcMap requires you to choose which factor between slope and land use would have a greater effect on the surface runoff capabilities. This is expressed as a percentage contribution. A ratio between slope and land use is then allocated together with the land use factors (Esri, 2016c). Weighted overlay was conducted by using two different influences from slope and percentage influence from land use. Figure 5-6d represents a ratio of slope:land use of 20% : 80% and Figure 5-6c one of 60% : 40%. This resulted in very coarse (pixelated) weighted overlay raster layers which were then smoothed with the filter tool (Spatial analyst, low, ignore data is ticked). The darker blue colour indicates a weighted class 9 of higher resistance and the lighter the blue colour, the less resistant the land use becomes. As these pixelated raster datasets were not of good enough quality for use within this study, another solution was attempted. Rather to create a map to describe the probability of a polluter reaching the river, it was assumed that the polluter will reach the river and that its distance to sampling point should just be calculated. This was to retrieve the distance of the point polluter to the sampling locality through network analyst (see above).

76

Figure 5-6: Weighted overlay (b), cost analysis (c, d) and slope (f) of the western

part of the Sabie-Sand catchment (watershed 1) used in combination with the land cover point and river point layers (a and e) to obtain a final cost path (g) for this watershed. 77 5.1.2.5 Ancillary data

River flow data: River flow data for gauging stations was obtained from DWS, 2017. These stations were representatives of the surface water flow through the river nodes closest to the station. Unfortunately, 1) the datasets for these gauging stations were not updated for the sampling periods and only historical data were available for some of the stations; 2) not all of the sampling localities had stations close by and 3) those that had, did not have updated surface flow data because they were out of order due to floods or poor maintenance leaving the stations non- operational. The river flow through each node could therefore not be analysed in terms of the most up to date or working gauging station closest to the node. The river flows through the highest and lowest nodes were originally for comparison against this data to determine whether possible over abstraction of river water was taking place. On reflection, such an approach ignores in- catchment abstractions and discharge into the stream. To compensate for these types of in and outflows of the system, a model of catchment hydrology would be required, which is beyond the scope of this study.

Results and Discussion

The tables and figures that follow summarise the physical and spatial parameters of the Sabie- Sand catchment and serve as the results of the land use analysis. The combined 18 class land use data is represented in the pie charts for each watershed (see Figures 5-7, 5-8 and 5-9 and Table 5-4 for detailed land use results).

5.1.3 Land use and tenure

The highest percentage land use for watershed 1, 2 and 3 as seen in Table 5-4, 5-5 is the commercial practices of plantations and woodlots at 74%, 71% and 66% respectively, followed by grasslands at 11% for watershed 1, and woodland – open bush for watershed 2 and 3 both at 15.7%. This is supported by the land use influence mentioned by Roux and Selepe (2011) for the area in their state of the rivers report that correlates to watershed 1 and 2 of the present study. The possible threats from extensive forestry that they highlighted for these watersheds, included extensive siltation, erosion and abstraction. For this area Roux and Selepe (2011) also mentioned alien riparian vegetation that may further reduce biodiversity and pose a threat to the system functioning of natural vegetation. Their study area that correlates with the present study’s watershed 3, 10 and 11 has the same forestry and alien vegetation impacts as mentioned above, but includes banana plantations as a form of monoculture, which they state causes a high erosion and sedimentation risk to the land surface. Plantation-woodlot percentages for watershed 3 is 66%, for watershed 10 is 56.6% and for watershed 11 is 64.9%.

78 Dominant land use for watersheds 4 to 9 and 12 were woodland – open bush ranging from 39.40% to 95.40%. Confirmed by Roux and Selepe (2011) a large part of watersheds 4, 10 and 12 exist as communal occupancy or tenure (Table 5-5). Activities related to rural communities include small scale and subsistence farming of fruit and livestock. Again, erosion and sedimentation pose a threat to the water quality, as this type of farming is prone to over-grazing of land. Livestock farming (cattle, goats, pigs and poultry), crop farming (maize, vegetables, etc.) are the dominant kinds of subsistence farming in the rural area. Even though as seen from Table 5-4, watershed 5 of the present study has a high woodland-open bush land use percentage, Roux and Selepe (2011) as well as personal site identification, view it as part of the region subjected to the rural community activities mentioned above.

To the east of the commercial farming and forestry areas and to the west of the KNP are former “self-governing territories” also known as homeland areas (in the present study referred to as a communal tenure class). It is evident from the study conducted by Woodhouse (1995) that farming in these areas is poorly developed. The reason is uncertain but Woodhouse (1995) says that this may be due to low and irregular rainfall in the area as well as lower potential soils. This means that the either stony or very sandy characteristics of these soils can have cultivation constraints in these flatter areas. Possible cropping system alternatives on these flatter slopes are cultivation of tobacco, grain and vegetables (Woodland, 1995). Steep slopes can also account for cultivation constraint as experienced in the commercial farming area.

The boundaries of Watersheds 5 to 10 form part of the conservation tenure class. Sampling locality 5 marked the first site within the Sabie River, monitored by the present study subjected to conservation by a National Park. The western boundaries of watersheds 11 and 12 are adjacent to a biosphere reserve called the Blyde River Canyon Nature Reserve; and to the east of watershed 11, Bosbokrand Nature Reserve is also classified as part of a biosphere reserve (Figure 9-1 Appendix C). Even though protected areas are located in the central to eastern side of the catchment, watersheds 5, 6, 7, 8 and 9, rural areas still occur adjacent to these conservation areas within this study area.

Parts of the river tributaries within this study area, especially the Sand River, flow through rural settlements and are subjected to non-conservational elements before reaching the protected areas of the Kruger National Park (KNP) and its surrounding private- and game reserves. This raised initial concern as the Sand River flowing directly through a rural community converges with the Sabie River within the KNP. The KNP is one of South Africa’s Protected Areas mainly established as a scenic tourist destination. Apart from tourism, other factors mentioned by Coetzer et al. (2010) also contribute to the existence of a Protected Area, namely; the land being unsuitable for agriculture or formed part of residual areas not belonging to any specific major land use. Non-conservational influences are inevitable as Coetzer et al. (2010) mentions that land

79 uses and development proposals limit conservation initiatives through continuing population growth and development of the socio-economic sector prioritised by the National Government. Population growth is highest in watersheds 11 and 12 (urban village land use percentage of 22.17% in watershed 12) with increasing Bushbuckridge and Thulamahashe populations. For progression, conservation should explicitly incorporate socio-economic development, especially in the former homeland areas. Such conservation decisions can no longer be made in isolation; therefore, the highest priority of current Protected Area systems is to address unavoidable land cover modifications associated with economic growth, whilst still maximising biodiversity protection (Coetzer et al., 2010).

Watershed 6, 8, 9 with dominant land use percentages comprising woodland-open bush, are all within the conservation boundaries of the KNP. Even though this confirms the low impact land uses mentioned by Roux and Selepe (2011) which were eco-tourism and conservation activities, they are of opinion that erosion still occurs in parts of the KNP. The present study identified points regarded as abandoned mines or bare ground in Figure 5-5 within these watersheds as well as surrounding the outer boundaries of the KNP.

The abandoned mines found in this study area were either small or not concerning regarding pollution of water sources. Operational mines were found to be sand mining for constructing houses near the villages and townships. It was decided that most of the points on the point pollution layer that occur further away from the sampling locality would experience a “remedial” effect before reaching the sampling localities. This left only WWTP points as sources of pollution in the main river stream (Table 5-6). Further discussion of the point pollution layer will commence in Chapter 6.

Eucalyptus and pine plantations dominate the upper Middleveld. According to Kimble et al. (2014) the middleveld is a region of “veld” with an intermediate altitude generally ranging between 600 and 1200mamsl. The Middelveld comprises the upper part of the Sabie River to just before Hazyview town, occupying about half to two-thirds of the land surface drained by the Sabie River. Exotic forest plantations include pine (Pinus patula Schlecht. & Cham., P. elliottii Engelm. and P. taeda L) and Eucalyptus (Eucalyptus grandis Hill ex Maiden). Pine and eucalyptus plantations are most successful in areas with average rainfall of more than 700mm/a, which is true for this region (Forest Owners Association, 1994, as cited by Dye, 1996). According to Woodhouse (1995) about 26% or more of the lower Middleveld (north, east and south of Hazyview, including the town) in the present study, was at that time (1995) occupied by irrigated agriculture which is still true for watersheds 1, 2, 3, 10, 11 and most of 12. Land use is a dynamic entity, so these numbers have gradually changed through the years. Mainly tropical and subtropical varieties cultivated are mangos, avocados, bananas, pecan and macadamia nuts (Woodhouse, 1995). The increased amount of water use from forestry and plantation is a great concern (Dye, 1996;

80 Smits et al., 2004; Tlou, 2011) and to some extent Roux and Selepe, 2011). Woodhouse (1995) verifies this by stating that the yearly fluctuation in water availability leads to a fluctuation in the irrigated areas of especially the Middleveld.

Figure 5-7: Pie charts representing land use results in percentages of watershed areas 1-4. Land use abbreviation meanings: CuCom - Cultivated Commercial; CuOrc – Cultivated Orchards; CuSub – Cultivated Subsistence; EBare – Erosion-Bare; GrasL – Grassland-Low Shrubland; IndFo – Indigenous Forest; PlanW – Plantations-Woodlots; UrBuU – Urban Built-Up; UrRes – Urban Residential; UrVil – Urban Village; Water – Water; WdlnO – Woodland; Wetld – Wetlands.

81

Figure 5-8: Pie charts representing land use results in percentages of watershed areas 5-8. Land use abbreviation meanings: CuOrc – Cultivated Orchards; CuSub – Cultivated Subsistence; EBare – Erosion-Bare; GrasL – Grassland-Low Shrubland; UrBuU – Urban Built-Up; UrRes – Urban Residential; UrVil – Urban Village; WdlnO – Woodland; Wetld – Wetlands.

82

Figure 5-9: Pie charts representing land use results in percentages of watershed areas 9-12. Land use abbreviation meanings: CuCom - Cultivated Commercial; CuOrc – Cultivated Orchards; CuSub – Cultivated Subsistence; GrasL – Grassland-Low Shrubland; IndFo – Indigenous Forest; PlanW – Plantations-Woodlots; UrTsh – Urban Townships; UrVil – Urban Village; Water – Water; WdlnO – Woodland; Wetld – Wetlands.

83 5.1.4 Soil, Geology and Slope

According to Gertenbach (1983) soil patterns for this catchment correspond closely with the position in the topography. The dominant slope for the catchment area has a general west to east downward gradient with the highest altitude being 1559.5mamsl and the lowest 178.6mamsl. The watershed with the steepest river slope is watershed 1 with 24% rise, whereas watersheds 2 to 3 have a steady downward percentage slope reaching a plateau at watershed 5 – 9. Slope analysis as a spatial characteristic on its own was mainly conducted as a means to include runoff influences. Slope influences runoff and can therefore also influence the number of surface constituents that enter the river.

The main geological rock type (Table 5-8) for this study area is granite as 59% of lithological layer 1 occurring at every watershed except for watershed 1. It is followed by shale and tonalite at 12% and 13.6% respectively. Shale is a finely stratified consolidated sedimentary rock while tonalite is a plutonic rock with the composition of diorite, but with considerable quartz (SCWG, 1991). The land type as seen in Table 5-7 is used for description of soil types that occur in the area.

The dominant land type (Table 5-7) with a percentage of 40% for the entire catchment is Ab consisting predominantly of dystrophic and/or mesotrophic red- or yellow-brown pedal soils. Red and Yellow-brown Apedal soils are freely drained weakly structured soils that are homogenous in colour through the soil profile (SCWG, 1991). Dystrophic diagnostic soil means that the soil has suffered marked leaching, so much so that the sum of exchangeable Ca2+, K+, Na2+ and Mg2+ is less than 5 cmol(+) per kg clay compared with mesotrophic soils at 5 – 15 cmol(+) per kg clay (see cation exchange capacity in definitions) (SCWG, 1991). The Ab land type occurs predominantly in watersheds 1 – 4, 10 and 11 on the lower slopes of the escarpment. Clay percentages are generally high (26 – 24%) due to their siliceous parent material in parts containing high percentages of 1:1 layered clays e.g. shale. (Refer to Figure 2-5 Chapter 2 – intercalated assemblage of compact sedimentary and extrusive rocks as example).

The second dominant land type (Table 5-7) with a percentage of 30% for the entire catchment is Fb characterised by soils of the Glenrosa and/or Mispah forms, although more developed soils may be present in specific landscape positions. Glenrosa soil forms are soils with an orthic topsoil A horizon and lithocutanic B subsoil which develops in situ from underlaying parent rock material (SCWG, 1991). The cutanic character develops from heterogenous weathering expressed as tongues of colour variegation due to illuviation (SCWG, 1991). The Mispah soil form is characterised by an orthic topsoil and hard rock subsoil. According to SCWG (1991) this hard rock cannot be cut with a spade, as it is a continuous layer of silcrete or metamorphic, igneous and indurated (hardened) sedimentary rocks. The Fb land type is associated within the Sabie- Sand catchment with poorly developed soils on Archaean Granitoid intrusions (Johnson et al.,

84 2006) along the bottom plains of the catchment and covers 16 % of the catchment area within watersheds 5 – 9 (Table 5-7).

The third most dominant land type (Table 5-7) is the Hb land type (9.3% of the catchment) characterised by grey regic sands and other soils characterised by low to very low clay percentages (7.6%) as seen for example in watershed 12. A regic sand is characteristic of being a recent deposit, showing little to no evidence of pedogenesis after deposition (SCWG, 1991). There could possibly be a darkening of topsoil brought about by organic matter from an orthic A horizon. These regic sands are either massive or single grained, have little to no macroscopic structure and are coarse in texture with loose, friable or soft consistency. They do not have a large difference in mineral composition from that of the parent material and as a result of minimal pedogenesis little to no clay formation has occurred (SCWG, 1991). Within the catchment the grey regic sand formation could be the result of deep weathering of granites. This is supported by the statement of Gertenbach (1983) that the low-lying areas along the river, are underlain by Archaean granite and gneiss which are intersected by dolerite intrusions. Gertenbach (1983) also mentions that soils from the low-lying areas are normally shallow, and where deeper, usually saturated with sodium. This is mainly due to mineral and clay accumulation within the low-lying areas.

85 Table 5-4: Land use results for watersheds 1 – 12 as the extracted 18 class land use categories. The table depicts the watersheds’ first and second most dominant land use as percentages in mustard and faded yellows respectively. Watershed Land use as percentage of each watershed CuCom CuOrc CuSub EBare GrasL IndFo Mines PlanW UrBuU UrInf UrLaw UrRes UrSma UrTsh UrVil Water WdlnO Wetld 1 1.157 0.000 0.000 0.011 11.314 3.489 0.000 74.351 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 9.526 0.152 2 0.133 0.054 0.000 0.056 5.371 4.167 0.010 71.583 0.387 0.000 0.139 0.655 0.031 0.355 0.148 0.001 15.798 1.111 3 0.165 8.769 0.000 0.047 3.267 4.277 0.004 66.162 0.271 0.013 0.014 0.196 0.009 0.033 0.116 0.215 15.724 0.719 4 1.696 11.041 0.775 0.078 1.524 0.789 0.024 31.082 0.873 0.000 0.262 0.627 0.489 0.051 8.895 0.505 39.711 1.578 5 0.000 0.000 0.422 0.056 6.127 0.000 0.018 0.000 0.200 0.000 0.073 0.275 0.000 0.000 1.832 0.088 90.908 0.000 6 0.154 0.735 2.220 0.053 4.501 0.000 0.010 0.053 0.499 0.000 0.004 0.000 0.000 0.039 11.032 0.090 80.495 0.114 7 0.004 0.012 1.537 0.135 3.622 0.000 0.005 0.000 0.350 0.000 0.000 0.000 0.000 0.027 5.650 0.053 88.242 0.363 8 0.000 0.000 0.000 0.117 4.388 0.000 0.001 0.000 0.010 0.000 0.000 0.000 0.000 0.000 0.000 0.067 95.416 0.000 9 0.000 0.000 0.000 0.127 12.033 0.000 0.001 0.000 0.021 0.000 0.000 0.000 0.000 0.000 0.000 0.104 87.693 0.021 10 0.039 0.959 0.136 0.049 0.370 0.241 0.007 56.611 0.038 0.000 0.000 0.000 0.037 0.000 6.227 0.130 34.568 0.588 11 1.151 1.214 0.044 0.243 1.468 10.360 0.000 64.992 0.258 0.000 0.006 0.000 0.070 0.266 3.823 3.375 12.060 0.669 12 0.033 0.669 1.773 0.046 1.903 3.937 0.070 10.914 0.291 0.000 0.030 0.000 0.000 1.239 22.172 0.184 54.173 2.564

Table 5-5: Land tenure table used within the multivariate analysis. Table 5-6: The distance in meters to the nearest waste water Watershed Tenure Commercial Communal Conservation treatment plant above the sampling locality. 1 Forestry 1 0 0 Watersheds Distance (m) 2 Forestry 1 0 0 1 5456 3 Forestry 1 0 0 2 6727 4 Communal 0 1 0 3 31555 4 1264 5 Conservation 0 0 1 5 38496 6 Conservation 0 0 1 6 54366 7 Conservation 0 0 1 7 121330 8 Conservation 0 0 1 8 139346 9 Conservation 0 0 1 9 174866 10 Communal 0 1 0 10 17931. 11 Forestry 1 0 0 11 36098 12 Communal 0 1 0 12 477 86 Table 5-7: Soils summary statistics of average soil depth, percentage clay and percentage occurrence of broad land types (Land Type Survey Staff, 2002) within the Sabie-Sand catchment. (Symbology description can be seen on the following page). Land type percentage of watershed Watershed Soil Depth (mm) Soil Clay% Ab Ac Ae Dc Ea Fa Fb Hb Ib (%) 1 918 39.10 27.067 36.395 - - - 36.538 - - - 2 969 42.20 45.385 44.689 - - - 9.925 - - - 3 650 33.70 80.488 17.354 ------2.157 4 421 26.20 91.325 0.417 6.325 - - - 0.399 - 1.535 5 318 16.00 7.577 - 17.983 - - - 45.481 - - 6 318 16.00 - - 2.147 - 6.694 - 90.553 0.605 - 7 318 16.00 - - - - 2.676 - 70.756 26.568 - 8 318 16.00 - - - 0.754 0.000 - 99.246 - - 9 347 16.80 - - - 14.679 24.301 - 61.021 - - 10 650 33.70 100.00 ------11 836 38.50 91.427 2.228 ------6.345 12 933 7.60 41.766 ------55.209 3.025

Table 5-8: Percentage occurrence of main lithological forms within the Sabie-Sand catchment, representing the geology spatial characteristic used for multivariate analysis. Lithological layer 1 of watershed Watershed Andesite Arenite Basalt Dolerite Dolomite Gabbro Gneiss Granite Granodiorite Granophyre Quartzite Rhyolite Sedimentary Shale Tonalite 1 ------100 - 2 0.29 - - - 44.72 - - 10.94 0.96 - 11.2 - 0.1 31.78 - 3 - - - 17.6 12.88 - - 61.22 1.56 - 5.6 - - 1.14 - 4 - - - 9.43 0.11 0.17 - 71.82 16.89 - 0.9 - - 0.69 - 5 - - - 0.04 - 6.39 - 50.17 ------43.4 6 - - - 0 - 1.65 - 74.76 ------23.59 7 - - - 0 - 0 - 59.1 ------40.9 8 - - - 0.95 - 3.33 - 95.72 ------9 - 3.16 25.11 0 - 0.46 1.61 51.36 - 1.04 - 9.91 - 7.35 - 10 ------90.63 ------9.37 11 - - - - 0.44 - - 83.89 - - 1.34 - - 4.76 9.56 12 - - - - 0.01 - - 61.31 - - 0.26 - - 2.04 36.38

87 Table 5-9: The mean slope for the Sabie-Sand catchment is shown as derived from the SRTM90 DEM. The mean elevation of the watersheds in meters was obtained. Also shown are the percentage rise in slope for the river and the percentage rise in slope for the watersheds (Jarvis et al., 2008). Watershed Mean Elevation of watershed Mean slope of river Mean slope of watershed (mamsl) % 1 1559.51 24.07 29.69 2 1024.50 10.04 21.66 3 715.98 13.63 18.05 4 481.80 7.14 13.14 5 345.70 4.23 5.88 6 267.14 4.23 4.14 7 327.41 4.10 4.43 8 232.99 4.40 4.01 9 178.59 3.50 4.50 10 652.90 9.27 12.63 11 887.19 10.06 16.21 12 416.63 2.73 9.85

Table 5-10: Land type descriptions (Land Type Survey Staff, 2002) within the Sabie-Sand catchment. Land types (soil) Soil description occurring in catchment Ab Red-yellow apedal, freely drained soils; red, dystrophic and/or mesotrophic Ac Red-yellow apedal, freely drained soils; red and yellow, dystrophic and/or mesotrophic Ae Red-yellow apedal, freely drained soils; red, high base status, > 300 mm deep (no dunes) Dc Prismacutanic and/or pedocutanic diagnostic horizons dominant. In addition, one or more of: vertic, melanic, red structured diagnostic horizons Ea One or more of: vertic, melanic, red structured diagnostic horizons, undifferentiated Fa Glenrosa and/or Mispah forms (other soils may occur), lime rare or absent in the entire landscape Fb Glenrosa and/or Mispah forms (other soils may occur), lime rare or absent in upland soils but generally present in low-lying soils Hb Grey regic sands and other soils Ib Miscellaneous land classes, rocky areas with miscellaneous soils

88 CHAPTER 6 LAND USE INFLUENCE ON WATER QUALITY Measuring water quality with regards to land use influences has widely become an accepted method to establish an integrated view of water quality. Several studies (Chen et al., 2016; Ding et al., 2016; Kändler et al., 2017 and Xizhi et al., 2017) have determined land use as a quantitative measure in relation to water quality indicators. Variables associated with water pollution can be significantly correlated with land uses (Donohue et al., 2006 and Xizhi et al., 2017). According to Donohue et al. (2006) the principal legislative tool for evaluating aquatic system integrity, has for decades been facilitated internationally through only chemical monitoring. However, within South Africa, a set of guidelines exist that strengthens not only chemical but biological, physical and aesthetic measurements of water quality for different land uses (DWAF, 1996). Until at least a decade ago this integrated approach, is sometimes lacking in monitoring programmes abroad, as most of them have for decades been focusing only on chemical water quality analyses (Donohue et al., 2006). The broad water uses that the South African Water Quality Guidelines (DWAF, 1996) takes into account are Domestic, Recreational, Industrial, Agricultural: Irrigation, Livestock Watering and Aquaculture and Aquatic Ecosystems.

Van Wyk et al. (2001) identified the historical and, the still existing research initiatives for the Sabie- Sand catchment and the way it enhanced monitoring programs and the filling of gaps within water quality analysis for the catchment. They realised that sufficient knowledge existed on monitoring and measurement, while ongoing research is still needed with regards to information that enhances the knowledge pertaining to: “needs of resource-poor people and the links between land activities and river health”. Ongoing research is thus needed to refine a more complete set of tools for resource management in the catchment. Therefore, the aim of this chapter was to investigate how land use modifies the water quality of the Sabie, Sand and Marite rivers as well as the Inyaka Dam.

Methodology

The data sources of this study were twofold. The first dataset consisted of biological, chemical and physical water quality parameters (discussed in chapter 4). The second data source related to spatial and non/spatial data representing the catchment area above each sampling point (discussed in chapter 5). Data analyses that were used to compare both datasets with each other included multivariate statistical analysis – redundancy analysis (RDA) and cluster analysis. The last mentioned was conducted with Statistica and the redundancy analysis was performed using CANOCO 4.5 software. The first question asked when doing the analysis was, does land use have an influence on water quality? Thereafter, the influence of soil on water quality was investigated.

6.1.1 Statistical Data Interpretation

Firstly, a Non-metric multidimensional scaling (NMDS), thereafter a redundancy analysis (RDA) and a hierarchal cluster analysis were conducted. 89 An NMDS of the percentage land use surrounding the rivers influencing each site indicates the similarity (or dissimilarity) of sites based on the land use (Figure 6-1). NMDS uses a numerical algorithm for a group of objects – in this case sample points – to best portray inter-sample distances. It analyses the matrix of dissimilarities between the samples to find a configuration where the distance between these samples within the ordination space corresponds to the dissimilarities (Lepš and Šmilauer, 2003). With interpretation – the sample points that are ordinated closer together are likely to be more similar to one another than those further apart (Buttigieg and Ramette, 2014). Sub- populations in the data will be indicated by clustered points (or samples) which are well-separated from the other clustered points.

A constrained form of the linear ordination method of principal components analysis, otherwise known as a redundancy analysis (RDA), was performed using the Canoco4.5 software programme (Ter Braak, and Šmilauer, 2002). According to Ding et al. (2016); Kändler et al. (2017) and Zhao et al. (2015), RDA is a key method used when determining environmental quality and landscape relationships. The multivariate correlation analyses which were used to assess associations between sites, different land use activities, and median measurements of water quality variables collected during the sampling time are represented in visual ordinations (Figures 6-2 and 6-3) (Lepš and Šmilauer, 2003 and Vrebos et al., 2017). The RDA method summarises the variation in a set of response variables (e.g. water quality variables) explained by a set of explanatory variables (e.g. catchment percentage land use) (Buttigieg and Ramette, 2014).

When analysing the RDA ordination plots, vectors represented by arrows indicate the degree of correlation between environmental variables (explanatory variables, predictors and independent variables) and species data (response variables and dependent variables). According to Dabrowski and De Klerk (2013) the orientation of, or angles between all arrows are indicative of their linear correlation (direction of maximum change of that variable). According to Lepš and Šmilauer (2003) the angle and degree of correlation is inversely proportional; a smaller angle thus represents a closer correlation between arrows. A 90-degree angle between arrows therefore indicates 0 or negative correlation (>than 90⁰ a weak or no correlation exists between arrows) and 0-degree angle between arrows indicates a numeric correlation of 1 which is a strong positive correlation (Buttigieg and Ramette, 2014). Points may indicate binary or nominal explanatory variables of which the centroids of these objects indicate a state of “1”. Projecting these explanatory points (environmental variables) (for instance sample points) at a 90-degree angle onto an arrow that represents a response variable, reflects the relationship between these two variables. To achieve this, the species arrow (or response variable) in a biplot from linear ordination can be lengthened straight through the biplot centre of origin. If the explanatory point is 90-degree projected near or beyond the arrow head of the response variable – these two variables are strongly positively correlated and if the explanatory point is projected near the origin, the variables will be weakly correlated and will be less influenced

90 by the arrow in question (Dabrowski and De Klerk, 2013 and Lepš and Šmilauer, 2013). The lengths of the arrows indicate the mutual similarity of contributions (Buttigieg and Ramette, 2014). For example, when analysing water quality values as arrows and site characterisations as points, a longer arrow indicates a greater influence on site characterization (Dabrowski and De Klerk, 2003). Arrows parallel to each other (e.g. pointing in the same direction) will indicate a positive correlation and those opposite each other on the same axis indicate a negative correlation (e.g. DO and TSi species variables in Figure 6-2). The direction of the arrow indicates the direction of increase for that variable (Buttigieg and Ramette, 2014).

A cluster analysis was used to determine the similarities between sites with regards to water quality (Figure 6-3). When conducting the (agglomerative) cluster analysis, a hierarchy of levels is used to form a dendrogram to indicate sub-groups within groups. The two most similar objects are joined first forming a new object. This happens repeatedly until all the objects (including the newly formed ones) join according to similarities or dissimilarities to form one final cluster (Lepš and Šmilauer, 2003). Used within the present study was the Ward’s and Complete-linkage (furthest neighbour) methods for clustering criteria (Kändler et al., 2017). One of the reasons for using the complete- linkage algorithm was that the result is more ecologically interpretable (Lepš and Šmilauer, 2003). When a new cluster forms during the complete-linkage, the method uses the dis-similarity between the farthest two members (points) of each group or single object – e.g. those with the smallest similarity. All members in the cluster therefore fall within a region of maximum similarity relative to one another (Buttigieg and Ramette, 2014). When a new cluster forms during the Ward’s method, a “cost” of merge is determined against an objective function. At each stage of the algorithm, merges with the minimum cost are performed. Minimum cost would be clusters with the smallest values when the sum of the squared deviations from the cluster centroids are calculated (Buttigieg and Ramette, 2014).

Results

6.1.2 NMDS (Non-metric multidimensional scaling)

The NMDS plot indicates similarity/dissimilarity between the different sampling sites of this study with regards to land use. Clearly grouped “closer” together are sites 5, 6, 7, 8 and 9 which are located in the Sabie – and Sand River that occur in the KNP which shows a clear separation of the conservation area. As the Inyaka Dam (site 11) is a still standing waterbody, it should be expected to stand alone, as the water parameters of a still standing waterbody will not experience the same environmental conditions as that of a flowing river’s water. Both sites 10 and 12 are close to communities with varying socio-economic status. They are however separated from each other. The Marite River (site 10) for example has higher runoff rate and erosion dongas where the Sand River’s urban township Thulamahashe experiences a flatter terrain. Sites 1, 2, 3 and 4 are all grouped to

91 the left quadrangles and indicate the grouping of sites closer to agricultural and forestry practices with less influence from rural settlements. The Inyaka Dam (site 11) is also surrounded by forestation.

Figure 6-1: Non-metric multidimensional scaling (NMDS) ordination plot of the percentage land use surrounding the river with regards to the sampling localities. The Bray Curtis 2D stress value was 0.02 and the Euclidian distance 2D stress 0.01, indicating a good fit. Sites shown are: 1 – Sabie River Headwaters; 2 – Sabie River After Waste Water Treatment; 3 – Sabie River Before Hazyview; 4 – Sabie River After Hoxane Water Treatment Works; 5 – Sabie River at Kruger National Gate; 6 – Sabie River Skukuza; 7 – Sand River Kruger National Park; 8 – Lower Sabie River after Sabie - Sand Confluence; 9 – Sabie River Bordering Mozambique; 10 – Marite River After Inyaka Dam; 11 – Inyaka Dam Outlet; 12 – Sand River Thulamahashe After Waste Water Treatment Plant.

6.1.3 RDA (Redundancy Analysis) and Cluster

6.1.3.1 Land use influence on water quality

The first RDA conducted was to see if the land use – predictor variables had a significant influence on the water quality – response variables. The sample localities in the RDA triplot (Figure 6-2) can be seen along with the land use and water quality indicators for this analysis. From the eigenvalues it is seen that the first axis describes 56.6% of the cumulative percentage variance of species- environmental relations and the second axis describes 19.3% thereof. Monte Carlo permutation (499 permutations under reduced model) was used for the test of significance (<0.05) of all canonical axes. From the test of significance of first canonical axis and the test of significance of all canonical

92 axes the p-value representing significance is equal to 1 for both. This answers the question above, whether the correlation of land use influence on water quality was significant, and it was not at p > 0.05. Even though the influence is not significant, the relationship between the predictor and response variables can still be explored.

Axis 1 describing 56.6% of the cumulative percentage variance is strongly associated with sites 1, 4, 5, 6, 7, 8, 9 and 12 and the predictor variables: mines, wetlands, urban township, urban village, cultivated subsistence, plantations-woodlots and urban smallholding. Axis 2 describing 19.3% of the cumulative percentage variance is strongly associated with sites 2, 3, 10, 11 and predictor variables water, erosion-bare, inland forestry, urban built-up, urban residential, urban informal and urban lawns. Equally associated with both axes are cultivated commercial, woodland open bush, cultivated orchards, grassland-low shrubland, urban informal. A strong positive correlation exists between sites 1, 10 and 11 and response variables SS, turbidity, Mn2+, Al3+ and Fe2+/3+. Plantation-woodlots; indigenous forestry; cultivated commercial and water land use classes were the predictor variables with strongest influence on these response variables. As mentioned before, site 11 represents the Inyaka Dam, the strong positive correlation observed is thus expected between site 11 and the predictor variable – water. However, the indigenous forestry, plantation-woodlots, cultivated commercial and erosion classes are still evident in these areas and still positively correlate with these sites. These positive correlations are due to the fact that plantations-woodlots and partly indigenous forestry make up the dominant land uses in these catchments (Table 5-4). This may result in organic accumulation, decomposition and leaching of metals through the forest floor into the groundwater that later connect with streams. What might contribute to high Mn2+, Al3+ and Fe2+/3+ metals are the main geological occurrences for these watersheds that are shale (site 1), granite and tonalite (sites 10 and 11). Granite contains minerals high in Ca2+, Na+ (see below for more detail) and importantly for this site Al3+. Tonalite, composed of various minerals including feldspar-plagioclase (containing

+ + + 2+ 2+ NaAlSi3O8 and CaAl2Si2O8); accessory minerals such as amphiboles (high in Na , K , Li , Ca , Mn ,

3+ 2+ 3+ 2+ 2+ Al , Fe / , Zn , Mg ); and pyroxenes, with general formula AB Si2O6, (where A is mainly

2+ 2+ Mg,Fe ,Ca,Na and B is mainly Mg,Fe or Al and Si may be replaced in part by Al) are associated with these sites (SCWG, 1991). This observation is supported by the high Fe2+/3+ concentrations identified in Chapter 4 for sampling sites 10 and 11, where it was first identified that Fe2+/3+ is in exceedance of the SANS241:2015 guideline posing an aesthetic risk. Also discussed in Chapter 4 was the high turbidity seen at site 11 (Inyaka Dam) which had a higher average turbidity (9.95 NTU) than the rest of the sites. The turbidity was positively associated with metal concentrations Fe2+/3+ and Mn2+ as supported by the RDA below. The water contained in the Inyaka Dam has been exposed to various geological layers within the area. This, in my opinion will have an influence on water quality, as the exposure of rock and bottom soil introduces new weathering opportunity for rocks. According to DWAF (2006a) residual granite soil is the main constituent of the valley flanks that are interspersed by completely weathered dolerite dykes; and the dam wall, founded on highly erodible 93 decomposed granite with intrusions of diabase dykes. The main characteristic minerals in diabase are plagioclase and pyroxene with lesser amounts of minerals highly associated with Mg2+ and Fe2+ namely magnetite, olivine, biotite, hornblende, ilmenite and also chlorite containing minerals.

Urban influence had strong correlations with sites 2 and 3. Strong positive responses were seen by

2+ 2+ 2- water quality variables Mg , pH, Hardness, Ca and SO4 with regards to urban lawns, urban built- up and urban informal. Hardness, increased as a result of higher concentrations of Mg2+ and Ca2+ that have their main source in urban watersheds from the natural geology of the catchment (DWAF, 1996a). In the case of watersheds 2 and 3, it is well explained by geology, as the two dominant geological rock types in the watersheds are dolomite and granite. Dolomite represents a direct

2+ 2+ source of Ca and Mg as its chemical formula is CaMg(CO3)2 and granite, a direct source of cations such as Ca+ and Na+ (King, 2013). Granite’s mineral composition varies, but one of its minerals

(plagioclase feldspar) have a chemical formula ranging from NaAlSi3O8 to CaAl2Si2O8 (SCWG, 1991). Sites 4 and 5 have strong positive correlations with predictor variables cultivated orchards, grasslands-low shrubland and urban informal. Related to these is the response variable DO. Directly opposite from DO and therefore relating negatively with sites 4, 5, cultivated orchard and grasslands- low shrubland are response variables TOC, DOC and TKN. This is an indication that urban influences (urban informal predictor variable) associated with watershed 4 and 5 did not have such a great effect on water quality, as TOC, DOC and TKN usually indicate organic pollution and bacterial activity associated with sewage pollution. Sites 6 and 8 have strong positive correlations with predictor variables grasslands-low shrubland, which is the second highest percentage land use at site 8, and third highest at site 6, urban informal, urban lawns, and urban built-up which is a representative of KNP staff and community housing in Skukuza. These sites correlate positively with response variables Mg2+ and pH and negatively with Zn2+. Site 7 correlated positively with the land use predominantly found within watershed 7 (Table 5-4) namely woodland-open bush which had a

2+ 2+ 2- strong positive influence on response variables Mg , pH, Hardness, Ca and SO4 where granite is once again the main geological occurrence (59.1% surface area) in watershed 7. It was noticed that all of the conservation sites grouped closely together (sites 6, 7, 8 and 9). Site 9 projected close to the origin, indicating a weak association in comparison to site 12 with all the variables. Originating from here are variables such as SPC, TDS, Cl-, Na+, K+, TOC, COD, DOC, Tsi, all of the biological indicators as well as all of the nutrients. Site 12 correlates strongly with predictor variables that represent rapidly-developing rural circumstances e.g. urban village, urban townships, cultivated subsistence, mines, wetlands and the above-mentioned water quality parameters (response variables). This high degree of correlation is supported by the cluster analysis (Figure 6-3) as this is the site that was the most dissimilar from all the rest. The positive correlations of these water quality variables were also seen and discussed in Chapter 4. This now gives further support to the negative effects associated with the lower income settlements or urban township/village at site 12 namely Thulamahashe. 94 Axes 1 2 3 4 Total variance Eigenvalues: 0.566 0.193 0.095 0.05 1 Species-environment correlations: 1 1 1 1 Cumulative percentage variance of species data: 56.6 75.9 85.4 90.4 of species-environment 56.6 75.9 85.4 90.4 relation: Sum of all eigenvalues 1 Sum of all canonical eigenvalues 1 All four eigenvalues reported above are canonical and correspond to axes that are constrained by the environmental variables. 1 *** Unrestricted permutation *** Seeds: 23239 945 **** Summary of Monte Carlo test **** Test of significance of first canonical axis: Eigenvalue = 0.566 F-ratio = 0 P-value = 1 Test of significance of all canonical axes: Trace = 1 F-ratio = 0 P-value = 1 (499 permutations under reduced model)

Figure 6-2: Redundancy Analysis (RDA) triplot of land use percentages (predictor) effect on mean measured water quality parameters (response), at the 12 different sampling sites (orange dots) for the Sabie-Sand catchment.

95 Figure 6-3: Dendrogram showing the cluster classification of the watersheds of the 12 sampling sites in the Sabie-Sand catchment, according to their land-use influences on water quality parameters. Break was inserted in the axis between 20 000 and 2 500 000. Inset: The smaller inset is the full Euclidean distances between the clusters before axis breakage. Interpretation of the large axis was difficult and therefore the above-mentioned break was introduced.

The smaller inset (Figure 6-3) of the land use and water quality cluster analysis can be seen as the full Euclidean distances between the clusters before axis breakage. The axis break was introduced between 20 000 and 2,5milion resulting in the larger figure 6-3 as interpretation was difficult. Supporting the correlations seen in the RDA is the cluster analysis depicting the land use influence on water quality variables (Figure 6-3). The first cluster could be distinguished between site 12 and all the other sites, forming a cluster with the biggest Euclidean distance. The group that had the biggest Euclidean distance from site 12 was the cluster formed by sites 1 and 11 (which are therefore also regarded as the other two sites that cause possible variance in the data). This difference in water quality was a direct result of the difference in land use as has been discussed. Related to broad land use classes, site 1 and 11 group closely together possibly as result of the majority plantation-woodlot dominated land use (also seen in the RDA). The second cluster distinguishes

96 between sites 8 and 9 and the rest. This similarity lies in the fact that sites 8 and 9 within the lower Sabie River are both subject to the conservation tenure class in the KNP. Site 9 is the Sabie River site which was most dissimilar from the other Sabie River sites. This is also seen in the RDA and is supported by the difference in water quality as discussed in chapter 4. The third cluster distinguishes between sites 1, 11, 4 and 7 and sites 2, 3, 5, 6, and 10. Sites 4 and 7 cluster closely together (due to similarity), possibly as a result of major land use as woodland-open bush. While site 4 and site 7 are separated on the RDA diagram, the cluster analysis shows that the influences experienced from commercial land use, such as cultivated orchard and plantations-woodlots at site 4 impact the water quality of this site, to a higher degree than woodland-open bush does.

Sites 2,3,5,6, and 10 cluster closely together, firstly as similarly influenced by major plantation- woodlot classes and secondly possibly as a result of woodland-open bush. Only site 10 of this cluster, close to a site of still standing water on the RDA did not group closely to the other sites during the RDA analysis.

6.1.3.2 Soil and slope combination influence on water quality

The second RDA conducted was to see if soil – predictor variables had an influence on the water quality response variables. The sample localities can be seen in the RDA triplot (Figure 6-4) along with the soil variables and the water quality indicators with co-variable class namely slope. From the eigenvalues it can be seen that the first axis describes 47% of the cumulative percentage variance of species-environmental relations and the second axis describes 19.2% thereof. From the test of significance of first canonical axis and the test of significance of all canonical axes the p-value representing significance is equal to 0.208 and 0.056 respectively. This answers the question above, that the correlation of soil influence on water quality was not significant as p>0.05. Even though the influence is not significant, there can still be remarked that a correlation exists between the species and environmental variables. Watershed slope as percentage rise was incorporated by use of slope classes: class 1 (0-5%); class 2 (>5-15%); class 3 (>15-25%); class 4 (>25%).

Axis 1 describing 47% of the cumulative percentage variance is strongly associated with sites 1, 4, 6, 7, 8, 12 and predictor variables Hb, soil depth, soil clay, mean river slope and the second axis describing 19.2% of the cumulative percentage variance is strongly associated with sites 2, 3, 5, 11 and predictor variables Ac, Fa, Fb, Ea, Dc and Ib. Equally associated with both axes are site 10 and predictor variables Ab and Ae. The watersheds that would receive the most influence from the slope factor would possibly be those watersheds indicating signs of increased runoff. A strong positive correlation exists between site 1 and mean watershed slope. This is validated by the fact that site 1 is the highest point measured in the study and is situated in the headwaters for the Sabie River. Within river profiles the headwaters testify of higher stream runoff (due to steep slope). This will usually increase the material load entrained by a river, but as the negative association with site 1

97 and TDS, SS and Mg2+ exist, this might not be true for a watershed with slope of class 4. This therefore indicates that runoff as a magnitude of slope did not have a large influence on water quality at site 1 as expected. The opposite is true for site 11. The slope class 3 of watershed 11 correlated strongly with predictor variable Ib (rocky areas with miscellaneous soils) and response variables turbidity, SS, Ca2+ as well as Mg2+. The stronger turbidity influence of site 11 could therefore be affiliated with higher runoff due to the higher percentage rise in slope. Sites 4 and 5 have strong positive correlations with predictor variables Ae (freely drained soils), Fb (low lying soils), mean river slope and slope co-variable class 2. Slope had an influence on watersheds’ 4 and 5 water quality as class 2 slope indicate low percentage rise and was negatively correlated with TDS, possibly indicating low amounts of runoff. This is also true for site 8. A strong positive correlation exists between site 8 and predictor variables Ae, Fb, mean watershed slope and slope co-variable class 2 and response variables DO. These variables do not indicate a high amount of runoff, typical of the lower percentage rise as seen at sites 4 and 5. Site 9 positively relates to Dc (Prismacutanic and/or pedocutanic soils and red structured soils), Ea (red structured soils) predictor variables and slope co-variable class 3 and Mg2+, Turbidity, SS and Ca2+ response variables. The higher slope class 3 association with turbidity again testifies that the co-variable influences the water quality. The lower percentage rise of sites 7 and 12 (association with slope class 2) is not indicative of water quality for these two sites. As they are strongly positive correlated with predictor variables Hb and response

2+ + 2+ 2- variables TDS, Mn , Na , all of the nutrients, Zn , SO4 , TOC, and SPC it should have rather indicated association with a higher slope class. TDS response variable should rather indicate strong association with higher amount of runoff, but as this is not true for sites 7 and 12, it cannot be said that slope influenced these sites. The water quality at site 12 according to this study is more strongly modified by land use than soil type and slope. Site 12’s positive association with Hb land type and soil depth could be an indicator of influence on water quality as the soil depth at site 12 was 933.2mm indicating thick soils. This land type related to grey regic sand type soils, which are very loose soils characterised by a low to very low clay percentage (7.6%). Soil infiltration is therefore suspected to be high within this watershed, which may result in the increased ion and metal concentrations seen for watershed 12 derived from the surrounding soils. Site 10 has strong positive correlation with predictor variables Ab (a freely drained soil), soilclay, slope co-variable class 2 and COD response variable. Negative association with TDS at this site indicates that the lower percentage rise of class 2 did have an influence on water quality. Furthermore, the clay percentage in this watershed was 33.7% indicating higher amounts of clay, unfortunately this does not indicate why COD had a strong positive association with site 10.

98

Axes 1 2 3 4 Total variance Eigenvalues: 0.47 0.192 0.041 0.025 1 Species-environment correlations: 1 1 1 1 Cumulative percentage variance of species data: 61.8 87.1 92.4 95.7 of species-environment relation: 61.8 87.1 92.4 95.7 Sum of all eigenvalues 0.76 Sum of all canonical eigenvalues 0.76 The sum of all eigenvalues is after fitting covariables Percentages are taken with respect to residual variances i.e. variances after fitting covariables All four eigenvalues reported above are canonical and correspond to axes that are constrained by the environmental variables. 1 *** Unrestricted permutation *** Seeds: 23239 945 **** Summary of Monte Carlo test **** Test of significance of first canonical axis: Eigenvalue = 0.47 F-ratio = 0 P-value = 1 Test of significance of all canonical axes: Trace = 0.76 F-ratio = 0 P-value = 1 (499 permutations under reduced model)

Figure 6-4: Redundancy Analysis (RDA) triplot of soil percentages (predictor) effect on mean measured water quality parameters (response), at the 12 different sampling sites (black dots) with regards to a co-variable namely slope for the Sabie-Sand catchment. Watershed slopes in percentage rise represented as the co-variables were divided into classes: Class 1 (0-5%) (not shown in RDA); class 2 (>5-15%); class 3 (>15-25%); class 4 (>25%).

99 Discussion

Based on a one-year sampling period, the correlation between land use proportions and water quality indicators for the Sabie-Sand catchment was analysed. The results showed that rural land use indicators and most of the environmental stress indictors of water quality were highly positively correlated in the Sand River in watershed 12 (as seen in Figure 6-2); and commercialised agriculture such as cultivated commercial fields and plantation-woodlots and the nutrient enrichment water quality indicators were negatively correlated in the most western watersheds. None of these correlations were however significant. This is possibly due to the relatively good quality of the water found within this catchment except site 12.

In terms of “lack” in nutrient enrichment from DIN and degree of conductivity (SPC), forest plantations and cultivated commercial fields had a positive effect on water quality (Figure 6-2), during a dry rainfall period or low baseflows of the upland area in this specific catchment. This type of positive effect was also seen by Xizhi et al. (2017). They describe an increase in water quality with the increase in forested areas, as a result of increase in forest undergrowth. Undergrowth serves as a buffer for soil erosion during rainstorm runoff and can also reduce nutrient inflow by serving as adsorption medium (Xizhi et al., 2017).

- In terms of organic enrichment indicated by TOC, DOC, NH3 and coliforms, rural influence and more specifically the point source pollution of wastewater treatment plant at site 12 had a negative influence on water quality. This is also substantiated by the negative correlation with DO concentrations at this site.

The negative correlation of turbidity with most of the Sabie River sites occurring within conservation areas suggests a positive influence on water quality from this land use/ tenure class. As a result of the negative correlation between site 11 and pH, it can be said that indigenous forestry, which makes up a large land use portion of watershed 11, has a possibly “acidifying” effect on water quality. It also had an effect on water turbidity, because turbidity, Mn2+, Fe2+/3+, Al3+ and SS correlate positively with site 11, slope class 3 and indigenous forestry. The same is true for watershed 1 where indigenous forestry as well as commercial plantations contribute to the watershed’s main land uses.

The site near an urban village, site 10, provides evidence of high runoff, as erosion-bare ground has a strong positive correlation with this site and water quality parameters turbidity, Mn2+, Fe2+/3+, Al3+ and SS (Figures 6-2) as well as indicating a strong correlation with slope class 3 which may contribute to this strong correlation with erosion and SS. Attention should be drawn to the heavy metals at sites 10 and 11 as high Al3+ concentrations in the presence of Mn2+ and Fe2+/3+, (as mentioned before) can cause discoloration effects and make the water less suitable for domestic uses.

100 The communal tenure class makes up a large percentage of land use area in this study especially within the central part of the catchment. Especially in watershed 12, this gives rise to excess nutrients within the river waters, probably through soil infiltration and runoff. Areas practising a basic degree of sanitation (e.g. pit latrines), slurry tanks and cattle grazing along riverbanks are suspected of discharging nutrients intermittently into streams (Preedy et al., 2001 and Withers et al., 2003 as cited by Jarvie et al., 2008). Septic tanks and sewage treatment works are also rural sources of nutrient effluent (Neal et al., 2005; Neal and Jarvie, 2005 as cited by Jarvie et al., 2008).

Livestock farming (cattle, goats, pigs and poultry), and crop farming (maize, vegetables, etc.) are the dominant subsistence farming practices observed in this rural area. In certain cases, livestock farming can introduce nitrogen and phosphorus into the streams (Haygarth et al., 1998 and Withers and Lord, 2002 as cited by Jarvie et al., 2008). Withers and Bailey (2003) (as cited by Jarvie et al., 2008) is also of opinion that a risk of soil erosion and its associated nutrient losses can occur from growing fodder crops such as maize. As these livestock farmers are mostly small-scale subsistence farmers they are not suspected of increasing river nutrient concentrations at such a high level within the catchment, as would be the case with increased manures, fertilizers and feed concentrates for animals in larger farm operations. What might result in higher nutrient values for the catchment is the large woodlot-open bush land use of the catchment. The exportation of nutrients from grasslands can be facilitated via runoff, either in the form of a solution or in association with particles and colloids (Haygarth et al., 1998, 2005a, 2006; Heathwaite and Dils, 2000; Turner and Haygarth, 2000 and Heathwaite et al., 2005, as cited by Jarvie et al., 2008).

Using soil as water quality indicator could be substantiated for most of the catchment except for sites 1, 2, 3 and 12 as sites 1-3 were associated with a higher slope but not with high turbidity and site 12 was associated with high TDS but not a high percentage rise. Co-variable slopes of class 3 had positive associations with sites indicating higher runoff and the negative associations of slope class 2 (lower percentage rise) indicated the inverse relationship with runoff.

101 Conclusion

Once again, land use on its own is a dynamic factor that can either contribute to a water system’s quality deterioration, stability or in rare cases enhancement. The main aim of this study was to assess the land use – water quality association of the catchment. A secondary aim, was the assessment of whether expansion of rural settlements has taken a toll on water quality.

Plantations-woodlots influence water quality, to a lesser extent than urbanization. The largest negative influence on water quality within the catchment is the rural township and associated influences in watershed 12. The influence of land use on water quality was therefore shown by the degree of degrading influence from a rural settlement and its associated ineffective infrastructure at for example the WWTP (site 12); as well as the remedial effect that woodland-open bush of the conservation tenure class had on the water quality, as observed at sites 6, 7, 8 and 9. The decreases in water quality parameters of concern from site 12 to site 7 suggested that there was “remedial” effect due to conservational practices within the Sand River. It was established that although urbanization and surrounding land uses had a negative effect on the water quality of the Sand River, they had less influence on the quality of the Marite and Sabie rivers. This is important because 6% of the land use for watershed 10 is classified as urban village, known as Bushbuckridge.

A need existed to enhance the knowledge pertaining to: “needs of resource-poor people and the links between land activities and river health” (Van Wyk et al., 2001). As the links have now been made between land activities that have the most influence on water quality, within the present study, further research now needs to be defined in a way for remediation towards the needs of resource- poor people.

Support should be given to this urban settlement in the form of adequate water supply and sanitation facilities, starting with the known wastewater treatment plant point polluter. Adequate maintenance and monitoring should be insisted on at the plant to ensure effluent that conforms to national water quality guidelines.

102 References

Abu-hashim, M., Mohamed, E. and Belal, A. 2015. Identification of potential soil water retention using hydric numerical model at arid regions by land-use changes. International Soil and Water Conservation Research, 3:305–315.

Africa Check. 2016. Frequently asked questions about South Africa’s drought. https://africacheck.org/2016/02/03/frequently-asked-questions-about-south-africas-drought/ Date of access: 3 May 2018.

APHA. 2013. Standard Methods for the Examination of Water and Wastewater. 22nd ed. American Public Health Association, Washington D.C.

Ashton, P.J., Earle, A., Malzbender, D., Moloi, M.B.H., Patrick, M.J. and Turton A.R. 2006. A Compilation Of All The International Freshwater Agreements Entered Into By South Africa With Other States. WRC Report No. 1515/1/06.

Atkinson, J.F., De Pinto, J.V. & Lam, D. 1999. Water quality. Potential Climate Change Effects on the Great Lakes Hydrodynamics and Water Quality, D. Lam and W. Schertzer, eds., American Society of Civil Engineers, Reston, Virginia.

Ayres, P. 2012. Continuous riparian vegetation change following a large, infrequent flood along the Sabie River, Kruger National Park. Potchefstroom: NWU. (Dissertation – MSc).

Baishya, B. 2006. Modelling Hydrological Components and Impact of Land Use/ Land Cover Change on Hydrological Regime. Dehradun: Andhra University. (Thesis – PhD).

Baker, A. 2003. Land use and water quality. Hydrological Processes, 17:2499-2501.

Barsi, J.A., K., Lee, G., Kvaran, B.L., Markham and J.A., Pedelty. 2014. The Spectral Response of the Landsat-8 Operational Land Imager. Remote Sensing, 6:10232-10251.

Bartram, J. and Helmer, R. 1996. Elements of a water quality monitoring programme. (In Bartram, J. and Ballance, R., ed. Water Quality Monitoring: A Practical Guide to the Design and Implementation of Freshwater Quality Studies and Monitoring Programmes. London: UK. p. 1- 366).

Beare, S. and Heaney. A. 2002. Climate change and water resources in the Murray Darling Basin, Australia: Impacts and possible adaptation. ABARE Conference Paper 02.11. 103 Belcher, A. 2004. The Ecological Reserve and River Health Monitoring: A Practical Approach in Water Resource Management. Proceedings of the 2004 Water Institute of Southern Africa (WISA) Biennial Conference, Cape Town, South Africa, 2-6 May. Produced by: Document Transformation Technologies Organised by Event Dynamics ISBN: 1-920-01728-3. p. 248 -253.

Beumer, J. & Mallory, S. 2014. Water Requirements and Availability Reconciliation Strategy for the Mbombela Municipal Area. South Africa. Department of Water Affairs. Report Nr.: PWMA 05/X22/00/2012/6.

Bobba, A.G., Singh, V.P., Berndtsson, R. & Bengtsson, L. 2000. Numerical Simulation of Saltwater Intrusion into Laccadive Island Aquifers due to Climate Change. Journal of the Geological Society of India, 55(63):589-612.

Bonan, G.B. 2002. Ecological Climatology, Chapter 14. 1st ed. Cambridge University Press, Cambridge. http://www.cgd.ucar.edu/tss/aboutus/staff/bonan/ecoclim/index-2002.htm Date of Access: 1 April 2017.

Boorman, D.B. 2003. LOIS in-stream water quality modelling. Part 2. Results and scenarios. The Science of the Total Environment, 314-316:397-409.

Boulton, A.J. 1999. An overview of river health assessment: philosophies, practice, problems, and prognosis. Freshwater Biology, 41:469-479.

Bouraoui, F., Grizzetti, B., Granlund, K., Rekolainen, S. & Bidoglio, G. 2004. Impact of climate change on the water cycle and nutrient losses in a Finnish catchment. Climatic Change, 66:109- 126.

Brauman, K.A., Daily, G.C., Duarte, T. K. & Mooney, H.A. 2007. The Nature and Value of Ecosystem Services: An Overview Highlighting Hydrologic Services. Annual Review of Environment and Resources, 32:6.1–6.32.

Bredenkamp, G.J., Van Rooyen, N. & Theron, G.K. 1991. A survey of the riparian vegetation of the Sabie River in the Kruger National Park. Kruger National Park Rivers Research Programme. Database: Waters & Oceans Worldwide.

Brierley, G. P. 2010. Landscape memory: the imprint of the past on contemporary landscape forms and processes. Area, 42:76–85.

104 Burgoyne, P.M. 1995. Phytosociology of the north-eastern Transvaal high mountain grasslands. Pretoria: UP. (Dissertation – MSc). p.212.

Buttigieg, P.L. & Ramette, A. 2014. A Guide to Statistical Analysis in Microbial Ecology: a community-focused, living review of multivariate data analyses. FEMS Microbiology Ecology, 90(3):543–550. https://mb3is.megx.net/gustame/constrained-analyses/rda Date of Access: 1 Sept. 2017.

Callaway, J.M. 2004. Adaptation benefits and costs: are they important in the global policy picture and how can we estimate them. Global Environmental Change, 14:273–282.

Census. 2011. Census 2011 Community Profile Databases. https://census2011.adrianfrith.com/place/877069 Date of Access: 17 Nov. 2017.

Chambers, P., DeKimpe, C., Foster, N., Goss, M., Miller, J. & Prepas, E. 2001. Agricultural and Forestry Land Use Impacts. (In Environment Canada. Threats to Sources of Drinking Water and Aquatic Ecosystems Health in Canada: National Water Research Report Series No.1. Canada: National Library of Canada Cataloguing). (in Publication). p. 69.

Chen, Q., Mei, K., Dahlgren, R.A., Wang, T., Gong, J., & Zhang, M. 2016. Impacts of land use and population density on seasonal surface water quality using a modified geographically weighted regression. Science of the Total Environment, 572:450–466.

Chen, Q., Wu, W., Blanckaert, K., Ma, J., Huang, G. 2012. Optimization of water quality monitoring network in a large river by combining measurements, a numerical model and matter- element analyses. Journal of Environmental Management, 110:116–124.

Chen, Z., Grasby, S.E. & Osadetz, K.G. 2004. Relation between climate variability and groundwater levels in the upper carbonate aquifer, southern Manitoba, Canada. Journal of Hydrology, 290:43-62.

CNES. 2009. Centre National d’Etudes Spatiales. https://web.archive.org/web/20131006213713/http://www.cnes.fr/web/CNES-en/1415-spot.php Date of access 3 May 2018.

Coetzer, K.L., Erasmus, B.F., Witkowski, E.T. and Bachoo, A.K. 2010. Land-cover change in the Kruger to Canyons Biosphere Reserve (1993– 2006): A first step towards creating a conservation plan for the subregion. South African Journal of Science, 106(221):1-10.

105 Cousins, T., Dlamini, V., Smits S. & Maluleke, N. (n.d). Planning for a Multiple Use System Approach at Local Level: experiences from Bushbuckridge, South Africa. http://award.org.za/wp- content/uploads/sites/2/2014/03/Planning-for-a-Multiple-Use-System-Approach-at-Local-Level.pdf Date of access: 24 May 2016.

Dabrowski, J.M. & De Klerk, L.P. 2013. An assessment of the impact of different land use activities on water quality in the upper Olifants River catchment. Water SA, 39(2):231-244.

Dalas, H.F. & Day, J.A. 2004. The Effect of Water Quality Variables on Aquatic Ecosystems: A Review. WRC Report No: TT 224/04.

DEA. The Department of Environmental Affairs. 2016. EGIS – Environmental Geographic Information Systems. https://egis.environment.gov.za/gis_data_downloads Date of access: 1 February 2016.

Dell Inc. 2016. Dell Statistica (data analysis software system), version 13. software.dell.com.

Dempsey, C. 2017. What is GIS? https://www.gislounge.com/what-is-gis/ Date of access: 3 Sept. 2017.

Di Gregorio, A & Jansen, L.J.M. 2000. Land Cover Classification System (LCCS): Classification Concepts and User Manual. Food & Agriculture Organization. http://www.fao.org/docrep/003/x0596e/X0596e00.htm#P-1_0 Date of access 27 Sept. 2017.

Ding, J., Jiang, Y., Liu, Q., Hou, Z., Liao, J., Fu, L. & Peng, Q. 2016. Influences of the land use pattern on water quality in low-order streams of the Dongjiang River basin, China: A multi-scale analysis. Science of the Total Environment, 551-552:205-216.

Donohue, I., McGarrigle, M. L. & Mills, P. 2006. Linking catchment characteristics and water chemistry with the ecological status of Irish rivers. Water Research, 40:91-98.

Du Plessis, A., Harmse, T. & Ahmed, F. 2014. Quantifying and predicting the water quality associated with land cover change Blesbok Spruit catchment. Water, 6:2946-2968.

DWAF. Department of Water Affairs and Forestry. 1996a. SOUTH AFRICAN WATER QUALITY GUIDELINES 2nd Ed. Volume 1: Domestic Water Use.

106 DWAF. Department of Water Affairs and Forestry. 1996b. SOUTH AFRICAN WATER QUALITY GUIDELINES 2nd Ed. Volume 7: Aquatic ecosystems.

DWAF. Department of Water Affairs and Forestry. 1996c. SOUTH AFRICAN WATER QUALITY GUIDELINES 2nd Ed. Volume 8: Field guide.

DWAF. Department of Water Affairs and Forestry. 1999. The Water Management Areas of South Africa - Catchment Management Report. http://www.dwaf.gov.za/CM/Docs/WORD/finalwma.doc Date of Access: 10 July 2017.

DWAF. Department of Water Affairs and Forestry. 2004. Internal Strategic Perspectives: Inkomati Water Management Area – Version 1. Tlou & Matji (Pty) Ltd. DWAF Report No.: P WMA 05/000/00/0303.

DWAF. Department of Water Affairs and Forestry. 2006a. Resource Management Plan Inyka Dam – Draft 2. Pretoria. p.10-12.

DWAF. Department of Water Affairs and Forestry. 2006b. Report of the Department of Water Affairs and Forestry to the Minister of Water Affairs and Forestry, 1 April 2005 to 31 March 2006.

DWAF. Department of Water Affairs and Forestry. 2006c. Resource Quality Services (was IWQS) and Chief Directorate of Surveys and Mapping – Water and Sanitation. Water quality data exploration tool: https://www.dwa.gov.za/iwqs/wms/data/000key2data.asp Date of access: 1 February 2016.

DWAF. Department of Water Affairs and Forestry – Water and Sanitation. 2012. Water quality data exploration tool: the nine South African water management areas. https://www.dwa.gov.za/iwqs/wms/data/000key2data.asp Date of access: 1 February 2016.

DWS. Department of Water and Sanitation. 2017a. Hydrological Services – Surface Water (Data, Dams, Floods and Flows). http://www.dwa.gov.za/Hydrology/Verified/hymain.aspx Date of access: 26 May 2017.

DWS. Department of Water and Sanitation. 2017b. Resource Quality Information Services – Water quality data exploration tool: layers. Supporting data for use in Google Earth http://www.dwa.gov.za/iwqs/wms/data/000key2data.asp Date of access: 1 February 2016.

Dye, P.J. 1996. Climate, forest and streamflow relationships in South African afforested catchments. The Commonwealth Forestry Review, 75(1):31-38.

107 Erasmus, A. 2017. Spatial changes in algal assemblages promoted by water quality in the Sabie River catchment. (Unpublished). [Personal correspondence]. 30 Oct., Potchefstroom.

Esri, 2015. ArcGIS Desktop: ArcMap Release 10.4. Redlands, CA: Environmental Systems Research Institute.

Esri. 2016a. ArcGIS for Desktop. ArcMap – Surface toolset concepts: How Slope works. http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/how-slope-works.htm Date of access: 11 Sept. 2017.

Esri. 2016b. ArcGIS for Desktop. ArcMap – Overlay toolset concepts: How Weighted Overlay works. http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/how-weighted- overlay-works.htm Date of access: 11 Sept. 2017.

Esri. 2016c. ArcGIS for Desktop. ArcMap – Distance toolset concepts: Creating the least-cost path. http://pro.arcgis.com/en/pro-app/tool-reference/spatial-analyst/creating-the-least-cost-path.htm Date of access: 11 Sept. 2017.

Ferrier, R. & Edwards, A. 2002. Sustainability of Scottish water quality in the early 21st century. The Science of the Total Environment, 294:57-71.

Fisher, A. 2000. Preliminary findings from the mid-Atlantic regional assessment. Climate Research, 14(3):261-269.

Foley, J.A., De Fries, R., Asner, G.P., Barford, C., Bonan, G., Carpenter, S.R., Chapin, F.S., Coe, M.T., Daily, G.C., Gibbs, H.K., Helkowski, J.H., Holloway, T., Howard, E.A., Kucharik, C.J., Monfreda, C., Patz, J.A., Prentice, I.C., Ramankutty, N. & Snyder, P.K. 2005. Global Consequences of Land Use. Review Article. Science, 309:570- 574.

Food and Agriculture Organization of the United Nations (FAO). 2002. Land Tenure Studies. Chapter 3. Land tenure and rural development. pp.7.

GeoSience. 2003. South Africa. 1996. Department of Environmental Affairs and Forestry. Geology layer. Council for GeoScience, M. Roos. Pretoria. http://www.geoscience.org.za/index.php/publication/ Date of access: 2003.

108 GeoTerraimage (South Africa). 2015. 2013 – 2014 South African National Land-Cover Dataset. DEA/CARDNO SCPF002: Implementation of Land-Use Maps for South Africa. Project Specific Data User Report and MetaData.

Gertenbach. W.P. 1983. Landscapes of the Kruger National Park. Koedoe, 26(1):9-121.

GlobalMapper. 2016. http://www.bluemarblegeo.com/products/global-mapper-download.php Date of access: 1 February 2016.

GoogleEarthPro. 2015. Satellite imagery from various satellites and providers: Landsat / Copernicus; Digital Globe and CNES / Airbus https://www.google.com/earth/download/gep/agree.html Date of access: 1 February.

Goudie, A.S. 2004. Encyclopedia of Geomorphology. Vol.1, 1A-I London: Routledge. Taylor & Francis e-Library, 2006.

Haidary, A., Amiri, B. J., Adamowski, J., Fohrer, N. & Nakane K. 2013. Assessing the Impacts of Four Land Use Types on the Water Quality of Wetlands in Japan. Water Resource Management, 27:2217–2229.

Hair, J.F., Jr., Black, W.C., Babin, B.J. & Anderson, R.E. 2009. Multivariate data analysis. 7th ed. Upper Saddle River, NJ: Pearson Prentice Hall. p. 116 - 120.

Hall, G.V, D’Souza, R.M & Kirk, M.D. 2002. Foodborne disease in the new millennium: out of the frying pan and into the fire? Medical Journal Australia, 177:614-618.

Hallett, C.S., Valesini, F., Scanes, P., Crawford, C., Gillanders, B.M., Pope, A., Udy, J., Fortune, J., Townsend, S., Barton, J., Ye, Q., Ross, D.J., Martin, K., Glasby, T. & Maxwell, P. 2016. A review of Australian approaches for monitoring, assessing and reporting estuarine condition: II. State and Territory programs, Environmental Science & Policy, (66) p. 270-281.

Harding, W.R. and Paxton, B. 2001. Cyanobacteria in South Africa: A review. WRC Report. No. TT 153/01.

Harker, B., Lebedin, J., Goss, M.J. Madramootoo, C., Neilsen, B.P & Van der Gulik, T. 2004. Land-Use Practices and Changes – Agriculture. (In Brannen, L. and Bielak, A.T., eds., Environment Canada. Threats to Water Availability in Canada: NWRI Scientific Assessment Series

109 Report No. 3 and ASCD Science Assessments Series No. 1. Canada: National Library of Canada Cataloguing in Publication p. 49-54).

Haron, M. & Dragovich, D. 2010. Climatic influences on dryland salinity in central west New South Wales, Australia. Journal of Arid Environments, 74:1216-1224.

Heritage, G.L. & Van Niekerk, A.W. 1995. Drought conditions and Sediment transport in the Sabie River. Koedoe, 38(2):1-9.

Heritage, G.L., Large, A.R.G., Moon, B.P. and Jewitt, G. 2004. Channel hydraulics and geomorphic effects of an extreme flood event on the Sabie River, South Africa. Catena, 58:151– 181.

Hurd, B.H., Callaway, M., Smith, J. & Kirshen, P. 2004. Climatic change and US water resources: From modelled watershed impacts to national estimates. (JAWRA) Journal of the American water resources association, 40(1):129-148.

Jarvie, H.P., Haygarth, P.M., Neal, C., Butler, P., Smith, B., Naden, P.S., Joynes, A., Neal, M., Wickham, H., Armstrong, L., Harman, S. & Palmer-Felgate, E.J. 2008. Stream water chemistry and quality along an upland- lowland rural land-use continuum, south west England. Journal of Hydrology, (350):215-231.

Jarvis, A., Reuter, H.I., Nelson, A. & Guevara, E. 2008. Hole-filled SRTM for the globe Version 4, available from the CGIAR-CSI SRTM 90m Database (http://srtm.csi.cgiar.org). Downloaded via Global Mapper: http://www.bluemarblegeo.com/products/global-mapper-download.php Date of access: 1 February 2016.

Jewitt, G.P., Weeks, D., Heritage, G., Gorgens, A.H. & O’Keeffe, J.H. 1997. The ecological significance of floods in the Sabie river, South Africa: A simulation exercise. In: Proceedings of the 3rd International Conference on River Flood Hydraulics,5–7 November 1997. University of Stellenbosch: Stellenbosch.

Jiménez, B. 2003. Health risks in aquifer recharge with recycle water. State of the Art Report: Health Risk in Aquifer Recharge Using Reclaimed Water. http://www.who.int/water_sanitation_health/wastewater/wsh0308chap3.pdf Date of Access: 1 March 2017.

110 Johnson, M. R., Anhaeusser, C. R., & Thomas, R. J. 2006. The geology of South Africa. Pretoria, Council for Geoscience.

Kändler, M., Blechinger, K., Seidler, C., Pavlů, V., Šanda, M., Dostál, T., Krása, J., Vitvar, T., Štich, M. 2017. Impact of land use on water quality in the upper Nisa catchment in the Czech Republic and in Germany. Science of the Total Environment, 586:1316–1325.

Kansiime, F., Saunders, M.J. & Loiselle, S.A. 2007. Functioning and dynamics of wetland vegetation of Lake Victoria: an overview. Wetlands Ecological Management, 15:443–451.

Karr, J.R. 1999. Defining and measuring river health. Freshwater Biology, 41:221-234.

Karr, J.R. & Chu, E.W. 1997. Biological Monitoring and Assessment: Using Multimetric Indexes Effectively. EPA 235-R97-001. University of Washington: Seattle. p.1-138.

Kempster, P.L. & Smith, R. 1985. Proposed aesthetic/physical and inorganic drinking–water criteria for the Republic of South Africa. National Institute for Water Research – Council for scientific and industrial research (CSIR), Pretoria. CSIR Research Report No. 628. ISBN: 0-7988 3411-0.

Khadse, G.K, Patni, P.M and Labhasetwar, P.K. 2015. Removal of iron and manganese from drinking water supply. Sustainable Water Resource Management, 1:157–165.

Kimble, G.H., Cooks, J. & Ferreira, C.E. 2014. Veld – Grasslands, Africa. (In Encyclopaedia Britannica, inc. https://www.britannica.com/science/veld Date of access: 11 Nov. 2017).

King, H.M. 2013. A sedimentary rock similar to limestone. Also known as "dolostone" and "dolomite rock." https://geology.com/rocks/dolomite.shtml Date of access: 3 May 2018.

Kressler, F.P. Kim, Y.S. and Steinnocher, K.T. 2003. Object-oriented land cover classification of panchromatic KOMPSAT-1 and SPOT-5 data. Geoscience and Remote Sensing, 19:263-269.

Kumagai, M., Ishikawa, K. & Chunmeng, J. 2002. Dynamics and biogeochemical significance of the physical environment in Lake Biwa. Lakes and Reservoirs: Research Management, 7:345-348.

Kundzewicz, Z.W., Mata, L.J., Arnell, N.W., Döll, P., Kabat, P., Jiménez, B., Miller, K.A., Oki, T., Sen, Z. & Shiklomanov, I.A. 2007. Freshwater resources and their management. (In Parry, M.L., Canziani, O.F., Palutikof, J.P., Van der Linden, P.J. & Hanson, C.E., eds., Climate Change 2007:

111 Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press. p. 173-210). https://www.ipcc.ch/publications_and_data/ar4/wg2/en/ch3s3-4-3.html Date of Access: 22 March 2017.

Kusangaya, S., Warburton, M.L., Archer van Garderen, E. & Jewitt, G.P.W. 2013. Impacts of climate change on water resources in southern Africa: A review. Physics and Chemistry of the Earth, 67–69:47–54.

Land Type Survey Staff 2002. Land type survey of South Africa: Modal profiles. ARC - Institute for Soil, Climate, and Water, Pretoria, South Africa.

Larned, S.T., Scarsbrook, M.R., Snelder, T.H., Norton, N.J. & Biggs, B.J. 2004. Water quality in low-elevation streams and rivers of New Zealand: recent state and trends in contrasting land-cover classes. New Zealand Journal of Marine and Freshwater Research, 38:347-366.

Lee, M.T. & White, D.C. 1992. Applications of GIS databases and water quality modeling for agricultural nonpoint source pollution control. WRC Research Report No. 214.

Lepš, J. & Šmilauer, P. 2003. Multivariate Analysis of Ecological Data Using CANOCO. 1st ed. Camridge: Camridge University Press.

Li, C., Liu, M., Hu, Y., Gong, J., and Xu, Y. 2016. Modeling the Quality and Quantity of Runoff in a Highly Urbanized Catchment Using Storm Water Management Model. Polish Journal of Environmental Studies, 25(4):1573-1581.

Lin, B., Chen, X., Yao, H. Chen, Y., Liu, M., Gau, L. & James, A. 2015. Analysis of land use change impacts on catchment runoff using different time indicators based on SWAT model. Ecological Indicators, 58:55-63.

Lipp, E., Kurz, R. Vincent, R. Rodriguez-Palacios, C., Farrah, S. & Rose, J. 2001. The effects of seasonal variability and weather on microbial faecal pollution and enteric pathogens in a subtropical estuary. Estuaries, 24:226-276.

Liu, Y., Yang, J., Hu, J., Tang, C., & Zheng, H. 2016. Characteristics of the surface–subsurface flow generation and sediment yield to the rainfall regime and land-cover by long-term in-situ observation in the red soil region, Southern China. Journal of Hydrology, 539:457–467.

112 LP DAAC. 2014. Land Processes Distributed Active Archive Centre – USGS Earth Explorer. https://lpdaac.usgs.gov/data_access/usgs_earthexplorer Date of access: 3 May 2018.

Mallory, S. & Beater, A. 2009. Department of Water Affairs and Forestry, South Africa. Inkomati Water Availability Assessment: Infrastructure and Operating Rules. Report No.: PWMA 05/X22/00/1208.

Mallory, S., Deacon, A., Huggins, G., Kotze, P., Louw, D., Mackenzie, J., Scherman, P. & Van Jaarsveld, P. 2013. The determination of water resource classes and associated resource quality objectives in the Inkomati Water Management Area: Status quo assessment, IUA and biophysical node delineation and identification. South Africa. Department of Water Affairs. Report Nr: RDM/WMA05/00/CON/CLA/0213.

Maplibrary. 2007. Maplibrary.org. Africa. http://www.maplibrary.org/library/stacks/Africa/index.htm Date of access: 30 Oct. 2016.

Matondo, J.I., Peter, G. & Msibi, K.M. 2005. Managing water under climate change for peace and prosperity in Swaziland. Physics and Chemistry of the Earth, 30:943–949.

Matthews, W.S. 1991. Phytosociology of the North-eastern Mountain Sourveld. Pretoria: UP. (Dissertation – MSc).

Matthews, W.S., Bredenkamp, G.J. & van Rooyen, N. 1991. The grassland-associated vegetation of the Black Reef Quartzite and associated large rocky outcrops in the north-eastern mountain sourveld of the Transvaal escarpment. South African Journal of Botany, 57(3):143-150.

Meehl, G.A., Stocker, T.F., Collins, W.D., Friedlingstein, P., Gaye, A.T., Gregory, J.M., Kitoh, A., Knutti, R., Murphy, J.M., Noda, A., Raper, S.C.B., Watterson, I.G., Weaver, A.J. & Zhao, Z.C. 2007. Global Climate Projections. (In Solomon, S., Qin, D. Manning, M. Chen, Z. Marquis, M. Averyt, K.B. Tignor, M.B. & Miller, H.L., eds., Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press. p. 748 – 845).

Meybeck, M., Kimstach, V. & Helmer, R. 1996. Strategies for water quality assessment. (In Chapman, D., ed. Water Quality Assessments: A Guide to Use of Biota, Sediments and Water in Environmental Monitoring. 2nd ed. Cambridge: University Press p.1-34). http://www.who.int/water_sanitation_health/resourcesquality/wqachapter2.pdf Date of access: 31 Oct. 2017. 113 Middleton, B.J & Bailey, A.K. 2005. Water Resources of South Africa 2005-Study. (WR2005). South African Water Research Commission. WRC Report No.: TT381/08.

Minitab Express Support (MES). 2017. Interpret the key results for Kruskal-Wallis test. http://support.minitab.com/en-us/minitab-express/1/help-and-how-to/modeling- statistics/anova/how-to/kruskal-wallis-test/interpret-the-results/key-results/ Date of access: 23 Oct. 2017.

Moodley, K., Pillay, S. & Pather, K. 2015. Spatiotemporal characterization of water chemistry and pollution sources of the Umhlatuzana, Umbilo and Amanzimnyama River catchments of Durban, KwaZulu-Natal, South Africa. Environmental Earth Sciences, 74:1273-1289.

Moulton, R.J. & Cuthbert, D.R. 2000. Cumulative Impacts/Risk Assessment of Water Removal or Loss from the Great Lakes–St. Lawrence River System. Canadian Water Resources Journal, 25:181-208.

Mucina, L. and Rutherford, M. C. 2006. The vegetation of South Africa, Lesotho and Swaziland. South African National Biodiversity Institute, Pretoria. Strelitzia, 19(3):31-50.

Mucina, L., Rutherford, M.C., Powrie, L.W., Rebelo, A.G., Camp, K.G.T., Lötter, M.C., Hoare, D.B., Boucher, C., Bredenkamp, G.J., Vlok, J.H.J., Euston-Brown, D.I.W., Jurgens, N., Du Preez, P.J., Le Roux, A., Schmiedel, U., Scott-Shaw, C.R., van Rooyen, N., Dobson, L., Palmer, A.R., Geldenhuys, C.J., Lloyd, J.W., van der Merwe, B., Bezuidenhout, H., Siebert, F., Siebert, S.J., Goodman, P.S., Winter, P.J.D., Helme, N., Smit, J.H.L, Desmet, P.G. Pfab, M., Mckenzie, B., Scholes, R.J., Manning, J.C., van Wyk, E., Zambatis, N., Lechmere-oertel, R.G. Eckhardt, H.C., Lubbinge, J.W., Matthews, W.S., Mcdonald, D.J., Smit, W.J., Bennet, R.G., Jonas, Z., Lombard, A.T., De Frey, W., Robesson, R., Oellermann, C., Grobler, A., van Niekerk, A., van der Merwe, & J.H. van der Merwe. 2006. Grassland Biome. p. 349-436. (In: L., Mucina and M.C., Rutherford. The vegetation of South Africa, Lesotho and Swaziland.) Strelitzia, 19:748-790.

Munzhedzi, S. & Mgquba S.K. 2013. Climate Change and The Water Sector. Climate and Impact Factsheet Series, Factsheet 3 of 7. www.sanbi.org/sites/default/files/documents/documents/ltas- factsheetclimate-change-and-water-sector2013.pdf Date of access: 20 Oct. 2017.

Nanson, G.C. & Gibling, M. 2004. Anabranching and Anastomosing River. Encyclopedia of Geomorphology: Vol. 1, 1A-I London: Routledge. Taylor & Francis e-Library, 2006.

114 National Ocean Service. 2015. National Oceanic and Atmospheric Administration U.S. Department of Commerce. What is the difference between land cover and land use? http://oceanservice.noaa.gov/facts/lclu.html Date of access: 16 March 2017.

Neff, R., Chang, H., Knight, C., Najjar, R., Yarnal, B. & Walker, H. 2000. Impact of climate variation and change on Mid-Atlantic Region hydrology and water resources. Climate Research, 14:207-218.

Norman, N. & Whitfield, G. 2006. Geological Journeys. A traveller’s guide to South Africa’s rocks and landforms. Cape Town, CT: Struik Nature. p. 10-320.

Nunes, A.N., de Almeida, A.C. and Coelho, C.O.A. 2011. Impacts of land use and cover type on runoff and soil erosion in a marginal area of Portugal. Applied Geography, 31:687-699.

NWRS. 2004. National Water Resource Strategy. 1st ed. Department of Water Affairs and Forestry, Pretoria. p.57-58.

O’Reilly, C.M., Alin, S.R., Plisnier, P.-D., Cohen, A.S. & McKee, B.A. 2003. Climate change decreases aquatic ecosystem productivity of Lake Tanganyika, Africa. Nature, 424:766-768.

O'Keeffe, J.H. 1985. The Conservation Status of the Sabie and Groot Letaba Rivers Within the Kruger National Park. IFWS Special Report No. 85/2. Rhodes University: Grahamstown. (South African Water History Archival Repository – C0424, Pretoria).

O'Keeffe, J.H., Weeks, D.C., Fourie, A. & Davies, B.R. 1996. A pre-impoundment study of the Sabie-sand river system, Mpumalanga with special reference to predicted impacts on the Kruger National Park. Volume Three: The effects of proposed impoundments and management recommendations. p.1-106.

Patz, J.A. 2001. Public Health Risk Assessment Linked to Climatic and Ecological Change. Human and Ecological Risk Assessment: An International Journal, 7(5):1317-1327.

Pieterse, A.J. and Janse van Vuuren, S. 1997. An investigation into phytoplankton blooms in the Vaal River and the environmental variables responsible for their development and decline. WRC Report No 359/1/97.

Pollard, S. & Du Toit, D. 2011. The Shared River Initiative Phase 1. Towards the sustainability of freshwater systems in South Africa: An exploration of factors that enable or constrain meeting the

115 Ecological Reserve within the context of Integrated Water Resources Management in the catchments of the Lowveld. WRC Report No.: TT 477/10. p.1-170.

Pollard, S. & Walker, P. 2000. Catchment management, water supply and sanitation in the Sand River Catchment, South Africa: Description and Issues. WHIRL Project Working Paper 1. UK Department for International Development (DFID). Project R7804. NRI, Chatham: UK. p.1-18.

Pollard, S. and Du Toit, D. 2011. The Shared River Initiative Phase I. Towards the sustainability of freshwater systems in South Africa: An exploration of factors that enable or constrain meeting the Ecological Reserve within the context of Integrated Water Resources Management in the catchments of the lowveld. WRC Report No.: TT 477/10.

Pollard, S., Mallory, S., Riddell, E. and Sawunyama, T. 2011. Towards improving the assessment and implementation of the Reserve: Real-time assessment and implementation of the Ecological Reserve. Final report WRC project K8/881/2.

Postel, S.L. 2000. Entering an Era of Water Scarcity: The Challenges Ahead. Ecological Applications, 10(4):941–948.

Quinn, G.P. and Keough, M.J. 2001. Experimental Design and Data Analysis for Biologists. Chapter 17 - Principal components and corresponding analysis. 1st ed. NJ: Cambridge University Press.

Resources Inventory Committee (RIC). 2017. British Columbia Ministry of Environment – Lands and Parks. Guidelines for Interpreting Water Quality Data. https://www.for.gov.bc.ca/hts/risc/pubs/aquatic/interp/interp-01.htm#2-7 Date of access: 26 Oct. 2017.

Rose, J.B., Daeschner, S., Easterling, D.R., Curriero, F.C., Lele, S. & Patz, J.A. 2000. Climate and Waterborne Disease Outbreaks. American Water Works Association, 92(9):77-87.

Rose, J.B., Epstein, P.R., Lipp, E.K., Sherman, B.H., Bernard, S.M. & Patz, J.A. 2001. Climate variability and change in the United States: Potential Impacts on Water and Foodborne Diseases Caused by Microbiologic Agents. Environmental Health Perspectives, 109(2):211-221.

Rountree, M.W., Rogers, K.H. & Heritage, G.L. 2000. Landscape State Change in the Semi-Arid Sabie River, Kruger National Park, in Response to Flood and Drought. South African Geographical Journal, 82(3):173-181. (DOI: 10.1080/03736245.2000.9713711).

116 Roux, D.J. 1999. Design of a National Programme for Monitoring and Assessing the Health of Aquatic Ecosystems, with Specific Reference to the South African River Health Programme. Environmental Science Forum, 96:13-32.

Roux, F. & Selepe, M. 2011. Sabie – Sand River: State of the Rivers Report. http://inkomaticma.co.za/images/Ecostatus_of_the_Sabie__Sand_Catchments.pdf Date of access 27 Aug. 2015.

Rutherford, M. C., Mucina, L. & Powrie, L. W. 2006. Biomes and Bioregions of Southern Africa. The vegetation of South Africa, Lesotho and Swaziland. Strelitzia, 19(3):31-50.

SANSA. 2015. FUNDISA disk – SPOT 6/7 mosaic and multispectral data via the North-West University Geo- and spatial department mapping facility.

Satellite Imaging Corporation. 2017. https://www.satimagingcorp.com/satellite-sensors/spot-6/ Date of access: 3 May 2018.

Schindler, D.W. 2001. The cumulative effects of climate warming and other human stresses on Canadian freshwaters in the new millennium. Canadian Journal of Fisheries and Aquatic Sciences, 58:18-29.

Schütte, S. & Schulze, R.E. 2017. Projected impacts of urbanisation on hydrological resource flows: A case study within the uMngeni Catchment, South Africa. Journal of Environmental Management, 196:527-543.

Smits, S., Pollard, S., Du Toit, D., Butterworth, J. and Moriarty, P. 2004. Modelling scenarios for water resources management in the Sand River Catchment, South Africa. WHIRL Project Working Paper 8 (DRAFT). https://scholar.google.co.za/scholar?hl=en&q=Modelling+scenarios+for+water+resources+manage ment+in+the+Sand+River+Catchment%2C+South+Africa&btnG=&as_sdt=1%2C5&as_sdtp= Date of access: 24 May 2016.

Soil Classification Working Group (SCWG). 1991. Soil classification – a taxonomic system for South Africa. Memoirs on the Agricultural Natural Resources of South Africa No. 15. Pretoria.

Soil quality. 2009. Soilquality.org.au. Fact Sheets – Cations and Cation Exchange Capacity http://www.soilquality.org.au/factsheets/cation-exchange-capacity Date of access: 3 May 2018.

117 Soulsby, C., Gibbins, C., Wade, A.J., Smart, R. & Helliwell, R. 2002. Water quality in the Scottish uplands: a hydrological perspective on catchment hydrochemistry. The Science of the Total Environment, 294:73-94.

South Africa. 2003. (Resource Quality Services). PDF files of South African landcover from the CSIR ARC national 1:250 000 land cover data set segmented by secondary drainage region X3.

South Africa. 2008. Department of Water Affairs and Forestry. DWAF hydrology. http://www.drdlr.gov.za/ Date of Access: 1 Jul. 2016.

South Africa. 2013. Spatial Planning and Land Use Management Act 2013 (Act No. 16 of 2013). (Notice 36730). Government gazette, 578:2, 5 Aug.

South Africa. 2016. National Water Act, 1998 (Act no. 36 of 1998): Classes of water resources and resource quality objectives for the catchments of the Inkomati. (Government notice no. 1616). GOVERNMENT GAZETTE, 40531(618), 30 Dec. 2016.

South African National Standard (SANS). 241-1:2015 Drinking water, Part 1: Microbiological, physical, aesthetic and chemical determinants. 241-2:2015 drinking water, Part 2: Application of SANS 241-1.

Stack Exchange. 2011. Geographic Information Systems. Calculating average slope along line using ArcGIS Desktop. https://gis.stackexchange.com/questions/15623/calculating-average-slope- along-line-using-arcgis-desktop Date of access: 11 Sept. 2017.

Strydom, W.F., Hill, L. and Eloff, E. 2006. The National Aquatic Ecosystem Health Monitoring Programme (River Health Programme) Achievements of the River Health Programme 1994 - 2004: A national perspective on the ecological health of selected South African rivers. Report contributors: The Department of Water Affairs and Forestry; The Department of Environmental Affairs and Tourism; Water Research Commission (WRC) and CSIR. Report No. ISBN: 0-620- 36168-9.

Swanepoel, A. 2015. Early warning system for the prediction of algal-related impacts on drinking water purification. Potchefstroom: NWU. (Thesis – PhD).

Ter Braak, C.J. & Šmilauer, P.N. 2002. CANOCO Reference Manual and CanoDraw for Windows User’s Guide: Software for Canonical Community Ordination (version 4.5). Ithaca, NY: Microcomputer Power, 500 pp.

118 Tlou, T. 2011. Bushbuckridge Local Municipality: First order Reconciliation Strategy for the Inyaka Regional Water Supply Scheme Area. Department of Water Affairs. Contract: WP 9712. p.2-27.

USGS. 2015. Shuttle Radar Topography Mission (SRTM) 1 Arc-Second Global. https://lta.cr.usgs.gov/SRTM1Arc Date of access 21 Aug. 2017.

USGS. 2018a. Earth Explorer Distribution Download. https://earthexplorer.usgs.gov/distribution Date of access 3 May 2018.

USGS. 2018b. Landsat missions – What are the band designations for the Landsat satellites? https://landsat.usgs.gov/what-are-band-designations-landsat-satellites Date of access 3 May 2018.

UW-Madison. 2018. University of Wisconsin-Madison Libraries. Information on Maps/Mapping & Geographic Information Systems (GIS) https://researchguides.library.wisc.edu/GIS Date of access: 3 May 2018.

Van der Hoven. C., Ubomba-Jaswa, E., van der Merwe, B., Loubser, M., Abia, A.L. 2017. The impact of various land uses on the microbial and physicochemical quality of surface water bodies in developing countries: Prioritisation of water resource management areas. Environmental Nanotechnology, Monitoring and Management, 8:280-289.

Van Ginkel, C.E., Hohls, B.C., Belcher, A., Vermaak, E. and Gerber, A. 2000. Assessment of the Trophic Status Project Main Report. Report number N/0000/00/DEQ/1799 of the Institute for Water Quality Studies, Department of Water Affairs. Pretoria, 314pp.

Van Wyk, E. Van Wilgen, B.W. & Roux, D.J. 2001. How well has biophysical research served the needs of water resource management? Lessons from the Sabie-Sand catchment. South African Journal of Science, 97:349-356.

Venter, F., Scholes, R. & Eckhardt, H. 2003. The abiotic template and its associated vegetation pattern. (In: Du Toit JT, Rogers KH, Biggs H, eds., The Kruger Experience: Ecology and management of savanna heterogeneity). Washington DC: Island Press, 2003. p. 83–117.

Vrebos, D., Beauchard, O. & Meire, P. 2017. The impact of land use and spatial mediated processes on the water quality in a river system. Science of the Total Environment, 601-602:365- 373.

119 Wade, A.J., Whitehead, P.G., Hornberger, G. M. & Snook, D.L. 2002. On modelling the flow controls on macrophyte and epiphyte dynamics in a lowland permeable catchment: the River Kennet, southern England. The Science of the Total Environment, 282:375-393.

Wang, Z., Jiao, J., Rayburg, S., Wanga, Q. & Su, Y. 2016. Soil erosion resistance of “Grain for Green” vegetation types under extreme rainfall conditions on the Loess Plateau, China. Catena, 141:109–116.

Warburton, M.L., Schulze, R.E. & Jewitt, G.P. 2012. Hydrological impacts of land use change in three diverse South African catchments. Journal of Hydrology, 414-415:118-135.

Water Research Commission. 2001. State of the Rivers Report. Crocodile, Sabie-Sand and Olifants River Systems. WRC Report No.: TT 147/01.

Wells, J. 1992. A Pre-Impoundment Study of the Biological Diversity of the Benthic Macro- Invertebrate Fauna of the Sabie-Sand River System. Cape Town: UCT. (Thesis – M.Sc).

Wetherald, R.T. & Manabe, S. 2002. Simulation of hydrologic changes associated with global warming. Journal of Geophysical Research, 107(D19, 4379):1-15.

White K. 2004. Encyclopaedia of Geomorphology: PEDIMENT. Vol.1, 1A-I London: Routledge. Taylor & Francis e-Library, 2006.

WHO. 2003. Total dissolved solids in drinking-water. Background document for preparation of WHO Guidelines for drinking-water quality. Geneva, World Health Organization, (WHO/SDE/WSH/03.04/16).

WHO. 2004. Guidelines for Drinking Water Quality: Third Edition Incorporating the First and Second Addenda. World Health Organization. Geneva, 1:1-515.

Wilbanks, T.J., P. Romero Lankao, M. Bao, F. Berkhout, S. Cairncross, J.-P. Ceron, M. Kapshe, R. Muir-Wood & R. Zapata-Marti. 2007. Industry, settlement and society. (In Parry, M.L., Canziani, O.F., Palutikof, J.P., Van der Linden, P.J. & Hanson, C.E., eds., Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press. p. 357- 390).

120 Woodhouse, P. 1995. Water Rights and Rural Restructuring in South Africa: A Case Study from Eastern Transvaal. Water Resources Development, 11(4):527-544.

WR 1990. Water Resources of South Africa, 1990 study. Simplified geology. (Source Unknown). www.sanbi.org/sites/default/files/documents/documents/ltas-factsheetclimate-change-and-water- sector2013.pdf Date of access: 16 Nov. 2017.

WR. 2012. Water Resources of South Africa, 2012 study. http://www.waterresourceswr2012.co.za Date of access: 16 Nov. 2017.

Xizhi, L., Jing, H., Peiqing, X., Pan, Z. & Jiaoxuehui. 2017. Correlation Analysis Between the Water Quality and Land Use Composition in Chaobai River Basin. Nature Environment and Pollution Technology, 16(1):307-314.

Yang, F., Zhang, G., Yang, J., Li, D., Zhao, Y., Liu, F., Yang, R. & Yang, F. 2014. Organic matter controls of soil water retention in an alpine grassland and its significance for hydrological processes. Journal of Hydrology, 519:3086–3093.

Yira, Y., Diekkrüger, B., Steup, G. & Bossa, A.Y. 2016. Modeling land use change impacts on water resources in a tropical West African catchment (Dano, Burkina Faso). Journal of Hydrology, 537:187–199.

Yu, S., Xu, Z., Wu, W. & Zuo, D. 2016. Effect of land use types on stream water quality under seasonal variation and topographic characteristics in the Wei River basin, China. Ecological Indicators, 60:202-212.

Zhao, J., Lin, L., Yang, K., Liu, Q. & Qian, G. 2015. Influences of land use on water quality in a reticular river network area: A case study in Shanghai, China. Landscape and Urban Planning, 137:20-29.

121 METADATA

Metadata table representing the sources used within ArcGIS to obtain the information used within the mapping Name Spatial Reference Data Format Type of Data Spatial Representation Source reference AfricaContinent_WGS1984_UTM Projected Coordinate System: Shapefile Vector Polygon Maplibrary (2007) 36S WGS_1984_UTM_Zone_36S http://www.maplibrary.org/library/sta (Africa_Base) cks/Africa/index.htm Projection: Transverse_Mercator

Linear Unit: Meter

Geographic Coordinate System: GCS_WGS_1984

Datum: D_WGS_1984

Prime Meridian: Greenwich

Angular Unit: Degree commDRDLR2008SRS Projected Coordinate System: Shapefile Vector Polygon South Africa (2008). WGS_1984_UTM_Zone_36S (Source Unknown). http://www.drdlr.gov.za/ Projection: Transverse_Mercator

Linear Unit: Meter

Geographic Coordinate System: GCS_WGS_1984

Datum: D_WGS_1984

Prime Meridian: Greenwich

Angular Unit: Degree

122 Metadata table representing the sources used within ArcGIS to obtain the information used within the mapping Name Spatial Reference Data Format Type of Data Spatial Representation Source reference Coordinates_Study_Project Projected Coordinate System: Shapefile Vector Point Personal data generated in this WGS_1984_UTM_Zone_36S project: Taken with handheld GPS in situ: Projection: Transverse_Mercator Garmin. GPSmap 60CSx

Geographic Coordinate System: GCS_WGS_1984

Datum: D_WGS_1984

Prime Meridian: Greenwich

Angular Unit: Degree clip2_ProjectRasterWGS1984_U Projected Coordinate System: TIFF Raster Grid Jarvis et al. (2008). TM36S WGS_1984_UTM_Zone_36S SRTM90 Data Downloaded via (SRTM90 DEM) Global Mapper Geographic Coordinate System: http://www.cgiar-csi.org/data/srtm- GCS_WGS_1984 90m-digital-elevation-database-v4-1 dams500g_wgs84_Project Projected Coordinate System: Shapefile Vector Polygon DWAF (2006c) WGS_1984_UTM_Zone_36S Downloaded from http://www.dwaf.gov.za/iwqs/gis_dat Projection: Transverse_Mercator a/river/rivs500k.aspx as obtained from NGI – National Geo-Spatial Geographic Coordinate System: Information GCS_WGS_1984 http://www.ngi.gov.za/index.php/wha t-we-do/maps-and-geospatial- Datum: D_WGS_1984 information

Prime Meridian: Greenwich

Angular Unit: Degree dea_cardno_2014_sa_lcov_utm3 Projected Coordinate System: TIFF Raster Grid GeoTerraimage (2014) 5n_vs2b_pivot-corr as well as WGS_1984_UTM_Zone_35S https://egis.environment.gov.za/natio SABIE_NLC2014_72CLS_UTM3 nal_land_cover_data_sa 6S Geographic Coordinate System: (Land use 18 class) GCS_WGS_1984 (NLC)

123 Metadata table representing the sources used within ArcGIS to obtain the information used within the mapping Name Spatial Reference Data Format Type of Data Spatial Representation Source reference InkomatiWMAdwafWGS1984_U Projected Coordinate System: Shapefile Vector Polygon DWAF (2012). TM36S WGS_1984_UTM_Zone_36S (Inkomati WMA) Projection: Transverse_Mercator

Linear Unit: Meter

Geographic Coordinate System: GCS_WGS_1984

Datum: D_WGS_1984

Prime Meridian: Greenwich

Angular Unit: Degree PPS Projected Coordinate System: Shapefile Vector Point Personal data generated in this Major_Point_Pollution_SourceW WGS_1984_UTM_Zone_36S project: GS1984_UTM36S Data points digitized using SPOT Projection: Transverse_Mercator 2015 natural colour mosaic (SANSA, Geographic 2015).

Geographic Coordinate System: GCS_WGS_1984 Datum: D_WGS_1984 Prime Meridian: Greenwich Angular Unit: Degree

124 Metadata table representing the sources used within ArcGIS to obtain the information used within the mapping Name Spatial Reference Data Format Type of Data Spatial Representation Source reference RSAgeolUnionWS_UTM36S Projected Coordinate System: Shapefile Vector Polygon GeoScience (2003) (Geology) WGS_1984_UTM_Zone_36S http://www.geoscience.org.za/index. php/publication/ Projection: Transverse_Mercator (WR, 1990).

Geographic Coordinate System: GCS_WGS_1984

Datum: D_WGS_1984

Prime Meridian: Greenwich

Angular Unit: Degree SoilGeolWatershedsUTM36S Projected Coordinate System: Shapefile Vector Polygon Land Type Survey Staff (2002). (Soil) WGS_1984_UTM_Zone_36S

Projection: Transverse_Mercator

Geographic Coordinate System: GCS_WGS_1984

Datum: D_WGS_1984

Prime Meridian: Greenwich

Angular Unit: Degree

125 Metadata table representing the sources used within ArcGIS to obtain the information used within the mapping Name Spatial Reference Data Format Type of Data Spatial Representation Source reference Sabie_Homelands1994_UTM36 Projected Coordinate System: Shapefile Vector Polygon Source Unknown S WGS_1984_UTM_Zone_36S

Projection: Transverse_Mercator

Linear Unit: Meter

Geographic Coordinate System: GCS_WGS_1984

Datum: D_WGS_1984

Prime Meridian: Greenwich

Angular Unit: Degree ShreveClip Projected Coordinate System: Shapefile Vector Polyline Generated using SRTM90 DEM WGS_1984_UTM_Zone_36S (Jarvis et al., 2008).

Projection: Transverse_Mercator

Geographic Coordinate System: GCS_WGS_1984

Datum: D_WGS_1984

Prime Meridian: Greenwich

Angular Unit: Degree

126 Metadata table representing the sources used within ArcGIS to obtain the information used within the mapping Name Spatial Reference Data Format Type of Data Spatial Representation Source reference SAPADor2015Q_WGS1984_UT Projected Coordinate System: Shapefile Vector Polygon DEA (2016) M36S WGS_1984_UTM_Zone_36S https://egis.environment.gov.za/

Projection: Transverse_Mercator

Geographic Coordinate System: GCS_WGS_1984

Datum: D_WGS_1984

Prime Meridian: Greenwich

Angular Unit: Degree Secondary_Catchment_Mpumal Projected Coordinate System: Shapefile Vector Polygon South Africa (2008) anga_reproj WGS_1984_UTM_Zone_36S http://www.dwaf.gov.za/iwqs/gis_dat a/river/rivs500k.aspx Projection: Transverse_Mercator

Geographic Coordinate System: GCS_WGS_1984 Datum: D_WGS_1984 Prime Meridian: Greenwich Angular Unit: Degree Watersheds_All_WGS_1984_UT Projected Coordinate System: Shapefile Vector Polygon Watersheds created using SRTM 90 M_Zone_36S WGS_1984_UTM_Zone_36S DEM (Jarvis et al., 2008).

Projection: Transverse_Mercator

Linear Unit: Meter

Geographic Coordinate System: GCS_WGS_1984

Datum: D_WGS_1984

Prime Meridian: Greenwich

Angular Unit: Degree 127 Metadata table representing the sources used within ArcGIS to obtain the information used within the mapping Name Spatial Reference Data Format Type of Data Spatial Representation Source reference Wriall500_primary_WGS1984_U Projected Coordinate System: Shapefile Vector Polyline DWAF (2006c) TM36S WGS_1984_UTM_Zone_36S http://www.dwaf.gov.za/iwqs/gis_dat and a/river/rivs500k.aspx Wriall500_secondary_WGS1984 Projection: Transverse_Mercator _UTM36S (Rivers) Linear Unit: Meter

Geographic Coordinate System: GCS_WGS_1984

Datum: D_WGS_1984

Prime Meridian: Greenwich

Angular Unit: Degree Clipped_SPOT_Raster Geographic Coordinate System: JPEG Raster Grid SANSA (2015) GCS_WGS_1984 http://www.sansa.org.za/earthobserv ation/services Datum: D_WGS_1984

Angular Unit: Degree

128 APPENDIX A

Figure 7-1: Site 1. Sabie River Headwaters. Left photo captured by Michaela Stolz during October 2016 and right captured by Dr. Annelie Swanepoel during July 2016 sampling.

Figure 7-2: Site 2. Sabie River below Wastewater treatment plant. July, (A. Swanepoel).

129

Figure 7-3: Site 3. Sabie River Below Waste Water Treatment. July 2016, (A. Swanepoel).

Figure 7-4: Site 4. Sabie River Below Hoxane Water Treatment Works. July 2016,

(A. Swanepoel).

130

Figure 7-5: Site 5. Sabie River KNP Gate. July 2016, (A. Swanepoel).

Figure 7-6: Site 6. Sabie River Skukuza KNP. July 2016, (A. Swanepoel).

Figure 7-7: Site 7. Sand River KNP. July 2016, (A. Swanepoel).

131

Figure 7-8: Site 8. Sabie River Lower. July 2016, (A. Swanepoel).

Figure 7-9: Site 9. Sabie River close to Mozambique. July 2016, (A. Swanepoel).

Figure 7-10: Site 10. Marite River after Inyaka Dam Outlet. July 2016, (A. Swanepoel).

132

Figure 7-11: Site 11. Inyaka Dam Outlet. July 2016, (A. Swanepoel).

Figure 7-12: Site 12. Sand River Thulamahashe Below Waste Water Treatment. July 2016, (A. Swanepoel).

133 APPENDIX B

Table 8-1: Spearman rank correlation test was used to determine whether there was significant (p < 0.05) differences between water quality indicators for the 12 sites. Spearman Rank Order Correlations (MD pairwise deleted). Marked correlations are significant at p <0.05 Variables Chl-a Coli E,coli Turb TDS M Alk Hard pH Temp MMHg DO SPC COD Al3+ Ca2+ Fe2+/3+ K+ Mg2+ Mn2+ Na+ Chl-a 1,000 0,235 -0,013 0,005 0,233 0,114 0,066 0,288 0,239 -0,048 -0,353 0,139 0,556 -0,105 0,089 -0,108 0,339 0,052 -0,250 0,288 Coli 0,235 1,000 0,715 0,231 0,216 0,366 0,203 0,266 0,343 0,343 -0,431 0,456 0,327 0,434 0,178 0,182 0,413 0,235 -0,249 0,508 E,coli -0,013 0,715 1,000 0,329 0,270 0,562 0,361 0,108 0,073 0,492 -0,163 0,620 0,297 0,171 0,316 0,014 0,319 0,411 -0,081 0,474 Turb 0,005 0,231 0,329 1,000 0,308 0,209 0,017 -0,038 0,201 0,476 -0,170 0,362 0,168 0,514 -0,011 0,538 0,363 0,016 0,261 0,580 TDS 0,233 0,216 0,270 0,308 1,000 0,581 0,459 0,214 0,196 0,339 -0,243 0,628 0,334 -0,034 0,443 -0,009 0,460 0,455 0,072 0,544 M Alk 0,114 0,366 0,562 0,209 0,581 1,000 0,818 0,260 -0,069 0,559 0,092 0,910 0,345 -0,177 0,809 -0,107 0,591 0,782 0,101 0,559 Hard 0,066 0,203 0,361 0,017 0,459 0,818 1,000 0,294 -0,211 0,350 0,247 0,730 0,253 -0,258 0,982 -0,170 0,359 0,934 0,180 0,357 pH 0,288 0,266 0,108 -0,038 0,214 0,260 0,294 1,000 0,541 0,205 -0,347 0,306 0,126 0,064 0,284 -0,081 0,128 0,360 -0,197 0,221 Temp 0,239 0,343 0,073 0,201 0,196 -0,069 -0,211 0,541 1,000 0,255 -0,771 0,138 0,139 0,565 -0,192 0,308 0,222 -0,150 -0,301 0,441 MMHg -0,048 0,343 0,492 0,476 0,339 0,559 0,350 0,205 0,255 1,000 -0,048 0,668 -0,003 0,347 0,292 0,281 0,466 0,421 0,028 0,718 DO -0,353 -0,431 -0,163 -0,170 -0,243 0,092 0,247 -0,347 -0,771 -0,048 1,000 -0,093 -0,319 -0,417 0,221 -0,250 -0,351 0,194 0,264 -0,369 SPC 0,139 0,456 0,620 0,362 0,628 0,910 0,730 0,306 0,138 0,668 -0,093 1,000 0,342 0,050 0,712 -0,016 0,660 0,728 0,038 0,718 COD 0,556 0,327 0,297 0,168 0,334 0,345 0,253 0,126 0,139 -0,003 -0,319 0,342 1,000 -0,111 0,283 -0,208 0,358 0,242 -0,172 0,349 Al3+ -0,105 0,434 0,171 0,514 -0,034 -0,177 -0,258 0,064 0,565 0,347 -0,417 0,050 -0,111 1,000 -0,250 0,684 0,163 -0,186 -0,006 0,392 Ca2+ 0,089 0,178 0,316 -0,011 0,443 0,809 0,982 0,284 -0,192 0,292 0,221 0,712 0,283 -0,250 1,000 -0,154 0,373 0,874 0,220 0,332 Fe2+/3+ -0,108 0,182 0,014 0,538 -0,009 -0,107 -0,170 -0,081 0,308 0,281 -0,250 -0,016 -0,208 0,684 -0,154 1,000 0,210 -0,231 0,433 0,384 K+ 0,339 0,413 0,319 0,363 0,460 0,591 0,359 0,128 0,222 0,466 -0,351 0,660 0,358 0,163 0,373 0,210 1,000 0,335 0,069 0,751 Mg2+ 0,052 0,235 0,411 0,016 0,455 0,782 0,934 0,360 -0,150 0,421 0,194 0,728 0,242 -0,186 0,874 -0,231 0,335 1,000 0,032 0,354 Mn2+ -0,250 -0,249 -0,081 0,261 0,072 0,101 0,180 -0,197 -0,301 0,028 0,264 0,038 -0,172 -0,006 0,220 0,433 0,069 0,032 1,000 0,059 Na+ 0,288 0,508 0,474 0,580 0,544 0,559 0,357 0,221 0,441 0,718 -0,369 0,718 0,349 0,392 0,332 0,384 0,751 0,354 0,059 1,000 M:D 0,294 0,126 -0,023 0,418 0,215 -0,162 -0,399 -0,121 0,417 0,078 -0,558 -0,004 0,143 0,312 -0,394 0,424 0,442 -0,422 0,091 0,532 P3- 0,347 0,289 0,342 0,213 0,269 0,347 0,347 0,091 -0,113 0,007 -0,154 0,346 0,558 -0,232 0,347 -0,194 0,451 0,277 0,080 0,347 - Si 0,116 0,217 0,228 0,107 -0,094 0,136 0,052 0,068 0,193 0,040 -0,211 0,180 0,059 0,168 0,079 0,430 0,344 0,222 0,365 0,0005 Tsi 0,126 0,213 0,224 0,106 -0,097 0,129 0,047 0,063 0,185 0,021 -0,211 0,166 0,061 0,157 0,076 0,427 0,346 -0,010 0,233 0,352 Cl 0,120 0,498 0,454 0,486 0,478 0,548 0,261 0,287 0,388 0,752 -0,236 0,646 0,164 0,357 0,232 0,356 0,582 0,291 0,124 0,807 2- NO3 + - -0,065 0,348 0,164 0,068 0,152 0,353 0,403 -0,011 -0,125 -0,045 0,097 0,230 0,350 -0,005 0,453 -0,124 0,027 0,331 0,024 -0,037 NO2 - NH3 0,432 0,378 -0,037 0,162 0,172 -0,150 -0,218 0,239 0,574 -0,161 -0,647 -0,062 0,465 0,204 -0,176 0,103 0,259 -0,246 -0,210 0,236 TKN -0,013 0,030 0,214 0,504 0,231 0,260 0,168 -0,322 -0,200 0,285 0,170 0,271 0,074 0,138 0,162 0,214 0,314 0,123 0,392 0,415 TN 0,060 0,111 0,228 0,494 0,267 0,309 0,186 -0,335 -0,166 0,249 0,108 0,304 0,206 0,130 0,207 0,208 0,341 0,112 0,360 0,421 3- PO4 -0,107 -0,054 0,323 -0,061 -0,183 0,255 0,229 -0,020 -0,311 0,049 0,184 0,183 -0,084 -0,231 0,195 0,023 0,162 0,201 0,215 0,049 TP -0,107 -0,054 0,323 -0,061 -0,183 0,255 0,229 -0,020 -0,311 0,049 0,184 0,183 -0,084 -0,231 0,195 0,023 0,162 0,201 0,215 0,049 134 Spearman Rank Order Correlations (MD pairwise deleted). Marked correlations are significant at p <0.05 Variables Chl-a Coli E,coli Turb TDS M Alk Hard pH Temp MMHg DO SPC COD Al3+ Ca2+ Fe2+/3+ K+ Mg2+ Mn2+ Na+ TN/TP 0,045 0,095 0,177 0,475 0,261 0,267 0,161 -0,357 -0,175 0,230 0,142 0,255 0,210 0,141 0,184 0,197 0,265 0,090 0,359 0,365 DIN:DIP 0,182 0,486 0,081 0,145 0,206 0,202 0,210 0,029 0,055 -0,179 -0,141 0,130 0,522 0,025 0,281 -0,049 0,127 0,093 0,019 0,041 DOC 0,371 0,446 0,294 0,406 0,433 0,273 0,055 0,104 0,338 0,376 -0,572 0,402 0,295 0,303 0,050 0,411 0,663 0,046 -0,007 0,680 Sulp 0,048 0,365 0,443 0,078 0,445 0,787 0,693 0,318 0,018 0,349 0,101 0,735 0,279 -0,009 0,704 -0,119 0,408 0,675 0,001 0,334 MIB 0,242 0,272 0,217 -0,116 0,026 0,184 0,016 0,058 0,195 0,026 -0,247 0,195 0,309 0,060 0,063 0,005 0,284 -0,026 -0,113 0,205 Geos -0,008 0,302 0,342 0,210 -0,053 0,303 0,258 0,050 -0,190 0,284 0,049 0,233 0,050 -0,024 0,256 0,328 0,234 0,155 0,206 0,260 TOC 0,352 0,458 0,383 0,502 0,449 0,304 0,050 0,236 0,492 0,464 -0,641 0,452 0,272 0,378 0,039 0,489 0,681 0,041 0,076 0,776

Table 8-2: Spearman rank correlation test was used to determine whether there was significant (p < 0.05) differences between water quality indicators for the 12 sites (continued). Spearman Rank Order Correlations (MD pairwise deleted). Marked correlations are significant at p <.05000 2- 3- - NO3 + - 3- Variables M:D P Si Tsi Cl - NH3 TKN TN PO4 TP TN/TP DIN:DIP DOC Sulp MIB Geos TOC NO2 Chl-a 0,294 0,347 0,116 0,126 0,120 -0,065 0,432 -0,013 0,060 -0,107 -0,107 0,045 0,182 0,371 0,048 0,242 -0,008 0,352 Coli 0,126 0,289 0,217 0,213 0,498 0,348 0,378 0,030 0,111 -0,054 -0,054 0,095 0,486 0,446 0,365 0,272 0,302 0,458 E,coli -0,023 0,342 0,228 0,224 0,454 0,164 -0,037 0,214 0,228 0,323 0,323 0,177 0,081 0,294 0,443 0,217 0,342 0,383 Turb 0,418 0,213 0,107 0,106 0,486 0,068 0,162 0,504 0,494 -0,061 -0,061 0,475 0,145 0,406 0,078 -0,116 0,210 0,502 TDS 0,215 0,269 -0,094 -0,097 0,478 0,152 0,172 0,231 0,267 -0,183 -0,183 0,261 0,206 0,433 0,445 0,026 -0,053 0,449 M Alk -0,162 0,347 0,136 0,129 0,548 0,353 -0,150 0,260 0,309 0,255 0,255 0,267 0,202 0,273 0,787 0,184 0,303 0,304 Hard -0,399 0,347 0,052 0,047 0,261 0,403 -0,218 0,168 0,186 0,229 0,229 0,161 0,210 0,055 0,693 0,016 0,258 0,050 pH -0,121 0,091 0,068 0,063 0,287 -0,011 0,239 -0,322 -0,335 -0,020 -0,020 -0,357 0,029 0,104 0,318 0,058 0,050 0,236 Temp 0,417 -0,113 0,193 0,185 0,388 -0,125 0,574 -0,200 -0,166 -0,311 -0,311 -0,175 0,055 0,338 0,018 0,195 -0,190 0,492 MMHg 0,078 0,007 0,040 0,021 0,752 -0,045 -0,161 0,285 0,249 0,049 0,049 0,230 -0,179 0,376 0,349 0,026 0,284 0,464 DO -0,558 -0,154 -0,211 -0,211 -0,236 0,097 -0,647 0,170 0,108 0,184 0,184 0,142 -0,141 -0,572 0,101 -0,247 0,049 -0,641 SPC -0,004 0,346 0,180 0,166 0,646 0,230 -0,062 0,271 0,304 0,183 0,183 0,255 0,130 0,402 0,735 0,195 0,233 0,452 COD 0,143 0,558 0,059 0,061 0,164 0,350 0,465 0,074 0,206 -0,084 -0,084 0,210 0,522 0,295 0,279 0,309 0,050 0,272 Al3+ 0,312 -0,232 0,168 0,157 0,357 -0,005 0,204 0,138 0,130 -0,231 -0,231 0,141 0,025 0,303 -0,009 0,060 -0,024 0,378 Ca2+ -0,394 0,347 0,079 0,076 0,232 0,453 -0,176 0,162 0,207 0,195 0,195 0,184 0,281 0,050 0,704 0,063 0,256 0,039 Fe2+/3+ 0,424 -0,194 0,430 0,427 0,356 -0,124 0,103 0,214 0,208 0,023 0,023 0,197 -0,049 0,411 -0,119 0,005 0,328 0,489 K+ 0,442 0,451 0,344 0,346 0,582 0,027 0,259 0,314 0,341 0,162 0,162 0,265 0,127 0,663 0,408 0,284 0,234 0,681 Mg2+ -0,422 0,277 -0,0005 -0,010 0,291 0,331 -0,246 0,123 0,112 0,201 0,201 0,090 0,093 0,046 0,675 -0,026 0,155 0,041 Mn2+ 0,091 0,080 0,222 0,233 0,124 0,024 -0,210 0,392 0,360 0,215 0,215 0,359 0,019 -0,007 0,001 -0,113 0,206 0,076 Na+ 0,532 0,347 0,365 0,352 0,807 -0,037 0,236 0,415 0,421 0,049 0,049 0,365 0,041 0,680 0,334 0,205 0,260 0,776 M:D 1,000 0,273 0,410 0,408 0,261 -0,366 0,473 0,267 0,259 0,017 0,017 0,191 -0,065 0,643 -0,396 0,100 0,032 0,666 P3- 0,273 1,000 0,238 0,239 0,039 0,097 0,383 0,347 0,346 0,437 0,437 0,239 0,334 0,347 0,078 -0,030 0,264 0,347 Si 0,410 0,238 1,000 0,997 0,164 -0,145 0,114 0,084 0,072 0,644 0,644 -0,047 -0,060 0,372 -0,003 0,237 0,238 0,423 Tsi 0,408 0,239 0,997 1,000 0,155 -0,150 0,115 0,087 0,073 0,644 0,644 -0,045 -0,065 0,353 -0,004 0,238 0,222 0,414 135 Spearman Rank Order Correlations (MD pairwise deleted). Marked correlations are significant at p <.05000 2- 3- - NO3 + - 3- Variables M:D P Si Tsi Cl - NH3 TKN TN PO4 TP TN/TP DIN:DIP DOC Sulp MIB Geos TOC NO2 Cl- 0,261 0,039 0,164 0,155 1,000 -0,114 0,036 0,279 0,277 -0,051 -0,051 0,280 -0,069 0,492 0,517 0,184 0,135 0,645 2- NO3 + - -0,366 0,097 -0,145 -0,150 -0,114 1,000 0,005 0,107 0,244 -0,174 -0,174 0,280 0,783 -0,123 0,350 0,122 0,123 -0,177 NO2 - NH3 0,473 0,383 0,114 0,115 0,036 0,005 1,000 0,060 0,148 -0,188 -0,188 0,125 0,524 0,279 -0,118 0,250 -0,098 0,329 TKN 0,267 0,347 0,084 0,087 0,279 0,107 0,060 1,000 0,967 0,120 0,120 0,939 0,195 0,155 0,062 0,106 0,169 0,208 TN 0,259 0,346 0,072 0,073 0,277 0,244 0,148 0,967 1,000 0,041 0,041 0,982 0,352 0,173 0,123 0,184 0,182 0,205 3- PO4 0,017 0,437 0,644 0,644 -0,051 -0,174 -0,188 0,120 0,041 1,000 1,000 -0,095 -0,215 0,117 0,038 0,066 0,337 0,128 TP 0,017 0,437 0,644 0,644 -0,051 -0,174 -0,188 0,120 0,041 1,000 1,000 -0,095 -0,215 0,117 0,038 0,066 0,337 0,128 TN/TP 0,191 0,239 -0,047 -0,045 0,280 0,280 0,125 0,939 0,982 -0,095 -0,095 1,000 0,380 0,097 0,145 0,174 0,113 0,131 DIN:DIP -0,065 0,334 -0,060 -0,065 -0,069 0,783 0,524 0,195 0,352 -0,215 -0,215 0,380 1,000 0,038 0,268 0,218 0,057 -0,0010 DOC 0,643 0,347 0,372 0,353 0,492 -0,123 0,279 0,155 0,173 0,117 0,117 0,097 0,038 1,000 0,096 0,227 0,357 0,913 Sulp -0,396 0,078 -0,003 -0,004 0,517 0,350 -0,118 0,062 0,123 0,038 0,038 0,145 0,268 0,096 1,000 0,249 0,007 0,129 MIB 0,100 -0,030 0,237 0,238 0,184 0,122 0,250 0,106 0,184 0,066 0,066 0,174 0,218 0,227 0,249 1,000 0,195 0,227 Geos 0,032 0,264 0,238 0,222 0,135 0,123 -0,098 0,169 0,182 0,337 0,337 0,113 0,057 0,357 0,007 0,195 1,000 0,382 TOC 0,666 0,347 0,423 0,414 0,645 -0,177 0,329 0,208 0,205 0,128 0,128 0,131 -0,001 0,913 0,129 0,227 0,382 1,000

Table 8-3: Kruskal-Wallis ANOVA and multiple test results. Correlation value Multiple Comparisons P-value (2-tailed) Chl-a Do not really show a significant difference between sites. H (11, N= 48) =16.70826 p =.1168 Coli Significant difference between sites: 1 and 12 of: 0.034844; 11 and 12 of: 0,019033 H (11, N= 35) =24.13933 p =.0122 E,coli Significant difference between sites: 1 and 12 of: 0,014957; 3 and 12 of: 0,045995; 8 and 11 of: 0,031413; 11 and 12 of: 0,000920 H (11, N= 47) =34.07156 p =.0004 Turb Significant difference between sites: 1 and 9 of: 0,022159 H (11, N= 48) =27.94974 p =.0033 TDS Do not really show a significant difference between sites. M Alk Significant difference between sites: 1 and 7 of: 0,039178; 1 and 12 of: 0,001037 pH Do not really show a significant difference between sites. MMHg Significant difference between sites: 1 and 6 of: 0,039178; 1 and 7 of: 0,041043; 1 and 8 of: 0,005678; 1 and 9 of: 0,000581; 2 and 8 H (11, N= 48) =44.16423 p =.0000 of: 0,028172; 2 and 9 of: 0,003519; 9 and 11 of: 0,021110 DO Do not really show a significant difference between sites. SPC Significant difference between sites: 1 and 7 of: 0,001824; 1 and 12 of: 0,001163; 7 and 10 of: 0,042991; 7 and 11 of: 0,014247; 10 H (11, N= 48) =41.56029 p =.0000 and 12 of: 0,029544; 11 and 12 of: 0,009523 COD Do not really show a significant difference between sites. Al3+ Do not really show a significant difference between sites. Ca2+ Significant difference between sites: 1 and 2 of: 0,018237; 1 and 3 of: 0,032478; 1 and 12 of: 0,025603 H (11, N= 48) =35.10423 p =.0002 Fe2+/3+ Significant difference between sites: 1 and 4 of: 0,039178; 1 and 9 of: 0,041043; 1 and 10 of: 0,006995 H (11, N= 48) =32.37298 p =.0007 K+ Do not really show a significant difference between sites. Mg2+ Significant difference between sites: 1 and 2 of: 0,019151; 2 and 10 of: 0,041043; 2 and 11 of: 0,006995 H (11, N= 48) =35.44401 p =.0002 Mn2+ Do not really show a significant difference between sites.

136 Correlation value Multiple Comparisons P-value (2-tailed) Na+ Significant difference between sites: 1 and 7 of: 0,011089; 1 and 9 of: 0,025603; 1 and 12 of: 0,001302; 2 and 12 of: 0,012264; 3 and H (11, N= 48) =42.14207 p =.0000 12 of: 0,045025 Tsi Do not really show a significant difference between sites. Cl- Significant difference between sites: 1 and 7 of: 0,014247; 1 and 9 of: 0,047147 H (11, N= 48) =35.03810 p =.0002 2- - NO3 +NO2 Do not really show a significant difference between sites. - NH3 Do not really show a significant difference between sites. TKN Do not really show a significant difference between sites. TN Do not really show a significant difference between sites. TP Do not really show a significant difference between sites. Sulp Do not really show a significant difference between sites. MIB Do not really show a significant difference between sites. Geos Do not really show a significant difference between sites. TOC Significant difference between sites: 1 and 12 of: 0,001542; 2 and 12 of: 0,005387; 3 and 12 of: 0,019151 H (11, N= 48) =35.54135 p =.0002 Temp Do not really show a significant difference between sites. P3- Do not really show a significant difference between sites. Si Do not really show a significant difference between sites. 3- PO4 Do not really show a significant difference between sites.

Table 8-4: Component-variable correlations (factor correlations), based on correlations of water quality variable concentrations. (Component heading abbreviated to C).

Factor-variable correlations (factor correlations), based on correlations

Variables C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 11 C 12 C 13 C 14 C 15 C 16 C 17 C 18 C 19 C 20 C 21 C 22 C 23 C 24 C 25 C 26 C 27 C 28

Chl-a -0,643 0,180 0,439 -0,500 -0,171 0,068 -0,063 -0,026 0,199 0,008 -0,002 0,117 -0,033 0,050 0,010 -0,041 -0,045 0,026 -0,045 0,055 -0,0106 -0,074 -0,012214 0,009449 -0,008419 0,003747 0,000125 -0,000031

Coli -0,919 -0,133 0,035 -0,338 -0,0007 0,012 0,008 -0,038 0,065 0,025 0,048 0,083 -0,017 -0,006 0,024 -0,009 -0,035 0,010 0,052 0,034 0,024 0,020 0,001705 -0,014659 0,009553 -0,009680 0,006267 -0,000594

E,coli -0,812 -0,370 -0,376 -0,092 0,174 0,037 0,101 -0,046 -0,055 0,038 0,025 -0,003 -0,012 0,005 0,005 0,009 -0,017 0,008 0,048 0,037 -0,004 -0,005 -0,017771 -0,010396 0,000443 -0,006958 -0,000799 0,002340

Turb 0,064 -0,666 0,436 0,390 -0,288 0,128 0,105 -0,050 0,073 0,015 -0,011 0,086 0,149 -0,027 0,165 0,117 0,068 0,120 0,027 0,012 0,0108 0,006 -0,012890 0,015751 -0,003656 0,002206 0,000276 0,000008

TDS -0,204 0,275 0,129 0,576 0,337 0,410 -0,178 0,095 0,239 -0,015 0,288 0,086 -0,039 0,217 0,070 -0,054 0,049 -0,043 0,052 -0,020 0,007 -0,0108 0,003452 -0,008659 0,000358 0,000410 -0,000141 0,000014

M Alk -0,790 0,258 -0,297 0,332 -0,221 -0,133 -0,021 0,049 0,004 0,070 -0,090 -0,034 -0,042 -0,098 0,019 -0,016 -0,029 -0,028 0,063 -0,048 0,070 -0,034 -0,005873 0,033652 0,017587 0,002494 -0,000234 0,000006

pH -0,063 0,477 -0,084 -0,054 0,121 -0,175 0,679 0,146 -0,045 0,398 0,197 0,096 0,098 0,043 0,074 0,031 -0,031 -0,011 -0,050 -0,020 0,0004 -0,0007 0,007131 0,003075 0,000219 -0,001321 0,000112 0,000031

MMHg -0,350 0,222 0,152 0,567 0,088 -0,496 0,023 -0,053 0,248 0,107 -0,334 -0,025 -0,143 -0,003 0,126 -0,025 0,005 -0,006 -0,034 0,047 0,002 0,005 0,002466 -0,026305 -0,001455 -0,000350 -0,000299 0,000016

DO 0,300 -0,378 -0,570 0,159 -0,376 -0,306 -0,224 0,089 -0,011 -0,135 0,020 0,134 0,191 0,150 0,076 0,009 -0,089 -0,108 -0,025 0,044 -0,0009 0,003 0,002640 0,007201 -0,001796 -0,001548 0,000087 -0,000023

SPC -0,879 0,272 -0,042 0,281 -0,064 -0,035 -0,091 0,118 -0,033 -0,0006 -0,052 -0,019 -0,065 0,031 -0,036 0,099 -0,127 0,050 0,048 -0,051 -0,034 0,007 0,015851 0,012989 -0,025612 -0,007926 0,000582 -0,000058

COD -0,631 0,122 -0,110 -0,075 -0,067 0,417 0,208 -0,257 -0,008 0,024 -0,456 0,079 0,151 0,186 -0,070 -0,0006 0,045 -0,029 -0,003 -0,040 0,009 0,0001 0,008682 -0,005237 -0,000120 -0,000068 0,000061 -0,000002

Al3+ 0,050 0,016 0,475 0,318 0,487 -0,384 0,201 -0,315 0,046 -0,267 0,048 0,161 0,140 -0,064 -0,124 0,001 -0,046 -0,0001 0,048 0,015 -0,020 -0,014 0,017164 0,005920 0,005927 0,001479 -0,000144 0,000036

137 Factor-variable correlations (factor correlations), based on correlations

Variables C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 11 C 12 C 13 C 14 C 15 C 16 C 17 C 18 C 19 C 20 C 21 C 22 C 23 C 24 C 25 C 26 C 27 C 28

Ca2+ -0,566 0,325 -0,335 0,333 -0,424 -0,190 0,088 0,094 -0,094 -0,032 0,112 0,143 -0,074 0,075 -0,195 -0,040 0,122 0,065 -0,015 0,066 0,014 0,013 0,002497 0,006499 -0,003315 0,001901 0,000249 0,000028

Fe2+/3+ 0,092 -0,653 0,577 0,317 -0,133 0,030 0,248 0,006 -0,015 0,031 0,035 -0,024 -0,043 0,030 -0,113 -0,134 -0,084 -0,047 -0,024 -0,029 0,051 0,013 -0,038754 -0,005483 -0,010376 0,000380 0,000309 -0,000074

K+ -0,889 0,057 0,242 -0,006 0,091 -0,038 -0,071 0,038 0,047 -0,108 0,096 -0,054 0,027 -0,057 -0,061 0,281 0,059 -0,092 -0,059 -0,002 0,051 -0,006 -0,006836 -0,012928 -0,004762 -0,000147 0,000025 0,000018

Mg2+ -0,063 0,327 -0,277 0,015 -0,342 0,234 0,540 -0,054 0,230 -0,493 0,073 -0,189 -0,049 -0,019 0,071 -0,007 -0,033 -0,005 -0,005 0,015 -0,006 0,005 0,002426 -0,001485 0,000447 0,000152 -0,000043 0,000000

Mn2+ 0,051 -0,665 0,313 0,228 -0,371 0,265 0,141 0,316 0,135 0,170 -0,043 -0,008 -0,050 -0,080 -0,096 0,038 -0,010 -0,060 0,019 0,023 -0,048 -0,016 0,038780 -0,007678 0,010105 -0,001136 -0,000191 0,000120

Na+ -0,866 0,133 0,244 0,189 0,119 0,090 -0,104 0,194 -0,095 -0,052 -0,024 -0,088 0,027 0,136 -0,035 0,060 -0,099 0,067 -0,063 0,011 -0,042 0,020 -0,022599 0,002162 0,023184 0,008164 -0,000309 -0,000036

Tsi -0,338 -0,111 0,576 -0,171 -0,107 -0,246 0,148 0,214 -0,498 -0,263 -0,051 0,088 -0,122 0,102 0,113 -0,014 0,060 -0,033 0,026 -0,035 -0,012 -0,014 0,005586 -0,005954 0,002702 -0,003439 -0,000280 0,000035

Cl- -0,778 0,140 0,164 0,174 0,171 0,114 -0,092 0,300 -0,182 -0,060 -0,014 -0,204 0,250 -0,093 0,024 -0,147 0,008 0,031 -0,025 0,038 0,038 -0,011 0,024202 -0,009273 -0,006912 -0,002951 0,000035 0,000022

2- - NO3 + NO2 -0,147 0,349 0,098 0,384 -0,273 0,412 -0,098 -0,482 -0,393 0,096 0,111 0,092 -0,063 -0,112 0,062 0,009 -0,068 -0,026 -0,032 0,038 0,0004 -0,0002 0,008994 -0,010738 0,001200 0,000627 -0,000384 -0,000008

- NH3 -0,860 -0,350 -0,262 -0,177 0,128 0,039 0,105 -0,019 -0,001 0,045 0,035 0,030 -0,007 0,003 -0,001 0,010 -0,016 0,013 0,052 0,042 0,006 0,002 -0,011435 -0,009106 0,002014 -0,011989 -0,005149 -0,001447

TKN -0,756 -0,510 -0,304 0,068 0,070 -0,037 -0,054 -0,070 0,093 -0,093 0,070 0,102 -0,037 -0,060 0,022 -0,062 0,021 0,028 -0,095 -0,071 -0,011 -0,003 0,013973 0,000864 0,001805 -0,001396 0,000504 0,000138

TN -0,798 -0,474 -0,298 0,038 0,070 -0,005 -0,028 -0,082 0,059 -0,063 0,069 0,094 -0,035 -0,054 0,021 -0,048 0,011 0,024 -0,070 -0,049 -0,008 -0,002 0,009670 -0,001543 0,001935 -0,003473 -0,000611 -0,000170

TP -0,790 -0,412 -0,378 -0,122 0,145 -0,037 0,104 0,003 -0,089 0,035 0,026 -0,008 -0,028 -0,011 0,029 0,002 -0,015 -0,013 0,054 0,008 -0,001 0,002 0,008433 -0,012548 -0,008900 0,031651 0,000379 -0,000324

Sulp -0,716 0,496 0,083 0,030 -0,358 -0,087 -0,087 0,061 0,088 0,018 0,034 0,075 0,155 -0,169 0,002 -0,063 0,050 -0,050 0,039 -0,046 -0,072 0,007 -0,036396 -0,018635 -0,001162 0,001370 -0,000087 -0,000051

MIB -0,422 0,282 0,582 -0,485 -0,267 -0,038 -0,143 0,026 0,207 -0,012 0,041 0,141 0,001 -0,008 0,026 -0,037 -0,044 0,005 0,010 -0,012 0,047 0,052 0,026752 0,002065 0,002308 0,004992 -0,002800 0,000647

Geos -0,488 -0,207 0,236 -0,077 -0,370 -0,376 -0,074 -0,410 0,022 0,161 0,201 -0,325 0,053 0,163 -0,014 -0,014 0,031 0,009 0,018 -0,026 -0,015 -0,010 0,014427 -0,005130 0,002786 -0,001348 -0,000166 0,000004

TOC -0,892 -0,115 0,248 -0,038 0,264 0,082 0,040 -0,096 0,004 0,056 -0,026 -0,085 -0,045 -0,023 0,044 -0,053 0,064 -0,101 -0,018 0,051 -0,040 0,032 -0,002387 0,050032 -0,004877 -0,001044 0,000467 -0,000044

Table 8-5: Hardness of water is classified as follows by Kunin (cited by DWAF, 1996a).

Hardness Range (mg CaCO3/l) Description of Hardness 0 - 50 Soft 50 - 100 Moderately soft 100 - 150 Slightly hard 150 - 200 Moderately hard 200 - 300 Hard > 300 Very hard

138 APPENDIX C

Figure 9-1: Map indicating the three main land tenure classes for the Sabie-Sand catchment (as seen in Table 5-5). To the west is a dominant commercial tenure, centre of the catchment is predominantly communal as indicated with the red and light blue polygons and to the east, the dominant land tenure class is conservational as represented by the SAPAD colour legend.

139

Figure 9-2: Map indication and description of the soil types that occur in the Sabie- Sand catchment as seen in Tables 7 and 12 (Land Type Survey Staff, 2002).

140 ANNEXURE

Figure 10-1: South Africa: PDF files of South African landcover from the CSIR ARC national 1: 250 000 land cover data set segmented by secondary drainage region X3 (Resource Quality Services, 2003).

141