- i - SNOW COVER ANALYSIS FOR THE HIGH DRAKENSBERG THROUGH REMOTE SENSING: ENVIRONMENTAL IMPLICATIONS By Nicholas Andrew Maurits Mulder B. Sc. Hons (Wits) A dissertation submitted to the Faculty of Science, University of the Witwatersrand, in fulfillment of the requirements for the Degree of Master of Science Johannesburg 2007 - ii - DECLARATION I declare that this dissertation is my own, unaided work. It is submitted for the Degree of Master of Science in the University of the Witwatersrand, Johannesburg. It has not been submitted before for any degree or examination in any other University. __________________________ Nicholas Mulder __________ day of _______________________ 2007 - iii - ABSTRACT Snow occurs in the High Drakensberg of southern Africa approximately eight times per annum. Snow cover is frequently captured by Landsat satellite imagery, which provide data for the monitoring of snow cover in other regions of the world. Together with a digital elevation model, repetitive snow cover data are used to analyse the distribution of snow cover in the High Drakensberg study area. The effect that the regional and local topography, latitude, and climatic conditions have on the spatial distribution of snow and the function that temperature, wind, altitude, aspect and slope gradient play in the preservation of snow cover are examined. The results of the spatial study allow for the identification of sites that support the accumulation of snow. Specific active and relict geomorphological features were surveyed and correlated spatially to the contemporary snow cover. Among such features are linear debris ridges on south-facing valley slopes in the High Drakensberg. These appeared similar to glacial features found elsewhere in the world and are thus significant in a long-standing and highly conjectured debate over the validity of possible plateau, cirque and niche glaciation in the region. Late-lying snow cover favours gently sloping south- and southeast-facing aspects at altitudes from 3000 m ASL to just below the highest peaks in the region near 3450 m ASL, above which higher insolation levels on the flat mountain summits provides unfavourable conditions. Snow cover immediately adjacent to the Drakensberg escarpment ablates quickly whilst snow cover at high altitudes in the Lesotho interior experiences better preservation conditions. Latitude has no obvious impact on the distribution of snow cover due to the dominant role of topography in the High Drakensberg other than a limiting of snowfall to regions south of 29?S in late spring. Various synoptic conditions produce snowfall in the region, with cold fronts associated with mid- latitude cyclones producing the majority of snow cover. A strong correlation exists between the spatial distribution of snow cover and specific geomorphological features. Observed linear debris ridges are located on slopes that experience frequent contemporary snow cover, lending credence for a glacial origin of the ridges during a period of colder environmental conditions. - iv - ACKNOWLEDGEMENTS This dissertation would not have been possible without support and assistance from various individuals and institutions. I would like to acknowledge the help and assistance received from Helen Every, Stephanie Mills and my mother, Liz Mulder during various fieldwork visits to the High Drakensberg. I would also like to thank these individuals, as well as Joel LeBaron and Christine Deschamps for their good company both at the Sani Top Chalets as well as in remote locations in Lesotho. The School of Archaeology, Geography and Environmental Studies at the University of the Witwatersrand has been exceedingly supportive of my dissertation. They have provided facilities as well as funding for this dissertation, particularly the purchase of the high-resolution Landsat imagery used in the study, for which I am exceedingly grateful. During a fieldwork trip accompanied by Liz in the central section of the High Drakensberg, we were repeatedly attacked and fired at by a Basutho herder whilst camped for the night near Redi Summit. Thanks to the combined effort of the Himeville Police Station, the Loteni forest wardens, the South African National Defence Force and particularly the local farming community of Himeville, we were evacuated by helicopter at daybreak. Throughout the night, we were in constant contact by cell phone with the police and wardens, who gave us psychological support as a group of rangers raced through the night on foot to reach us. The incident was the most traumatic situation I have had to endure, and I?d like to thank the numerous anonymous faces that went the night without sleep to ensure that we arrived safely back in Himeville. My parents, Liz and Pieter Mulder, have been exceedingly supportive and understanding, both personally and financially during the years it has taken me to complete this dissertation. Thank you for your encouragement. - v - Finally and most importantly, I?d like to acknowledge my sincere appreciation for my supervisor, Professor Stefan Grab. He has been immensely supportive and enthusiastic about my research and has played a major guiding role in the preparation of this dissertation. He has been pivotal in suggesting many new directions for my dissertation when various problems manifested themselves. I would like to thank him for his time and patience, his guidance and co-authorship in having my results published in journals and presented at conferences, but most importantly, for encouraging and galvanising me into completing this dissertation. - vi - CONTENTS DECLARATION ii ABSTRACT iii ACKNOWLEDGEMENTS iv LIST OF FIGURES x LIST OF TABLES xv ACRONYMS xvi CHAPTER 1 - INTRODUCTION 1.1 Settings 1 1.2 The High Drakensberg 3 1.2.1 The contemporary periglacial environment 6 1.2.2 The palaeo-periglacial environment 7 1.2.3 The palaeo-glacial environment 8 1.3 The importance of snow cover in periglacial and glacial environments 11 1.4 Remote sensing 14 1.5 Motivation 16 1.6 Aims and objectives 18 1.6.1 Hypotheses 19 CHAPTER 2 ? ENVIRONMENTAL SETTING 2.1 Introduction 20 2.2 Study area 23 2.3 Geology and geomorphology 23 2.3.1 Geomorphic evolution 26 2.4 Soils and weathering 27 2.5 Vegetation 28 2.6 Climate 31 2.6.1 Precipitation 31 - vii - 2.6.2 Wind 34 2.6.3 Insolation 36 2.6.4 Temperature 37 2.6.5 Frost 38 2.7 Land management and water resources 39 CHAPTER 3 ? LITERATURE REVIEW 3.1 Introduction 42 3.2 Periglacial and glacial geomorphology in southern Africa 43 3.3 Geomorphological phenomena of the High Drakensberg 45 3.3.1 Soil erosion forms 46 3.3.2 Thufur 47 3.3.3 Sorted patterned ground 48 3.3.4 Block deposits 49 3.3.5 Stone-banked lobes 49 3.3.6 Valley asymmetry 50 3.3.7 Basalt terraces 50 3.3.8 Glacial and nival phenomena 52 3.4 Snow cover and glacial ice in geomorphology 54 3.4.1 The importance of snow cover 55 3.4.2 The importance of glacial ice 58 3.5 Remote Sensing and Geographical Information Systems 61 3.5.1 Satellite sensing 62 3.5.2 Aerial photography 66 3.5.3 Digital elevation models 67 3.5.4 Problems and considerations in remote sensing 67 CHAPTER 4 - METHODOLOGY 4.1 Introduction 69 4.2 Constraints 69 4.3 Topographic data 71 4.4 Geomorphological ground-truthing 73 - viii - 4.5 Landsat satellite data 76 4.5.1 Satellite imagery 76 4.5.2 Snow identification in high-resolution imagery 81 4.5.3 Application of snow algorythm to high-resolution satellite data 83 4.5.4 Analysis of snow cover in high-resolution images 84 4.5.5 Snow cover change analysis from high-resolution images 86 4.5.6 Low-resolution satellite imagery 87 4.6 Climatic data 88 4.7 Aerial photographic imagery 89 4.8 Geomorphological phenomena and the relationship to snow cover 90 4.9 Summary 91 CHAPTER 5 - RESULTS 5.1 Introduction 92 5.2 Satellite imagery data 92 5.2.1 Snow cover identification from high-resolution images 92 5.2.2 Spatial analysis of snow cover from high-resolution images 97 5.2.3 Snow cover change analysis from high-resolution images 111 5.2.4 Snow cover analysis of low-resolution images 130 5.3 Climate data 149 5.4 Aerial photographic data 153 5.5 Distribution of geomorphological phenomena 154 5.6 Summary 172 CHAPTER 6 ? DISCUSSION AND CONCLUSION 6.1 Introduction 173 6.2 The use of satellite imagery 173 6.3 Factors contributing to snow cover distribution in the High Drakensberg 176 6.3.1 Distance from the escarpment 177 6.3.2 Altitude 178 6.3.3 Aspect 179 6.3.4 Slope gradient 180 - ix - 6.3.5 Latitude 181 6.3.6 Topography 182 6.3.7 Seasonal trends 184 6.4 Climatic considerations in snow cover distribution 185 6.5 Snow cover distribution and geomorphological phenomena 187 6.5.1 Block fields and block streams 188 6.5.2 Patterned ground 188 6.5.3 Stone-banked lobes 189 6.5.4 Solifluction lobes 189 6.5.5 Thufur 189 6.5.6 Debris ridges 190 6.6 Snow cover implications for the High Drakensberg environment 192 6.7 Conclusion 195 REFERENCES 199 APPENDIX 1 ? Weather data 225 APPENDIX 2 ? Google Earth data 247 APPENDIX 3 ? Discussion of weather data in relation to snow cover 263 - x - LIST OF FIGURES 1.1 Southern Africa, indicating the Great Escarpment and High Drakensberg 1 1.2 Temperature fluctuations over the last 450?000 years. Mean values derived from the Vostok Ice Core 3 1.3 The High Drakensberg and Lesotho Highlands region 4 2.1 Topographic map of the High Drakensberg region 21 2.2 Typical cross-profile of the High Drakensberg and Lesotho Highlands region 23 2.3 Landsat TM band 4 image showing relief of a section of the study area between 29?00?S - 29?50?S and 28?56?E - 29?32?E 23 2.4 Geological time frame of the earth 24 2.5 The geological sequence of the High Drakensberg and Lesotho Highlands region 24 2.6 Map of southern Africa showing the position of the Great Escarpment, the coastal plinth, and location of the study area in the High Drakensberg region 27 2.7 Relationship between altitude and rainfall between Mothelsessane, Lesotho and Bergville, KwaZulu-Natal 32 2.8 Air temperature and frost data for Letseng-la-Draai (3050m ASL), Lesotho Highlands 37 4.1 Digital Elevation Model for the High Drakensberg, showing elevations between 2600m and 3450m ASL 72 4.2 Topographic map indicating the area covered by ground-truthing 73 4.3 Topographic map showing location of observed debris ridges 74 4.4 A photo taken looking north, showing the larger of two debris deposits on a south-facing slope of the Leqooa mountain ridge 75 4.5 A photo taken looking south, showing the larger of two debris deposits on a south-facing slope of the Leqooa mountain ridge 75 4.6 A photo taken looking south, showing shadowing created by the Leqooa debris deposit in the mid-afternoon in winter 76 4.7 Topographic map showing the test area for snow cover algorithms 81 4.8 Distance operator showing distance away from the escarpment towards the Lesotho interior 85 - xi - 5.1 False colour 3-5-4 composite image of the high Drakensberg from 03 August 1990, with snow visible in pink 93 5.2 Boolean image of snow cover in the high Drakensberg from 03 August 1990, with snow cover represented in cyan 94 5.3 True colour 1-2-3 composite image of the high Drakensberg from 03 August 1990, with snow visible in white 95 5.4 Histogram of 3-5-4 composite image of 19 August 1990 96 5.5 Histogram of 1-2-3 composite image of 19 August 1990 96 5.6 Boolean snow cover images of 03 August 1990 and 19 August 1990 98 5.7 The distribution of snow cover on 03 and 19 August 1990 with respect to altitude 99 5.8 The minimum, lower quartile, median, upper quartile and maximum altitudes of snow cover on 03 and 19 August 1990 100 5.9 The distribution of snow cover on 03 and 19 August 1990 in relation to aspect 102 5.10 The distribution of snow cover on 03 and 19 August 1990 with respect to slope gradient 103 5.11 The aspect of snow cover on 03 and 19 August 1990 at different slope gradients 104 5.12 The distribution of snow cover on 03 August and 19 August 1990 with respect to distance away from the escarpment and towards the Lesotho interior 105 5.13 The distribution of snow cover with respect to distance away from the escarpment towards the Lesotho interior 106 5.14 Snow covered area with respect to latitude 108 5.15 The area of snow cover > 3000m ASL with respect to latitude 109 5.16 Change in snow covered area between 03 and 19 August 1990 110 5.17 The minimum, lower quartile, median, upper quartile and maximum altitudes of snow cover in 5? latitudinal sectors on 03 and 19 August 1990 112 5.18 The mean aspect of snow cover for 5? latitudinal sectors 113 5.19 The minimum, lower quartile, median, upper quartile and maximum slope gradient of snow cover in 5? latitudinal sectors on 03 and 19 August 1990 114 5.20 The mean displacement of snow cover from the escarpment in 5? latitudinal Sectors 115 5.21 Surface graph of snow cover in relation to distance from the escarpment for 03 August 1990 116 - xii - 5.22 Surface graph of snow cover in relation to distance from the escarpment for 19 August 1990 117 5.23 Cross-tabulation of 3?5?4 composite images between 03 and 19 August 1990 showing snow cover change 118 5.24 Altitudinal distribution of residual snow, new snow and snowmelt area from 03 to 19 August 1990 in km2 120 5.25 Altitudinal distribution of residual snow, new snow and snowmelt area from 03 to 19 August 1990 as a percentage of the total area 121 5.26 The minimum, lower quartile, median, upper quartile and maximum altitudes of snow cover on 03 and 19 August 1990 122 5.27 Altitudinal distribution of snowmelt, as a percentage of snow cover on 03 August 1990 123 5.28 The aspect of new snow, residual snow and snowmelt between 03 and 19 August 1990 124 5.29 The aspect of snowmelt between 03 and 19 August 1990 125 5.30 The distribution of new snow, residual snow and snowmelt between 03 and 19 August 1990 with respect to slope gradient 127 5.31 Distribution of residual snow, new snow and snowmelt area between 03 and 19 August 1990 with respect to distance from the escarpment towards the interior of Lesotho 128 5.32 Distribution of residual snow, new snow and snowmelt area between 03 and 19 August 1990 with respect to latitude 129 5.33 Snow cover changes between 03 and 19 August 1990 with respect to topography, showing residual snow and snow melt 131 5.34 Snow cover changes between 03 and 19 August 1990 with respect to topography, showing new snow 132 5.35 The north-western part of the study area near the Amphitheatre showing residual snow and snowmelt in relation to topography 133 5.36 The Amphitheatre area showing new snow in relation to topography 134 5.37 The Sani Pass area showing residual snow and snowmelt in relation to topography 135 5.38 The Leqooa area showing residual snow and snowmelt in relation to topography 136 - xiii - 5.39 Boolean snow cover images for 41 dates where low-resolution images were available 137-141 5.40 The recurrence of snow cover over 41 images dating between 1989 and 2004 144 5.41 The regression of repeated snow cover against altitude 145 5.42 The minimum, lower quartile, median, upper quartile and maximum number of incidences where a pixel was covered by snow across 41 images (1989-2004) in 50m altitudinal classes 146 5.43 The correlation between altitude and the number of snow covered images between 1989 and 2004 occurring in each pixel 147 5.44 The occurrence of snow per month based on data from 1989 to 2004 147-148 5.45 The occurrence of snow per season, based on data from 1989 to 2004 148 5.46 The occurrence of snow cover for different snow-producing weather types 152 5.47 The occurrence of block fields noted from ground-truthing and published literature 155 5.48 The occurrence of block streams noted from ground-truthing and published literature 156 5.49 The occurrence of sorted and non-sorted patterned ground noted from ground-truthing and published literature 157 5.50 The occurrence of stone banked lobes noted from ground-truthing and published literature 158 5.51 The occurrence of thufur noted from ground-truthing and published literature 159 5.52 The occurrence of solifluction lobes noted from ground-truthing and published literature 160 5.53 The occurrence of debris ridges noted from ground-truthing and published Literature 161 5.54 The occurrence of block fields in the Mafadi summit region in relation to snow cover on 03 August 1990 162 5.55 The occurrence of block streams in the Mafadi summit region in relation to snow cover on 03 August 1990 162 5.56 The occurrence of block fields in the Sani Pass and Thabana-Ntlenyana regions in relation to snow cover on 03 August 1990 163 5.57 The occurrence of sorted and non-sorted patterned ground in the Mafadi Summit region in relation to snow cover on 03 August 1990 164 - xiv - 5.58 The occurrence of sorted and non-sorted patterned ground in the Thabana-Ntlenyana region in relation to snow cover on 03 August 1990 164 5.59 The distribution of various geomorphological phenomena and snow cover around the Thabana-Ntlenyana summit 165 5.60 The occurrence of stone banked lobes in the Mafadi summit region in relation to snow cover on 03 August 1990 165 5.61 The occurrence of thufur in the Sani Pass and Thabana-Ntlenyana regions in relation to snow cover on 03 August 1990 166 5.62 The occurrence of solifluction lobes in the Leqooa River region in relation to snow cover on 03 August 1990 167 5.63 The occurrence of debris ridges in the Sani Pass region in relation to snow cover on 03 August 1990 168 5.64 The occurrence of debris ridges in the Leqooa River valley region in relation to snow cover on 03 August 1990 169 5.65 A debris deposit on the Tsatsa-La-Mangaung ridge just north of Sani Pass, showing its relation to topography and snow cover on 03 August 1990 170 5.66 A photo taken looking north-westwards at the Tsatsa-La-Mangaung debris ridge, showing the preferred south-east facing orientation of late-lying snow 170 5.67 A debris deposit on the south side of the Leqooa Range, showing its relation to topography and snow cover on 03 August 1990 171 5.68 A photo taken looking northwards at the Leqooa debris deposit, early morning shadowing in winter and different colluvial mantles on south-east and south-west facing slopes 171 5.69 A photo taken looking northwards at the western Leqooa debris ridge, showing the exposed bedrock underneath the upper part of the deposit 172 - xv - LIST OF TABLES 2.1 Characteristics of synoptic rain-producing system types, event frequency and contribution to mean annual KZN rainfall 35 3.1 The confirmed presence of periglacial and glacial landforms in the High Drakensberg 46 3.2 Landsat satellites and their sensors 63 3.3 Spectral Thematic Mapper band widths and their uses 63 3.4 Advantages and disadvantages of snow cover monitoring techniques 64 4.1 Comparative assessment between high and low resolution Landsat TM and ETM+ images 79 4.2 Low-resolution Landsat images where snow cover is present 80 4.3 Broad and unsupervised classification of composite images with resultant number of classes 82 4.4 Five land-cover classes developed for a supervised classification 83 4.5 The number of low resolution images with snow cover (1989-2004) and their classification in early, middle and late snow-seasons 88 4.6 Synoptic rain-producing system types and their reclassification for snow producing weather categories in the High Drakensberg and Lesotho Highlands 89 4.7 List of literature used in the identification and mapping of select periglacial and glacial geomorphological features in the High Drakensberg and Lesotho Highlands 91 5.1 Snow-covered area on 03 August 1990 107 5.2 Cross-tabulation of 3?5?4 composite images between 03 and 19 August 1990 showing snow cover change 119 5.3 Snow covered area derived from low-resolution images 142 5.4 Weather conditions between 28 July and 19 August 1990 for Underberg, KwaZulu-Natal and Bethlehem, Free State 150 5.5 Climate data for the 16 days prior to each satellite data capture 151 - xvi - ACRONYMS ASL Above Sea Level BP Before Present (the year 1950) CAD Computer Aided Design DAC Drakensberg Alpine Centre (floral region) DEM Digital Elevation Model EDM Electronic Distance Meter ELA Equilibrium Line Altitude ENSO El Ni?o Southern Oscillation ETM+ Enhanced Thematic Mapper GIS Geographical Information System GPS Geographical Positioning System KZN KwaZulu-Natal (province of South Africa) LGM Last Glacial Maximum LHWP Lesotho Highlands Water Project Ma Million years ago MAAT Mean Annual Air Temperature MALR Mean Annual Lapse Rate NDSI Normalised-Difference Snow Index RMS Root Mean Square SAWB South African Weather Bureau SWE Snow Water Equivalent TM Thematic Mapper UTM Universal Transverse Mercator WRS-2 World Wide Reference System - 2 Chapter 1 - 1 - CHAPTER 1 INTRODUCTION 1.1 SETTINGS The Drakensberg is the highest and longest mountain range in southern Africa, constituting a major component of southern Africa?s Great Escarpment, a scarp remnant of the fragmentation of the Gondwanaland super-continent, that runs parallel to the eastern and southern edge of the present African continent (Partridge, 1997a). The Drakensberg section of this escarpment is aligned in a general north-south orientation and is situated approximately 200km inland from the Indian Ocean (Figure 1.1). The central part of this mountain chain is often referred to as the ?High Drakensberg?, due to its notable elevations for large sections of the range, including the highest mountaintops in Africa south of Killimanjaro, Tanzania. Figure 1.1: Southern Africa, indicating the Great Escarpment and High Drakensberg. Chapter 1 - 2 - The High Drakensberg experiences a wide variety of climatic conditions, being particularly well known for its summer thunderstorms and winter cold frontal systems. The mountainous relief has consistently high elevations, with large areas more than 3000m ASL, with the highest peak (Thabana-Ntlenyana) reaching 3482m ASL. Snow is most common above 3000m ASL during the colder months of the year (Killick, 1963; Mulder and Grab, 2002). The High Drakensberg has been repeatedly described as a contemporary periglacial environment (Lewis, 1998a, 1998b; Boelhouwers, 1991a; Hanvey and Marker, 1992; Grab, 2000a), with snow often present after severe cold front events (Mulder and Grab, 2002). Snowfall, prolonged snow cover and snowmelt are all known to be important factors in the development of many types of mountain geomorphological phenomena, be it by initiating or hindering the development of features (Thorn, 1978, 1979; Fitzharris and Owen, 1984; Rapp, 1984; Str?mquist, 1985; Ballantyne, 1987; Raczkowska, 1990; Berrisford, 1991; Thorn, 1992; Palacios and S?nchez-Colomer, 1997; Nelson, 1998; Sene et al., 1998; Grab, 1998a; Shakesby et al., 1999; Thorn and Hall, 2002; Palacios et al., 2003; Grab et al., 2005). Snowmelt is generally rapid in the High Drakensberg, but occasional prolonged snow cover may remain for several months during winter (Hanvey and Marker, 1992). Recent palaeogeomorphic studies have suggested that sufficient snow may have accumulated during the Last Glacial Maximum (LGM) (14?000 to 20?000 yrs BP ? Figure 1.2) to produce geomorphological features resulting from perennial snow packs and small glaciers (Hall, 1994; Grab, 1996a, 2000a; Mulder and Grab, 2002; Mills and Grab, 2005). However, there have been few studies investigating the distribution and longevity of snow patches in the High Drakensberg (Mulder and Grab, 2002). Remotely sensed images provide a suitable means for snow mapping on a repetitive basis (Dozier, 1989). Such an analysis would thus indicate areas of preferred contemporary snow cover and provide indications of the likely distribution of palaeo- snowpacks and small glaciers. This dissertation thus uses remote sensing as a tool to map the distribution of snow patches in the High Drakensberg and correlate such findings to contemporary and palaeo-periglacial and possible glacial landforms. Chapter 1 - 3 - Figure 1.2: Temperature fluctuations over the last 450?000 years. Mean values derived from the Vostok Ice Core (after Petit et al., 1999). 1.2 THE HIGH DRAKENSBERG ENVIRONMENT The High Drakensberg is characterised by a steep escarpment that forms the border between the eastern part of Lesotho and the KwaZulu-Natal province of South Africa (Figure 1.3). To the west of the escarpment is the predominantly grassy highland region of Lesotho, referred to as the ?Lesotho Highlands?, where a significant portion of the alpine terrain is located at over 3000m ASL. The KwaZulu- Natal side is characterised by montane grassland and wooded ravines at significantly lower altitudes (generally lower than 2200m ASL) (Killick, 1963, 1978a, 1978b; Morris et al., 1993; Carbutt and Edwards, 2004). This hilly region with steep incised valleys and long spurs is often referred to as the ?Little Berg? (Boelhouwers, 1991a). The escarpment elevation of the High Drakensberg generally remains above 2900m ASL between the Leqooa River valley and the Amphitheatre (Figure 1.3), with the escarpment face ranging in height from 300m to 1100m. Mountain peaks regularly Chapter 1 - 4 - reach over 3000m ASL on the escarpment and the adjacent interior of Lesotho (Boelhouwers and Meiklejohn, 2002). Figure 1.3: The High Drakensberg and Lesotho Highlands region. The high altitude topography of Lesotho and the Drakensberg present the coldest regional weather conditions in southern Africa (Tyson and Schulze, 1976), with a mean annual air temperature (MAAT) of 4?C recorded on the highest elevations (Grab, 1999a). The Drakensberg itself is a major orographic barrier on a sub-continental scale, affecting the weather and climate patterns over a large part of southern Africa (Preston-Whyte and Tyson, 1997). Regions immediately east and above the escarpment experience increased rainfall due to orographic blocking, particularly in the form of thunderstorm activity, whilst the adjacent coastal belt experiences coastal lows formed when air descending over the escarpment undergoes cyclonic vorticity (Preston-Whyte and Tyson, 1997). Conditions towards the central plateau of South Africa are considerably drier by comparison, as a consequence of the Chapter 1 - 5 - escarpment blocking the movement of moist air from the warm Indian Ocean into the interior of the country (Killick, 1978c; Sene et al., 1998; Nel and Sumner, 2005). A common rain-shadow situation occurs when ridging anticyclones producing extensive cloud along the southern and eastern coastal belts do not have sufficient cloud depth to rise over the Drakensberg (Preston-Whyte and Tyson, 1997). As the High Drakensberg forms an ecotone (a significant climatic and ecological boundary) between the eastern coastal belt and the interior, it is an important region for studying both contemporary and palaeo-landforms (Marker, 1998). In many respects, the importance of cryospheres is now widely recognised. They are recognised as highly sensitive systems that impact on global heat balances and climates, water resources, ecology and geomorphology, whilst also storing information about past climates (Marcus et al., 1992; Hecht, 1997; Liston, 1999). The High Drakensberg is a marginal periglacial environment where ?Periglacial? is lately understood to mean ?the conditions, processes and landforms associated with cold, non-glacial environments? (Harris et al., 1988; Boelhouwers, 1991a). These areas are sensitive indicators to climatic change and a depository of palaeoclimatic data such as through periglacial phenomena (Hanvey and Marker, 1992; Marcus et al., 1992; Grab, 2000a; Boelhouwers and Sumner, 2003; Sumner, 2004a). With mounting concern over contemporary accelerated climate change, research fields such as periglacial geomorphology have received increasing scientific attention. The study of active periglacial landforms, as well as the interpretation of relict features, both of which have been recorded in the High Drakensberg (Lewis, 1988b; Boelhouwers, 1991a; Grab et al., 1999; Grab, 2000a), is of great importance. The presence and absence of relict and contemporary features and the knowledge of the climatic conditions in which they develop allows us to determine past environmental changes. The application of scientific models with the results from these processes allows us to interpret and predict geomorphic responses to future environmental change (Barsch, 1993; Hecht, 1997). South African periglacial studies began with Troll (1944) and have slowly shifted from a descriptive approach through most of the 20th Century to a more process based approach in recent years, paralleling a world-wide trend (Harris, 1988; Grab, 2000a). Most of the early and recent work in the High Drakensberg has been criticised for being ?qualitative in nature, with an apparent lack of scientific rigour? Chapter 1 - 6 - (Boelhouwers and Meiklejohn, 2002). The early Drakensberg work focused primarily on describing the periglacial landscape and process origins were assumed on the basis of morphology (e.g. Troll, 1944; Sparrow, 1967a, 1967b; Marker and Whittington 1971; Hastenrath and Wilkinson, 1973; King, 1974; Dyer and Marker, 1979). More recently, there has been a focus on monitoring and understanding the geomorphic process origins of particular periglacial phenomena such as thufur (Grab, 1994, 1997a, 1998a), sorted patterned ground (Boelhouwers, 1994; Grab, 1996b, 1997b, 2002a, 2004; Sumner, 2000, 2004b), stone-banked lobes (Boelhouwers, 1994; Grab, 2000b; 2004) and the occurrence of needle-ice (Grab, 1999b, 2002b). Various geomorphological processes may produce similar features, thus necessitating thorough investigations into the underlying processes that may have created them (e.g. Sumner, 1995). Numerous debates have emerged on the apparent misinterpretation and misrepresentation of landform types and associated process origins (Lewis 1988a, 1994; Lewis and Hanvey 1993; Hall 1994, 1995; Sumner 1995; Grab 2000a) with criticisms often noting that periglacial features and models from temperate northern hemisphere regions are adopted without proper evaluation. Other problems that have emerged are a lack of meaningful data sets for conclusive studies and debates of whether periglacial studies are significantly meaningful in southern Africa at all, with a belief that the periglacial climate in the country is too marginal or not present at all (Boelhouwers and Meiklejohn, 2002). . 1.2.1 The contemporary periglacial environment The term ?periglacial? was a term historically meant to refer to environments that are ?peripheral to an ice sheet or glacier?, where frost action was the predominant geomorphic force (Lozinsky, 1910). This term has since been more widely used in geomorphology to include any cold climate where snow, frost or freeze-thaw action have a significant impact on the environment (Thorn, 1992). Inherent in this wider definition are mountainous regions in temperate areas such as the High Drakensberg in southern Africa, where all three factors are noted (Marker and Whittington, 1971; Hanvey and Marker, 1992; Grab, 1997a, 2004; Mulder and Grab, 2002). The High Drakensberg is presently considered to be a marginal periglacial environment, hosting a variety of active periglacial features (Lewis, 1988a, 1988b; Boelhouwers, 1991a, 1994; Grab, 1997b, 2002b). The region is characterised by Chapter 1 - 7 - frequent winter diurnal freezing cycles at the ground surface, with no instances of permafrost (Grab, 2004). However, as the MAAT in the High Drakensberg varies between 4 and 7?C (Grab, 1999a; Boelhouwers and Sumner, 2003), the thin soil layer on shaded south-facing slopes at higher altitudes may remain continually frozen for several months (Grab, 2004). Although some typical periglacial phenomena that require permafrost conditions are absent from the region (e.g. pingos, rock glaciers and ice wedges) (Lewis, 1988b), a variety of active features, such as needle-ice, thufur, sorted and non-sorted patterned ground and solifluction lobes occur (Lewis, 1988a, 1988b; Boelhouwers, 1991a, 1994; Hanvey and Marker, 1992; Grab, 1994, 1996b, 1997a, 1997b, 1999a, 1999b, 2000a, 2002b, 2004; Grab et al., 1999). Research has indicated that the distribution of these features is dependent on moisture, soil, aspect and altitudinal factors (Boelhouwers, 1991a). However, the recording of the spatial and temporal variations of these features is extremely sporadic and incomplete, due to the vast size and inaccessibility of major sections of the High Drakensberg and Lesotho Highlands (Grab, 2000a; Boelhouwers and Meiklejohn, 2002). Results are thus confusing and sometimes contradictory. On the few occasions when research in the High Drakensberg and Lesotho Highlands has focused on the macro-scale, rather than at site-specific localities (e.g. Dyer and Marker, 1979), such work has relied on remotely sensed data such as aerial photography. 1.2.2 The palaeo-periglacial environment Relict periglacial phenomena reported from the High Drakensberg include blockfields, blockslopes and blockstreams (Linton, 1969; Hastenrath and Wilkinson, 1973; Lewis, 1988b; Boelhouwers, 1999; Grab, 1999a, 2000a; Grab et al., 1999; Sumner and Meiklejohn, 2000; Boelhouwers et al., 2002; Boelhouwers and Sumner, 2003; Sumner, 2004a), turf and stone-banked lobes (Boelhouwers, 1991a, 1994; Grab, 2000b, 2004), sorted patterned ground (Dardis and Granger, 1986; Boelhouwers, 1991a, 1994; Hanvey and Marker, 1992; Sumner, 2000, 2004b; Grab, 2002a, 2004), cryoplanation terraces (Grab et al., 1999, 2005), and solifluction lobes (Sparrow, 1967a; Hastenrath and Wilkinson, 1973; Lewis, 1988b; Boelhouwers, 1991a; Meiklejohn, 1992; Grab et al., 1999). A further debate as to the periglacial origin of asymmetric valleys has also taken place, (Boelhouwers, 1988, 2003; Meiklejohn, 1992, 1994; Grab et al., 1999). Grab (2002a) proposes that based on relict sorted Chapter 1 - 8 - patterned ground on mountain summits, temperatures must have been depressed sufficiently for the occurrence of permafrost to occur during the LGM. Some of the noted periglacial data have produced contradictory interpretations, including suggestions for arid and wet palaeo-periglacial conditions (Boelhouwers and Meiklejohn, 2002), whilst other features have been met with scepticism, such as the presence of nivation cirques (Sparrow, 1967b, 1974; Marker and Whittington, 1971; Hastenrath and Wilkinson, 1973; Dyer and Marker, 1979; Lewis, 1988b; Marker, 1991; Hanvey and Marker, 1992), protalus ramparts, rock glaciers and ice-wedge casts in the nearby Eastern Cape Drakensberg (Lewis and Dardis, 1985; Lewis, 1988b, 1994; Lewis and Hanvey, 1993). Although the present High Drakensberg periglacial climate can be considered marginal, it is also feasible to suggest that the climate was much cooler during the Late Pleistocene (LGM), allowing for the development of many forms of periglacial activity (Boelhouwers, 1991a; Boelhouwers and Meiklejohn, 2002). A temperature drop of 5.5?C to 9?C has been proposed for the Lesotho Highlands during the Late Pleistocene (Harper, 1969), which concurs with a 6?C drop suggested for the Western Cape, based on the Cango cave speleothem (Talma and Vogel, 1992). However, the formation of rock glaciers and ice-wedge casts require a highly anomalous temperature drop of up to 17?C (Grab, 2000a) for the features described by Lewis (1988b) and Lewis and Illgner (2001). To obtain a better understanding of the High Drakensberg palaeoenvironment, and to determine whether conditions during the LGM were conducive to a periglacial and / or glacial environment, further detailed sedimentological, climatological, and age-dating analyses are required for the high altitude region (Grab, 1998b; Boelhouwers and Meiklejohn, 2002). 1.2.3 The palaeo-glacial environment Claims for post-Quaternary glaciation have been made for various regions in southern Africa. Some of these claims, particularly those in the Western and Eastern Cape, have been questioned and criticised. It has been suggested that the Western Cape mountains were subjected to cryonival and gelifluvial activity (Linton, 1969), as well as apparent glaciation on the evidence of cirques and moraine (Borchert and Chapter 1 - 9 - S?nger, 1981). In the Eastern Cape Drakensberg, glacial moraine ridges were described and attributed to glaciers based on their morphology and sedimentology. Located at 2000m ASL, these ridges would have required a 10?C drop in MAAT (Lewis and Illgner, 2001). In the High Drakensberg meanwhile, it has been suggested that limited glaciation has taken place during the Pleistocene, with cirque and niche glaciation being commonly proposed (e.g. Marker and Whittington, 1971; Dyer and Marker, 1979; Marker 1991, 1998; Hall, 1994; Grab, 1996a). Recently, elongate debris ridges that resemble glacial moraine in more temperate climates have been examined in the High Drakensberg, with the proposition that these indicate the former presence of small glaciers (Mulder and Grab, 2002; Mills and Grab, 2005). A central tenet to all periglacial claims is that the cooler temperatures of the LGM and earlier Pleistocene temperature depressions resulted in a poleward expansion of the circumpolar vortex and the subsequent increased severity of cold fronts and strength of westerly winds (Budin, 1985; Preston-Whyte and Tyson, 1997). Such colder conditions would most likely increase the frequency of snowfall and the preservation of late-lying snow cover, thus encouraging snow pack development. Some valley forms in the High Drakensberg are said to exhibit morphological characteristics of nivation hollows and glacial cirques (Dyer and Marker, 1979; Marker, 1991, 1998). The hollows have their greatest concentration near the escarpment, with their frequency diminishing towards the Lesotho interior. The hollows have a preferred north-facing orientation, which has been attributed to a process of strong southerly and south-westerly winds blowing snow onto the leeward side (north and northeast) of ridges during the Pleistocene (Marker, 1991, 1998). However, this is contradictory to Nelson (1998), who asserts that cirque orientation in areas of continental climates is primarily polewards, together with a superimposed tendency for snow fencing. The formation of cirques is also contradictory to Aizen et al.?s (1997) findings, which suggest that less glacial development is found on equatorward slopes, despite such slopes having a greater recorded snowfall. Further to this, Evans (1977) concurs that there is a global trend for cirque glaciers to favour shaded and east-facing aspects. A wind-monitoring programme at Ben Macdui (3001m ASL) in the southern Drakensberg concluded that the high altitude sites generate gradient winds (Freiman et al., 1998). As the predominant climatic condition in winter is anti-cyclonic flow Chapter 1 - 10 - over the interior during the passage of westerly lows (mid-latitude cyclones), the gradient wind is usually southwesterly, affecting all exposed high altitude areas of the Drakensberg (Preston-Whyte and Tyson, 1997). Snow fencing, or the build up of deposited snow during a storm on north-facing slopes, is thus plausible. However, the ability of the snow cover to develop sufficiently against the onslaught of constant and direct incoming solar radiation is debatable. Glaciers on such an aspect would be fully exposed to solar heating during the warmest part of day. Cirque formation itself is subject to a multitude of diverse influences. Numerous factors need to be considered for the formation of glaciers, including macroclimatic (temperature and precipitation), geomorphic (landscape topography) and meso-climatic conditions (aspect) (Evans, 1977; Klimaszewski, 1993). Several large debris deposits have been located in a few cutbacks along the Drakensberg escarpment (Grab, 1996a). Although the deposits do not provide unequivocal evidence for past glaciation, it was suggested that such features may be the product of possible plateau, niche and cirque glaciation (Grab, 1996a). This hypothesis is supported by Hall (1994), who initially proposed niche glaciation within certain cutbacks along the High Drakensberg. It was further proposed that glacial activity took place ?18000 years BP, during the LGM (Grab, 1996a). Linear debris ridges, described as glacial moraine have been noted on south- facing slopes in the High Drakensberg at altitudes exceeding 3000m ASL (Mills and Grab, 2005). Topographic and spatial positioning in relation to present late-lying snow cover and geomorphological and sedimentological data confirm that the ridges may be attributed to nival and glacial processes. The results confirm that topographic controls are important for the capture and accumulation of wind blown snow, which ultimately may aid glacier development (Brown and Ward, 1996). The record of glacial and periglacial features in the High Drakensberg indicates that the climatic state of the mountain palaeoenvironment at the end of the LGM is a contentious issue, but pivotal to any claim for glacial activity. Widely accepted climate data for the latter part of the Pleistocene period, determined from cave speleothem hydrochemistry (Talma and Vogel, 1992), which is conducive for marginal glacial activity. The Cango Cave speleothem data indicate a 5?C temperature drop, which is consistent with findings elsewhere over the subcontinent (Tyson, Chapter 1 - 11 - 1999). Temperature data derived from the High Drakensberg can be inferred or obtained via two means; namely by looking at the palaeogeomorphology or at the palaeoclimatology. The present MAAT of the High Drakensberg is presently between 4?C and 6?C (Grab, 1997a, 1999a), but selective geomorphological evidence, such as patterned ground at Mafadi Summit and other summits (Grab, 1997b, 2002a), suggest that sub-zero MAATs may have occurred. Rainfall conditions during the late Pleistocene were generally drier than at present, with a 40% drop over the Kalahari, but a smaller diminution over the eastern coastal areas (Tyson, 1999). The cooling of the southern African sub-continent during this period, brought about by an expansion of the circumpolar vortex towards the equator (Preston-Whyte and Tyson, 1997), produced a set of combined effects. The westerly wind belt was displaced northwards and strengthened, resulting in stronger cold fronts occurring over southern Africa. At the same time, anticyclonic circulation over the centre of the subcontinent was weakened. The High Drakensberg and Lesotho highlands however have poor palaeoclimatic records. After analysing changing lapse rates with cold front events, Grab and Simpson (2000) note that if the intensity or frequency of cold fronts increased, a strengthening of lapse rates would occur, suggesting a cooler temperature for the High Drakensberg relative to the surrounding regions. A greater temperature depression would thus have been experienced during a period of pronounced cold front activity, encouraging the development of heavier snowfalls and better snow preservation. The combined data, showing stronger cold fronts and a decrease in temperatures, suggests that conditions for the accumulation of snow were significantly improved during the LGM. 1.3 THE IMPORTANCE OF SNOW COVER IN PERIGLACIAL AND GLACIAL ENVIRONMENTS Little research has focused on the contemporary occurrence and patterns of snow cover in the High Drakensberg. Snowfalls in the Drakensberg firstly require sufficient cold climatic conditions, and secondly, the presence of a snow producing cold front (Preston-Whyte and Tyson, 1997). As the passage of cold fronts occurs approximately on a six to seven day cycle in winter, a week of ameliorating weather conditions associated with snow thaw is provided before the passage of the next front. Chapter 1 - 12 - Consideration must be given as to the initial depth of the snow pack as well as the amount of insolation received during this amelioration period, both of which can affect the longevity of the snow-covered area significantly. Snowfall in the High Drakensberg is not always continuous and widespread. A snowfall on 4 and 5 May 2001 was recorded over large sections of the Western and Eastern Cape, the southern Drakensberg and the Maluti Mountains, yet the northern Drakensberg between Cathedral Peak and the Amphitheatre remained snow free (Rae, 2001). The distribution and volume of the majority of snowfalls is generally not extensive, with late-lying snow cover seldom lying for long periods of time (Killick, 1963). However, Grab (2002a) noted that snow cover might last up to five months in some years on well-shadowed topographic positions. Snowfall and snow cover data are poor and non-existent for the majority of the High Drakensberg (Nel and Sumner, 2005). Snow cover is a good parameter in helping understand global warming and climate change (Sokol et al., 1999). It is an important factor impacting on the geomorphology, ecology and microclimatology of an area. Thus, an examination of the occurrence, distribution, size and survival of contemporary snow patches will provide insight for palaeo-environmental reconstructions (Grudd, 1989; Dozier, 1991). Local environmental factors such as aspect, moisture, topography, soil depth and texture affect the regional distribution of periglacial features (Boelhouwers, 1991a). Many of these environmental factors, such as aspect and topography, also constrain snow cover (Brown and Ward, 1996; Aizen, 1997; Baral and Gupta, 1997, Etienne, 1999), indicating a direct causal link between snow cover and periglacial landform distribution. The association of contemporary snow cover and the distribution of active periglacial features (e.g. Andr?s et al., 2005) would also provide some insight into palaeo-nival and palaeo-geomorphic spatial patterns. Snow cover and snow packs are important environmental attributes for the formation of several types of palaeogeomorphic features such as nivation hollows and cryoplanation terraces (Nelson, 1989; Raczkowska, 1990; Thorn, 1992; Czudek, 1995; Shakesby, 1997, Christiansen, 1998; Shakesby et al., 1999; Thorn and Hall, 2002). Pronival ramparts are common landforms in mountain environments associated with long-lasting snow packs, where the snow surface acts as a transportational surface upon which falling debris is translocated (Ballantyne and Kirkbride, 1986; Ballantyne, 1987; Hall and Meiklejohn, 1997; Shakesby et al., 1999). Other effects of Chapter 1 - 13 - snow cover include a suppression of needle-ice and freeze-thaw processes through insulation effects, as well as decreasing lichen growth and plant growing seasons and constraining the burrowing activity of animals (Thorn, 1979; Str?mquist, 1985; Berrisford, 1991; Palacios and S?nchez-Colomer, 1997; Christiansen, 1998; Liston, 1999; Sancho et al., 2001). The melting of snow cover also has significant geomorphic impacts, thawing frozen ground and acting as an erosive factor when it has accumulated over sedimentary and weathered soil, where it acts as a substrate upon melting for processes such as gelifluction and solifluction, sheetwash and mudflows (Thorn, 1979; Czudek, 1995; Palacios and S?nchez-Colomer, 1997; Fukai, 2003). Snow is known to enhance solifluction processes that produce turf-banked and stone-banked terraces (Thorn, 1978; Grab, 2004). Perennial snow cover can lead to the formation of glaciers, where the topoclimatic effects encourages their formation on shady (poleward), leeward, and generally east-facing slopes (Evans, 1977; Klimaszewski, 1993). Analysis of aspect and other morphometric conditions for present day glaciers can give us a better understanding of palaeoglacier locations and palaeoclimates (Evans, 2006). Glacial research and associated work on snow cover, generally focuses on issues of morphometry (Garc?a-Ruiz, 2000) and genesis (Liebling and Scherp, 1981; Etienne, 1999), the determination of mass balance and equilibrium line altitudes (ELAs) (Adams et al., 1998; Benn and Lehmkuhl, 2000), and snow water equivalents (SWEs) (Cline et al., 1998; Liston, 1999; Molotch et al., 2001). ELA determination for palaeoglacial environments is of critical importance in the determination of past environmental conditions and the subsequent climate change (Trenhaile, 1977; Meierding, 1982; Grudd, 1989; Ohmura et al., 1996; Benn and Lehmkuhl, 2000; Porter, 2001), whilst SWE data are used in estimating the contribution that snow makes to runoff and water resources in spring and summer (Gil?ad and Bonne, 1990; Molotch et al., 2001). Accumulated snow cover contributes to cirque and other glacier formation, a factor that is very important in ecotone locations such as the High Drakensberg and Lesotho Highlands. Past research has focused on the presence of cirque glaciers in High Drakensberg, which includes an examination on their morphometry and associated sedimentary sequences (Marker and Whittington, 1971; Marker, 1994, 1995). Findings suggest a predominance of north-facing cirques (Dyer and Marker, 1979; Marker, 1991, 1998), which contradicts the majority of glacial research which indicates that radiation receipt and the occurrence of shade are some Chapter 1 - 14 - of the major determining factors for cirque glacier development, ahead of the leeward snow-blow factors contributing to the accumulation of snow proposed by Marker (Evans, 2006). Most recent glacial and some periglacial studies have started making use of remote sensing, particularly satellite data in the determination of ELA, SWE, morphometric mapping and snow cover distribution for climatological and hydrological processes (Cline et al., 1998; Hall et al., 1998; Derksen and LeDrew, 2000; Gao and Liu, 2001; Molotch et al., 2001). Remote sensing is particularly useful in the monitoring of seasonal and variable snow cover as well as reconstructing past glacial stages (Peterson et al., 2004). 1.4 REMOTE SENSING Aerial photography and remote sensing by means of satellite imagery has proven to be an excellent opportunity to precisely monitor snow cover over a large area (Piesbergen et al., 1997; Cline et al., 1998; Hall et al., 1998; Derksen and LeDrew, 2000; Molotch et al., 2001). With satellite imagery, the advantages of having a broad synoptic overview, multi-spectral data and repetitive satellite coverage allows for the recording of beneficial data on spatial and temporal snow variables (Baral and Gupta, 1997; Derksen and LeDrew, 2000). This would be particularly so in areas such as the High Drakensberg, where isolated snow patches, found over a sizeable region, can be exposed to sudden and significant micro-climatic changes that dominate such regions (Tyson et al., 1976). Aerial photography, often used prior to as well as in conjunction with the advent of satellite imagery, can complement satellite data through the accurate, high-resolution visual analysis and identification of large and medium scale geomorphological features (Dyer and Marker, 1979; Marker, 1991; Rosenthal and Dozier, 1996; Derksen and LeDrew, 2000). Recent technical advances on the Internet, particularly the satellite image map viewer Google Earth, may further aid aerial photography. The complex mathematical data derivable from remotely sensed data are catered for through the use of a Geographical Information System (GIS) database (Vitek et al., 1996; Zietsman et al., 1996; Gao and Liu, 2001). Landsat satellites, Chapter 1 - 15 - which use a TM sensor to record data in seven bands, each with separate wavelengths, are often used to study aspects related to snow (Hall et al., 1995a; Rosenthal and Dozier, 1996; Baral and Gupta, 1997; Piesbergen et al., 1997; Cline et al., 1998; Hall et al., 1999; Schaper et al., 1999; Derksen and LeDrew, 2000; Pinkham, 2000; Molotch et al., 2001; Peterson et al., 2004). Through manipulation of these data images, sometimes in conjunction with ground-based and other data, factors such as total snow covered area, snow melt and run-off rates, SWE and climatic modelling may be determined once problems such as mixed pixels, variable illumination and detector saturation have been eliminated (Rosenthal and Dozier, 1996; Derksen and LeDrew, 2000). When combined with a Digital Elevation Model (DEM), snow cover factors can be linked to factors such as slope gradient and aspect (Baral and Gupta, 1997; Gao and Liu, 2001; Molotch et al., 2001). A DEM can thus further help by compensating for shadow effects created by the topographic environment, specifically for high relief areas such as the High Drakensberg (Hall et al., 1995a; Rosenthal and Dozier, 1996). Other limiting factors in the use of remotely sensed data include the availability of such data. With Landsat imagery, scenes are only available at a sixteen- day interval, with an eight-day interval between April and August 2001 when data from both the LANDSAT 5 and 7 satellites were available. Occasional cloud cover in the region compounds the low image-frequency problem, leaving gaps in the temporal aspect of the data and potentially affecting time series analyses (Derksen and LeDrew, 2000). Snow cover influences both climatological and hydrological processes (Derksen and Le Drew, 2000). A complete analysis of snowfall, snow covered area and snow depletion rates in the High Drakensberg is not only dependent on repetitive satellite data, but also ground-truthed data. By applying geomorphological ground- truthed findings to snow cover patterns, assertions can be made over how the present snow cover area links to the palaeogeomorphology, and thus inferences can be made on the likely palaeoclimatology (Costin and Polach, 1973; Peterson et al., 2004). Climatological data also play an integral role in the development of any findings, as present climatological conditions, particularly those in winter, may provide a means for extrapolating and predicting climatic conditions during the LGM. As mentioned earlier, it has been proposed that the strengthening of the circumpolar vortex brought colder and more severe weather conditions to the sub-continent (Preston-Whyte and Tyson, 1997; Tyson, 1999; Grab and Simpson, 2000). Sedimentary sequences from Chapter 1 - 16 - the High Drakensberg have concurred and support a warming only after 13490 BP (Marker, 1995). Climate modelling for KwaZulu-Natal and the High Drakensberg has shown that cold front events produce increased lapse rates and reduce temperature recovery rates following such events (Grab and Simpson, 2000). The lapse rate modelling has indicated that temperature depressions in the High Drakensberg may be higher than previously assumed, and that contemporary lapse rates are not as severe as those of colder palaeoclimates (Grab and Simpson, 2000). With nival effects on surface temperatures, such as a further temperature depression due to a higher albedo, and decreased solar radiation due to increased cold front events generating increased cloud cover (Derksen and LeDrew, 2000), it has been proposed that temperatures may have been sufficiently low to permit perennial accumulations of snow and localised glacial ice development (Grab, 1996a; Grab and Simpson, 2000). 1.5 MOTIVATION The High Drakensberg is a unique research area, hosting the vast majority of periglacial features in the sub-continental region. There is still uncertainty regarding palaeoenvironments in this high altitude area, particularly over the presence of glacial conditions during the LGM (Boelhouwers and Meiklejohn, 2002). As the High Drakensberg is an ecotone, this has implications for climatic modelling (Cline et al., 1998; Marker, 1998; Gao and Liu, 2001). The inaccessibility of the High Drakensberg and Lesotho Highlands area, coupled with a harsh climate and poor civil infrastructure, has left the region with a dearth of scientific research (Grab and N?sser, 2001). Until recently, fieldwork has thus generally been site-specific and restricted to areas of easier access, such as the areas around Sani Pass and high mountain peaks along the escarpment. A spatial analysis of researched sites indicates that significant sections of the High Drakensberg have not yet been geomorphologically surveyed. Climatic data for much of the mountainous area are also limited, both spatially and temporally (Boelhouwers and Meiklejohn, 2002; Nel and Sumner, 2005). Much of the geomorphological research that has thus been documented has focused on describing geomorphological forms and has avoided time-consuming process studies (Boelhouwers and Meiklejohn, 2002). Chapter 1 - 17 - Remote sensing, be it either through aerial photography or satellite imagery, is well positioned for in-depth periglacial research in the High Drakensberg. Research on the contemporary spatial distribution of snow cover, as well as accumulation and ablation patterns, may help validate any claims for or against Pleistocene glaciation. Through spatial analysis, snow cover can be accurately recorded over vast areas on a repetitive basis. Snow cover has attracted little research in the High Drakensberg, with no specific research or data recorded for it. At the same time, snow cover is intricately linked to many periglacial processes (Raczkowska, 1990; Christiansen, 1998; Thorn and Hall, 2002; Andr?s et al., 2005). Significant steps can be made through the spatial analysis of snow cover in the High Drakensberg to obtain a better understanding of periglacial features in the region, with a further understanding of processes that generate these features through their spatial and temporal link with snow cover. Repeat snow cover analysis through remote sensing, in combination with ground-truthed data, would be useful in the High Drakensberg in verifying the extent of present late-lying snow patches and thus, potential locations for palaeo-glacial activity. Small glaciers are sensitive indicators of climatic fluctuations (Grudd, 1990) and as such, any possible evidence for glaciation in the High Drakensberg should correlate closely with late-lying snow patches and may have a profound impact on our understanding of the climatic conditions in the region. The implications for vegetational, faunal, pastoral and hydrological issues are also associated with snow cover and climate change (Meakins and Duckett, 1993; Grab and Morris, 1999; Makhoalibe, 1999; Grab and N?sser, 2001; Carbutt and Edwards, 2004). Field mapping in the High Drakensberg has identified numerous periglacial landforms and processes, which include thufur, sorted and non-sorted patterned ground, stone-banked lobes and solifluction lobes (Boelhouwers and Meiklejohn, 2002). The presence of blockfields, blockslopes, blockstreams and stone-banked lobes allude to periods of intensified periglacial activity, whilst linear debris ridge deposits are of potential glacial origin (Mills and Grab, 2005). The location of these debris ridges in specific south-facing valleys in the Lesotho Highlands appears to have a close visual relationship to areas of prolonged contemporary snow cover. The relationship between the periglacial landforms and contemporary snow cover, as determined through remote sensing, needs to be investigated with the aid of a GIS database. An important consideration in any snow cover analysis is the length of time Chapter 1 - 18 - over which snow cover totally ablates at such sites, as the duration has a profound effect on vegetation and fauna, as well as the geomorphological features and processes themselves (Derksen and LeDrew, 2000; Thorn and Hall, 2002; Andr?s et al., 2005). Several factors determining the spatial occurrence of snow patches and their continuing survival need to be considered. These include latitude, distance from the escarpment edge, topographic orientation, altitude and the type of climatological events preceding and following the snowfall event. Correlation between specific cold front events and their late lying snow patches is essential and should provide further insight into the variables determining snow accumulation. These parameters would differ depending on the type of cold front event and season / month (Preston-Whyte et al., 1991). Given the proposal that the High Drakensberg palaeoclimate was favourable for the accumulation of snow into ice packs and / or glaciers (Marker, 1991, 1998; Hanvey and Marker, 1992; Hall, 1994; Grab, 1996a; Grab and Simpson, 2000; Mills and Grab, 2005), it is hypothesised that contemporary areas of prolonged snow cover would be preferential sites for such phenomena. Together with geomorphological analyses of these sites, the results may help improve our present understanding of the palaeogeomorphology and palaeoenvironmental conditions for the High Drakensberg region. 1.6 AIMS AND OBJECTIVES An attempt is made to quantify the spatial and temporal distribution of late- lying snow cover in the High Drakensberg and Lesotho Highlands through remote sensing. Ground-truthing was performed to verify the existence and location of pertinent geomorphological features, such as stone banked lobes, thufur, block fields and block streams, solifluction lobes and debris ridges. Data on approximate melt out rates of late-lying snow patches are determined from satellite images. A further objective is to gain a better understanding of how latitude, altitude, topography, aspect and slope gradient affect snow cover distribution on a mesotopographic scale. This study also aims to demonstrate that snowfall events and distribution is closely associated with particular synoptic weather patterns. The decisive objective is to Chapter 1 - 19 - determine any associated patterns of snow patch distribution with that of periglacial and glacial landform distribution. 1.6.1 Hypotheses The following hypotheses will be tested for the High Drakensberg and Lesotho Highlands environment: (1) Using remotely sensed data over multiple snow events, it is proposed that: - snowfall occurs preferentially in southern regions due to latitude and climate patterns; - less snow falls north of Giants Castle due to a macro-scale topographic snow- shadow effect; - a westward increase in distance from the escarpment will lead to a decrease in snowfall due to the orographic snow-shadow effect of the escarpment; - snow patches will be confined to predominantly high altitude south-facing aspects for longer periods of time; - aspect and slope gradient are significant factors in the duration of snow cover. (2) Snow cover trends will be affected by different types of synoptic weather conditions, as well as the season during which such events take place. (3) Through ground truthing of geomorphological data and its correlation with remotely sensed snow cover, it is proposed that there is a spatial relationship between contemporary snow cover and active / inactive palaeogeomorphic, periglacial / glacial phenomena. Chapter 2 - 20 - CHAPTER 2 ENVIRONMENTAL SETTING 1.1 INTRODUCTION The Drakensberg Mountains, together with the Lesotho Highlands, form one of the most prominent topographic features in southern Africa, with the High Drakensberg section of the Great Escarpment forming a physical barrier between the elevated interior and the coastal lowlands to the east. The High Drakensberg watershed, which separates catchments flowing eastward to the Indian Ocean and westward to the Atlantic Ocean, also provides a natural border between the Kingdom of Lesotho and the province of KwaZulu-Natal in South Africa, complicating trans- border issues of land management and environmental conservation (King, 1974; Meakins and Duckett, 1993; Grab and N?sser, 2001) (Figure 2.1). The main focus area of this thesis will cover the High Drakensberg with its scarp slope as well as the adjoining high-lying areas of Lesotho. In general, areas above ca. 2900m ASL along the Drakensberg range are referred to as the ?High Drakensberg? and ?Lesotho Highlands? (Figures 2.1 and 2.2). The High Drakensberg has a general north-south alignment, but more precisely a southwest to northeast alignment from Giant?s Castle (29?21?S) to the Leqooa River valley (29?44?S) in the southern part of the High Drakensberg, and a southeast to northwest alignment from Giant?s Castle (29?21?S) to the Amphitheatre (28?44?S), in the northern part of the High Drakensberg. Mountain peaks frequently reach heights of over 3200m ASL along or beside the escarpment line between 28?40?S and 29?50?S, reaching a maximum altitude at Thabana-Ntlenyana (3482m ASL; 29?28?S, 29?16?E). The height of the watershed that forms the border between Lesotho and South Africa never falls below 2820m ASL between the Leqooa River valley and the Amphitheatre and generally remains higher than 2900m ASL. West of the watershed, the Lesotho Highlands are characterised by a fluvially dissected mountainous topography, which reaches altitudes of over 3000m ASL. Towards the east, steep escarpment cliffs immediately below the watershed give way to the ?Little Chapter 2 - 21 - Berg?, which is characterised by undulating terrain and ravines between 1800m and 2300m ASL (Figure 2.3). The High Drakensberg region is remote and isolated with Sani Pass offering the only transport link between eastern Lesotho and KwaZulu-Natal. On the South African side of the border, the 243?000 hectare uKhahlamba-Drakensberg Park was proclaimed a world heritage site in December 2000, in recognition of its unique natural beauty and San rock art (Derwent et al., 2001). The majority of the South African region is thus a protected natural and wildlife reserve, with little or no farming and herding. Settlements are only found outside the park borders. On the Lesotho side, the land is communal, falling under the domain of tribal chiefs. Rural settlements are found in the valleys whilst isolated herding posts can be found on the mountain slopes at higher altitudes. Wildlife is scarce due to game hunting, but cattle, horse and sheep herding is widespread (Grab and N?sser, 2001). Figure 2.1: Topographic map of the High Drakensberg region (after Dyer and Marker, 1979). Chapter 2 - 22 - Figure 2.2: Typical cross-profile of the High Drakensberg and Lesotho Highlands region (not to scale). Figure 2.3: Landsat TM band 4 image showing relief of a section of the study area between 29?00?S - 29?50?S and 28?56?E - 29?32?E. Chapter 2 - 23 - 2.2 STUDY AREA The study area incorporates a significant part of the High Drakensberg, from 28?40'S to 29?55'S and includes the adjoining areas of the Lesotho Highlands. The length of the Great Escarpment in this area is approximately 170km. The Main Focus Area (Figure 2.1) was identified from detailed remote sensing and geomorphological analysis. The area includes the majority of the highest elevations in southern Africa, which has attracted much periglacial and glacial geomorphological research. The extent of the study area is set so that almost all elevations greater than 3000m ASL along the Great Escarpment in southern Africa are included. This would maximize the probability of yielding good data on snow cover, which appears preferential to higher altitudes. The selected focus area is also generally remote and difficult to access, limiting the volume and spatial distribution of scientific research that has previously been conducted in the area. The Main Focus Area was set as the parameters for remotely sensed images and geomorphological ground-truthing was undertaken in a variety of selected regions within this focus area. 2.3 GEOLOGY AND GEOMORPHOLOGY The underlying geology of the High Drakensberg and Lesotho Highlands has a relatively uncomplicated stratigraphy when compared to other mountain ranges of the world. The lithology is essentially dominated by thick, volcanic, amygdaloidal, flood basalt of Upper Jurassic to Lower Triassic age, laid down approximately 182 million years ago (+/- 2 Ma), overlying aeolian sedimentary sandstones (King, 1974; Pickles, 1985; Mitchell et al., 1996) (Figures 2.4 and 2.5). The uppermost stratigraphic layer consists of Drakensberg Flood Basalts and associated Dolerite Dykes of igneous origin, the dykes indicating locations where channels of lava flowed to the surface, allowing the lava to flood over the landscape and heralding the break-up of the Gondwana super-continent (King, 1974; Van Rooy and Van Schalkwyk, 1993; Partridge, 1997a). The thickness of this flood basalt layer varies, with depths of up to 1500m in places (Mitchell et al., 1996), with an original thickness of 1600 to 1800m prior to weathering and erosion (Dunlevey et al., 1993). The basalt layer can be sub- divided into six different sections, differentiated by the amount of amygdales and disseminated clay minerals found within each type (Van Rooy and Nixon, 1990). At Chapter 2 - 24 - Figure 2.4: Geological time frame of the earth (after San Diego Natural History Museum, 2006). Figure 2.5: The geological sequence of the High Drakensberg and Lesotho Highlands region (not to scale) (after Pickles, 1985). Chapter 2 - 25 - Sani Pass (Figure 2.1), the basalt lavas are only 800m thick (Mitchell et al., 1996). The lower 175m of this basalt is subdivided into four separate units (lowest to uppermost: the Giant?s Cup, Agate Vale, Sakeng and Mkhomazana), with tuffaceous sedimentary lenses occasionally found between units, deposited during periods of geological inactivity (Mitchell et al., 1996). The upper 625m of basalt, the Lesotho Unit, is comparatively geochemically uniform (Mitchell et al., 1996). This regular sequence of basaltic rock is found to maintain its thickness for great distances along the length of the escarpment (King, 1974). Basalt terraces are commonly formed in this area, with distinct zonation for each period of lava flow easily identifiable and linked to the height of individual scarp faces. The separate lava flows average about 6m in depth and are each identifiable by the high number of pipe amygdales at the base of each flow where the colder surface generated gas bubbles which started rising through the cooling deposit (Van Rooy and Van Schalkwyk, 1993; Grab et al., 2005). Underlying the basalts are sandstone groups comprising the Clarens, Elliot and Molteno formations that represent an epoch of Late Triassic desertification (Van Rooy and Van Schalkwyk, 1993; Mitchell et al., 1996; Carbutt and Edwards, 2004) (Figure 2.5). The formations consist of rock types such as siltstones, mudstones and sandstones and were formed through deposition on alluvial fans, flood plains and flood fans in the Molteno and Elliot formations. The Clarens formation is relatively drier, mainly consisting of sandstones, with most deposition being the product of aeolian activity (Van Rooy and Van Schalkwyk, 1993). Intrusions are common within the basalts, which include dolerite dykes as well as kimberlite pipes (Partridge, 1997a). Many doleritic intrusions manifest themselves as cutbacks along the escarpment, where enhanced weathering takes place at the basalt / dolerite contact boundary (Hall, 1994; Grab et al., 1999). Together with the doleritic intrusions, the east-west fracture nature of the basalts encourages the formation of valleys that dominate in an east-west direction (Boelhouwers, 2003). Surface run-off is very high, due to the steep relief of the area, with disintegrated rock and weathered minerals being removed almost as quickly as it forms, resulting in very shallow regolith (Van Rooy and Van Schalkwyk, 1993). The continental watershed running down the length of the High Drakensberg is continually undergoing backward erosion. Stream incision to the east of the escarpment has created many cutbacks into Chapter 2 - 26 - the escarpment, producing steep gorges. By comparison, the streams in the Lesotho Highlands are less steep and more graded, resulting in a more undulating topography (Sumner, 2004b). Weathering depths of the basalt layer vary between 10m on the higher slopes to more than 25m in these valleys as a result of the increased fluvial activity (Van Rooy and Van Schalkwyk, 1993). Valleys in the Little Berg and the High Drakensberg generally have steep south-facing and gentler north-facing mountain slopes, producing asymmetric valleys that have been attributed to periglacial processes associated with a cooler climate during the Pleistocene (Meiklejohn, 1992). Gradients on south-facing basaltic slopes are generally over 30? whilst north-facing slopes are usually between 15? to 30? (Boelhouwers, 1988). 2.3.1 Geomorphic evolution The geomorphic evolution of the High Drakensberg area is primarily a consequence of an event sequence initiated by the break-up of the Gondwanaland super-continent (Partridge, 1997a). Southern Africa is a remnant of the central part of the ancient continent. The subcontinent has a high interior plateau, bounded on the west, south and east by a horseshoe-shaped escarpment, commonly referred to as the ?Great Escarpment? (Figure 2.6). Since the break-up of Gondwanaland in the late Jurassic and early Cretaceous, the movement of South America, Australia and Antarctica away from Africa has created a plinth, 50 to 200km wide, which has eroded backwards from the continental shelf towards the present line of the escarpment (King and King, 1959; King, 1963; Partridge and Maud, 1987; Partridge, 1997a). The Lesotho Highlands are a remnant of the pre-rift surface, with the High Drakensberg being the most prominent part of the scarp formed during rifting (King and King, 1959; Partridge, 1997a). Analysis of Kimberlite pipes in the highlands has indicated that the basalt surface, which has been the highest land surface in southern Africa since it was laid down, has eroded 300 - 330m in elevation over time (King, 1963; King, 1974; Partridge, 1997a). Most of the geomorphic evolution has occurred along the marginal zone of the subcontinent below the escarpment, in a process of Cretaceous erosion. Analysis of basalt clasts in rivers and marine sediment indicate that the escarpment receded almost entirely during this period of humid, tropical conditions which resulted in enhanced weathering and higher runoff rates (Partridge, Chapter 2 - 27 - 1997a). Pliocene uplift of 900m, one of four major continental uplifts that have raised the High Drakensberg by 2760m (King, 1974), encouraged fluvial incision through rejuvenated stream erosion, particularly below the escarpment. By comparison, slope development in the Lesotho Highlands above 2900m ASL dates back to the Tertiary and shows little impact of stream rejuvenation (Patridge, 1997), although basalt terrace formation is attributed to fluvial action during different erosional cycles as a result of the uplift (Van Rooy and VanSchalkwyk, 1993, Grab et al., 2005). The Lesotho Highlands above the escarpment thus exhibit a landscape history extending from beyond the Quaternary (King and King, 1959; King, 1963; King, 1974; Boelhouwers, 2003). Figure 2.6: Map of southern Africa showing the position of the Great Escarpment, the coastal plinth, and location of the study area in the High Drakensberg region. 2.4 SOILS AND WEATHERING Higher moisture levels in the environment lead to increased basalt weathering (Weinert, 1961). The basalt layers of the Lesotho Highlands and High Drakensberg are highly susceptible to such weathering, particularly in warmer, wetter conditions such as certain epochs during the Tertiary (Partridge and Maud, 1987; Boelhouwers Chapter 2 - 28 - and Sumner, 2003). Present soil mantles on the valley floors and slopes of the Lesotho Highlands are predominantly colluvial and shallow in nature, with the coarse, granular loamy material, containing spheroidal corestones reaching thicknesses of 2.5m (Boelhouwers, 1999; Grab et al., 1999; Mills and Grab, 2005). Mineral content is high, with a high base status and organic matter content up to 20% (Schmitz and Rooyani, 1987). These thin soils and exposed basalt bedrock leads to rapid run-off in the region (Sene et al., 1998). South-facing slopes above the escarpment have higher levels of moisture and show greater mantle thickness than north-facing slopes. The south-facing slopes, dominated by surface-mass-wasting processes, also have higher organic matter and clay content than north-facing slopes, which are warmer, drier and the product of in situ chemical weathering (Boelhouwers, 2003; Boelhouwers and Sumner, 2003). The varying rates of soil formation and removal caused by differing north and south- facing aspects has been argued as the cause for notable valley asymmetry, where south-facing slopes are steeper than north-facing aspects in the High Drakensberg (Klug et al., 1989). Wetlands are also found in the High Drakensberg and Lesotho Highlands. Cool temperatures, high soil moisture and dense vegetation covering fertile soils results in organic matter accumulation in the regolith (Schmitz and Rooyani, 1987; Klug et al., 1989) The higher organic matter content in turn results in decreased soil erosion compared to other areas due to higher soil stability (Klug et al., 1989; Morris et al., 1993), but this is undermined by increasing soil erosion due to vegetation loss and soil break-up as a result of increased animal husbandry and fuel burning by local communities (Meakins and Duckett, 1993). 2.5 VEGETATION The vegetation in the High Drakensberg and Lesotho Highlands is controlled by altitudinal and topographic factors. It is also controlled by seasonality, such as in winter when low temperatures and low rainfall stress the ecology (Killick, 1978c). These factors create a unique environment in southern Africa, resulting in a rich species flora and high level of endemism (Morris et al., 1993). The region forms part Chapter 2 - 29 - of the Drakensberg Alpine Centre (DAC), a floral region stretching across 40?000km2 of high-altitude environment, which contains over 2800 specific and intra-specific native taxa, a mosaic of Afro-Alpine flora of the eastern African mountain archipelago and the flora of the southern Cape Mountains, of which approximately 16% are endemic (Morris et al., 1993; Carbutt and Edwards, 2004). The flora of the region can be categorized into alpine, sub-alpine and montane belts (Killick, 1990). The alpine belt is distinguished by its climax heath communities, made up predominantly of Erica dominans. These isolated communities are dotted across a windblown treeless area of sparse and stunted vegetation featuring short grassland and evergreen dwarf shrubs including Festuca caprina, Merxmuellera disticha, Pentaschistis oreodoxa, Erica sp. and Helichrysum sp. On basalt rock outcrops, Euryops decumbens, Crassula sp., Zaluzianskya sp., Craterocapsa sp., Romulea sp., Helichrysum sp., and Merxmuellera sp. may be found, whilst dwarf shrubs such as Helichrysum trilineatum, Cutia nana, Basutica propinqua and Eumorphia sericea, are most common on the mountain summits. Interspersed among the grassland are aquatic / hygrophilous communities found in bogs on flat land and seepage points at the base of mountain slopes and river heads, such as the moss Bryum aulacomnioides, as well as higher plants such as Scirpus diabolicus, Crassula vaillantii and the sedges of the Limosella sp. Merxmuellera drakensbergensis, Senecio sp, Scirpus ficinioides, Colpodium sp. and Erica alopecurus are found in stream bank communities (Killick, 1978b; 1990). Some species have shown preferences for certain topographic positions in the alpine belt. Merxmuellera disticha, Helichrysum sessiloides and Festuca caprina are found predominantly on south-facing slopes, whilst Harpochloa falx and Pentaschistis oreodoxa prefer warmer, drier north-facing slopes (Morris et al., 1993). The wetlands of the alpine belt, which act as sponges for river water, attract a large quantity of floral life, resulting in a high organic content for the underlying soil and moderately acidic waters. Carbon dating of these wetlands has shown that they are of post-LGM Holocene age (Grobbelaar and Stegmann, 1987). The peat wetlands are generally classified as either bogs or fens, neither of them generally found below 2750m ASL (Schwabe, 1995). Bogs are found on the cooler steeper and moisture south-facing slopes, whilst fens, which are much larger in size, are located on warmer and gentler north-facing slopes. The bogs and fens are further differentiated into Chapter 2 - 30 - raised bogs (where they are situated at the base of basalt terraces and show a slight dome profile), slope bogs (at the footslope of mountains with a less pronounced dome), valleyhead fens (large and extensive, with meandering streams, small pools and hollows at the source of valleys and rivers) and slope fens (small wetlands on the foot or mid-slopes of mountains) (Schwabe, 1995). The bogs have a higher organic component in their soil due to their south-facing aspect, with a high water table that encourages short sedges and grasses to predominate, particularly Haplocarpha nervosa and Isolepis angelica. The fens, which have a higher mineral content to their soils, have a deeper water table, only having sporadic sections where the water table is near the surface. They feature large lawns of sedges and grasses, including Isolepis angelica (Schwabe, 1995). The alpine belt covers altitudes ranging from 2740m ASL and upwards (Killick, 1990). However, this value is dependent on slope aspect and a more average altitude of c. 2900m ASL and higher is suggested in places (Morris et al., 1993). Below this altitude is the sub-alpine belt, a sub-tropical grassland of Harpochloa falx, Themeda triandra and Merxmuellera distichiaon on south-facing slopes and Themeda triandra, Chrysocoma ciliata on north-facing slopes. The margin between temperate and sub-tropical grasslands is also notable for the differentiation between grasslands of predominantly C3 (non-Kranz) and C4 (Kranz) species for the alpine and sub-alpine belts respectively (Morris et al., 1993). The C3 and C4 species have different photosynthetic pathways as a result of different solar radiation receipts. It has been proposed that low soil temperatures of less than 8?C at night in the growing season limit the growth of C4 species to warmer altitudes (Vogel et al., 1978). The montane belt is found below the sub-alpine belt at altitudes less than c. 2290m ASL. The vegetation typifies the Little Berg, a steeply incised region below the High Drakensberg escarpment where pockets of indigenous forest, mainly dominated by Podocarpus sp., may occur in protected valleys (Killick, 1963; Morris et al., 1993). Chapter 2 - 31 - 2.6 CLIMATE The High Drakensberg and Lesotho Highlands experience some of the most severe climatic conditions in southern Africa, with some of the coldest temperatures and strongest inland winds on the subcontinent (Tyson et al., 1976; Freiman et al., 1998). The high altitude, orographic controls of the escarpment and mountainous highlands, as well as the marked seasonality of the regional climate, all contribute towards producing a complex mountain climate (Killick, 1963). However, the High Drakensberg climate remains understudied due to a lack of climate records (Killick, 1963; Grab, 1997c; Sene et al., 1998; Grab and Simpson, 2000; Nel and Sumner, 2005). 2.6.1 Precipitation Precipitation data in the High Drakensberg and adjoining Lesotho Highlands are particularly poor due to the inaccessibility of large areas of the escarpment (Nel and Sumner, 2005), with snow precipitation data even less forthcoming than rainfall data. Precipitation in the region is strongly affected by orographic factors, with as much as 80% of precipitation resulting from orographic thunderstorms, where moist maritime air is blown in from the Indian Ocean in the east against the High Drakensberg (Schulze, 1965; Sene et al., 1998). A cross-section from Mothelsessane in the Lesotho interior to Bergville in KwaZulu-Natal demonstrates how the regional distribution of precipitation is dictated by such orographic controls (Schulze, 1979; Figure 2.7). A general lack of data however, has led to significant uncertainty as to the distribution and volume of precipitation within the mountain range. A general consensus is that the greatest rainfall is located on the escarpment and that the scarp creates a rain-shadow that extends into Lesotho (Killick, 1963, 1978c; Schulze, 1979; Grobbelaar and Stegmann, 1987). Chapter 2 - 32 - Figure 2.7: Relationship between altitude and rainfall between Mothelsessane, Lesotho and Bergville, KwaZulu-Natal (after Schulze, 1979). The village of Mokhotlong, located in a deep valley in eastern Lesotho, clearly is impacted by the effect of rainfall shadowing along the High Drakensberg, where mean annual precipitation was recorded as 562mm during the period 1963 to 1978 (Killick, 1978b). Letseng-la-Draai, a Lesotho diamond mine closer to the escarpment and at a high elevation (3050m ASL), records 714mm per annum. Sani Pass, at 2865m ASL, records 996mm for these years, whilst 1609mm is recorded at 2927m ASL at Organ Pipes Pass Summit (29?01?S, 29?11?E) (Killick, 1963, 1978b, 1978c). More recent data confirms this trend where rainfall varies in Lesotho from 500mm in low-lying areas to over 1000mm (and potentially up to 1600mm) in the northeast Highlands (Sene et al., 1998). Flow gauge measurements in rivers draining the northeast Lesotho Highlands suggest an average mean annual precipitation for these watersheds of 1006mm (Sene et al., 1998). The number of rainy days and the frequency of rainfall events also increase with an increase in altitude (Nel and Sumner, 2005). The rain shadow effect may also be observed on a local scale, with some evidence that northeast, east and southeast facing mountain slopes have more rainfall than southwest, west and northwest slopes of similar elevation (Sene et al., 1998). Along the escarpment, annual precipitation values of 742mm have been recorded at Sani Pass in 2002 and 765mm at Sentinel Peak (S28.74?, E28.89? 3165m) in 2003. Both these values were in drier than normal years, but suggest that predictive values approaching 2000mm in the early years of research for the top of the escarpment were over-estimated (Tyson et al., 1976; Schulze, 1979; Nel and Sumner, 2005). It is noted however that rain-catch deficiencies, estimated at up to 8%, may be Chapter 2 - 33 - exaggerated by severe windy conditions such as those observed regularly during thunderstorms on the escarpment (Schulze, 1979). The increase in rainfall across the Lesotho Highlands from the southwest to the northeast is consistent with an increase in precipitation against an increase in altitude (Sene et al., 1998). This is true of transects running from the Little Berg up the High Drakensberg as well, with areas to the east of the escarpment experiencing an increase in precipitation with increasing altitude towards the escarpment summit. (Killick, 1963; Tyson et al., 1976; Schulze, 1979; Grab et al., 1999). The zone of maximum precipitation, with 1800mm, is purported to be located between 2287m and 2927m ASL, immediately to the east and slightly upwind of the escarpment (Killick, 1963; Schulze, 1979, Sene et al., 1998). Precipitation is seasonal in the High Drakensberg and Lesotho Highlands, with a distinct dry period in winter whilst heavy downpours associated with thunderstorms and other instability-driven precipitation, are common in summer (Tyson et al., 1976; Sene et al., 1998). The summer rainfall season occurs between October and March, when 77% of the precipitation is recorded (Killick, 1978c), whilst the driest months are June and July (Sene et al., 1998). In mountain environments, the fraction of precipitation falling as snow increases with an increase in altitude (Barry, 1992). Although there is no accurate recording of snowfall in the High Drakensberg (Nel and Sumner, 2005), precipitation in the form of snow is thought to occur approximately eight times per annum, usually associated with cold-frontal cyclonic weather systems in autumn, winter and early spring (Mulder and Grab, 2002). Snow cover seldom survives for any length of time, but a record from the winter of 1957 shows the snow cover surviving for two months (Killick, 1963). The importance of snow cover with regard to run-off is less important in the High Drakensberg than other mountainous regions around the world due to the low volume of snow cover and its generally thin depth. However, no studies have taken place to accurately determine the overall contribution of snow to the water resources of the region (Makhoalibe, 1999). For the 2002 / 2003 seasons, less than 0.5m of snow is thought to have fallen on the High Drakensberg, a water-equivalent of approximately 50mm, or less than 5% of the mean annual precipitation (Nel and Sumner, 2005). Even on horizontal surfaces, the snow depth is felt unlikely to exceed Chapter 2 - 34 - 1m or less than 10% of the mean annual precipitation (Sene et al., 1998; Nel and Sumner, 2005). Eight separate precipitation-producing synoptic types have been identified by Preston-Whyte et al. (1991) for KwaZulu-Natal (Table 2.1), with some of these bringing precipitation to the High Drakensberg and Lesotho Highlands. Four synoptic types were found to produce 81% of the precipitation for the province over a 20 year period in the following descending order of their contribution: tropical temperate trough, westerly wave, ridging high and east coast low (Preston-Whyte et al., 1991). Due to the lack of climate data for the High Drakensberg, their exact influence on precipitation in the region is unknown. Tropical-temperate troughs create large cumulus afternoon thunderstorms, which are regularly recorded as affecting the High Drakensberg in summer. Westerly waves extend cloud with a deep moist layer over the High Drakensberg and therefore also contribute, but Ridging Highs tend to create precipitation in mainly coastal areas, with the cloud deck generally not ascending over the High Drakensberg. East coast lows may also bring precipitation to the mountains through the help of orographic forcing. High Pressures, Easterly Flow and Tropical Cyclones are not expected to generate precipitation on the escarpment, but Mid- Latitude Cyclones, predominantly in winter may bring cold fronts to the High Drakensberg at a frequency of 4 or 5 events a month, with the potential for snow (Killick, 1963; Preston-Whyte et al., 1991; Sene et al., 1998; Grab and Simpson, 2000; Mulder and Grab, 2002). 2.6.2 Wind The wind direction on the High Drakensberg is variable, depending on a series of factors that may influence it. These include the gradient wind, regional valley and mountain winds as well as local katabatic and anabatic flow (Preston-Whyte and Tyson, 1997). The strength of these factors change seasonally and diurnally, with the gradient wind having a greater affect at higher altitudes (Barry, 1992; Freiman et al., 1998). Observations at Sani pass have revealed marked diurnal variations of wind direction under clear weather conditions, which are presumed to occur along the entire High Drakensberg (Preston-Whyte, 1971). At night, gradient winds predominate, which generally conform to the pressure gradient of the general circulation over South Chapter 2 - 35 - Table 2.1: Characteristics of synoptic rain-producing system types, event frequency and contribution to mean annual KZN rainfall (Preston-Whyte et al., 1991). Synoptic type Characteristics Frequency Rainfall Tropical- temperature trough Link between tropical easterly wave and mid-latitude low. Well-defined cloud band. Axis of surface wind convergence over KZN. Heavy summer rainfall. 15.0% 28% Westerly Wave (including Cut-off Low) Strong ridging high and upper air westerly wave. Deep layer of moist air. Cloud extends beyond Drakensberg. A cut-off low is the intense form of this type. Extended rain over 2/3 days. Can produce floods. 7.2% 24% Ridging High Ridging of South Atlantic high south of subcontinent. Absence of upper air forcing. Advection of cool unstable air below 700hPa level. Stratiform cloud deck below Drakensberg. Orographic rainfall. 21.1% 15% East Coast Low Absence of link between low off coast and interior trough. Convection less spatially organised than for trough. Convection dynamically induced. Sever thunderstorms if anticyclonic circulation over interior. 14.5% 14% High Pressure Dominant anticyclone over subcontinent. Subsidence throughout troposphere. Weak thermal advection. Slack pressure gradient. No rain. 33.8% 9% Easterly Flow Advance if high to the east. Winds back to northeasterly. Drying in upper air over most of KZN except n. sectors. Winds onshore along northern coast advecting moist tropical air. Rain in northern coastal sector. 5.1% 3% Mid-latitude Cyclone Absence of ridging high. Rapid succession of cold fronts. Southwesterly winds. Low temperatures and cloud cover. Short period rainfall in winter. 3.1% 1% Tropical Cyclone Intense tropical low advancing from Mozambique channel. Intense rain if on coast and suppresses rain if further away. 0.2% 0% Chapter 2 - 36 - Africa (Freiman et al., 1998). These winds are channelled by valley alignments in the Lesotho Highlands as well as below the escarpment in the Little Berg. At Sani Pass, the local valleys thus increase the frequency of winds from a northwesterly direction at night (Preston-Whyte, 1971). During the day, easterly winds predominate due to overflow across the escarpment from a valley ? mountain wind model occurring in the Little Berg (Preston-Whyte, 1971). In winter months, the prevailing wind direction is generally westerly as a result of a strengthened gradient wind (Killick, 1963; Preston- Whyte and Tyson, 1997). In the early 1960s, a plantation of Pinus patula on the escarpment at Cathedral Peak showed the effect of strong winds, with highly stunted trees, affected by the strong winds, leaning eastwards (Killick, 1963). No records have been made regarding the potential for windblown snow in the High Drakensberg. 2.6.3 Insolation The High Drakensberg and Lesotho Highlands are located between 28?S and 30?S. Due to these latitudes, pronounced differences in insolation between north and south-facing slopes occur, with consequences to the surface energy budget and local soil and air temperatures, soil moisture balance, soil weathering processes, vegetation and snow patch survival (Barry and Van Wie, 1974; Granger and Schulze, 1977; Morris et al., 1993). These factors are further influenced by the seasons, with less insolation received in winter when the sun is at a lower angle in the sky. Winter also presents a greater difference between north and south facing slopes than summer (Boelhouwers, 1988), creating harsher environmental conditions on south-facing winter slopes that favour long-lasting snow cover. Winter radiation in the Little Berg show a 4.5 x 106 Jm-2 day-1 energy receipt on steep south-southeast slopes, whilst north-facing slopes receive 21.0 x 106 Jm-2 day-1 of energy (Granger and Schulze, 1977). A similar study showed that at midday in winter, south facing slopes of 10? receive about 50% of the radiation that opposite valley slopes receive, whilst for 30? slopes, this is only about 17% of the level for north-facing slopes (Boelhouwers, 1988). Chapter 2 - 37 - 2.6.4 Temperature The climate of the High Drakensberg is dominated by significant diurnal and seasonal temperature variations (Grab, 1998; 1999a). The absolute maximum temperature recorded in the Little Berg area below the escarpment is approximately 35?C, whilst the lowest is approximately ?12.5?C (Tyson et al., 1976). Above the escarpment, at the Letseng-la-Draai diamond mine (3050m ASL) in the Lesotho Highlands, an absolute minimum temperature of ?20.4?C was recorded at on 12 June 1967 (Grab, 1994, 1997b), whilst an absolute high of 31?C was recorded on 29 January 1972 (Killick, 1978c) (Figure 2.8). Seasonal climatic records for Letseng-la- Draai indicate that the Lesotho Highlands have a mean summer air temperature of 11?C and mean winter air temperature of 0?C. The average minimum temperature for this area during the winter months (June to August) is ?5.6?C (Grab, 1994). Figure 2.8: Air temperature and frost data for Letseng-la-Draai (3050m ASL), Lesotho Highlands. Average values are based on ten years of data (after Grab, 1994). Along the escarpment, the MAAT at 3000m ASL is estimated at 5?C to 7?C (Boelhouwers and Sumner, 2003), whilst at Mafadi Summit (3450m ASL), it may be as low as 4?C (Grab, 1999a). Climatic monitoring for the northern part of the High Drakensberg, has recorded a MAAT of 7?C at 3200m ASL along the escarpment (Sumner, 2004a). Altitudinal changes have a strong effect on temperature; however, lapse rates are not consistent across all altitudes. Whilst the mean annual lapse rate (MALR) for the Lesotho Highlands is 5.5?C/km, an area of the sub-alpine zone Chapter 2 - 38 - between 2400 ? 2600m ASL experiences a MALR of 19.4?C/km (Grab, 1997c). This is even stronger in the winter months when the MALR is 19.9?C/km. Altitudes above this boundary only experience MALRs of 3.0?C/km (Grab, 1997c). Cold front events have a significant impact on temperature in the High Drakensberg. Before the passage of a front, a strong north-westerly blows over the escarpment, warming the air mass as it descends and creating a dry, warm ?Berg wind? with significantly elevated temperatures at lower altitudes (Preston-Whyte and Tyson, 1997). Upon the passage of a front, temperatures drop quickly with the new colder air mass and stay depressed due to cloud cover after the immediate passage. The postfrontal phase is accompanied by a slow temperature recovery during the day, but cloudless nights allow the temperatures to plummet further (Preston-Whyte and Tyson, 1997; Grab and Simpson, 2000). A strengthening of the lapse rate associated with colder temperatures is also found during cold front events. Lapse rates increased from 5.5?C/km in the pre-frontal phase to 6.6?C/km during its passage (Grab and Simpson, 2000). This has significant implications for the LGM climate, where steeper lapse rates due to more frequent cold front events would have produced colder MAATs at higher elevations in the High Drakensberg than those at lower altitudes. Snow cover would further exacerbate this cooler trend through its higher albedo and reflectance of insolation (Grab and Simpson, 2000). 2.6.5 Frost Frost is an important geomorphic agent, and may occur in the High Drakensberg and Lesotho Highlands throughout the year (Marker and Whittington, 1971; Van Zinderen Bakker and Werger, 1974) (Figure 2.8). Present maximum ground frost penetration may exceed 50cm depth and last for over 3 months at altitudes above 3400m ASL (Grab 2004). Frost frequency is estimated to range from 120 to 180 days per annum (Tyson et al., 1976; Killick, 1978c) and is most common from May to September (Grab, 1994). Frost frequency is also dependent on altitude, with the number of frost days increasing significantly between 2400 and 2600m ASL, which corresponds with the increase of MALRs (Grab, 1997c). Cryogeomorphic phenomena should therefore also increase significantly above this narrow altitudinal belt. Chapter 2 - 39 - 2.7 LAND MANAGEMENT AND WATER RESOURCES The Drakensberg?s status as a world heritage site has put significant pressure on the region to maintain its scenic, ecological and cultural heritage (Derwent et al., 2001). The adjoining Lesotho Highlands are also notable for their unique floral diversity as well as their potential water resource. However, whilst most of the KZN side of the watershed is protected by a series of parks and conservation areas, the region within Lesotho is under severe pressure from uncontrolled land resource management, leading to the degradation of its ecosystem (Makhoalibe, 1999). The Lesotho Highlands endure problems such as wildlife hunting, over grazing by domestic livestock, burning, fuelwood collection, wetland degradation, soil erosion, and degradation of its water resources (Garland, 1987; Grobbelaar and Stegmann, 1987; Schwabe, 1989; Grab and Morris, 1999; Makhoalibe, 1999; Grab and N?sser, 2001; N?sser and Grab, 2002). A high population density with limited land resources has led to socio-economic and cultural problems. Two-thirds of the Kingdom of Lesotho is covered by the highlands of the Maloti and Drakensberg mountain systems. Human settlement in the region is fairly recent, with herders only arriving in the 19th and early 20th centuries to practice transhumance pastoralism. Grazing in the mountains was initially restricted to the summer, but permanent settlements in the mountain valleys were soon established. The agricultural and pastoral intensification of the area following a population explosion saw the ploughing of valleys and lower slopes near these settlements, with little regard to adequate land-management practises, whilst grazing pushed higher up into the mountains in summer, returning to the lower valleys in winter (Schwabe, 1995; Grab and Morris, 1999; Makhoalibe, 1999). The communal grazing and uncontrolled land resource management has accelerated degradation of a fragile ecosystem through erosion. The breakdown of resources has further aggravated the situation, with increasing numbers of livestock invading wetlands and remaining waterholes in search of greener vegetation during dry periods (Makhoalibe, 1999; Grab and N?sser, 2001). The effect of overgrazing on wetlands may have a significant impact on vegetation and water-retention capacity. Erosion, through the removal of vegetation Chapter 2 - 40 - and the trampling of soils by livestock, encouraging soil break-up and aeration, is becoming increasingly more common, increasing run-off and creating short, sharp peaks in river flow (Grobbelaar and Stegmann, 1987; Schwabe, 1995; Makhoalibe, 1999; N?sser and Grab, 2002). An explosion in the Sloggett?s Ice Rat population (Otomys sloggetti) (Lynch and Watson, 1992), as a result of the drying out of wetlands, coupled together with increased burning of the wetlands to encourage new vegetation growth for livestock by herders, has exacerbated the problem by encouraging conditions for further soil moisture loss and subsequent erosion (Grobbelaar and Stegmann, 1987; Meakins and Duckett, 1993; Grab and Morris, 1999; N?sser and Grab, 2002). The wetlands are soon unable to continue their role as ?sponges? and flood flow attenuators, accumulating water from run-off and seepage, filtering it and releasing it gradually over an extended time period into the streams (Grobbelaar and Stegmann, 1987; Schwabe, 1995). The Lesotho Highlands provide a fundamental water resource for much of the subcontinent. Lesotho only occupies 3% of the Senqu and Orange River catchment (the largest catchment area in the subcontinent), yet contributes more than 47% of the total water flow (Makhoalibe, 1999). To harness these water resources, the Lesotho Highlands Water Project (LHWP), a joint initiative between the Lesotho and South African governments, was started in the mid 1980s and saw the building of large reservoirs and transfer tunnels as a hydro electrical power scheme and to supply water to the sprawling urban centres around Johannesburg. The first phase, the building of the Katse Dam and transfer tunnels, was completed in 1997 (Grab and Morris, 1999; Makhoalibe, 1999). Accelerated erosion, higher flood flow levels and the decreased ability of wetlands to retain water will all have an adverse effect on the LHWP and its surrounding communities (Grobbelaar and Stegmann, 1987), whilst the Katse Dam itself may affect local climatic conditions. In the interests of the regional environment, water resources as well as the local community?s socio-economic welfare, there have been numerous calls for improvement in grazing management at a community level (Grobbelaar and Stegmann, 1987; Meakins and Duckett, 1993; Schwabe, 1995; Grab and Morris, 1999; Makhoalibe, 1999; Grab and N?sser, 2001). Increasing population density, drought and the initial degradation of the environment has made this a complex task and severe flooding and erosion through large landslides are now affecting large areas Chapter 2 - 41 - of the Lesotho Highlands (Pers. obs., Malibimatso Valley, Nov. 2006). Transboundary initiatives are underway to link the world heritage site of the uKhahlamba-Drakensberg Park to portions of the escarpment and Lesotho Highlands (Carbutt and Edwards, 2004), providing the opportunity for linking the needs of a sustainable environment to those of the livelihoods of the local communities (Grab and N?sser, 2001). Chapter 3 - 42 - CHAPTER 3 LITERATURE REVIEW 3.1 INTRODUCTION Satellite imagery has been used extensively in glacial geomorphological studies around the world, particularly the mapping of glaciers for their spatial distribution (e.g. Brown and Ward, 1996; Baral and Gupta, 1997; Gao and Lui, 2001), the estimation of glacier equilibrium line altitudes (ELAs) (Gao and Lui, 2001), mapping of snow patches and glaciers for snow water equivalence (SWE) and runoff modelling (e.g. Gil?ad and Bonne, 1990; Cline et al., 1998; Schaper et al., 1999; Derksen and LeDrew, 2000; Gao and Lui, 2001; Molotch et al., 2001), avalanche prevention and the determination of LGM ice extents (e.g. Peterson et al., 2004). The use of topographic maps (Garc?a-Ruiz et al., 2000), aerial photography (Palacios et al., 2003), GIS and DEMs (e.g. Cline et al., 1998; Gao and Liu, 2001) and visual ground truthing is still regularly used in glacial and periglacial research. In South Africa, given the relatively recent advent in the technology, remote sensing from satellite platforms for geomorphological research has yet to be used to its full potential. Aerial photography, which has been used for many decades, has only been applied to isolated studies in the High Drakensberg (e.g. Dyer and Marker, 1979), with satellite imagery only being used in a few instances so far (e.g. Mulder and Grab, 2002; Grab et al., 2005). The studies that make use of topographic maps and aerial imagery in the analysis of geomorphological data are mostly dated (e.g. Marker, 1991), and have generally focused on the identification of features by form, paying little attention to the ramifications of their spatial distribution. Satellite imagery has been commonly used for land cover classification in southern Africa (e.g. Zietsman et al., 1996) and even wetland identification and monitoring in the High Drakensberg (Schwabe, 1995). Together with data obtained from the use of topographic data, aerial photography, DEMs and periglacial features observed through ground-truthing, a wealth of data can be collated for our understanding of the present and palaeogeomorphic and climatic conditions of the High Drakensberg. Chapter 3 - 43 - 3.2 PERIGLACIAL AND GLACIAL GEOMORPHOLOGY IN SOUTHERN AFRICA Southern African periglacial and glacial geomorphological research has taken place in sporadic locations in most of the main mountain ranges of the subcontinent (Boelhouwers and Meiklejohn, 2002). The High Drakensberg has unsurprisingly received the majority of this interest, as the mountain range is more significant, with a longer length and significantly higher elevations than other ranges. The mountains of the Eastern and Western Cape, however, have also offered some highly debated research (Boelhouwers and Meiklejohn, 2002). Although not as high as the main Drakensberg, they are located slightly further south, incurring a higher number of cold front events due to their closer proximity to the circumpolar vortex (Preston-Whyte and Tyson, 1997). It is postulated that colder palaeoclimatic conditions during the LGM may have created cooler local climates in these areas than those in the more northerly latitudes of the Drakensberg (S?nger, 1987; Lewis, 1994; Lewis and Illgner, 2001). These regions have thus seen various calls for glacial and periglacial phenomena not found in the High Drakensberg. In the mountains of the Western Cape, mountain tops reach heights of up to 2249m ASL. High levels of winter rainfall may reach 2000mm and create snow cover up to 31 days a year (Boelhouwers and Meiklejohn, 2002). Linton (1969) proposed gelifluction processes for various debris slopes down to sea level. This interpretation is widely rejected and is instead attributed to sheetwash, creep and other gravitational mass movement processes by Butzer and Helgren (1972) and Boelhouwers (1991b). Needle-ice growth, micro-patterned ground and stone- and turf-banked lobes have been recorded at high altitudes (S?nger, 1988; Boelhouwers, 1995), with Borchert and S?nger (1981) also claiming evidence for blockfields above 1500m ASL. Such features have been refuted and suggested to be the result of severe frost action rather than Pleistocene glaciation (Grab, 2000a). S?nger (1987) also proposed glacial activity resulting in blockfields and moraines in the region, whilst Boelhouwers (1991b) noted the occurrence of solifluction phenomena, debris covers, cirque-like hollows and nivation niches in the Waaihoek Mountains. The claims for glaciation have received little support, primarily due to concerns that a glacial climate in the Western Cape mountains during the LGM is unlikely, given the expected temperature Chapter 3 - 44 - depressions proposed for the late Pleistocene and early Holocene, based on palaeoclimatic records, particularly those obtained from the Cango Cave speleothem data (Talma and Vogel, 1992). The proxy data record shows that the MAAT at sea level for the Western Cape during the height of the LGM was probably around 10?C (Partridge, 1997b), suggesting that the nearby mountain environment was probably more conducive to periglacial rather than glacial phenomena (Grab, 2000a). The Eastern Cape mountains, forming the southern end of the High Drakensberg range, have also had claims for glacial and periglacial activity. Rock glaciers, ice-wedge casts, protalus ramparts and moraine ridges have variously been described (Lewis and Dardis, 1985; Lewis and Hanvey, 1993; Lewis, 1994; Lewis and Illgner, 2001). As per the claims of certain types of glacial activity in the Western Cape mountains, many of these claims have been disputed. Grab (2000a) notes that the rock glaciers at Bottelnek near Barkly Pass, as described morphologically and sedimentologically, are identical to solifluction lobes in the High Drakensberg. Environmental conditions conducive to the formation of rock glaciers would require a MAAT of about -2?C, a temperature decrease of 17?C between present conditions and the LGM (Grab, 2000a), thus significantly below the 6?C drop proposed by the speleothem data from the nearby Western Cape (Talma and Vogel, 1992). Similarly, claims for protalus ramparts (Lewis, 1994) have been suggested to be the result of pre-existing structural control rather than glacial conditions. The suggestion that these features, at 1800m ASL in the Eastern Cape, as well as those at similar altitudes in the Western Cape, be the result of such a low-temperature origin is inconsistent with geomorphological research in the High Drakensberg where such phenomena have not been recorded at significantly higher elevations and in significantly colder present-day conditions (Grab, 2000a). The study of periglacial and glacial geomorphology in the High Drakensberg has received considerable attention over the past few decades. There has been much debate and argument on the interpretation of geomorphological features, especially the Western and Eastern Cape mountains; those which are said to have a glacial origin (Marker and Whittington, 1971; Dyer and Marker, 1979; Marker, 1991; Hall, 1994, 1995; Sumner, 1995; Grab, 1996a; Grab and Hall, 1996; Mulder and Grab, 2002; Mills and Grab, 2005). A slow shift has been evident in the development of a more comprehensive understanding of the High Drakensberg in the last few decades, Chapter 3 - 45 - coinciding with a general research trend from geomorphological form to a process analysis approach (Harris, 1988; Grab, 2000a). Sparrow (1967b) describes nivation cirques in the High Drakensberg and Lesotho Highlands, but does not account for the causal process. Sparrow (1967a) also assumed that angular rocks were the product of freeze-thaw shattering, processes that were long thought to be the sole forms of weathering in cold environments. However, other weathering processes such as biological, chemical, thermal fatigue and wetting and drying are now considered to be important contributors to rock weathering in periglacial environments (Hall, 1991). A better understanding of the contemporary and palaeoenvironment of the High Drakensberg is thus developing. 3.3 GEOMORPHOLOGICAL PHENOMENA OF THE HIGH DRAKENSBERG The High Drakensberg and Lesotho Highlands have a fairly wide range of geomorphological phenomena, both active and relict, that are the result of periglacial and suspected glacial activity (Fitzpatrick, 1978; Lewis, 1988a; Boelhouwers, 1991a, 1994; Grab et al., 1999; Boelhouwers and Meiklejohn, 2002) (Table 3.1). Many geomorphic features and processes are found in the region that may also be affected by a variety of environmental factors including fauna and vegetation, topographic setting, lithological controls and climate (Lewis, 1988b; Boelhouwers, 1991a). Snow cover is also a proposed factor, with deep layers of snow cover on south-facing aspects limiting the development of periglacial phenomena reliant on frost and freeze- thaw processes. Such phenomena may thus be more common on mountain summits and interfluves (Grab, 2004). The High Drakensberg also has strong seasonality, from cold dry winters to mild wet summers that aid the formation of many geomorphic phenomena (Grab, 2002c). There is an ongoing debate over the spatial distribution and intensity of the past and present periglacial activity in the High Drakensberg and Lesotho Highlands (Garland, 1979; Boelhouwers, 1988), as is the case for the mountains of the Eastern and Western Cape. Insufficient data on the morphology and composition of phenomena has caused controversy over their origin and related palaeoenvironmental conditions (Butzer, 1973; Hall, 1992; Boelhouwers, 1994). Chapter 3 - 46 - Table 3.1: The confirmed presence of periglacial and glacial landforms in the High Drakensberg (Grab, 2000a). Landform Feature Active Relict No recording Needle ice X Deflation hollows X Sorted patterned ground X X Mudboils X Stripes X Circles ? Steps X X Thufur X X Earth Hummocks X Palsas X Pingos X Blockfields and Blockslopes ? X Blockstreams ? X Asymmetric valleys ? X Stone pavements X Cryoplanation terraces ? X Nivation hollows ? Protalus ramparts X Slush flows / thaw slumps X Turf-banked lobes X X Stone-banked lobes X X Solifluction lobes ? X Ploughing blocks X Rock glaciers X Ice-wedge casts X Sand-wedge casts X Ventifacts ? X Moraine X 3.3.1 Soil erosion forms Various forms of soil erosion phenomena are found in the High Drakensberg and Lesotho Highlands. Erosion is most commonly the result of sheet and surface erosion and gullying, but other causes are needle-ice processes, desiccation, cracking, deflation and surface and subsurface flow (Grab and Morris, 1999). Chapter 3 - 47 - Terracettes are found on dry shallow soils on steep grassy slopes. They are effectively steps in the soil surface running parallel to the contour and are a common feature of the Little Berg, but extend up to the higher elevations of the High Drakensberg and Lesotho Highlands. Their formation processes are poorly understood, but initial animal disturbance may trigger development that is complemented by a variety of processes such as soil slippage, soil flow and soil creep (Watson, 1988). A grazing test in California using a flock of sheep showed that well developed terracettes of 0.8m in width and 0.4m in height formed after only 6-weeks on slope gradients greater than 15? (Higgins, 1982). The back riser of well-developed terracettes often has soil exposed to weathering elements and as a result, terracettes at altitudes where periglacial activity is active are also subject to turf exfoliation (Grab, 2002c). Here the removal of soil that is exposed along the terrace front is primarily through needle ice action, desiccation and aeolian deflation, with occasional zoogenic help from livestock and ice rats. Winter is potentially the most important season for weathering, when cold, dry and windy conditions, particularly during strong westerly winds, are highly conducive to aeolian erosion and freeze action. Rain-splash and rain-wash would be primary erosional factors in the wet summer months (Grab, 2002c). 3.3.2 Thufur Thufur, or frost hummocks are the most common type of non-sorted patterned ground found in the High Drakensberg and Lesotho Highlands and are found in wetlands and short grasslands with a suitable soil horizon and soil moisture availability (Grab, 1998a). Lynch and Watson (1992) suggested that ice rat burrows were the originating cause for the hummock formation, as ice rats also primarily inhabit the bogs and fens where thufur are most common (Schwabe, 1995). However, Grab (1994) showed that the formation process is frost induced rather than zoogeomorphic, with the thufur profile showing a frozen region favouring the south- side of the hummock. Thufur have been widely described within the region, appearing in an altitudinal zone above 2750m ASL, below which temperature conditions are too mild for their formation (Grab, 1998a). Thufur are closely associated with the wetlands of the region. South-facing bogs often have large concentrations of thufur covering their surfaces (Schwabe, Chapter 3 - 48 - 1995), with the bog centre having a higher concentration than the drier periphery (Grab, 1998a). North-facing fens have fewer thufur, restricted to areas where the water table is near the surface (Schwabe, 1995). Closely spaced thufur encourage the preservation of snow cover in the inter-thufur spaces for several weeks due to the protection they offer the snow through shading and ?snow-fencing?. Such snow often in turn insulates the base whilst the hummock apex is exposed to freezing conditions (Grab, 1997a; 1998a). 3.3.3. Sorted patterned ground Sorted stone polygons / circles and stripes, both active and relict, are located in many high altitude sites in the High Drakensberg. The dominant process appears to be frost heaving, with angular debris sorted into different sizes by the process, leaving coarser material at the periphery and finer material in the centre of the sorted circle (Lewis, 1988b; Sumner 2003). Smaller polygons, of about 15cm in diameter, may also be affected by needle ice and other processes such as desiccation and thermal contraction (Grab, 1997b; Grab et al., 1999). Larger circles may vary in width from 0.5m to 1.3m on slope gradients of 0? to 5?, but will form into sorted stripes on steeper slopes (Dardis and Granger, 1986; Grab, 1996b). Areas of large sorted patterned ground, such as those found on the slopes of Giant?s Castle, comprise material of larger clast sizes. These relict features are indicative of a colder palaeoenvironment, as the contemporary freeze-thaw environment is not severe enough to displace clasts of such size (Boelhouwers, 1994). Such relict periglacial landforms indicate a distinct lowering of the MAAT in the past. Circles developing a diameter of 80cm or more indicate a MAAT of ?1.6?C or less, suggesting Late Quaternary permafrost at altitudes above 3400m (Grab 2002a). Other studies have concurred in part by noting that winter frost penetration in sorted circles has reached 1m depths during the LGM, but that no evidence is present for permafrost conditions (Sumner, 2004b). Sorted circle process origins meanwhile infer an absence of extensive snow cover or glacial ice in such locations (Boelhouwers, 1994). A difference in the effect snow cover has on north-facing and south-facing stone circles is noted, with higher insolation levels contributing to ground moisture and creating a stronger freezing effect in circles on north-facing aspects, whilst circles on south- facing aspects are insulated by the snow cover (Grab, 2004). Chapter 3 - 49 - 3.3.4 Block deposits Block fields are sheet-like accumulations of angular block-like material on flat or slightly sloping surfaces (Embleton and King, 1975). When they form on slope gradients steeper than 10?, they are generally referred to as block slopes. When the elongated rock body extends further downslope than across it, they are referred to as block streams (Boelhouwers, 1999). The formation of these phenomena is generally associated with active frost processes, generally freezing and thawing, on igneous rocks such as the High Drakensberg basalts (Lewis, 1988b). However, subsurface weathering for block origin also needs to be considered (Boelhouwers, 1999). Allochthonous blockfields, where the material weathered in situ, may be the result of severe seasonal freezing and not necessarily permafrost, as is evidenced by the distribution of allochthonous blockfields in various non-periglacial environments in southern Africa (Boelhouwers, 1994; Boelhouwers and Sumner, 2003). Blockstreams, some up to 1.2km long, commonly associated with first-order catchments on cooler south-facing aspects, have moved downstream through frost action (Hastenrath and Wilkinson, 1973; Boelhouwers, 1994; Grab, 1999a, 2000a; Boelhouwers and Meiklejohn, 2002; Boelhouwers et al., 2002). Block debris occurs on the upper colluvium layer in one study (Sumner, 2004a), indicating block production on the upper scarps and deposition over existing soil mantles. This suggests that the soil mantle predates the onset of the LGM and precludes against deep snow accumulations and glacier development at these locations (Boelhouwers et al., 2002; Sumner, 2004a). The movement mechanics of block deposits in general are still poorly understood in periglacial environments, limiting the ability of their use in reconstructing periglacial environments (Boelhouwers and Sumner, 2003). 3.3.5 Stone-banked lobes Stone-banked lobes are rocky crescent-shaped, V-shaped and U-shaped debris lobes overlaying existing soil horizons (Boelhouwers, 1994). Various types have been described in the High Drakensberg, resulting from different controlling mechanisms and environmental conditions, such as bedrock and regolith depth. Lobe lengths may extend up to 30m and have banks up to 2.2m in height (Boelhouwers, 1994; Grab, 2000b). They are restricted to cooler, south facing rock-strewn slopes with sparse vegetation and are thought to be the result of gelifluction and frost creep processes Chapter 3 - 50 - (Van Steijn et al., 1995; Grab, 2000a). Smaller lobes at higher altitudes may still be active, but larger lobes are relict, thought to date from the LGM when enhanced gelifluction activity was possible (Grab, 2000b). 3.3.6 Valley asymmetry Valley asymmetry and asymmetrical slope development has been proposed for sites both above and below the escarpment (Garland, 1979; Boelhouwers, 1988, 2003; Meiklejohn, 1992, 1994; Grab, 1999a). These valleys have north-facing slopes with longer lengths and gentler gradients than south-facing ones with steeper, shorter elements. North-facing slopes, which are warmer and drier, have concave profiles with slopes that have a shallow regolith and are exposed predominantly to wash- dominated wasting processes. Comparatively, south-facing slopes with a deeper saprolite are dominated by surface mass-wasting phenomena (Boelhouwers, 1998; 2003). Although periglacial activity was originally argued in their formation (e.g. Sparrow, 1964, 1971), it is presently thought to have little effect on valley asymmetry (Garland, 1979). Due to differences in insolation and the subsequent net radiation balance, the north-facing slopes experience higher chemical weathering and rapid removal of products by wash and creep processes, compared to the relatively inactive south-facing slope (Meiklejohn, 1994; Boelhouwers, 2003). Meiklejohn (1992, 1994) argues that these north-facing slopes would, to some extent, have experienced accelerated weathering during cooler palaeoclimates. North-facing slopes have relatively young Holocene carbon dates (Marker, 1994; 1995), whilst south-facing aspects are pre-Holocene, which depicts a distinct sedimentary asymmetry (Grab, 1999a; Boelhouwers and Meiklejohn, 2002; Boelhouwers et al., 2002). 3.3.7 Basalt terraces A characteristic of mountainous periglacial regions is the development of cryoplanation terraces. These occur in mountainous regions with continental climatic conditions that are altitudinally close to the snowline (Nelson, 1989). The terraces have a planar tread, often with geometric sorted patterned ground, with a steep scarp immediately upslope. The presence of deep snow cover on the tread against the scarp is required for long periods of the year, providing moisture for frost wedging and solifluction (Nelson, 1989; Czudek, 1995; Thorn and Hall, 2002). Cryoplanation Chapter 3 - 51 - terraces preferentially occupy pole-facing and windward slopes, as dictated by the prevailing conditions during the Pleistocene glaciations (Nelson, 1998). Czudek (1995) also describes cryoplanation summit flats, which undergo the same processes, but with the parallel retreat of the scarp slope segment exposing a flat plateau. The origin of cryoplanation terraces has been frequently debated, with a tentative hypothesis presented that these landforms develop through localised erosion due to the spatial patterns of snow accumulation and ablation (Nelson, 1998). Processes involved in their development include frost weathering, gravitational processes, gelifluction, sheet-wash, suffosion and frost heaving, with snow patches playing a role through their various roles as insulators, as well as mechanical and chemical agents during melt out. These processes would occur in zones of existing geological structural weakness (Czudek, 1995). Thorn and Hall (2002) have thus called for nivation and cryoplanation to be viewed as a single process spectrum given that the same weathering and erosional processes are common to each. The geology and slope gradient of the region remain underlying factors. Permafrost zones are commonly seen as sites for development, but areas without permafrost are also viable as long as there is deep seasonal ground freezing (Czudek, 1995). Nelson (1989, p31), describes cryoplanation terraces as ?a tiered series culminating in a summit flat, but rarely encircle an entire mountain or hill continuously?, a description that fits the High Drakensberg and Lesotho Highlands region well. The slopes of the High Drakensberg and Lesotho Highlands are dominated by basalt terrace formations that correspond to individual lava flow strata (Grab et al., 2005). Although found on all aspects in the region, south-east facing scarps predominate, encouraging the accumulation of late-lying snow cover, which may last up to 5 months in some years on well-shadowed positions (Grab, 2002a). Scarp face directions are 89% similar to the major joint strike direction, suggesting that lithological control is the predominant factor in their initial formation. (Grab et al., 2005). Given the similarities to cryoplanation terraces, Grab et al. (2005) propose that cryoplanation may have contributed to the modification and extension of these basalt terraces. Chapter 3 - 52 - 3.3.8 Glacial and nival phenomena There have been numerous claims for small-scale late-Pleistocene and Holocene glaciation in the form of cirque, niche and plateau glaciation in the High Drakensberg (e.g. Sparrow, 1964, Marker and Whittington, 1971; Dyer and Marker, 1979; Marker, 1991; Hall, 1994; Grab, 1996a; Mulder and Grab, 2002; Mills and Grab, 2005). With the expansion of the circumpolar vortex and strengthening of cold frontal systems with their northward displacement, colder conditions than at present would have prevailed during the LGM (Budin, 1985; Preston-Whyte and Tyson, 1997). It is generally accepted that the MAAT of southern Africa would have been up to 6?C cooler than at present (Talma and Vogel, 1992). Given that the MAAT is presently between 5?C and 7?C, depending on exact location within the High Drakensberg or Lesotho Highlands (Grab, 1997c; Boelhouwers and Sumner, 2003; Sumner, 2004a), a LGM MAAT of approximately 0?C is suggested, a value very near that required for permafrost development at higher elevations (Boelhouwers, 1994; Grab 2002a). Simultaneously, precipitation values for the LGM are observed to drop to 70% of current values (Partridge, 1997b), creating a case for a relatively dry, geomorphologically inactive periglacial climate. Marker (1991) proposed that weak cirque glaciation occurred in the Lesotho Highlands during the Pleistocene, with carbon dating of organic sediment showing no accumulation of material prior to 13?500 BP in cirque hollows (Marker, 1994; 1995). The distribution of these hollows, determined through the use of hollow flatness indices from topographic maps, shows a greater concentration next to the escarpment, with their frequency diminishing towards the Lesotho interior. Some cirques were noted as having been subsequently truncated by the retreating escarpment. Hollow orientation meanwhile, showed a strong preference for north-facing slopes, particularly on ridges running northwest to southeast and east to west. These preferences are attributed to strong southerly and southwesterly winds during the Pleistocene, that would have blown snow onto the leeward (north- and northeast- facing) ridges (Marker, 1991, 1998). The sites were analysed using Derbyshire and Evans? (1976) formula to distinguish between nivation and glacial cirques, of which half were glacial (Marker, 1998). Marker's hypothesis for process origin runs contradictory to the established literature, especially the suggestion that snow cover is Chapter 3 - 53 - more likely to survive on north-facing slopes where the highest levels of insolation are found (Granger and Schulze, 1977; Boelhouwers, 1988). Nival and glacial processes are touted as potential models for debris ridge formation in some cutbacks of the High Drakensberg escarpment (Hall, 1994). Debris ridges with distinctive channel systems are found at the lower end of these cutbacks, bearing strong resemblances to deposits on Alexander Island in the Southern Ocean, that are the result of large glacial and nival body meltout (Hall, 1994). Sumner (1995) and Hall (1995) note that the attribution of this model to the cutbacks in the High Drakensberg is based purely on qualitative form and not process origin. The morphology and sedimentology of these deposits in Nhlangeni Cutback near Thabana-Ntlenyana are recorded by Grab (1996a). These deposits, with slope gradients of approximately 24? ? 29? along the crest and reaching lengths of 747m and heights of 18m are dominant on south-facing slopes within the cutback and consist of randomly orientated angular clasts. Several hypotheses are postulated and potential cryogenic processes analysed, with the call for localized plateau, niche and cirque glaciation on higher summits (Grab, 1996a). Erosion by ice on the slope immediately above the cutback as well as the cutback walls would have dumped significant debris into the cutback, where rapid warming after the LGM would have resulted in sudden ablation and melting of glacial ice and snow, producing such debris flows (Grab, 1996a). Debris ridges are also proposed for the south-facing slopes of mountains in the Lesotho Highlands, such as the Tsatsa-La-Mangaung (near Sani Pass) and Leqooa Ranges (Mills and Grab, 2005). Various hypotheses pertaining to their origin were tested, with the proposal that they are moraines relating to the former presence of small glaciers in the south-facing valleys, which are supported by the morphology and sedimentology (Every, 2002). The ridges are only found at high altitude locations (>3000m ASL) and have been linked to areas with high frequencies of snow cover (Mills and Grab, 2005). Other claims for glacial and near-glacial environments have also been made. Fitzpatrick (1978) recorded that certain buried and compacted soil profiles found above 2280m ASL could only have formed under perennially frozen conditions, whilst Hastenrath and Wilkinson (1973) recorded cirque morphology between 2900 Chapter 3 - 54 - and 3100m ASL. Boelhouwers (1994) noted that the possibility for permafrost at Giants Castle could not be excluded given the evidence for significant mass wasting processes on the slopes of the summit. Despite high winds on the escarpment inhibiting extensive snow accumulations through snow blow and airflow friction ablation, it is proposed that limited snow cover would have given a sufficient source of moisture for such phenomena (Boelhouwers, 1994). For a present periglacial environment such as the High Drakensberg and Lesotho Highlands, claims for glaciation will need to prove that during the LGM there was sufficient snow cover and sufficiently depressed temperatures to encourage the formation of perennial snow masses and subsequent glacier formation. The distribution and effect of snow cover within both glacial and periglacial environments would need to be examined to help in this regard. 3.4 SNOW COVER AND GLACIAL ICE IN GEOMORPHOLOGY The volume of research outputs in periglacial and glacial geomorphology around the world is immense. Snowfall, snow cover and snowmelt are important considerations in cold climate geomorphology, affecting many types of periglacial phenomena as well as the larger periglacial environment. Snow cover affect climatological (the earth surface energy balance), biological and ecological (plant growth and species habitats) and hydrological (runoff flow control and frozen storage) processes through its unique physical properties: a high surface albedo; and low thermal conductivity (Derksen and LeDrew, 2000). Snow, as the progenitor of glacial ice, also plays significant roles in the mountainous regions of the world, affecting many aspects with importance to climate change observation and modelling, avalanche dangers, recreational sports, run-off and meltout predictions (Thorn, 1978, 1979, 1992; Fitzharris and Owens, 1984; Rapp, 1984; Str?mquist, 1985; Ballantyne, 1987; Raczkowska, 1990; Berrisford, 1991; Palacios and S?nchez-Colomer, 1997; Nelson, 1998; Sene et al., 1998; Grab, 1998a; Shakesby et al., 1999; Thorn and Hall, 2002; Palacios et al., 2003; Grab et al., 2005). Chapter 3 - 55 - 3.4.1 The importance of snow cover From a geomorphic perspective, snow is temporarily and spatially as important as glacial ice, with the presence and characteristics of snow fundamentally affecting weathering and mass wasting (Thorn, 1978). Significant differences exist between geomorphologically effective snowfall and rainfall precipitation, with the amount of liquid water that is released onto the surface from rainfall and made available to geomorphological processes as snow, differing due to snowdrift and sublimation processes (Thorn, 1978). Late-lying snow patches of greater depths normally protect the underlying surface from erosion, weathering and temperature changes, making it geomorphologically inactive through insulation (Str?mquist, 1985; French, 1996; Palacios and S?nchez-Colomer, 1997; Vieira et al., 2003). Snowfalls may lag behind deep freezing of the soil however, allowing deep freezing in winter before a snow cover is in place. Such snow cover may remain till late summer, preventing rainwater percolation from reaching the surface for a while (Fukai, 2003). Other effects of snow cover may include a suppression of needle-ice activity; a preclusion of lichen growth under prolonged seasonal snow cover and a constraint on burrowing animals, all of which depress both chemical and mechanical weathering of the surface (Thorn, 1978; Sancho et al., 2001). Snow cover depth is often accentuated as a result of snow blow in windy environments, with three types of snow transport: creep (rolling of snow particles), saltation (bouncing of snow particles) and suspension (turbulent uplift of particles into the atmosphere) (Barry, 1992). The need for shelter from the existing topography is paramount to the capture and protection of wind blown snow. During the melt-out phase, the water supply becomes an erosive factor immediately downslope of the snowpatch, altering geomorphic processes as the substrate experiences radical changes in temperature and moisture regimes (Thorn, 1979; French, 1996). The process is particularly strong over sedimentary and weathered material, notably morainic ridges, where gradual snowmelt encourages the removal of loose material through creep by inclusion into the base of the snowpack, frost creep, solifluction and gelifluction, small-scale mudflows and overland flow (Thorn, 1979; Palacios and S?nchez-Colomer, 1997). Thorn (1976) showed that surficial transport rates from a melting snow pack inside a nivation hollow were twice as great as adjacent snow-free surfaces. The exposed surface is subsequently prone to sheetwash and solifluction downslope of the nivation hollow (Thorn, 1979; Christiansen, 1998). Snow patches have also been shown to contribute to mechanical Chapter 3 - 56 - weathering, where surface fracturing and splitting of boulders underwent a 3 to 5 fold increase due to annual freeze and increased moisture when exposed at retreating snow margins (Berrisford, 1991). Chemical weathering by comparison, may be increased by a factor of two to four fold (Thorn, 1976, 1998). Snow packs also offer themselves as transportation surfaces, allowing movement of weathered material from above the snow pack over a further distance and with less fragmentation than would be possible without the snow (Thorn, 1978). Protalus ramparts are the only form widely attributed to such debris free-fall across snow-covered surfaces (Thorn, 1978; Ballantyne and Kirkbride, 1986; Shakesby, 1997). These features are described as ridges or ramps that form where clasts falling from a cliff face slide or roll down the surface of a perennial snow patch and accumulate at its foot (Ballantyne and Kirkbride, 1986; Shakesby, 1997). As such, they are restricted to locations where a steep slope with a reduced gradient at the base is overlooked by a cliff face. The conventional supranival processes that build up protalus ramparts are augmented by other factors as well. Shakesby et al. (1999) recorded that snow-pushing mechanisms may take place, with the snowpack weight bulldozing the substrate in addition to the conventional supranival processes. Avalanche debris may also contribute to development, bringing in a mixture of finer particles and clastic material (Ballantyne, 1987). Strong development of protalus ramparts may eventually result in situations where the stationery firn would eventually compact into glacial ice, with basal shear and internal creep of the snow / ice pack destroying or modifying as the glacier grows into protalus type moraines (Ballantyne and Benn, 1994; Hall and Meiklejohn, 1997). The landscape morphometry that is a prerequisite for protalus rampart formation would suggest that the High Drakensberg and Lesotho Highlands are well suited for the development of such features, yet none have been recorded so far (Grab, 2000a). Transportation is also possible by the movement of the snow pack itself, varying in magnitude from a gentle snow creep up to avalanches. Examples of snow pack creep at Mount Seymour in British Columbia and on Mount Twynam in Australia has found that stones in a de-glaciated cirque moved a median distance of 155m during a single winter (Thorn, 1978). Such movement is significant, inferring that snow creep can, in instances, have impacts that fall within the range expected of glaciers. Avalanches are themselves distinguished between surface avalanches and Chapter 3 - 57 - full-depth avalanches, where the entire snow pack moves as a sheet resulting in significant displacement of debris. Types of geomorphic forms resulting from avalanches include avalanche tracks, mounds, levees, pits, tarns, debris tails, and both fan and road-bank avalanche boulder tongues (Fitzharris and Owens, 1984; Fitzharris et al., 1999; Jomelli and Bertran, 2001). Altitude is the primary factor influencing the spatial distribution of snow precipitation and snow cover in mountainous areas (Aizen et al., 1997). Orographic effects encourage the accumulation of snow through increased precipitation, but high mountain environments discourage ablation due to lower temperatures at such elevations (Brown and Ward, 1996). Vegetation and geological structures are also considered controls for snow cover and nival activity (Raczkowska, 1990). As such, evenly distributed snow packs are rare on the landscape. Research has shown that snow cover is often discontinuous, occupying topographic lows and influenced by wind, vegetation, topography, precipitation, and solar radiation (Barry, 1992; Christiansen, 1998; Liston, 1999). Snow is then subject to further drifting due to these prevailing winds, thus making certain topographical sites more prone to accumulating snow packs (Thorn, 1978; Barry, 1992). Plant community distribution is also affected by such uneven distribution (Liston, 1999). Aizen et al. (1997) noted that storms in the central Tien Shan mountains in Kazakhstan deposited the thickest layers of fresh snow cover on the middle part of the windward slopes and the upper parts of leeward slopes. Fukai (2003) observed in the Japanese Alps that snow covered the ground in topographic depression near the summits, but mountain crests remained snow free and prone to permafrost. Snow accumulated to more than 16m on leeward sides of ridges as a result of wind drift versus 1m on windward slopes (Fukai, 2003). A study by Palacios et al. (2003) in Spain has shown that nivation is presently effective on exposed soil mantles on leeward slopes. Other studies in Spain have confirmed strong correlations between snow cover and leeward wind situations and geomorphological activity (Andr?s et al., 2005). In Colorado, turf- and stone-banked terraces, the result of solifluction associated with frost creep, have orientations differing from contemporary snow accumulation patterns. The results indicate differences in the prevailing wind directions between a palaeoclimate and the present (Thorn, 1978). There are many instances where snow cover exists in environments that are marginal to its requirements for long lasting cover. Snow cover in the Snowy Chapter 3 - 58 - Mountains of Australia covers an area of approximately 1400km2 for about two months per annum (Slater, 1995). The Interannual variability of snow depth is high, but no consistent or cyclic pattern is observable. However, high snowfalls are observed in years proceeding El Nino Southern Oscillation (ENSO) events, whilst low falls are recorded for non-ENSO events. Snow depth itself is related to the strength of the low-level westerly cold fronts (Budin, 1985). A general trend of decreasing snow depth over the period 1954-1993 is evident (Slater, 1995). This is confirmed by trends in alpine temperature, precipitation and snow depth records, where a 35-year warming trend and weak decline in maximum snow depth across four sites at higher elevations are also found (Hennessy et al., 2003). As the Snowy Mountains are an ecotone, a small change in environmental conditions may have a significant environmental impact. In this case it is on snow cover, where decreases of 10 to 39% are predicted by 2020 and by 22 to 85% by 2050 due to global warming (Whetton et al., 1996; Hennessy et al., 2003). This will have a significant impact on the Australian recreational skiing industry as well as the Snowy Mountain Hydro-electrical scheme (Slater, 1995). In the meantime, studies have looked into the scientific feasibility of cloud seeding for winter snowpack enhancement (Warburton and Wetzel, 1992). 3.4.2 The importance of glacial ice As per the preferential distribution of late-lying snow patches, topoclimatic effects on glaciers encourage their formation on poleward, leeward, and generally east-facing slopes (Evans, 1977). Due to the nature of the recent Pleistocene, where significant periods of severe glaciation have recently occurred, most present glaciers occur in re-glaciated environments. Such glaciers are in a geomorphic feedback loop where the glacier is protected by the cirque it forms (Evans, 1977). Instances of initial glaciation occurrences around the world are rare. The literature on initial glacial formation suggest that the need to shelter and protect wind blown snow is a principal requirement, with the ability to foster significant depths of accumulation being the overriding factor (Graf, 1976; Brown and Ward, 1996). Investigations into initial glaciation in southern Iceland provide insight into the difference of initial glacial processes from subsequent re-glaciations (Brown and Ward, 1996). Remotely sensed data were used to compare snow patch distribution between a previously glaciated site and a young volcanic landscape. Although both sites are presently well glaciated, different erosional processes were occurring. A conventional 'top-down' model for Chapter 3 - 59 - glacier growth is applicable for the re-glaciated site, but a 'bottom-up' model is more suited for the site experiencing initial glaciation. In initial glaciation, basal ice may be found in snow patches as shallow as 1m, due to melt and refreeze, allowing larger snow patches to form glacier ice in their basal layers. A large wetted perimeter exists, allowing for active erosion on all interfaces rather than linearly, as per more developed valley glaciers (Brown and Ward, 1996). In re-glaciated environments, the macroclimatic conditions for glacier formation (temperature and precipitation) are more dominant than geomorphic conditions (topography) and meso-climatic conditions (aspect) (Klimaszewski, 1993). The world-wide distribution of glaciers shows a few significant general trends: 1) glacier snow lines (the lower altitude of snow accumulation and glacier formation) rises in altitudes from polar regions towards the tropics. 2) Snowline altitudes are lower on continental sides of mountain ranges due to the controlling effect on the continental climate. 3) Poleward facing slopes have a lower snowline altitude. 4) Snowlines will have lower altitudes on east-facing slopes where the morning insolation has less effect than those on west-facing afternoon slopes. 5) Mid-latitude regions will have lower snowlines on eastern leeward slopes due to the predominant westerly winds (Embleton and King, 1968; Evans, 1977; Klimaszewski, 1993; Benn and Lehmkuhl, 2000). South Island in New Zealand for example, where the Southern Alps reach over 3000m ASL, glaciers have an average aspect of 175?, with lower glaciers orientated significantly east of south (154?) due to prevailing wind conditions (Fitzharris, et al., 1999; Chinn, 2001). However, the degree of importance of each trend is variable depending on other factors such as pre-glacial relief (Klimaszewski, 1993). Cirques are erosional hollows, with a gently sloping floor, open downstream and bounded upstream by a headwall. Unlike mountain / valley glaciers, there is little or no ice that has flowed over the top of the headwall to contribute to the erosion of the landform (Evans and Cox, 1974). Any ice has therefore accumulated in situ in the hollow and the hollow is eroding through the weight of the firn and ice (Evans, 1977). The topoclimatic characteristics of these particular forms of glaciers are similar to mountain / valley glaciers, in that they favour similar aspects and are influenced by the same set of factors (Graf, 1976; Evans, 1977; Trenhaile, 1977). However, cognisance needs to be given to the fact that many valley glaciers are of Pleistocene Chapter 3 - 60 - age, when different climatic conditions prevailed, and have subsequently been reformed by glacial erosion. Cirques are dependent on the availability of suitable topographic depressions for initial accumulation (generally a depression previously eroded by stream action), with environmental factors such as altitude, aspect and lithology playing a lesser role (Graf, 1976; Evans, 1977; Trenhaile, 1977; Garc?a-Ruiz et al., 2000). The accumulation of firn and its preservation for long periods during the summer is important in cirque development. As such, they are often more prevalent on cooler and east-facing slopes (Liebling and Scherp, 1981). This underscores the importance of nivation modification in the genesis of cirque hollows. Trenhaile (1977) identifies three main types of hollows that may allow the development of cirques: sites developed through stream erosion; sites initiated by nivation; and sites cut by streams, but subsequently modified by nivation. Through the analysis of glacier morphometry, including aspects and altitudes, glacier mass balances and erosional debris, we can provide an understanding of palaeoclimatic conditions in glacial regions (Derbyshire, 1971; Evans, 2006). The determination of equilibrium-line altitudes (ELAs) has been a vital part of using glaciation in the determination of Pleistocene conditions by comparing present ELAs to those during the time of maximum glaciation (Meierding, 1982). The ELA of a glacier is the average elevation where accumulation and ablation balance over 1 year (Benn and Lehmkuhl, 2000). Various methods are used to determine this level on larger sized glaciers, including the calculation of: altitudes of remnant cirque floors; the altitudes of glaciation thresholds; the maximum altitudes of lateral moraines; the median altitudes of reconstructed glaciers; and the areal relationship between accumulation and ablation zones of a reconstructed glacier (Meierding, 1982; Benn and Lehmkuhl, 2000). When comparing palaeo-ELAs to present ones, sea level changes need to be taken into account. The LGM saw a fall in sea levels by 120m, which effectively raises the mountains by 120m ASL, affecting temperature levels at specific locations and thus snow and glacier altitudes (Porter, 2001). ELAs are often used to compare glacial locations against altitude, aspect, latitude and location. A greater numbers of glaciers, with larger sizes at lower altitudes are shown to occur on favoured aspects. Poleward aspects once again dominated with a slight eastward tendency due to radiation receipt levels and westerly wind drift in mid-latitude regions, whilst glaciers where strong winds were effective were limited to moderate relief areas (Evans, 2006). Some factors may distort the determination of ELAs. A Chapter 3 - 61 - thin debris cover on a glacier may increase albedo, creating a higher ablation rate on the glacial surface. For thicker (>2cm) debris layers however, ablation is lower than clean ice due to the insulating effect, and glaciers may thus survive to lower altitudes (Benn and Lehmkuhl, 2000; Porter, 2001). The use of moraines in determining palaeoglacial morphology may also be affected by meltwater discharges across a broad front, which may erode moraine ridges or even prevent their initial formation (Wilson and Clark, 1998). Glaciers are good indicators of climatic change, with small glaciers and cirques being most effective as they have smaller response times to environmental conditions than larger ones (Grudd, 1989). Data from the ice record of temperate and polar glaciers has shown that climates may fluctuate more in temperate zones (Hecht, 1997), whilst glaciers in the tropics may experience even greater stresses (Klein and Isacks, 1999; Porter, 2001). In tropical regions, glaciers have formed during the Pleistocene where high altitudes gave sufficiently low temperatures for their genesis (Grab, 2002d). The Ethiopian highlands, with altitudes over 4000m ASL, held 20 glaciers with moraine ridges still evident down to 3760m ASL (Porter, 2001). The Pleistocene glaciation is estimated to have covered approximately 600km2, with glaciers extending 10 to 12km down valleys (Grab, 2002d). On Killimanjaro in Tanzania, glaciation is still present (<5km2) at altitudes above 5000m ASL, whilst Pleistocene moraines are recorded as low as 3250m when the glacial extent was 153km2 (Porter, 2001). In Papua New Guinea, where current glaciation is limited to altitudes above 4700m ASL, evidence still remains for mountain glaciation above 3700m ASL (Peterson et al., 2004). Snow lines are shown to have been 1000 to 1500m lower than present during the LGM, indicating a 6 to 8?C lowering of temperatures in the regional highlands (Webster and Streten, 1978). 3.5 REMOTE SENSING AND GEOGRAPHICAL INFORMATION SYSTEMS Remote sensing, specifically from satellites, has proven to be an indispensable tool for snow and glacier mapping and research. Although remote sensing is limited by wavelength issues such as cloud penetration and land-cover dependence, it can complement in situ snow measurement procedures which are deceptively difficult to Chapter 3 - 62 - measure and spatially constrained (Derksen and LeDrew, 2000; Gao and Liu, 2001). Remote sensing has allowed for glacial features such as spatial extent, transient snowlines, ELAs, accumulation and ablation zones, and differentiation of snow and ice around glaciers to be analysed (Gao and Liu, 2001). Remote sensing has also been used for the identification of Pleistocene glacial extents (Peterson et al., 2004). This is primarily due to the flexibility of a GIS, which can store a wide variety of inputs. Such data are stored and manipulated by the GIS in variables of location (x, y), attribute (z) and time (t) (Vitek et al., 1996). In the case of remotely sensed data, the pixels each have a x and y value and a reflectance value as z. If repetitive images are available, then t also becomes a variable. DEMs meanwhile allow for additional attributes to be assigned to pixels such as altitude, and subsequently aspect and slope gradient (Baral and Gupta, 1997). Remote sensing, with the aid of a GIS, is an extremely useful tool for snow and glacier mapping, as it allows for spatially continuous data and the manipulation of images to identify pixels with snow / ice attributes. Factors such as area, rates of snow / ice cover depletion and surface albedo can be measured (Dozier, 1991). 3.5.1 Satellite sensing Various remote sensing techniques exist, with satellite sensors becoming available in 1972 (Mulders, 2001; Rango et al., 2002; Schowengerdt, 2006). These include visible and infrared sensors, radar, gamma surveys and passive microwave sensors, with technology allowing for better monitoring through finer channels and better resolution (Derksen and LeDrew, 2000; Mulders, 2001). Some of the most commonly used forms of remote sensing for the mapping of snow has been recorded on-board Landsat satellites carrying visible and infrared Thematic Mapper (TM), MSS and ETM+ sensors (Table 3.2). These sensors have been used in many fields of study such as land cover determination (Zietsman et al., 1996), visual analysis of land use and land degradation (N?sser and Grab, 2002), the mapping of soil erosion features (Metternicht and Zinek, 1998) and wetland monitoring (Schwabe, 1995). This is possible due to the different spectral signals that soil moisture, vegetation and organic matter reflect (Schwabe, 1995). Data from the Landsat 5 sensor have been publicly available since 1984, with data from Landsat 7 since 2001. The TM, MSS and ETM+ sensors are advanced, multi-spectral, earth-resource scanning instruments that record data in spectral bands or channels. The data bands are scanned simultaneously, each Chapter 3 - 63 - Table 3.2: Landsat satellites and their sensors (NASA, 2006). Satellite Launch Date Sensors Status Landsat 1 7-23-72 MSS Expired 1-6-78 Landsat 2 1-22-75 MSS Expired 2-5-82 Landsat 3 3-5-78 MSS Expired 3-31-83 Landsat 4 7-16-82 MSS, TM TM Sensors no longer operational since 7-87 Landsat 5 3-1-84 MSS, TM Operational Landsat 6 10-5-93 MSS, ETM Lost at launch Landsat 7 4-15-99 ETM+ Operational Table 3.3: Spectral Thematic Mapper band widths and their uses (after Zietsman et al.,1996; Hall et al., 1999). Channel Wavelength (?m) Part of Spectrum and Primary Use 1 0.451 ? 0.521 Visible blue: Coastal water mapping, soil vegetation differentiation, deciduous / coniferous differentiation 2 0.526 ? 0.615 Visible green: Green reflectance by healthy vegetation 3 0.622 ? 0.699 Visible red: Chlorophyll absorption for plant species differentiation 4 0.771 ? 0.905 Reflective infrared: Biomass surveys, water body delineation 5 1.564 ? 1.790 Mid-infrared: Vegetation moisture measurement, snow / cloud differentiation 6 10.450 ? 12.460 Thermal infrared: Plant heat stress measurement, other thermal mapping 7 2.083 ? 2.351 Mid-infrared: Hydro-thermal mapping through a different wavelength (Table 3.3) and are displayed visually as a set of images. There are seven TM bands with a resolution of 30 metres for bands 1 to 5 and 7, and 120 metres for band 6 on board Landsat 5 (Sabins, 1986; Dozier, 1991; Schowengerdt, 2006). The ETM+ sensor on board Landsat 7 differs in that it has an additional eighth panchromatic band with 15m resolution whilst the MSS sensor only had a spatial resolution of 80m (NASA, 2006). The Moderate Resolution Imaging Spectroradiometer (MODIS), is another sensor commonly used for snow mapping. MODIS, which uses the Earth Observing System (EOS) satellite platform, is also different in that it uses 36 spectral bands with resolutions of between 250m to 1km (Hall et al., 1995b; Hall et al., 1998; Rango et al., 2002). The French SPOT satellite, Chapter 3 - 64 - operating since 1984, has a finer spatial resolution at 20m in multi-spectral mode and 10m in panchromatic mode. However, Dozier (1991) notes that the selection of a preferred sensor is a trade-off between spatial and temporal resolution, with SPOT having the finer spatial resolution, but TM having the better spectral coverage, making it preferable for snow mapping in specific watersheds (Dozier, 1989) (Table 3.4). Table 3.4: Advantages and disadvantages of snow cover monitoring techniques (after Derksen and LeDrew, 2000). Snow monitoring technique Advantages Disadvantages Snow course Feasible in regions where remote sensing is not operational. Long time series. Allows investigation of topographical controls, etc. Labour intensive. Point data only. Limited spatial domain. Systemic underestimation of SWE. Gauge measurements Long time series. Automatic data collection. Recent improvements in defining necessary gauge corrections. Sparse and irregular networks. Gauge catchment inaccuracy. High potential for time-series in-homogeneities. NOAA snow charts Longest satellite data record. Quality controlled and standardized. No quantitative information. Subjectively derived. Time-series discontinuity due to cloud cover. Poor accuracy with some types of surface cover and lighting conditions. MODIS Daily imagery. Automated snow mapping routine performs accurately on regional data. Transfer of regionally developed and assed algorythms to global product generation. Time-series discontinuity due to cloud cover. No quantitative SWE information. Landsat TM Suitable for basin-scale studies. Applicable in mountainous regions. Need extensive field information for verification. Potential cloud cover problems. 16-18 day revisit time. Radar All-weather imaging. Suitable for basin-scale studies. Dependent on snow wetness. 16-18 day revisit time. Gamma Quantitative derivation of SWE. Not sensitive to phase of water in snowpack. Limited to calibrated flight lines. Requires knowledge of soil conditions. No systematic measurement. Passive microwave All-weather imaging. Rapid revisit time. Ability to derive SWE. 20-year time series. Limited regional use of ground validated algorythms. Problems mapping wet snow. Sensitivity to snowpack structure. Chapter 3 - 65 - Significant research has made use of these and other sensor data sets to map snow- covered area. William and Ferrigno (1998) used satellite data to create a glacier atlas of South America. Rosenthal and Dozier (1996) used TM data from the Sierra Nevada to make quantitative estimates of the fractional snow-covered area within each pixel, whilst Sokol et al. (1999) compare passive and active microwave sensing of snow cover. Dozier (1989) successfully mapped basic-scale snow cover using surface reflectance thresholds and ratio techniques for TM bands 1, 2 and 5. TM band 5 was further used to mask cloud cover where it is highly reflective (Dozier, 1989). Hall et al. (1998) utilised MODIS data at a 500m resolution to map snow cover on a daily basis to a near 100% accuracy in tundra and grassland and within 95% accuracy in alpine areas. Hall et al. (1999) utilised TM images to perform snow mapping in the Prince Albert National Park, Saskatchewan, Canada. The imagery was used to test the SNOMAP algorithm for use with MODIS data. A Normalised-Difference Snow Index (NDSI), similar to a Normalised-Difference Vegetation Index (NDVI) was used for the identification of snow and ice, as well as the separation of snow and ice from most cumulus clouds. NDSI = (TM Band 2 - TM Band 5) / (TM Band 2 + TM Band 5) The NDSI is favourable in that it does not rely on single band thresholds. Pixels with 50% or greater snow coverage have NDSI values greater than or equal to 0.4 and a reflectance in TM band 4 greater than 11%. These results were determined through comparisons with supervised classification techniques (Hall et al., 1995b). These snow and ice determination techniques have been applied to analyse temporal and spatial patterns. Ellis and Paul (2001) have used temporal analyses of snow cover to identify long term trends of snow cover change, a valuable monitor in global climate change as the radiative properties of snow cover can create a positive feedback loop of increasing winter temperatures. Snow covered area quantifications are often used to obtain snow water equivalent (SWE) estimates. Repeated snow cover analysis, particularly in mountainous regions with heavy snowfall allows for the prediction of runoff and groundwater recharge (Cline et al., 1998; Molotch et al., 2001). A similar study has shown that snowmelt estimates in drier regions such as Israel are also valid (Gil?ad and Bonne, 1990). Chapter 3 - 66 - The use of snow and ice mapping for climate variability and climate change studies has also been undertaken. The effect that ENSO events have on snow cover has provided ambiguous results, such as with Brown (1998) where a twenty-year span of satellite data differed from a sixty-year set from station data, highlighting the complex relationship between snow cover and global climate systems (Derksen and LeDrew, 2000). For climate change, the identification of ablation and accumulation zones on glaciers as well as the transient snowline has given proxy data for ELA estimates (Klein and Isacks, 1999). Landsat TM imagery is extremely valuable in the determination of well-preserved terminal moraines and other glacial geomorphological features from the Late Pleistocene glaciation in the central Andes (Klein and Isacks, 1998). Finally, in Papua New Guinea, a variety of SPOT and Landsat, as well as aerial photography were used to map glacial moraines in mountainous highlands dating from the LGM. The imagery was able to allow the calculation of the former ice extent using the subjective analysis of moraine ridges, rock tarns, rock basin lakes and previous research in the literature (Peterson et al., 2004). 3.5.2 Aerial photography Aerial photography has received less attention since the advent of satellite platforms and particularly the recent improvements in wavelength accuracy and pixel resolution (Mulders, 2001). It is noted in the prediction of SWE through joint LANDSAT and aerial photographic data, that Rosenthal and Dozier?s (1996) spectral mixture model using TM imagery provides equal accuracy to that obtainable using high-resolution aerial photography (Cline et al., 1998). Most aerial photographs are now used to cross check reliability of satellite data or as primary data when satellite imagery is not available. In some circumstances, aerial photography may give better image resolution for the identification of small-scale features. Gil?ad and Bonne (1990) used aerial photography in conjunction with Landsat TM imagery for the determination of snow-covered area, whilst N?sser and Grab (2002) have used it for the monitoring of vegetation, land use and land cover change. Schwabe (1995) used aerial photography for change analysis of wetlands in the High Drakensberg. Detail such as the broadening of rivers, stream-bank erosion, an increase in size of pools in wetlands and their siltation were notable (Schwabe, 1995). Chapter 3 - 67 - 3.5.3 Digital Elevation Models Digital elevation models allow remotely sensed data to be seen in a quantitative altitudinal component through the use of GIS. Factors such as aspect and slope gradient may be calculated using DEM data (Baral and Gupta, 1997). Carroll (1990) initially used a DEM together with satellite data to show percentage run-off through snow melt-out rates, whilst Sene et al. (1998) used a DEM to determine the influence of topography on rainfall in the Lesotho Highlands. The use of DEMs in snow mapping has also been increasingly used to improve snow-mapping accuracy. Hall et al. (1995b) used a DEM with Landsat TM data to further develop the snow- mapping algorithm SNOMAP (incorporating the NDSI equation). Errors were noted in the initial snow mapping in mountainous areas due to the effects of topography. To overcome this, a DEM was co-registered to TM data for an area in the Glacier National Park in Montana (Hall et al., 1995a), where the terrain was so rugged, that only some of the snow was accounted for when conventional satellite data were used. It was noted that certain geometric characteristics are lost when 3-D geoids are projected onto a 2-D surface map. Hall et al.'s results showed that the DEM helped to significantly increase the accuracy of snow-mapping. Similarly, Baral and Gupta (1997) used a DEM to rectify shadow effects on repetitive images from the IRS-1B LISS-II sensor. With a 36.25m resolution, a small glacial basin in the Garhwal Himalayas was classified according to topographic facets as defined by slope gradient and aspect. Schaper et al. (1999) use Landsat TM and MSS scenes that were ortho- rectified through a DEM. The subsequent data were used to separately map snow and glacier areas in the Massa-Blatten Basin in the Swiss Alps for improved runoff modelling. 3.5.4 Problems and considerations in remote sensing Certain limiting factors need to be understood before a snow cover study through remote sensing can be undertaken. Dozier (1991) notes that snow is a collection of ice grains and air which at 0?C also has a significant fraction of liquid water. Soot, dust, pollen, plant material and trace minerals are also often contained and will to some degree affect the snow's reflectance back to the sensor. The bulk optical properties of water and snow are similar in the visible and near-infrared wavelengths, making it difficult to distinguish between cloud and snow from sensor Chapter 3 - 68 - data other than by texture and variable minor differences in wavelength reflectance. Rosenthal and Dozier (1996) warn of problems of mixed pixels, variable illumination and detector saturation. Mixed pixels occur when snow and other ground cover types are found in the same pixel area, altering reflectance values to variable degrees in different bands. This factor is especially significant in areas where snow cover is thin and where snow patches are widely distributed. Variable illumination is a result of shadowing caused by steep topography. Snow in shadows will reflect significantly lower values than snow in sunlight, leading to misclassifications of ground cover unless topography is accounted for - such as with a DEM. Dozier (1989) also noted reflectance oddities in mountainous regions due to changes in air mass with elevation. A final consideration, detector saturation, occurs when clouds or snow cover large sections of an image. This problem occurs frequently as sensors are generally developed to image factors of soils, vegetation and water. The high albedo of snow and clouds saturate the sensor when excessive light is reflected (Sabins, 1986; Rosenthal and Dozier, 1996; Pinkham, 2000). Clouds generally cover significant portions of remotely sensored images and are thus problematic, firstly because of the masking of underlying ground and secondly in the separation of snow cover from cloud (Dozier, 1991). Although snow and cloud can occasionally be distinguished through texture, this will not be possible if the sensor is saturated. In the SNOMAP equation, confusion of snow with cumulus cloud is overcome to a degree by noting that snow has a TM Band 4 reflectance greater than 11% (Hall et al., 1999; Pinkham, 2000). However, Peterson et al. (2004) note that a sufficient database of imagery now exists for a significant number of cloud free days to be easily obtainable in studies that don?t require specific dates. Chapter 4 - 69 - CHAPTER 4 METHODOLOGY 4.1 INTRODUCTION A variety of data forms are used to analyse snow cover and its geomorphological implications for the High Drakensberg within the study area. These data fall into two categories: those used in conjunction with the remotely sensed images to determine trends of snow cover, and those used to investigate geomorphological features at specific sites. High and low resolution satellite imagery, aerial photography, topographic maps, theodolite (EDM) mapping and ground-truthed geomorphological data are jointly used to test the various hypotheses. The compilation of all the data into an understandable format, together with the subsequent manipulation, analysis and display, has been made possible through the use of a Geographic Information System (GIS) database (Vitek et al., 1996). 4.2 CONSTRAINTS A variety of problems have limited the collection of more extensive data sets. These range from financial, technological, to logistical and physical safety issues. Remote sensing, specifically from satellites, is an invaluable tool in snow mapping (Dozier, 1991) due to its scope, resolution and temporal characteristics. However, the satellite platform imagery is very costly due to the nature of the space industry. These costs, although heavily subsidised by some governments, are still passed on to the end user. The scope of the research and the purchasing of high-resolution Landsat satellite data sets has thus been limited to two separate dates. To offset the inadequate temporal quality, low-resolution satellite imagery was used to provide better time series analyses of snow cover. Although freely available, the disadvantage of these low-resolution images is that poorer spatial resolution, data registration and land classification is attained. The primary shortcoming with using the low-resolution Chapter 4 - 70 - imagery is that images are single band composite images, whilst the high-resolution imagery from the satellites provide information along seven different wave lengths. IDRISI for Windows ver. 2.01 GIS software was used to process the remotely sensed data and subsequently to integrate it with the Digital Elevation Model (DEM) and ground-truthed data. The software version, released in 1998 was thus fundamentally simplistic and easy to use, but lacked the computational power of more modern GIS software. Processing of data was slow and laborious, due to the large file size of the remotely sensed and DEM data. The software was unable to process commands on many occasions, resulting in system failures. Problems were slowly overcome by using computers with more memory and by partitioning spatial data for processing into smaller segments, with the results subsequently being merged into larger data images. Ground-truthing of the main focus area has also been hampered by logistical and personal safety factors. The length of the Great Escarpment within the main focus area is approximately 170km, most of which is difficult to access. Sani Pass is the only vehicular access route into the mountains, and is only possible with a 4-wheel drive vehicle. The remainder of the escarpment region is only accessible via a minimum of a full days hike through the Little Berg. Due to the large area, comprehensive ground-truthing of the entire escarpment region and immediate interior is not practical. Specific regions for ground-truthing were identified and prioritised, based on their altitude and topographic setting. The author was physically assaulted whilst ground-truthing in the Redi Summit area along the Great Escarpment, forcing the abandonment of a particular expedition. As this was not an isolated incident (IOL, 2004; Sunday Times, 2004), further ground-truthing in the area was discontinued for personal safety reasons. To compensate for the lack of ground- truthing in these and other areas, aerial photographs were used in an attempt to map slope deposits and any other geomorphological phenomena that would not usually be identifiable on the Landsat satellite imagery. Chapter 4 - 71 - 4.3 TOPOGRAPHIC DATA Ten 1:50 000 topographical maps covering the study area (2929CC Bushman?s Nek; 2929CA Sani Pass [West]; 2929CB Sani Pass; 2929AD Giant?s Castle; 2929BC Kamberg; 2929AB Champagne Castle; 2929AA Champage Castle [West]; 2829CC Cathedral Peak; 2828DD Mont-aux-Sources; 2828DB Witsieshoek) were purchased from the Chief Directorate of Survey and Mapping, South Africa and used in combination with remotely sensed data to help with visual and GIS analysis. Contour data from the maps were digitised using a CAD program at 100m contour intervals from 2600m ASL to 2800m ASL, and at 50m contour intervals for altitudes above 2800m ASL up to 3450m ASL. The subsequent vector file was imported into the GIS software programme for further manipulation. This required splitting the vector file into 5? latitude sections for importation, as the IDRISI software was not powerful enough to import the larger file size in one process. The vector contour data of each 5? latitude file was firstly rasterised, then interpolated to form DEMs of the area. The DEMs of each 5? latitude section were subsequently refined using a 3x3 mode filter before being concatenated to generate a DEM of the entire area (Figure 4.1). This larger DEM was co-registered to all high- resolution LANDSAT TM and ETM+ images. The process made use of 28 fixed definite points, such as river junctions and road junctions. These points were easily identifiable on images and had precise co-ordinates. The images were subsequently georegistered using a bilinear interpolation resampling process with a cubic polynomial fit (Singh, 1989). Program modules in the GIS used the DEM to generate aspect (slope orientation) and slope gradient images of the study area. The topographic data were overlaid onto the snow cover images to analyse whether these factors affected snow patch distribution. By digitising the line of the escarpment top, distance operators in the GIS software could be used to determine distance of snow patches from the escarpment. Chapter 4 - 72 - Figure 4.1: Digital Elevation Model for the High Drakensberg, showing elevations between 2600m and 3450m ASL. Chapter 4 - 73 - 4.4 GEOMORPHOLOGICAL GROUND-TRUTHING Considerable ground-truthing was performed in the study region (Figure 4.2). Geomorphological features identified and mapped include thufur, stone banked lobes, large sorted patterned ground, block fields, block streams and debris ridges. These landforms were surveyed and mapped through the use of a Geographical Positioning System (GPS) device. Satellite reception on the device was dependably strong, allowing for positional accuracy within 5m on a consistent basis. Such data were imported into a GIS using their known co-ordinates as points and vectors. The translation of latitude and longitude into the GIS co-ordinate system required careful assessment in which the datum was noted. The incorporated DEM provided information on altitude, slope gradient and aspect positions for the recorded features. Spatial and temporal trends in snow cover could then be compared against the spatial distribution of geomorphological features. Figure 4.2: Topographic map indicating the area covered by ground-truthing. Chapter 4 - 74 - Debris ridges were recorded at three different locations within the study area (Figures 4.3 to 4.6). Initial investigations on the morphological and sedimentary characteristics of these deposits indicate that they are of glacial origin (Mills and Grab, 2005). Using an EDM, morphological investigations were undertaken on a debris ridge along the south-facing slope of the Tsatsa-La-Mangaung ridge, 3 km north of Sani Pass. A further two ridges in the Leqooa River area were mapped using manual survey equipment, so as to determine the geometry of the debris ridges. Morphological surveys were undertaken for the ridges and surrounding slopes, indicating significant breaks of slope, slope angles and slope directions. Lengths, heights and widths of the debris ridges were thus recorded. Six additional ridges along the Sekhokong Range (4 km southwest of Sani Pass) were mapped using a GPS. All results were incorporated into the GIS database to permit overlaying with snow- covered areas. Figure 4.3: Topographic map showing location of observed debris ridges. Chapter 4 - 75 - Figure 4.4: A photo taken looking north, showing the larger of two debris deposits on a south-facing slope of the Leqooa mountain ridge. Figure 4.5: A photo taken looking south, showing the larger of two debris deposits on a south-facing slope of the Leqooa mountain ridge (a person on the front right of the ridge provides scale). Chapter 4 - 76 - Figure 4.6: A photo taken looking south, showing shadowing created by the Leqooa debris deposit in the mid-afternoon in winter. Information recorded in the geomorphological literature of the High Drakensberg was incorporated into the study to complement the ground truthing and extend the range and resolution of geomorphological information. Where possible, particularly through the use of geomorphological maps available in the literature, such data were incorporated into the GIS. 4.5 LANDSAT SATELLITE DATA 4.5.1 Satellite imagery Landsat TM and ETM+ imagery was selected ahead of MODIS and other satellite platform sensors because of their higher resolution, good determination of snow cover in basin scale studies, as well as their suitability to mountainous environments (Dozier, 1989; Hall et al., 1998; Derksen and Le Drew, 2000; Schowengerdt, 2006) (Table 3.4). This is imperative in the High Drakensberg where the restricted and sporadic snow covers very small areas. The Landsat TM bands 1 Chapter 4 - 77 - through 5 and 7 have a resolution of 30m whilst band 6 has a resolution of 120m (Sabins, 1986; Dozier, 1991). However, given that bands 1, 2, 3 and 5 are considered the most important in snow mapping (Dozier, 1989; Hall et al., 1999; Derksen and Le Drew, 2000), they were used predominantly. Although images are generally available every 16 days at each satellite overpass, a short period in 2001 saw the LANDSAT 5 TM and LANDSAT 7 ETM+ sensors receiving data concurrently, allowing for data to be captured every 8 days. However, data availability depends on favourable reception conditions with little or no cloud or haze, thus resulting in not all satellite-overpass days presenting usable data. Two high resolution sets of TM images, obtained from the Satellite Application Centre, South Africa, with the full 7 spectral bands taken from the Landsat 5 satellite and corresponding to path 169 / row 80 of the World Wide Reference System 2 (WRS-2) were used to focus on specific components of the study. Cloudless images, with varying degrees of snow cover in the higher elevations of the Drakensberg were preferentially selected. The data fitted the Universal Transverse Mercator (UTM) map projection and the Clarke 1880 spheroid, and were subjected to Level 4 processing. As unprocessed satellite data are influenced by electronic, geometric, mechanical and radiometric distortions resulting from sensor defects, the motion of the satellite and variations in topography (Zietsman et al., 1996), radiometric and systematic geometric corrections of the original images were processed (Piesbergen et al., 1997; Schowengerdt, 2006). Data for 03 and 19 August 1990 were selected as they had clear and definite areas of scattered snow cover, providing good data to analyse snow cover aspects such as slope, aspect, altitude, snow shadowing and latitudinal difference. The image for 19 August has some scattered cloud, but a full analysis of this set indicates that the cloud is confined to areas below the escarpment and thus outside the study area. A simple masking process using the GIS effectively excluded the cloudy areas and prevents confusion with snow cover (Baral and Gupta, 1997). It was also confirmed that the snow and cloud cover had not caused detector saturation. The data set of 03 August 1990 was cloud-free. Chapter 4 - 78 - The images are highly suitable due for their close temporal relationship, with little difference in environmental setting such as land cover use, seasonal atmospheric conditions and air moisture, sun angle and sensor calibration likely to influence snow cover change data (Singh, 1989; Schowengerdt, 2006). Both images have scattered snow cover over large sections of the High Drakensberg, with the later image showing a significant decrease. Initial climatic data indicate that no snowfall occurred between the two dates and thus the difference in snow cover portrayed in the images would provide snow ablation and snowmelt data for the intermediate sixteen days. The two data sets were each taken at 07h17 in late winter. As the images were taken during the early winter mornings, which limited the angle of the sun above the horizon, the steep topographic regions cast shadows upon some south-facing areas of the images. The high cost of Landsat images with full data sets has limited the use of high- resolution images as discussed earlier. However, lower resolution composite images, freely available over the Internet and compiled from the high-resolution images, were used to analyse snow cover spatially and temporally, thus complementing the main full data set TM images. The lower resolution images, which were taken from Landsat 5 and 7 satellites using the TM and ETM+ sensors respectively, have a pixel resolution of 360m and thus do not provide the high resolution (30 m) available from the full data sets. Further, these low-resolution images are single band composite images compared to the seven data bands obtained from the high-resolution images. Thus, a decrease in spectral data, similar to resolutions obtained by the MODIS satellite, is obtained from the low-resolution images. Although accurate snow mapping and topographic analysis is therefore not possible, snow cover research on macro-spatial components can still be effective. The effect that latitude, distance from the escarpment and shadowing has on snow cover can be determined. The comparative differences between the two types of data are summarised in Table 4.1. The main benefit from low-resolution images is the high number of historical images available. A total of 41 low-resolution snow covered images, with insignificant or no cloud cover, were identified from 1989 to 2004 (Table 4.2), thus allowing for good temporal data. Unfortunately, no usable snow-covered images were identified in the 2001 timeframe that would have allowed for data at an 8-day interval. Chapter 4 - 79 - Table 4.1: Comparative assessment between high and low resolution Landsat TM and ETM+ images. High resolution images Low-resolution images Spectral range 7 bands 1 composite image comprised of 3 bands Pixel resolution 30 m 360 m Costs High Free Number of image dates 2 41 Advantages High resolution and good spectral range Low costs Disadvantages High cost of purchase Poor resolution Poor data registration Non-original data Unable to select bands preferentially Uses Detailed spatial, aspect and slope studies Generalised temporal and spatial studies Importance is placed on the temporal aspect of these low-resolution images, as a large frequency of images from multiple winter seasons is necessary to determine trends in snow cover. These data were used to examine snow distribution from different snow forming climatic events during different seasons and months. Repetitive data were also used to examine the rate of snow ablation against these criteria. Cross-checking of remotely sensed data with climatic data has shown that not all snow events are evident in the satellite imagery due to the temporal staging of satellite overpasses. Occasional snowfall events have been recorded in weather reports (SAWB, 1999-2004), yet fail to feature on satellite images as the data were presumably only captured after the snow had ablated. Chapter 4 - 80 - Table 4.2: Low-resolution Landsat images where snow cover is present. Date of image Landsat platform Cloud cover Snow cover area: scale (0 to 5) 31 July 1989 5 None 1 16 August 1989 5 Light over southern part 1 03 August 1990 5 None 2 19 August 1990 5 None 1 04 September 1990 5 Below escarpment 2 22 October 1990 5 None 4 19 June 1991 5 Light scattered 3 08 August 1992 5 Scattered 2 08 June 1993 5 None 1 24 April 1994 5 Below escarpment 3 29 July 1994 5 None 5 30 August 1994 5 None 1 30 June 1995 5 Light 2 18 September 1995 5 Below escarpment 2 20 October 1995 5 Scattered 2 31 May 1996 5 None 1 18 July 1996 5 None 5 19 August 1996 5 Below escarpment 5 24 September 1996 5 Haze 3 06 October 1996 5 Scattered 3 05 May 1997 5 Below escarpment 4 21 July 1997 5 None 4 22 August 1997 5 None 3 25 August 1998 5 None 5 12 August 1999 5 None 2 10 May 2000 5 None 2 30 June 2001 5 Light 5 09 August 2001 7 Haze 1 10 September 2001 7 Haze 1 08 May 2002 7 None 2 24 May 2002 7 None 3 09 June 2002 7 Light Scattered 2 25 June 2002 7 None 4 11 July 2002 7 None 1 27 July 2002 7 None 4 12 August 2002 7 None 2 13 September 2002 7 None 1 22 June 2004 7 None 1 08 July 2004 7 None 2 09 August 2004 7 None 1 10 September 2004 7 Below escarpment 4 Chapter 4 - 81 - 4.5.2 Snow identification in high-resolution imagery Existing snow cover algorithms from the literature were used to identify snow cover on the high-resolution images (Dozier, 1989; Hall et al., 1995b; Derksen and LeDrew, 2000; Pinkham, 2000; Schowengerdt, 2006). A smaller test area in the Giants Castle region (Figure 4.7) was used to accurately determine the best algorithms and classifications for snow covered area in the images. The results were subsequently used to determine appropriate algorithms for the larger study area. Through the use of a smaller test area, faster and more accurate processing was possible, allowing multiple tools and algorithm combinations to be tested. Detailed visual observations of satellite images could also allow cross-checking to be performed. Figure 4.7: Topographic map showing the test area for snow cover algorithms. Given that TM bands 1, 2, 3 and 5 were reported to be the most essential in snow mapping, as they best record the optical characteristics of snow (Dozier, 1989; Chapter 4 - 82 - Hall et al., 1995b; Hall et al., 1999), they were alternatively combined to produce composite images with a 2.5% linear saturation stretch in each tail. Blue?Green?Red composites comprising TM bands 1?2?3, 3?5?4 and 2?3?4 were produced. The 1?2? 3 composite was beneficial in that it produced a real-colour image that is useful for visual confirmation of select data. A smaller sample area was used to test for land-cover classifications that would most accurately represent snow cover. A mask of areas below the High Drakensberg escarpment was initially created to nullify the effect of cloud and minimise the spatial coverage and variability of ground cover (Baral and Gupta, 1997). This would provide for a more accurate classification of ground cover. Various types of classifications were subsequently performed on the area. Broad and fine unsupervised classifications, generated through clustering modules, were performed on the composite images. The number of resultant classes was noted (Table 4.3) and a reclassification was performed to group similar land cover types. In all cases, the least significant clusters (< 1% of total area) were dropped and an adaptive box filtering process was performed to eliminate remote and insignificant pixels. Table 4.3: Broad and unsupervised classification of composite images with resultant number of classes. Composite image Broad unsupervised classification (classes) Fine unsupervised classification (classes) 3 ? 4 ? 5 7 18 1 ? 2 ? 3 5 12 2 ? 3 ? 4 5 15 A supervised classification using all 7 TM bands was performed for comparison with the unsupervised classifications. Five land-cover classes were developed from a visual analysis of the image data (Table 4.4) and training sites for the classes were identified and mapped. Subsequently, a maximum likelihood Chapter 4 - 83 - classification, using probability densities to assign pixels to the trained classes, was performed. Table 4.4: Five land-cover classes developed for a supervised classification. Class Land-cover 1 Snow 2 Fire ? burnt grassland 3 Shadow 4 Low elevation grassland and other vegetation 5 High elevation grassland and other vegetation Finally, the SNOMAP and NDSI equations (Hall et al., 1995b; Hall et al., 1998; Hall et al., 1999; Derksen and Le Drew, 2000; Pinkham, 2000) were preformed and compared to previous classifications. The NDSI has proven useful for the identification of snow and ice, for the separation of snow / ice and most cumulus cloud, and for the decreased reliance on single-band, universal thresholds (Hall et al., 1998; Hall et al., 1999; Pinkham, 2000; Dong et al., 2005). The NDSI equation applied to the raw data follows as: NDSI = (TM Band 2 - TM Band 5) / (TM Band 2 + TM Band 5) In the SNOMAP equation (Hall et al., 1995b; Hall et al., 1999), pixels of NDSI values greater than 0.4 were extracted, and if the corresponding pixel contained a TM band 4 reflectance greater than 11%, it was allocated a snow-covered value. The comparative classifications of the test area were analysed, with the fine unsupervised classification of the 3?5?4 composite image the most accurate based on the comparison of each result with the true-colour image. This classification was subsequently selected for the full study area for both high-resolution data sets. 4.5.3 Application of snow algorythm to high-resolution satellite data To enable the two high-resolution data sets for 03 and 19 August 1990 to be used in conjunction with the DEM, the data were registered through the use of rubber sheet resampling, so as to ensure a good spatial fit onto the DEM. A root-mean-square (RMS) error of 38.87m for the 03 August 1990 data and a RMS error of 42.98m for Chapter 4 - 84 - the 19 August 1990 data were obtained. The satellite data sets of 03 and 19 August 1990 were subsequently overlaid onto the DEM and all data below 2600m ASL excluded from the analysis, thus improving the computer processing speed. Further, an increased range of land types and land uses below this altitude decreased accuracy through the increased number of land cover classifications (Rosenthal and Dozier, 1996). A visual analysis confirmed that areas below 2600m ASL did not have an incidence of snow. Composite images of the data sets were created. Linear stretches with 2.5% saturation at tail points were performed, with zero omitted from the stretch calculation. A histogram and visual analysis of 1?2?3 composite images versus 3?5?4 and 1?2?5 composites reconfirmed that the fine unsupervised classification of the 3? 5?4 composite image was most suitable for the determination of snow cover. After reclassing the results to produce Boolean images of snow cover for each date, a median 3x3 filter was performed on the data. 4.5.4 Analysis of snow cover in high-resolution images Following the production of snow cover images, the DEM was used to analyse the snow cover data obtained for the 03 and 19 August 1990 dates. Altitudinal classes were developed at 50m intervals to analyse snow cover at different altitudes. Aspect / slope orientation (in 15? intervals) and slope gradient (in 2? intervals) were also determined for the whole study area and overlayed with the snow cover. Aspect was further analysed within 5? slope gradient categories between 0? and 40?. A distance operator was deployed by digitising the escarpment in order to develop a map identifying snow cover distance away from the escarpment and towards the interior of Lesotho (Figure 4.8). The distance from the escarpment was divided into 1 km intervals. Results were expressed as areas of snow cover, as well as a percentage when ratioed against available land at specific altitudes, aspects, slope gradients or distance from the escarpment. Finally, a statistical correlation between increasing snow cover and distance from the escarpment was performed. The study area was segmented into latitudinal zones of 5? to determine latitudinal effect on snow cover. Altitude, aspect, slope and distance from the escarpment were analysed for each latitudinal sector. Chapter 4 - 85 - Figure 4.8: Distance operator showing distance away from the escarpment towards the Lesotho interior. Chapter 4 - 86 - 4.5.5 Snow cover change analysis from high-resolution images Numerous techniques were investigated to analyse the spatial distribution of snow through time, using both raw and modified data. As there are no standard techniques that are always appropriate for change analysis, a multitude of parallel change analysis techniques is recommended (Singh, 1989; Eastman and McKendry, 1991). Image differencing, ratioing, cross-tabulation, change vector analysis and principal components analysis are all options in examining the difference in snow cover between the 03 and 19 August 1990. However, the study is only interested in changes in snow patch distribution and not general land cover change, making many of these techniques superfluous. The classification of composite images has also proven to be very successful, and for these reasons, cross tabulation of reclassified fine unsupervised classifications of 3?5?4 composite images was chosen as the preferred change analysis technique. In this examination, thresholding, which examines the degree or variation of change, was employed. Offset and gain factors were accounted for in the processing (Singh, 1989). The resulting cross tabulation produced images where snow cover had been recorded on both dates (?residual snow?), where snow had ablated (?snowmelt?), and where subsequent new snow had fallen (?new snow?). The data and process are however not able to determine instances where new snow has covered existing snow, nor where existing snow has completely melted before the ground was once again covered by snow. ?New snow? is thus an indication of area where snow cover has increased, and not necessarily an indication of the distribution of the snowfall event. The resultant images depicting change were analysed to record differences in snow patch distribution with altitude, slope gradient and aspect (slope orientation). The Aspect of snowmelt was analysed as a percentage of total area as well as a percentage of snow cover as per 03 August. The role that distance from the escarpment played in the change of snow distribution was analysed using a distance operator. Finally, specific areas in the high Drakensberg were examined in more detail by overlaying snow cover change with topographic images to determine site-specific preferences for various snow cover changes. Chapter 4 - 87 - 4.5.6 Low-resolution satellite imagery The 41 low-resolution images (Table 4.2) were imported into the GIS and georegistered against the DEM to ensure a clean overlay. Problems were encountered, as the pre-processed low-resolution images used as primary data did not accurately meet the given geo-references. On some occasions, two separate low-resolution images had to be merged to cover the full study area. Repeated attempts at geo- registering had to be made. Finally, random sampling confirmed that an error of less than 1 pixel (360m) was obtained across all 41 images. As the images were already in colour composite form, a fine unsupervised classification of the images was performed to identify snow cover once areas below 2600m ASL had been masked out. Boolean images of snow cover were created for each date and the area of snow cover noted. To analyse the repeated occurrence of snowfall in the study area, all 41 snow cover maps were overlayed to create a snow cover repetition image. To observe the relationship between repeated snow cover and topography, the DEM and repetition image were rescaled to percentages and ratioed. A regression of repeated snow cover against altitude was obtained. The number of incidences where snow covered a pixel was examined for altitudes in 50m intervals. However, analyses of slope gradient and aspect were not viable for the lower resolution images, due to the poorer resolution of the DEM after a rescaling of resolution was performed to match that of the satellite images. Simple cross-tabulation change analysis techniques were also performed on the low-resolution satellite images (Singh, 1989). Other change analysis techniques were discarded as the images were composites and as such no longer contained raw data. Whilst the high-resolution images from 03 and 19 August 1990 produced short- term ablation information and an accurate determination of snow cover with respect to altitude, aspect, slope gradient and distance from the escarpment, the low-resolution images provided information about the temporal variability of snow cover. Snow cover patterns for individual months over multiple years were developed and analysed to determine monthly and seasonal trends. Images were firstly grouped into months from May to September, with snow occurrences in April and October being rejected Chapter 4 - 88 - due to their low recurrence (1 and 3 instances respectively) (Table 4.5). Images were also grouped into early, middle and late snow-season scenarios. All images were further analysed by examining the distribution of snow cover in respect to latitude. Table 4.5: The number of low resolution images with snow cover (1989-2004) and their classification in early, middle and late snow-seasons. Month Snow-season class Number of images Month Snow-season class Number of images January - 0 July Middle 7 February - 0 August Middle 12 March - 0 September Late 6 April Early 1 October Late 3 May Early 5 November - 0 June Early 7 December - 0 4.6 CLIMATIC DATA Climatic data and synoptic weather charts obtained from the South African Weather Bureau (SAWB) bulletins were used to compare the spatial distribution of late-lying snow patches with specific weather conditions. The effect that different types of weather patterns have in producing precipitation was examined (Preston- Whtye and Tyson, 1997), with Preston-Whyte et al.?s (1991) analysis of rain producing synoptic systems (Table 2.1) being used to identify various categories. Only synoptic types with the potential to produce precipitation in the High Drakensberg region were considered, with certain synoptic types being split into two categories (Table 4.6). Climatological records were examined for the 16 days prior to and on the day of data capture for the 2 high-resolution and 41 low-resolution data images (SAWB, 1989-2004), so as to verify the most likely cause of snowfall. This permitted the recording of weather conditions over all the days between consecutive satellite image dates (16 days apart). Daily minimum and maximum temperatures and precipitation for each event were recorded for Underberg and Bethlehem. Average temperature and Chapter 4 - 89 - total precipitation were also calculated. The Underberg and Bethlehem weather station sites were chosen as they were the closest stations to the study region that consistently produced data for the national weather bureau. The two stations are also located in different areas adjacent to the High Drakensberg escarpment. Whilst Underberg is located in the foothills to the east of the central / southern part of the High Drakensberg, Bethlehem is situated on the interior plateau, north of the Lesotho Highlands and above the main escarpment. Dates of probable snowfall were determined, thus making it possible to calculate the period of ablation. The types of climatic events leading to probable snowfalls were noted for each of the 41 low- resolution images. The mean snow distribution and seasonal snow distribution patterns associated with each type of climatic event were determined. Table 4.6: Synoptic rain-producing system types (after Preston-Whyte et al., 1991) and their reclassification for snow producing weather categories in the High Drakensberg and Lesotho Highlands. Synpotic type Potential for snow precipitation Reclassified weather category Tropical-temperature trough No. Summer rainfall. - Westerly Wave (incl. Cut-off Low) Yes. Cloud extends beyond Drakensberg. A cut-off low is the intense form of this type. Low pressure over the interior. Cut-off low Ridging High Yes. Orographic precipitation. Onshore flow East Coast Low Yes. Severe thunderstorms over interior of KZN. Coastal low High Pressure No. Stable conditions. - Easterly Flow No. Rain in northern coastal KZN. - Mid-latitude Cyclone Yes. Low temperatures. Short period rainfall in winter. Broad cold front Tropical Cyclone Unknown due to the limited occurrence of these events in the KZN area. No snow noted in climatic record. - 4.7 AERIAL PHOTOGRAPHIC IMAGERY Aerial photographs covering the escarpment region from Nhlangeni Peak (4 km east of Thabana Ntlenyana) to Giants Castle were used to help map large-scale geomorphological features. The photographs comprise 9 images that overlap by 30%, Chapter 4 - 90 - thus enabling three-dimensional analysis through the use of a stereoscope. The photographs were taken at an altitude of 7900m at 12h45 on 17 August 2000 and are displayed at an approximate scale of 1:30 000. Shadowing in the aerial photographic series is kept to a minimum as the sun is near its apex, however given that the photographs were taken during the late winter season, the sun was still fairly low on the horizon. A visual analysis of the images indicates that shadowing is limited to the south-facing slopes of the escarpment, whilst the south-facing slopes in the Lesotho highlands are shadow free. No obscuring cloud cover is evident. Snow cover is also absent in the images, which would have favoured the identification of large-scale geomorphic features. A comprehensive spatial analysis of a significant section of the High Drakensberg study area was thus possible. Observed geomorphic features were to be digitised into the GIS using their observed co-ordinates as points and vectors and analysed in relation to snow cover, but no such features were identified. To improve the geomorphological data record, real colour composite satellite images were obtained through the Internet using the Google Earth map program. Technological advances have recently allowed for the viewing of imagery of specific regions, including large parts of the High Drakensberg and Lesotho Highlands, at high-resolution. A comprehensive spatial analysis was undertaken, with numerous debris ridges and occasional block streams identified. Topographic location and the upper and lower altitudes were noted for the debris ridges. 4.8 GEOMORPHOLOGICAL PHENOMENA AND THE RELATIONSHIP TO SNOW COVER Various geomorphological features were identified through ground truthing and recorded in the GIS. Data in the available literature were also included where the exact location of the phenomena could be identified through figures and photographs in the literature (Table 4.7). Block fields, block streams, patterned ground, stone banked lobes, thufur, debris ridges and solifluction lobes were recorded. The relationship between topography and specific types of geomorphological data was Chapter 4 - 91 - analysed through the production of images superimposing the resultant data on top of the topography. In order to analyse their relationship to snow cover, images at higher resolution were also produced for specific areas with a superimposition of snow patches from 03 and 19 August 1990. Detailed maps of the hollow topography surrounding significant debris ridges of a possible glacial origin were also produced. Table 4.7: List of literature used in the identification and mapping of select periglacial and glacial geomorphological features in the High Drakensberg and Lesotho Highlands. Geomorphological feature Literature used for identification and mapping Block fields Boelhouwers, 1994; Grab, 1999a; Grab et al., 1999; Boelhouwers et al., 2002; Sumner, 2004a. Block streams Grab, 1999a; Grab et al., 1999; Boelhouwers et al., 2002; Sumner, 2004a. Patterned ground Boelhouwers, 1994; Grab, 1996b, 2002a, 2004; Grab et al., 1999; Sumner, 2004b. Solifluction lobes Grab et al., 1999; Boelhouwers et al., 2002; Mills and Grab, 2005. Thufur Grab, 1994, 1997a, 1998a, 2002b; Grab et al., 1999; Boelhouwers et al., 2002. Debris ridges Grab, 1996a; Mills and Grab, 2005. Stone-banked lobes Boelhouwers, 1994; Grab, 1999a, 2000b, 2004; Mills and Grab, 2005. 4.9 SUMMARY The various sets of data available for the study have been outlined, with certain constraints highlighted. Additional data types have been included to reduce any short falls. Various methods of image data processing have been tested and preferred algorithms selected where applicable. Ground truthing has taken place to identify and survey specific geomorphological features and their spatial relationship to snow cover. Weather records have also been noted in order to analyse the conditions during periods of snow cover and snow ablation. Chapter 5 - 92 - CHAPTER 5 RESULTS 5.1 INTRODUCTION The construction of a Digital Elevation Model (DEM) for the study area is essential for an in-depth understanding of snow cover trends in the High Drakensberg. The impact of altitude, aspect (slope orientation), slope gradient, micro and macro topography and distance away from the Great Escarpment on snow cover, was analysed once the DEM had been correctly fitted to the low and high-resolution satellite imagery. Similar data are also available for geomorphological phenomena that were identified in the study area, mapped, and incorporated into the GIS database. Weather conditions were analysed for the period leading up to snow falls, allowing for the comparison of synoptic weather types with different snowfall patterns. 5.2 SATELLITE IMAGERY DATA 5.2.1 Snow cover identification from high-resolution images The accuracy of snow cover identification on the high-resolution satellite images using a multitude of snow cover algorythms suggested that a fine unsupervised classification of a 3-5-4 composite image (Figure 5.1) provides the most accurate results within a specific test area. This preferred algorythm was thus applied to the larger research area for the 03 August 1990 data set, resulting in the production of a Boolean image showing snow cover (Figure 5.2). An in-depth visual analysis of this Boolean image compared to a 1-2-3 real-colour composite image covering the same area (Figure 5.3) confirmed its suitability. A significant factor in this consideration is observed in the use of histograms, through an analysis of pixel brightness of the 3-5-4 and 1-2-3 composites of the 19 August 1990 data (Figures 5.4 and 5.5). Although the histogram of the 1-2-3 composite image, which analyses pixel Chapter 5 - 93 - Figure 5.1: False colour 3-5-4 composite image of the high Drakensberg from 03 August 1990, with snow visible in pink. Chapter 5 - 94 - Figure 5.2: Boolean image of snow cover in the high Drakensberg from 03 August 1990, with snow cover represented in cyan. Chapter 5 - 95 - Figure 5.3: True colour 1-2-3 composite image of the high Drakensberg from 03 August 1990, with snow visible in white. Chapter 5 - 96 - Figure 5.4: Histogram of 3-5-4 composite image of 19 August 1990. Figure 5.5: Histogram of 1-2-3 composite image of 19 August 1990. Chapter 5 - 97 - brightness, indicates a better breakdown of different land cover types into clearly defined groups or modes, snow and cloud cover are grouped into the same class. This makes separation of cloud cover from snow cover difficult. In contrast, although the 3-5-4 composite image has poorer separation of land cover classes, it is preferred because snow and cloud pixels are allocated to separate modal groups. The results of the identification of snow using the preferred fine unsupervised classification of the 3-5-4 composite images and its application to both the 03 and 19 August 1990 data sets indicate that considerable snow cover occurs in selected areas (Figure 5.6). The 03 August image recorded 191.93 km2 of snow whilst the 19 August 1990 image recorded 138.93 km2. Large areas of late lying snow occur in the interior of Lesotho on both dates, particularly to the southwest of Witsieshoek and the Amphitheatre. Other significant areas of snow cover are identifiable in the Leqooa River valley region, the Thabana-Ntlenyana region and the Mafadi Summit region of the High Drakensberg. Smaller accumulations of snow are observable on some of the altitudinally lower ranges between these regions. All the areas where large accumulations of snow are recorded coincide with predominantly high altitude areas (> 3000m ASL). It is worth noting that new snow cover is present in the central to northern part (Champagne Castle to Witsieshoek) of the High Drakensberg on the 19 August 1990 (Figure 5.6). This newer snow appears limited to the escarpment region between 28?40? and 29?05?, as it is not observed towards the interior of Lesotho. A detailed visual analysis indicates that this snow is limited to within 10 km from the escarpment. It is further limited to southerly facing slopes of ridges in the Lesotho Highlands. 5.2.2 Spatial analysis of snow cover from high-resolution images The distribution of snow cover in the High Drakensberg at the time that the two data sets were captured shows that the largest area of snow in square kilometres is recorded between 3150m and 3200m ASL (Figure 5.7). Data indicate that the mean Chapter 5 - 98 - Figure 5.6: Boolean snow cover images of 03 August 1990 (left) and 19 August 1990 (right). Chapter 5 - 99 - Figure 5.7: The distribution of snow cover on 03 and 19 August 1990 with respect to altitude. Chapter 5 - 100 - altitude of snow cover (the average height of all snow covered pixels) on 03 August 1990 was 3143m ASL whilst the mean altitude on 19 August 1990 was 3144m ASL. The quartile distribution of snow cover on either date showed little difference other than snow being recorded at a lower minimum on 19 August (Figure 5.8). More snow cover is measured on the 03 August at all altitudes, as opposed to the 19 August. When snow cover is expressed as a percentage of snow covered area within a particular altitudinal belt, a better understanding of snow cover near the mountain tops is obtained. Altitudes from 3150m to 3200m and from 3200m to 3250m record the highest percentages of snow for all altitudinal intervals, with 23% and 16% for both intervals on the 03 August and 19 August respectively. Percentage snow cover for altitudes above 3250m decreases with increasing altitude. Near the highest elevations above 3400m ASL, snow cover rapidly decreased to near zero. Figure 5.8: The minimum, lower quartile, median, upper quartile and maximum altitudes of snow cover on 03 and 19 August 1990. The spatial analysis of snow cover in relation to aspect (slope orientation) indicates that there is a strong preference to south-facing slopes. The macro-scale topographic aspect of the study region was analysed and indicates that the highest number of slopes were southeast and northeast facing. Greatest snow cover, when Chapter 5 - 101 - expressed as a percentage of the total area of each aspect, is favoured on slopes orientated between 127? and 203? (Figure 5.9). Only 26% of snow cover on 03 August and 16% of snow cover on 19 August is located on north-facing slopes (270? - 90?). There is also a general trend for greater snow cover on east-facing rather than west-facing aspects, with 63% of snow cover on 03 August and 69% of snow cover on 19 August on east-facing slopes (0? to 180?). Snow distribution in relation to slope gradient indicates that snow cover is predominantly confined to slopes between 2? and 36? (Figure 5.10). The mean slope gradient for 03 August 1990 is 16.5?, whilst the mean for 19 August 1990 is a steeper 18.3?. Snow cover declines on slopes steeper than 14? until no snow is observable on slopes steeper than 62?. An exception occurs on slopes between 78? and 86?, where snow covers up to 6% of the area available at such gradients. Snow cover for 03 August (expressed as a percentage of total area at that gradient) is considerably greater on slope gradients between 0? and 28?, than on the 19 August. This shows a trend for the earlier snow data to cover a constant 10% of gently sloped surface areas (2? - 12?) whilst the latter snow data shows a decrease in snow cover with decreased gradient. When aspect of the snow cover are analysed in 5? gradient intervals for each date, a few additional trends are noticeable (Figure 5.11). Snow cover is shown to predominate on north-facing slopes on 03 August on gradients from 0? to 15?. This snow cover reaches its largest extent on 5? to 10? slopes, where an average of 34% of the slope is covered between 307.5? and 53.5? bearing, reaching a maximum of 40% covering the 7.5? to 22.5? bearing. By the 19 August, a lesser area of snow cover favours south-facing slopes between 5? and 15? gradients, with a north-facing trend still predominant between 0? and 5? gradients. On slopes steeper than 15?, south- facing snow cover predominates on both the 03 and 19 August. The effect that the escarpment has on snow cover was analysed using a distance operator. Distance away from the escarpment towards the Lesotho interior is overlaid with snow cover for the 03 and 19 August 1990 images (Figure 5.12). The data were quantified and expressed as a percentage of total area for each kilometre distance interval (Figure 5.13). The data from 03 August 1990 show a gradual increase of snow cover from the escarpment to 21 km into Lesotho, with a mode Chapter 5 - 102 - Figure 5.9: The distribution of snow cover on 03 and 19 August 1990 in relation to aspect. Chapter 5 - 103 - 0 2 4 6 8 10 12 14 0? 10? 20? 30? 40? 50? 60? 70? 80? Slope gradient S n o w c o v e r ( % o f a r e a a t s p e c i f i c g r a d i e n t ) 03 August 1990 (% of area at specific gradient) 19 August 1990 (% of area at specific gradient) Figure 5.10: The distribution of snow cover on 03 and 19 August 1990 with respect to slope gradient. Chapter 5 - 104 - Figure 5.11: The aspect of snow cover on 03 and 19 August 1990 at different slope gradients. Chapter 5 - 105 - Figure 5.12: The distribution of snow cover on 03 August (left) and 19 August 1990 (right) with respect to distance away from the escarpment and towards the Lesotho interior. Chapter 5 - 106 - 0 5 10 15 20 25 30 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1 9 2 0 2 1 2 2 2 3 2 4 2 5 Distance away from the escarpment (km) S n o w c o v e r ( % o f a r e a a t s p e c i f i c d i s t a n c e ) 03 August 1990 (% of area at specific distance) 19 August 1990 (% of area at specific distance) Figure 5.13: The distribution of snow cover with respect to distance away from the escarpment towards the Lesotho interior. Chapter 5 - 107 - recorded at 16 km as well. Further away from the escarpment, snow cover data are erratic. By comparison, snow cover on 19 August occupies a marginally greater area near the escarpment (between 0 and 4 km into Lesotho), but thereafter remains constant with increasing distance from the escarpment. A sudden drop in snow cover between 21 and 23 km away from the escarpment interrupts this trend. Statistical correlations between snow cover as a percentage of area against distance away from the escarpment produced a correlation coefficient of 0.63 for the 03 August and 0.04 for 19 August. There was thus a fairly strong correlation between distance from the escarpment and higher snow cover on 03 August, whilst there was virtually no correlation on 19 August. When analysed in 5? latitudinal sectors, it becomes evident that the distribution of snow cover is confined to specific regions of the High Drakensberg for the 03 and 19 August 1990 dates (Table 5.1 and Figure 5.14). An analysis of snow cover in areas over 3000m ASL shows little difference to the analysis of snow cover at all altitudes (Figure 5.15 vs. Figure 5.14). Snow cover occupies a greater percentage of area when only altitudes greater than 3000m ASL are analysed. The greatest snow cover is in the 28?45? S to 29?00? S region, which includes areas from the Amphitheatre through to midway between Cathedral Peak and Champagne Castle (Figure 5.1). Other areas of notable snow cover are located between 29?10? S and 29?15? S (Mafadi Summit area), 29?30? S and 29?35? S (Thabana-Ntlenyana area), as well as a larger area at 29?40? S and 29?45? S (Leqooa River valley area). A Snow cover change comparison between 03 and 19 August 1990 indicates an increase in snow between 28?40? S and 28?50? S and less significantly, between 29?00? S and 29?10? S (Figure 5.16). All other latitudes experience a decrease in snow cover. Table 5.1: Snow-covered area on 03 August 1990. Latitude Snow-covered area 28?40? S ? 29?00? S 17 % 29?00? S ? 29?20? S 2 % 29?20? S ? 29?50? S 6 % 29?50? S ? 29?55? S 0% Total 8% Chapter 5 - 108 - 0 5 10 15 20 25 30 28?40' - 28?45' 28?45' - 28?50' 28?50' - 28?55' 28?55' - 29?00' 29?00' - 29?05' 29?05' - 29?10' 29?10' - 29?15' 29?15' - 29?20' 29?20' - 29?25' 29?25' - 29?30' 29?30' - 29?35' 29?35' - 29?40' 29?40' - 29?45' 29?45' - 29?50' 29?50' - 29?55' Latitude (? S) S n o w c o v e r ( % o f a r e a a t s p e c i f i c l a t i t u d e ) 03 August 1990 (% of area at specific latitude) 19 August 1990 (% of area at specific latitude) Figure 5.14: Snow covered area with respect to latitude. Chapter 5 - 109 - Figure 5.15: The area of snow cover > 3000m ASL with respect to latitude. Chapter 5 - 110 - -10 -8 -6 -4 -2 0 2 4 6 8 28?40' - 28?45' 28?45' - 28?50' 28?50' - 28?55' 28?55' - 29?00' 29?00' - 29?05' 29?05' - 29?10' 29?10' - 29?15' 29?15' - 29?20' 29?20' - 29?25' 29?25' - 29?30' 29?30' - 29?35' 29?35' - 29?40' 29?40' - 29?45' 29?45' - 29?50' 29?50' - 29?55' Latitude (? S) S n o w c o v e r c h a n g e ( % o f a r e a a t s p e c i f i c l a t i t u d e ) Change in snow cover between 03 and 19 August 1990 (% of area at specific latitude) D e c r e a s e i n s n o w c o v e r I n c r e a s e i n s n o w c o v e r Figure 5.16: Change in snow covered area between 03 and 19 August 1990. Chapter 5 - 111 - The quartile distribution of snow cover in each 5? latitudinal sector for each date was also analysed (Figure 5.17). It is noted that the correlation between average altitude of snow cover and the average altitude of the study area at that latitude is fairly strong (r=0.67) on 03 August and (r=0.67) on 19 August 1990. The comparison of the 03 and 19 August shows that snow cover on 19 August occurs on slightly higher altitudes to the south of 29?00? S, than on the earlier date. To the north of this latitude, snow cover has a wider range of altitudes than the earlier date. An analysis of mean aspect per latitudinal sector does not show any clear trend with respect to aspect or change in aspect between the two dates other than a preference for slightly more easterly-facing slopes north of 29?05? S on the 19 August compared to 03 August (Figure 5.18). By comparison, the analysis of mean slope gradient shows distinct results (Figure 5.19). Snow cover in the northern sections (28?40?S to 29?00?S) is found on gentler slope gradients than snow located in the central and southern sections. Analysis of snow distribution between the two dates also shows that snow cover is generally confined to steeper slopes by the 19 August 1990. When the mean distance of snow cover away from the escarpment is calculated, significant variations are noted at different latitudes (Figure 5.20). This is due to geometric distortions resulting from the uneven western boundary of the multi-rectangular study area. A comparison of the mean distance of snow cover from the escarpment on 03 and 19 August dates was more informative, showing that there is an escarpment-wards shift in the mean snow cover for almost all latitudes between the two dates, but particularly at latitudes north of 29?00? S (Figure 5.20). Three-dimensional surface graphs for the 03 and 19 August both show that snow cover as a percentage of total area per latitude is more significant in the northern areas, particularly from 28?45? to 28?50? (Figures 5.21 and 5.22). 5.2.3 Snow cover change analysis from high-resolution images The cross-tabulation of the high-resolution 3-5-4 composite images between the 03 and 19 August 1990 is used to determine change in snow cover during this period. The cross-tabulation allows for an image to be developed that indicates where snow cover has disappeared between the two dates (?snowmelt?), where previously snow-free areas are now covered by snow (?new snow?) and where snow cover has remained constant (?residual snow?) (Figure 5.23 and Table 5.2). Chapter 5 - 112 - Figure 5.17: The minimum, lower quartile, median, upper quartile and maximum altitudes of snow cover in 5? latitudinal sectors on 03 and 19 August 1990. Chapter 5 - 113 - Figure 5.18: The mean aspect of snow cover for 5? latitudinal sectors. Chapter 5 - 114 - Figure 5.19: The minimum, lower quartile, median, upper quartile and maximum slope gradient of snow cover in 5? latitudinal sectors on 03 and 19 August 1990. Chapter 5 - 115 - Figure 5.20: The mean distance of snow cover westwards from the escarpment in 5? latitudinal sectors. Chapter 5 - 116 - Figure 5.21: Surface graph of snow cover in relation to distance from the escarpment for 03 August 1990. 012 345 678 9101112 13141516 17181920 21222324 28?40' 28?50' 29?00' 29?10' 29?20' 29?30' 29?40' 29?50' 0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00 100.00 S n o w c o v e r ( % o f a r e a a t s p e c i f i c l a t i t u d e ) Distance from the Escarpment (km) L a t i t u d e ( ? S ) 0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100 Chapter 5 - 117 - Figure 5.22: Surface graph of snow cover in relation to distance from the escarpment for 19 August 1990. 012 345 678 9101112 13141516 17181920 21222324 28?40' 28?50' 29?00' 29?10' 29?20' 29?30' 29?40' 29?50' 0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00 100.00 S n o w c o v e r ( % o f a r e a a t s p e c i f i c l a t i t u d e ) Distance from the Escarpment (km) L a t i t u d e ( ? S ) 0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100 Chapter 5 - 118 - Figure 5.23: Cross-tabulation of 3?5?4 composite images between 03 and 19 August 1990 showing snow cover change. Chapter 5 - 119 - Table 5.2: Cross-tabulation of 3?5?4 composite images between 03 and 19 August 1990 showing snow cover change. Area with no snow 19 Aug 91.20 km2 - Area with snow 19 Aug 100.73 km2 38.20 km2 Area with snow 03 Aug Area with no snow 03 Aug The cross tabulation confirms that significant new snowfall had occurred in the central to northern part of the High Drakensberg between Witsieshoek and Champagne Castle. This occurred in an area where minimal late lying snow was recorded on 03 August 1990, thus confirming a new snowfall event. It is noted however, that the data only indicates the increased area of snow cover and not the full extend of the snow fall. An area of 38.2km2 was found to have snow cover on 19 August in areas where there was no snow cover on 03 August (Table 5.2). Ablation rates of snow cover are similar across the entire study area, whilst residual snow cover appears to be concentrated on the southern slopes at higher elevations in the Leqooa Valley, at Thabana-Ntlenyana and adjacent summits to the west, the Mafadi Summit area, and the high interior of Lesotho to the southwest of the Amphitheatre (Figure 5.23). The occurrence of new snow cover, residual snow cover and area of snowmelt is examined with respect to altitude, aspect and slope gradient, so as to determine further trends. Results of snow-covered area are expressed in both km2 and as a percentage of the whole study area (Figure 5.24 and Figure 5.25). Altitudinal analyses of residual snow, new snow and areas of snow melt between 03 and 19 August 1990 show that snow cover and ablation occurs predominantly above 3000m ASL (Figure 5.25). New snow has the lowest mean altitude (average altitude of all snow cover pixels) at 3105m, ablated snow has a mean of 3126m ASL whilst residual snow cover has the highest at 3159m ASL. A quartile distribution shows that new snow cover has the lowest minimum, lower quartile, median and upper quartile of the three categories (Figure 5.26). Analysis of snow cover, when expressed as a percentage of the total area at a given altitude, shows that new snow accumulation and residual snow increase with altitude to 3250m (Figure 5.25). At altitudes above this value, there is a considerable decrease in the relative volume of new and residual Chapter 5 - 120 - Figure 5.24: Altitudinal distribution of residual snow, new snow and snowmelt area from 03 to 19 August 1990 in km2. Chapter 5 - 121 - Figure 5.25: Altitudinal distribution of residual snow, new snow and snowmelt area from 03 to 19 August 1990 as a percentage of the total area. Chapter 5 - 122 - snow. This decrease continues to 3400m ASL for new snow, before there is a sudden increase at higher altitudes. No similar increase is noted for residual snow. For the period between the two images, snowmelt occurred most rapidly around 3250m ASL and again above 3450m ASL, the highest elevation class. The percentage area of snowmelt of the 03 August snow cover shows that the highest level of snowmelt (20%) occurs between 3100m and 3150m ASL (Figure 5.27). Figure 5.26: The minimum, lower quartile, median, upper quartile and maximum altitudes of snow cover on 19 August 1990, expressed as a difference from 03 August 1990. With regard to aspect (slope orientation), new, residual and ablated snow is most pronounced on the south and southeast-facing slopes, particularly at orientations between 127.5? and 202.5? (Figure 5.28). Minimal new snow or residual snow is observed on north-facing slopes, whilst snowmelt is more evenly distributed. Chapter 5 - 123 - Figure 5.27: Altitudinal distribution of snowmelt between 03 and 19 August 1990, as a percentage of snow cover on 03 August 1990. Chapter 5 - 124 - Figure 5.28: The aspect of new snow, residual snow and snowmelt (as a % of total area in 15? slope orientation classes) between 03 and 19 August 1990. However, a larger proportion of snowmelt is observed in the intermediate days on east-facing slopes than west-facing slopes. New snow is more strongly associated with the south-facing aspect than residual snow and snowmelt, with east and west- facing aspects being almost absent of new snow cover. When snowmelt is analysed as a percentage of snow cover on 03 August, rather than the total area, snowmelt indicates a strong relationship with north and northwest-facing slopes, with more than 76% of snow melting on aspects between 262.5? and 52.5?, whilst less than 40% of snow cover melts between 112.5? and 202.5? (Figure 5.29). Chapter 5 - 125 - Figure 5.29: The aspect of snowmelt (as a % of snow covered area on 03 August 1990 in 15? slope orientation classes) between 03 and 19 August 1990. An analysis of slope gradient with regards to new snow cover, late lying snow and snow melt areas on all aspects and at all altitudes produces varying results (Figure 5.30). Snowmelt occurs on slopes with low gradients (mean of 15.4?), whilst residual snow favours slightly steeper slopes (mean of 17.4?). New snowfall occupies Chapter 5 - 126 - the steepest gradient, with a mean of 20.8?. Very little snow (<1.5%) is recorded on near-flat slopes (0? to 2?). When analysed as a percentage of the area at that specific gradient, snowmelt occurs primarily on flatter slopes (0? to 14?). Residual snow occupies slightly steeper slopes (4? to 34?), whilst new snow was recorded evenly across most slope gradients from 2? to 50?. Residual snow and snowmelt also occur on very steep slopes of 78? to 86?. Snow cover change with distance into Lesotho from the Drakensberg escarpment was assessed for all latitudes across the study area (Figure 5.31). New snow predominates within 3 km from the escarpment. In comparison, residual snow and areas of snowmelt are mostly absent within the first kilometre of the escarpment, instead predominating only beyond 7 km from the escarpment. Thereafter, the percentage area of residual snow cover remains fairly constant whilst snowmelt continues to increase with increasing distance. The mean distance (average displacement of all pixels away from the escarpment) of residual snow is 10.0 km away from the escarpment, whilst that for snowmelt is further towards the Lesotho interior at 11.8 km. When snow cover change is categorised into 5? latitudinal components (Figure 5.32), it is evident that new snow is recorded primarily in the far northern regions (28?40? S to 28?50? S) and gradually decreases southwards until 29?15? S, where after it is negligible. Residual snow and snow melt occur primarily at specific latitudes, with the largest snow coverage between 28?45? S and 29?00? S (from the Amphitheatre to halfway between Cathedral Peak and Champagne Castle). Other areas for residual snow and snowmelt are located between 29?10? S and 29?15? S (Mafadi Summit area), 29?30?S and 29?35? S (Thabana-Ntlenyana and immediate areas to the south) and from 29?40? S to 29?45? S (Leqooa River area). A higher percentage of residual snow cover than snowmelt area is found in the Mafadi Summit region, whilst a higher percentage of snowmelt than residual snow area is recorded in the Thabana-Ntlenyana region. In order to determine the effect of topography on snow cover in the High Drakensberg, snow cover change results were overlaid onto a topographic Chapter 5 - 127 - Figure 5.30: The distribution of new snow, residual snow and snowmelt between 03 and 19 August 1990 with respect to slope gradient. Chapter 5 - 128 - Figure 5.31: Distribution of residual snow, new snow and snowmelt area between 03 and 19 August 1990 with respect to distance from the escarpment towards the interior of Lesotho. Chapter 5 - 129 - Figure 5.32: Distribution of residual snow, new snow and snowmelt area between 03 and 19 August 1990 with respect to latitude. Chapter 5 - 130 - representation (Figures 5.33 and 5.34). Four smaller images of specific areas reflecting snow cover change over topography are focused on. These include images from the far northwestern part of the study area (Figure 5.35), the Amphitheatre area (Figure 5.36), the Sani Pass region (Figure 5.37) and the Leqooa River valley region (Figure 5.38). In the northwestern Drakensberg image (Figure 5.35), residual snow cover favours south-facing slopes of mountain ridges with an east-west orientation. Snowmelt has occurred on mountain tops, north-facing slopes and valleys between the ranges. Areas near the escarpment have an absence of both residual snow cover and snowmelt. New snow cover is located on south-facing slopes of the ridges near the escarpment (Figure 5.36). Further into the Lesotho interior, the new snow has expanded the size of residual snow patches, with ?halos? surrounding such snow patches visible in Figure 5.36. Such increases in snow patch size show a tendency for snow increases on the flatter gradients near mountain tops. In the Sani Pass area, residual snow cover and snowmelt is found in south-facing, highly shadowed hollows of the east-west mountain ridges, particularly the Sekhokong and Tsatsa-La- Mangaung ridges (Figure 5.37). In these hollows, snow cover appears to favour southeast-facing slopes. The Leqooa River valley region shows large areas of snow cover and snowmelt on south-facing slopes of the ridges. The highest levels of cover appear on the Leqooa range immediately north of the Leqooa River, which is the highest range in the southern High Drakensberg region. Residual snow cover appears to favour well-shaded hollows, with snowmelt occurring particularly on the lower edge, and to a lesser extent on the upper edge of snow patches. 5.2.4 Snow cover analysis of low-resolution images Boolean images of snow cover were produced for the 41 low-resolution images (Figure 5.39). Area of snow cover was measured for each image and notes made for images where poor snow identification had taken place pdue to sensor saturation resulting from partial cloud cover (Table 5.3). It is noted that the snow- covered area for the 03 and 19 August 1990 images is 155.10 km2 and 50.88 km2 respectively. The values obtained from the high-resolution images for the same date were 192.29 km2 and 139.19 km2 (Table 5.2), a 19% and 63% reduction in recorded Chapter 5 - 131 - Figure 5.33: Snow cover changes between 03 and 19 August 1990 with respect to topography, showing residual snow and snow melt. Chapter 5 - 132 - Figure 5.34: Snow cover changes between 03 and 19 August 1990 with respect to topography, showing new snow. Chapter 5 - 133 - Figure 5.35: The north-western part of the study area near the Amphitheatre showing residual snow and snowmelt in relation to topography. Chapter 5 - 134 - Figure 5.36: The Amphitheatre area showing new snow in relation to topography. Chapter 5 - 135 - Figure 5.37: The Sani Pass area showing residual snow and snowmelt in relation to topography. Chapter 5 - 136 - Figure 5.38: The Leqooa area showing residual snow and snowmelt in relation to topography. Chapter 5 - 137 - (1) 31 July 1989 (2) 16 August 1989 (3) 03 August 1990 (4) 19 August 1990 (5) 04 September 1990 (6) 22 October 1990 (7) 19 June 1991 (8) 08 August 1992 (9) 08 June 1993 Figure 5.39 (1-9): Boolean snow cover images for 41 dates where low-resolution images were available. Chapter 5 - 138 - (10) 24 April 1994 (11) 29 July 1994 (12) 30 August 1994 (13) 30 June 1995 (14) 18 September 1995 (15) 20 October 1995 (16) 31 May 1996 (17) 18 July 1996 (18) 19 August 1996 Figure 5.39 (10-18): Boolean snow cover images for 41 dates where low-resolution images were available. Chapter 5 - 139 - (19) 20 September 1996 (20) 06 October 1996 (21) 05 May 1997 (22) 21 July 1997 (23) 22 August 1997 (24) 25 August 1998 (25) 12 August 1999 (26) 10 May 2000 (27) 30 June 2001 Figure 5.39 (19-27): Boolean snow cover images for 41 dates where low-resolution images were available. Chapter 5 - 140 - (28) 09 August 2001 (29) 10 September 2001 (30) 08 May 2002 (31) 24 May 2002 (32) 09 June 2002 (33) 25 June 2002 (34) 11 July 2002 (35) 27 July 2002 (36) 12 August 2002 Figure 5.39 (28-36): Boolean snow cover images for 41 dates where low-resolution images were available. Chapter 5 - 141 - (37) 13 September 2002 (38) 22 June 2004 (39) 08 July 2004 (40) 09 August 2004 (41) 10 September 2004 Figure 5.39 (37-41): Boolean snow cover images for 41 dates where low-resolution images were available. Chapter 5 - 142 - Table 5.3: Snow covered area (km2) derived from low-resolution images. No. Date of image Landsat platform Snow covered area (km2) Comments 1 31 July 1989 5 45.32 2 16 August 1989 5 32.48 3 03 August 1990 5 155.10 High-resolution = 192.29 km2 4 19 August 1990 5 50.88 High-resolution = 139.19 km2 5 04 September 1990 5 41.86 6 22 October 1990 5 477.90 7 19 June 1991 5 409.73 8 08 August 1992 5 242.78 9 08 June 1993 5 7.90 10 24 April 1994 5 97.56 11 29 July 1994 5 1908.76 12 30 August 1994 5 24.94 13 30 June 1995 5 75.70 14 18 September 1995 5 21.49 15 20 October 1995 5 99.16 16 31 May 1996 5 9.26 17 18 July 1996 5 2124.74 18 19 August 1996 5 1478.16 19 20 September 1996 5 81.63 20 06 October 1996 5 161.28 21 05 May 1997 5 666.34 22 21 July 1997 5 946.05 23 22 August 1997 5 408.50 24 25 August 1998 5 1713.77 25 12 August 1999 5 9.76 26 10 May 2000 5 205.24 Sensor saturation (cloud) 27 30 June 2001 5 1990.14 28 09 August 2001 7 6.79 29 10 September 2001 7 14.08 Sensor saturation (cloud) 30 08 May 2002 7 476.91 31 24 May 2002 7 802.43 32 09 June 2002 7 367.63 33 25 June 2002 7 1200.06 34 11 July 2002 7 357.75 35 27 July 2002 7 1663.64 36 12 August 2002 7 448.88 37 13 September 2002 7 24.32 38 22 June 2004 7 7.16 39 08 July 2004 7 302.79 40 09 August 2004 7 17.17 41 10 September 2004 7 744.64 Sensor saturation (cloud) Chapter 5 - 143 - snow cover for each date. The higher error level on the 19 August date is due to the combined effect of a smaller snow-covered area on lower resolution imagery allowing for increased misclassification of land cover. The combination of images, through an overlaying process, allowed for the production of an image that spatially identified the repeated occurrence of snow in the study area (Figure 5.40). The highest value for any pixel localities were 28 days of snow cover, generally on south-facing slopes of the more prominent mountain ridges in the Lesotho Highlands. As the repetition image appeared to correspond well with the DEM image (Figure 4.3), a ratioing of the two images was performed to analyse their relationship (Figure 5.41). A quartile distribution of the number of incidences of snow cover at 50m altitudinal intervals shows an increase in the number of snow incidences with increased altitude up to 3250m ASL. Above this altitude, the number of incidences is constant (Figure 5.42). A statistical correlation was performed between the altitude of snow-covered pixels and the number of occurrences that each pixel was covered in snow between 1989 and 2004. The correlation coefficient of 0.34 indicated a proportional relationship, where increased height led to an increase in the number of dates where snow cover was recorded (Figure 5.43). Repeated snow cover was also analysed in relation to the month and season in which the satellite image was taken (Figures 5.44 and 5.45). In the monthly analysis, only the months of May to September were included due to the low incidence of snow cover for other months. June, July and August recorded the majority of snow cover, with May and September showing an absence of snow cover outside the major elevation regions, particularly at the lower altitudes (2600m ? 2800m ASL). July has the most extensive snow cover, whilst September data showed a lack of notable snow cover in areas to the north of Mafadi Summit (Figure 5.44). The rough grouping of snow cover images into an early snow-season (April, May and June), mid snow- season (July and August) and late snow-season (September and October) showed little difference in the distribution of snow in the early and mid seasons (Figure 5.45). During the early season, a lower number of snowfalls occur to lower altitudes (<2800m ASL). During the late snow season a significant absence of snow cover is recorded in areas to the north of Champagne Castle, particularly the northwestern highlands, where snow was recorded on only one occasion. Chapter 5 - 144 - Figure 5.40: The recurrence of snow cover over 41 images dating between 1989 and 2004. Chapter 5 - 145 - Figure 5.41: The regression of repeated snow cover against altitude. Chapter 5 - 146 - Figure 5.42: The minimum, lower quartile, median, upper quartile and maximum number of incidences where a pixel was covered by snow across 41 images (1989-2004) in 50m altitudinal classes. Chapter 5 - 147 - Figure 5.43: The correlation between altitude and the number of snow covered images between 1989 and 2004 occurring in each pixel. (1) May (n=5) (2) June (n=7) (3) July (n=7) Figure 5.44 (1-3): The occurrence of snow per month based on data from 1989 to 2004. Other months were excluded from the analysis due to the low frequency of snow cover depicted on images. Altitude (m ASL) N u m be r o f s n o w co v er ed im a ge s (0- 41 ) Y = -22.646441 + 0.010426 R = 0.3462 Chapter 5 - 148 - (4) August (n=12) (5) September (n=6) Figure 5.44 (4-5): The occurrence of snow per month based on data from 1989 to 2004. Other months were excluded from the analysis due to the low frequency of snow cover depicted on images. (1) Early snow-season (2) Mid snow-season (3) Late snow-season (April, May, June) (July, August) (September, October) Figure 5.45: The occurrence of snow per season, based on data from 1989 to 2004. Chapter 5 - 149 - 5.3 CLIMATE DATA Climatological data for the period leading up to and between the dates of the two high-resolution images of 03 and 19 August 1990 were examined to determine likely snowfall dates and the climatic conditions contributing to snowmelt (Table 5.4). No cold front passed the region following 17 July 1990, until the satellite data were captured on 03 August. However, a coastal low associated with a mild cold front bringing onshore flow that is sometimes conducive to producing snowfall is noted for 29 July (Table 5.4). On 04 August, a low-pressure trough over the interior of South Africa brought cool conditions. Conditions thereafter remained mild until a snow bearing cold front arrived on 11 August. Temperatures were significantly depressed on 12 and 13 August, with temperatures only reaching 11.5?C and 14.5?C compared with 24.5?C on the previous two days. Such temperature drops follow immediately after the passage of cold fronts (Grab and Simpson, 2000). Milder conditions predominated thereafter, with temperatures generally between 15?C and 19.5?C, until the next satellite data capture on 19 August. Temperature and precipitation data were also graphed for the 16-day periods prior to each low-resolution data image (Appendix 1). Average temperature and total precipitation were also calculated (Table 5.5). Synoptic charts (SAWB, 1989-2004) were examined to determine probable snowfall events during the period of analysis. The dates of these events, as well as the climatic anomaly responsible for the snowfall were noted and the number of days during which snow ablated prior to the satellite overpass was determined (Table 5.5). Snow cover images were linked to snow forming weather conditions and analysed (SAWB, 1989-2004) (Figure 5.46). ?Broad cold fronts? (1), where the frontal system of a mid-latitude cyclone passes across all, or the greater part of southern Africa, account for the majority of snowfall events (66%). Snow cover is usually restricted to regions with large areas at higher elevations (>3000m ASL), but extensive snowfalls down to 2600m are also common. There is a preference for snowfall to occur only near the main escarpment during the passage of weaker frontal Chapter 5 - 150 - Table 5.4: Weather conditions between 28 July and 19 August 1990 for Underberg, Kwa-Zulu Natal (KZN) and Bethlehem, Free State (FS). Underberg Bethlehem Date (1990) High (?C) Low (?C) Precip (mm) High (?C) Low (?C) Precip (mm) Comments 28 July - -3.0 0 17.0 1.5 0.9 No cold front since 17 July 29 July 17.5 - 0 9.1 5.8 4.5 Cold front 30 July 19.5 -2.5 0 13.5 -2.3 0 31 July 20.0 -4.0 0 19.2 0.1 0 01 August 22.5 -4.0 0 19.2 -3.1 0 02 August 17.5 -1.0 0 - -2.5 0 03 August - 3.0 0 19.5 - 0 04 August 8.0 - 0.5 17.8 1.8 12.6 Low pressure trough over FS 05 August 12.5 1.0 0 14.2 7.2 0.3 06 August 17.6 -2.5 0 16.6 5.4 0 07 August 20.0 -2.4 0.4 17.5 2.8 0 Thundershowers over KZN 08 August 22.6 -1.5 0 19.7 0.7 0 09 August 23.7 -1.5 0 22.0 0.9 0 10 August 24.5 2.0 0 20.5 -0.1 0 11 August 24.5 2.2 4.0 20.3 5.9 0 12 August 11.5 1.5 0 - 4.9 0 Cold front over KZN ? snow 13 August 14.5 -2.0 0 14.7 - 0 14 August 24.4 -9.0 0 17.7 -4.5 0 15 August 19.5 -5.0 0 17.2 -3.7 0 16 August 18.6 -2.0 0 17.0 -2.5 0 17 August 15.5 1.5 0 16.3 2.0 0 18 August 19.0 -4.0 0 18.7 -1.9 0 19 August 15.0 -2.5 5.0 16.6 -2.3 0 Chapter 5 - 151 - Table 5.5: Climate data for the 16 days prior to each satellite data capture. Ave Temperature for last 16 days (?C) Total precipit. for previous 16 days (mm) Underberg Bethlehem No Date of image Date of latest probable snowfall Snowfall description Abla tion days Max Min Max Min Under berg Bethle hem 1 31 July 1989 18-19 Jul Broad cold front 11 17.6 -7.4 14.6 -3.2 1.4 0.1 2 16 August 1989 2-3 Aug Broad cold front 12 22.0 -1.9 20.6 -0.2 0.0 5.9 3 03 August 1990 28-29 Jul Low pressure over interior 4 19.5 -2.7 17.4 -0.7 0.0 5.4 4 19 August 1990 11-12 Aug Onshore flow 6 18.2 -1.6 16.7 1.1 4.9 12.9 5 04 September 1990 29-30 Aug Low pressure over interior 4 17.2 -0.3 19.2 1.6 46.5 0.4 6 22 October 1990 18-19 Oct Two broad cold fronts 2 17.6 4.6 21.1 6.8 32.0 56.6 7 19 June 1991 18-19 Jun Broad cold front 0 15.7 -3.1 15.1 -1.5 16.1 21.2 8 08 August 1992 3 Aug Low pressure over interior 4 18.2 -2.4 15.8 -0.2 0.0 34.8 9 08 June 1993 3 Jun Broad cold front 4 18.5 -1.8 17.6 -1.0 0.0 0.0 10 24 April 1994 23-24 Apr Broad cold front 0 19.8 4.6 22.0 5.1 2.0 36.9 11 29 July 1994 25-26 Jul Cut-off low and cold front 2 17.7 -3.9 16.6 -4.1 33.5 1.2 12 30 August 1994 19-20 Aug Broad cold front 9 20.7 -0.7 18.8 -0.2 6.5 0.0 13 30 June 1995 16-17 Jun Coastal front and Low pressure over interior 12 16.8 -3.7 6.5 17.7 -4.0 3.6 14 18 September 1995 17 Sep Broad cold front 0 21.9 2.3 22.8 1.8 13.5 1.9 15 20 October 1995 19 Oct Coastal front and Low pressure over the interior 0 21.8 9.4 23.9 9.2 52.9 39.5 16 31 May 1996 24-25 May Broad cold front 5 17.5 -0.7 16.6 2.4 5.5 15.2 17 18 July 1996 16 Jul Broad cold front 1 12.9 -5.7 9.1 -4.9 0.0 52.7 18 19 August 1996 17-18 Aug Broad cold front 0 15.4 -1.7 14.8 -0.3 5.6 13.6 19 20 September 1996 17-18 Sep Two broad cold fronts 1 24.1 5.2 22.1 2.2 0.7 0.4 20 06 October 1996 1-2 Oct Coastal cold front 3 23.0 7.2 23.9 6.2 18.9 16.5 21 05 May 1997 23 Apr Broad cold front 11 18.4 -0.6 18.1 3.3 3.2 41.8 22 21 July 1997 17-18 Jul Broad cold front 2 16.9 -2.6 15.8 -1.1 8.5 0.0 23 22 August 1997 20 Aug Coastal cold front, onshore flow 1 22.3 0.4 20.8 0.3 0.0 0.0 24 25 August 1998 22-23 Aug Broad cold front 1 19.1 -0.1 20.4 1.9 15.5 0.1 25 12 August 1999 29-30 Jul Cut-off low in interior 12 20.1 -0.9 18.2 -1.8 0.0 0.7 26 10 May 2000 5 May Cut-off low in interior, moist onshore NE winds 4 17.5 3.4 17.2 5.0 26.0 48.3 27 30 June 2001 29-30 Jun Broad cold front 0 17.5 -4.4 15.6 -2.0 28.2 2.4 28 09 August 2001 4 Aug Broad cold front 4 19.1 -4.9 18.4 -2.8 0.0 0.0 29 10 September 2001 30 Aug Broad cold front, low pressure in interior 10 22.3 2.1 20.7 2.4 10.4 55.8 30 08 May 2002 6-7 May Broad cold front 0 21.3 4.9 21.3 5.1 6.1 5.5 31 24 May 2002 23 May Broad cold front 0 21.8 -1.7 19.2 -0.1 6.0 9.6 32 09 June 2002 30-31 May Broad cold front 8 19.0 1.2 15.5 1.6 10.5 20.5 33 25 June 2002 13-15 Jun Cut-off low and cold front 10 16.8 -2.3 14.8 -0.4 9.0 25.8 34 11 July 2002 9-10 Jul Broad cold front 0 16.6 -6.0 15.0 -4.0 0.0 0.0 35 27 July 2002 15-21 Jul Cut-off lows and cold fronts 5 18.1 -2.3 15.1 -3.0 12.6 0.0 36 12 August 2002 4 Aug Upper air cut-off low 7 19.4 1.8 18.7 1.1 12.0 9.0 37 13 September 2002 10 Sep Broad cold front, low pressure in interior 2 19.3 3.5 17.5 3.3 5.5 27.3 38 22 June 2004 7-8 Jun Broad cold front 13 18.1 -5.0 16.9 -2.6 0.0 0.0 39 08 July 2004 27-28 Jun Cold front and low pressure in interior 10 16.8 -2.0 15.5 -0.5 13.8 31.9 40 09 August 2004 30-31 Jul Broad cold front 9 16.9 1.5 18.3 0.4 45.1 0.5 41 10 September 2004 8-9 Sep Broad cold front 0 16.2 2.4 18.6 2.5 30.9 1.2 Chapter 5 - 152 - (1) Broad cold front (n=27) (2) Coastal low (n=4) (3) Cut-off Low (6) (4) Low pressure over (5) Onshore flow (n=1) the interior (n=3) Figure 5.46: The occurrence of snow cover for different snow-producing weather types. Chapter 5 - 153 - systems, which do not push far into the Lesotho interior. Snowfall associated with low pressure systems along South Africa?s eastern coastline (2) are infrequently observed, with snow cover occurring preferentially in the Thabana-Ntlenyana and Mafadi Summit areas. ?Cut-off lows' (3) appear to deposit large quantities of snow over several days, and may cover extensive areas to lower altitudes (below 2600m ASL). Snow incidence is uniformly high across much of the area. However, the northern sections of the High Drakensberg north of Giant?s Castle appear to have a high incidence of snow cover, particularly in regions near the escarpment. Snow cover resulting from low-pressure systems (4) in the interior of the sub-continent appears to affect only the northwestern sections of the study area (Witsieshoek, the Amphitheatre and the adjoining Lesotho interior), with little snow cover recorded further south. Onshore flow (5) from the warm, moist Indian Ocean off the eastern coastline results in snow cover due to orographic uplift along a narrow section of the escarpment north of Champagne Castle, particularly in the northwestern sections of the study area. 5.4 AERIAL PHOTOGRAPHIC DATA The use of the aerial photographic data in conjunction with pocket stereoscopes was unsuccessful in identifying any large-scale geomorphological features in the Nhlangeni Peak to Giant?s Castle sector. Although the mountainous terrain in the photographs was striking when viewed in three dimensions, there was no indication of debris ridges or other features in this region when viewed at the scale in the photographs. However, image quality was high, with significant potential for analysing large-scale debris deposits and snow cover distribution should data sets covering such phenomena be available. There is significant potential for mapping of snow cover in relation to topography, insolation and wind effects. The use of Google Earth data (Google, 2006) was more successful in identifying geomorphological features, particularly debris ridges and block streams. Twenty-five debris ridges and 2 blocks streams were identified through the Google imagery, indicating that the imagery was comparable and more adaptable to broad analysis needs than high-quality aerial photographic data. Images were recorded of the debris ridges and other significant geomorphic phenomena (Appendix 2). Chapter 5 - 154 - 5.5 DISTRIBUTION OF GEOMORPHOLOGICAL PHENOMENA The occurrence of specific geomorphological features, as observed during ground truthing and noted from the published literature, was used to generate images showing their spatial distribution in relation to the topography of the region (Figures 5.47 to 5.53). However, this is by no means a complete distribution list, as it is limited to areas that were ground-truthed and for which available information exists. Block fields, block streams, patterned ground and stone banked lobes were all noted from high altitude areas above 3000m ASL. Thufur were noted from wetland regions in valleys, whilst debris ridges and solifluction lobes were noted on south facing slopes of mountain ranges at altitudes greater than 3000m ASL. A high incidence of block fields (Figure 5.47) and block streams (Figure 5.48) can be noted in the Mafadi Summit (Figures 5.54 and 5.55), Sani Pass and Thabana-Ntlenyana regions (Figure 5.56). The block fields in the Mafadi Summit area are not covered by snow on the 03 and 19 August 1990, which is more prominent further away from the escarpment in this region. The Thabana-Ntlenyana and Sani Pass block fields are also mostly snow free, although snow patches occur adjacent to the features at the majority of these locations. Sorted and non-sorted patterned ground (Figure 5.49) was also found primarily in the Mafadi summit and Thabana-Ntlenyana regions. The features in the Mafadi area are located primarily on mountain tops and on south-facing slopes where there is a significant absence of snow cover (Figure 5.57). Further south in the Thabana-Ntlenyana region (Figure 5.58), the patterned ground is located on south- facing slopes near snow cover, but closer analysis reveals that the features are mostly peripheral to these patches (Figure 5.59). Stone banked lobes (Figure 5.50) were recorded from the Leqooa and Mafadi ? Njasuthi Summit areas. The features on the south-facing slopes of the Mafadi Summit area are found in similar locations to the block deposits and patterned ground, so are absent of snow cover (Figure 5.60). Thufur (Figure 5.51) appear to be common throughout the High Drakensberg and Lesotho Highlands, with significant concentrations found along the upper catchments of high altitude valleys (Figures 5.59 and 5.61). A few occurrences of solifluction lobes (Figure 5.52) are recorded, which show a preference for well shaded south- facing slopes (Figure 5.62). Debris ridges (Figure 5.53) are found in escarpment Chapter 5 - 155 - Figure 5.47: The occurrence of block fields noted from ground-truthing and published literature. Chapter 5 - 156 - Figure 5.48: The occurrence of block streams noted from ground-truthing and published literature. Chapter 5 - 157 - Figure 5.49: The occurrence of sorted and non-sorted patterned ground noted from ground-truthing and published literature. Chapter 5 - 158 - Figure 5.50: The occurrence of stone banked lobes noted from ground-truthing and published literature. Chapter 5 - 159 - Figure 5.51: The occurrence of thufur noted from ground-truthing and published literature. Chapter 5 - 160 - Figure 5.52: The occurrence of solifluction lobes noted from ground-truthing and published literature. Chapter 5 - 161 - Figure 5.53: The occurrence of debris ridges noted from ground-truthing and published literature. Chapter 5 - 162 - Figure 5.54: The occurrence of block fields in the Mafadi summit region in relation to snow cover on 03 August 1990. Figure 5.55: The occurrence of block streams in the Mafadi summit region in relation to snow cover on 03 August 1990. Chapter 5 - 163 - Figure 5.56: The occurrence of block fields in the Sani Pass and Thabana-Ntlenyana regions in relation to snow cover on 03 August 1990. Chapter 5 - 164 - Figure 5.57: The occurrence of sorted and non-sorted patterned ground in the Mafadi summit region in relation to snow cover on 03 August 1990. Figure 5.58: The occurrence of sorted and non-sorted patterned ground in the Thabana-Ntlenyana region in relation to snow cover on 03 August 1990. Chapter 5 - 165 - Figure 5.59: The distribution of various geomorphological phenomena and snow cover around the Thabana-Ntlenyana summit. Figure 5.60: The occurrence of stone banked lobes in the Mafadi summit region in relation to snow cover on 03 August 1990. Chapter 5 - 166 - Figure 5.61: The occurrence of thufur in the Sani Pass and Thabana-Ntlenyana regions in relation to snow cover on 03 August 1990. Chapter 5 - 167 - Figure 5.62: The occurrence of solifluction lobes in the Leqooa River region in relation to snow cover on 03 August 1990. cutbacks near Thabana-Ntlenyana and Champagne Castle and on south-facing slopes in the Sani Pass, Sekhokong (Figure 5.63) and Leqooa regions (Figure 5.64). No significant snow cover is observable in the escarpment cutbacks, but significant snow cover is found in well-shaded valleys in the remaining debris ridge locations. Chapter 5 - 168 - Figure 5.63: The occurrence of debris ridges in the Sani Pass region in relation to snow cover on 03 August 1990. A detailed analysis of the topographic location of the debris deposits and their relationship to snow cover was performed on the two largest deposits through the use of GIS and ground-truthing and surveys. On the Tsatsa-La-Mangaung ridge (Figure 5.65), a debris deposit forms a slight arc to the side and beneath the southeast-facing slope. The preference for late-lying snow cover occurring on southeast, rather than southwest-facing slopes is indicated through remote sensing, with snow cover occurring on the upper part of the preferred slope on the 03 and 19 August 1990. Ground truthing after another light snowfall confirmed this trend (Figure 5.66). On Chapter 5 - 169 - the Leqooa ridge (Figure 5.67), two debris deposits form curved ridges, the westernmost, at the base of the southeast-facing slope, being significantly larger. Snow cover on this slope is extensive and may have a significant impact on slope processes, with different colluvial mantles apparent on southeast-facing and southwest-facing slopes (Figure 5.68). It was also observed that the debris ridge covered exposed bedrock in the upper part of the valley (Figure 5.69). Figure 5.64: The occurrence of debris ridges in the Leqooa River valley region in relation to snow cover on 03 August 1990. Chapter 5 - 170 - Figure 5.65: A debris deposit on the Tsatsa-La-Mangaung ridge just north of Sani Pass, showing its relation to topography and snow cover on 03 August 1990. Figure 5.66: A photo taken looking north-westwards at the Tsatsa-La-Mangaung debris ridge, showing the preferred south-east facing orientation of late-lying snow. Chapter 5 - 171 - Figure 5.67: A debris deposit on the south side of the Leqooa Range, showing its relation to topography and snow cover on 03 August 1990. Figure 5.68: A photo taken looking northwards at the Leqooa debris deposit, early morning shadowing in winter and different colluvial mantles on south-east and south-west facing slopes. Chapter 5 - 172 - Figure 5.69: A photo taken looking northwards at the western Leqooa debris ridge, showing the exposed bedrock underneath the upper part of the deposit. 5.6 SUMMARY Extensive use of remotely sensed imagery has yielded a vast array of information about the spatial distribution of snow cover on the landscape in the High Drakensberg and Lesotho Highlands. The use of two high-resolution data sets, recorded 16 days apart, has permitted the determination of snow cover changes in relation to altitude, aspect, slope gradient and distance from the escarpment edge. Changes in snow cover at different latitudes have also been examined. A series of 41 low-resolution images has also enabled the analysis of long-term snow cover characteristics such as those brought on by different seasons and different snow- forming synoptic weather systems. The spatial distribution of snow cover has also been compared to the spatial distribution of selected periglacial and glacial geomorphological features in the high mountain environment. A close spatial occurrence between large debris ridges in the Leqooa and Sani Pass regions and longer lasting snow cover distribution is recorded. Chapter 6 - 173 - CHAPTER 6 DISCUSSION AND CONCLUSION 6.1 INTRODUCTION Snow cover in the High Drakensberg has only received limited attention (Mulder and Grab, 2002), yet is an important environmental factor in periglacial and glacial environments (Thorn, 1978; Dozier, 1991; Marcus et al., 1992; French, 1996; Derksen and LeDrew, 2000; Evans, 2006). Snow cover is known to provide some control on vegetation, climate and geomorphology (i.e. weathering and erosion processes). An ongoing debate has focused on the potential for cirque, plateau and niche glaciation in the High Drakensberg (Marker, 1991; Hall, 1994; Sumner, 1995; Grab, 1996a; Mills and Grab, 2005). Thus, an understanding of the contemporary distribution of snow cover, in association with the distribution of palaeogeomorphological phenomena, could be fundamental to this debate. Satellite based imagery offers one of the best methods to analyse the distribution of snow cover (Gao and Lui, 2001). Its ability to immediately analyse the environment in spatial terms and its regular temporal frequency make it the preferred method over ground-truthing and gauge measurements (Derksen and LeDrew, 2000). Given the extensive area of the High Drakensberg environment, it is almost impossible to provide adequate data based on ground truthing and gauge measurements alone. This study has thus used a GIS approach to determine the spatial associations of snow cover and geomorphological features. 6.2 THE USE OF SATTELITE IMAGERY Landsat satellite imagery from the TM and ETM+ sensors was chosen over other types of remotely sensed data as they gave the best array of benefits for the High Drakensberg and Lesotho Highlands environment. Landsat data are suitable for basin- scale studies on snow cover as well as mapping of snow in mountainous terrain Chapter 6 - 174 - (Derksen and LeDrew, 2000). The most significant advantage is that the high- resolution imagery has a 30m pixel resolution in seven different wavelengths, as opposed to that of the MODIS satellite sensor, where resolution varies from 250m to 1km across 36 spectral bands (Hall et al., 1995b). The SPOT satellite meanwhile may yield a higher resolution at 10m and 20m, but poorer spectral and temporal coverage is obtained (Gao and Lui, 2001). Although Landsat has fewer spectral bands than MODIS, the seven bands allowed for the easy identification of snow cover. Thematic data captured by the Landsat satellites thus provides the best trade offs and makes it the preferable snow-mapping tool with which to spatially analyse snow cover in the High Drakensberg and Lesotho Highlands with great accuracy (Dozier, 1991). Temporal studies are possible within the 16-day time frame between Landsat images, thus allowing for long term changes to be studied. However, short- term changes such as for the analysis of daily melt out rates are not as viable (Derksen and LeDrew, 2000). The study area covered by the data is extensive and includes all high altitudes along the escarpment line of the High Drakensberg for approximately 200km. A large area of the Lesotho Highlands, up to distances of 25km westward from the escarpment, is also incorporated in the data. This allowed for the analysis of snow cover across almost the entire High Drakensberg as well as a considerable portion of the Lesotho Highlands, thus ensuring a representative spatial distribution of snow cover in the area. The data sets for the study area are all located on the same satellite path, resulting in standardised environmental conditions across the area for all images. Data are collected at the same time of day on each 16-day overpass cycle, further limiting reflective variability and the potential for distorted results (Rosenthal and Dozier, 1996). The need for good satellite data requires minimal cloud cover (Dozier, 1991; Pinkham, 2000). Data were selected preferentially to avoid cloud and haze, but the variable environments above and below the escarpment limits the number of usable image sets. However, in certain incidences, cloud is limited to below the escarpment. Through the use of GIS software, the application of a mask removed such cloud cover, allowing for easy identification and processing of snow cover above the escarpment line and increasing the number of data sets available to the study (Dozier, 1989). Chapter 6 - 175 - A Digital Elevation Model (DEM) was created to analyse altitudinal, slope gradient, slope orientation and micro and macro topographic factors in the study (Baral and Gupta, 1997; Cline et al., 1998). The creation of a DEM at 50m contour intervals was time consuming and frustrating, but required for the needs of the study. Using nine topographical maps, two contours per every 100m of elevation were digitised using a CAD program. File conversion and importation between this CAD program, the GIS software program and image programs were further frustrated by limited computer memory and formatting errors. Most of these are a result of the inadequate IDRISI GIS software program used, which on many occasions stalled in processing the data for such a large study area. This requires the data to be split into various sections for data processing and thereafter remerged. The high cost of Landsat satellite imagery restricted the study to the use of only two full high-resolution data sets. To maximise the temporal element of the research, two dates showing distinct snow cover, but in close proximity to each other, were selected to best analyse the environmental effect on the snow cover for a set period. A resolution of 30m on these high quality images allows for the identification and mapping of large blanketed areas of snow through to small, dispersed snow patches often isolated in regions shadowed by the mountainous topography. Change in snow cover between the 03 and 19 August 1990 provides a clearer understanding of the effect of topography on snow cover. Lower resolution images that are freely available were also selected and used to increase the data on spatial and temporal distribution of snow cover in the area. As the spatial resolution is not high (360m), detailed analysis of individual regions and on specific topographic faces was not possible. Analysis is constrained to studies of the region as a whole, of which temporal comparisons are of good value. Problems were encountered regularly with the spatial referencing of the original images, which would allow for the overlaying of the satellite data onto the DEM. Resampling through trial and error of the 2 high- resolution and 41 low-resolution images had to be done to ensure that the images were correctly registered. Eventually, a good fit was found between all data sets with a satisfactorily minimal spatial error of less than 1 pixel (30m in the high resolution and 360m in the low resolution images). Chapter 6 - 176 - A fine unsupervised classification of a 3-5-4 composite image was the most accurate in the detection of snow on the high-resolution images. This was established through the use of a smaller test area where the results of various snow cover algorythms were compared with a true colour image. This image was of high quality, with snow cover clearly visible, even on highly shadowed south-facing slopes. Other algorythms, such as the SNOMAP equation with the NDSI as well as an image ratioing technique using TM bands 1, 2 and 5, which are favoured in the international literature (Dozier, 1989; Hall et al., 1995b; Derksen and LeDrew, 2000; Pinkham, 2000; Dong et al., 2005), did not perform as effectively. This is attributed to the automatic determination of snow cover in these processes. The manual reclassification of the fine unsupervised classification allowed for the better recording of discontinuous snow cover, as thin layers of snow cover regularly allowed soil and vegetation to show through the snow surface. Mixed pixeling, where snow and other land cover occupy the same pixel in an image, is a very serious concern in marginal snow environments such as the High Drakensberg (Rosenthal and Dozier, 1996), leading in this case to the poor quality of automated snow detection processes. Other potential constraints however, such as variable illumination and detector saturation of imagery were not encountered in the High Drakensberg data. The fine unsupervised classification?s ability to identify snow correctly on steep, shadowed south-facing slopes negated the need for shadow rectification to be performed on the data through the use of the DEM (Hall et al., 1995b; Baral and Gupta, 1997). 6.3 FACTORS CONTRIBUTING TO SNOW COVER DISTRIBUTION IN THE HIGH DRAKENSBERG The results of the snow mapping on the 03 and 19 August 1990 images show an allotment of both larger areas and smaller patches of snow cover. The 03 August data has the highest level of snow cover, with 8% of all areas above 2600m ASL covered by snow. The snow coverage was 17% between 29?40? S and 29?00? S, 2% between 29?00? S and 29?20? S and 6% between 29?20? S and 29?50? S. No snow was recorded between 29?50? S and 29?55? S (Figure 5.14). The data for the whole area indicates 38% more snow cover than on the 19 August image (Table 5.2). The difference in the 16-day intermediate period was originally assumed to be entirely the Chapter 6 - 177 - result of snow melt and ablation, but more careful analysis of the central to northern section of the High Drakensberg (Champagne Castle to Witsieshoek), particularly using change analysis techniques, indicates the occurrence of new snow at high altitudes adjacent to the escarpment (Figure 5.19). The identification of snow and its quantification of aerial extent for the 03 and 19 August 1990, demonstrates a difference in accuracy between low and high- resolution images. Both low-resolution images indicated a smaller area of snow cover (Table 5.3) than the higher resolution images (Table 5.2), with low-resolution data only capturing 80.66% of the snow cover of the higher resolution data on 03 August and 36.56% on 19 August. A detailed analysis of specific snow patches, particularly in regions south of 29?00? S, suggests that due to the 19 August snow cover being smaller and fragmented in this region, owing to the extended melting and ablation period since the last snowfall over much of the region, the snow cover equations have not been as successful in recording the snow at the lower resolution. Conclusions made from the low resolution images therefore cannot rely on the quantitative accuracy of snow cover on specific slopes, but must rather make use of general trends in snow cover where such misidentification is negligible. 6.3.1 Distance from the escarpment Analysis of snow cover in respect to distance away from the escarpment (in the direction of the Lesotho interior) shows that there is a general trend of increasing snow cover with an increased distance away from the escarpment on 03 August 1990 when expressed as a percentage of the study area at a given distance (Figures 5.13), which is confirmed by a positive correlation coefficient of r=0.63. The snow cover appears to occur on high lying grounds in the interior, of which there are large areas at similar altitudes to those at the escarpment. Two possible scenarios need to be considered here: firstly that there was initially a more pronounced snow fall over the interior highlands, as opposed to the escarpment edge region, and secondly, that there is preferred snowmelt in areas adjacent to the escarpment due to the effect of orographic uplift of warmer air onto the High Drakensberg from the adjacent KwaZulu-Natal. The distribution of residual snow, new snow from a subsequent snowfall and snowmelt between the 03 and 19 August provides further indication of Chapter 6 - 178 - escarpment distance trends (Figures 5.23 and 5.31). The rate of snowmelt after 03 August appears to have increased with distance from the escarpment from less than 1% near the escarpment to over 12% of the area between 15km and 23km away, thus suggesting that the original snowfall favoured locations towards the interior of Lesotho in this instance, particularly the northwestern section of the study area, where there are large areas of land at high altitudes. However, it should also be noted that the recorded lower levels of snow melt in the escarpment region could have been skewed by fresh snow cover, which may have re-covered areas where snow melt had already taken place. It is therefore still possible that snowmelt may originally have been uniform across the region. The fresh snow cover occurred between the 03 and 19 August, possibly on 11 August (Appendix 1), near the escarpment between the Amphitheatre and Champagne Castle, indicating that the snowfall event was localised and restricted to the northern sections of the High Drakensberg. This snow cover is probably associated with onshore flow brining moist air from coastal regions where it was orographically uplifted by the escarpment, resulting in rain in the Low Berg and snow at higher elevations (Appendix 3). Snow cover change between the two dates also confirms a slight decrease in the mean distance of snow cover for the escarpment at latitudes south of Champagne Castle (Figure 5.20), indicating a preference for snow preservation closer to the escarpment due to warmer conditions at slightly lower altitudes towards the Lesotho interior. 6.3.2 Altitude An analysis of the altitudinal position of snow cover reveals that there is a general increase in snow cover with an increase in altitude, from 5% of the area between 2950m and 3000m ASL to 23% of the area between 3150m and 3250m ASL on 03 August 1990 (Figures 5.7, 5.25 and 5.42). This is consistent with expectations, as cooler conditions at higher elevations would permit better preservation of late-lying snow (Barry, 1992). However, at altitudes above 3250m ASL, a decrease in snow cover of 14.5% with an increase in altitude to 3450m ASL is noted. At the highest elevations in the High Drakensberg (between 3400m and 3482m ASL), snow cover is low (<8%) when expressed as a percentage of the available land surface area at such altitudes. These results are similar to slope gradient analyses of snow cover where lower snow cover percentages are found on gentle gradients such as mountain Chapter 6 - 179 - summits, indicating that insolation is a factor. Change analysis reveals that the new snow cover on the 19 August 1990 is generally located at lower altitudes (lowest level of snow cover at 2630m ASL) than residual snow cover (lowest level at 2790m ASL), with modes of 3125m and 3250m ASL respectively, showing a trend for the preservation of snow at higher altitudes (Figures 5.25 and 5.26). The rate of snowmelt also increases with altitude, with the single highest percentage of snowmelt recorded above 3450m ASL at 12% of the area, possibly due to reduced slope gradients on the mountain tops which allow for increased insolation and higher levels of snowmelt. In addition, it has been proposed that such mountain tops are subjected to strong winds where much of the snow would be deflated, thus reducing snow cover (Grab, 2002a). However, analysis of snowmelt as a percentage of existing snow from the 03 August shows that the highest level of snowmelt was between 3100m and 3150m ASL at 20% (Figure 5.27). This occurs primarily in the northwest part of the study area, where large areas of snow cover on high-lying ground at gentle gradients melts out before 19 August. An analysis of the quartile distribution of snow cover incidences for the 41 low-resolution images shows that higher altitudes (above 3200m ASL) experience the greatest number of snow events with a median of 12 recorded instances between 1989 and 2004 and a maximum of 28. Altitudes lower than 3000m ASL had a median of between 5 and 7 snowfalls during this period. Between 3250m and 3450m ASL, the median number of events remains stable at 12 (Figure 5.42). Finally, the quartile deviation in the number of snow cover incidences differs between lower altitudes (<2800m ASL), where 3 snow cover incidents differ between the lower and upper quartile, and higher altitudes (>3200m ASL), where up to 11 snow cover incidences can differ. This indicates that snow cover frequency is more variable at higher altitudes than lower. Snow cover image analysis (Figure 5.39) shows that this is a result of lower elevations only getting snow during severe snowfall events, where all locations are covered in snow at the same time. Higher elevations are receptive to more frequent snowfalls of lesser magnitude, but such snowfalls are more varied in distribution across the study area. 6.3.3 Aspect An analysis of the relationship between snow cover and aspect confirms that there is a preference for south- and southeast-facing slopes, with the preferred Chapter 6 - 180 - bearings for late-lying snow cover being 127.5? to 202.5? (Figure 5.9). This compares favourably with international trends (Klimaszewski, 1993), with snow cover preferentially occupying east-facing slopes (from 0? to 180?) to west-facing slopes (180? to 360?) by 63% on 03 August 1990 and 69% on 19 August due to reduced insolation and cooler temperatures on east-facing slopes in the afternoon. An analysis of the aspect of snow cover reveals little difference between residual snow, new snow and snow melt between the 03 and 19 August, other than a higher preference towards south-facing as opposed to southeast- and southwest-facing slopes for new snow (Figures 5.28). This could potentially be a result of the new snow on 19 August having a shallower depth than the snow cover prior to 03 August, with the newer and lighter snowfall undergoing rapid ablation. As residual snow cover already favours east-facing over west-facing aspects, new areas of snow cover on these slopes may not fully represent the snowfall event and bias subsequent melting processes. A more detailed analysis of snowmelt (Figure 5.29), where the snow that had melted by 19 August was shown as a percentage of that present on 03 August, shows a strong trend for ablation on north- and northwest-facing slopes, with approximately 80% of the snow cover on these slopes (262.5? to 52.5?) having disappeared within 16 days. South-facing slopes had a 30% reduction in snow cover. Southeast-facing slopes recorded a 38% reduction in snow cover whilst southwest-facing slopes were less suitable for snow preservation, with 48% of the original snow having ablated (Figure 5.29). 6.3.4 Slope gradient Snow cover predominates on slope gradients between 4? and 34?, with a significant absence of snow on flatter slopes between 0? and 4? (Figure 5.10). Change analysis indicates that snowmelt occurs most rapidly on the gentler gradients, with a mode at 2? gradient. Residual snow (mode of 12? gradient) and new snow (various modes between 14? and 28? gradient) is more pronounced on comparatively steeper slopes. However, new snow is found at a constant 1.5% of areas across all gradients between 10? and 34?, whilst residual snow has a mode at 12?, with almost 12% of the total area covered by snow (Figure 5.30). The absence of snow cover and the high level of snowmelt on flatter gradients concur with the lack of snow cover at higher altitudes above 3250m ASL. Insolation is a key factor, with the flat-topped mountain Chapter 6 - 181 - summits of the High Drakensberg receiving radiation levels that quickly melt and ablate any late-lying snow patches. An analysis of aspect with snow-covered slope gradients, shows a break from this domination by insolation (Figure 5.11). Although snow cover only occupies slightly over 2% of all north-facing slopes on 03 August 1990 (Figure 5.9), such snow occupies a large proportion of snow on slopes between 0? and 15?, occupying up to 40% of north-facing slopes at gradients between 5? and 10? (Figure 5.11). Meanwhile, steeper gradients (>15?) have a higher percentage of snow cover on south-facing aspects (approximately 20% for slope gradients between 10? and 25?). These figures may be affected by the tendency for south-facing slopes to be steeper due to valley asymmetry. The significant influence of insolation on snow cover has already been mentioned, so another factor must have contributed to the high percentages of snow cover on gentler (0? to 15? gradient) north-facing aspects. By the 19 August 1990, smaller percentages (8%) of snow cover occurred on north-facing slopes with gradients between 0? and 15?, hence the north-facing longevity of snow was also short on such gradients. Weather data (Appendixes 1 and 3) indicate that northwesterly winds brought snow-producing weather conditions that may have produced snowdrifts on the north- and northwest-facing slopes. Snow depths would have had to be deep enough for snow cover to remain on the high insolation slopes for the 3 days between the snowfall and the satellite data capture, but were obviously not sufficient to survive a further 16 days of insolation, before the 19 August image was taken. Finally, a strong occurrence of snow is recorded on very steep slopes (76? - 86?), but is lacking on more intermediate gradients (50? to 76?) at similar slope aspects (Figure 5.10). Such findings are a result of the misclassification of land types as snow, given the poor illumination conditions on very steep south-facing slopes. These errors are however easily identifiable and can be excluded from the snow cover analysis. 6.3.5 Latitude The new snow in the northern sections of the High Drakensberg between Witsieshoek and Champagne Castle is clearly visible between 20?40? and 29?05? in various altitudinal analyses. The new snowfall prior to the 19 August 1990, increased snow covered area (Figure 5.15 and 5.16), decreased the minimum and quartile distribution of snow cover altitude (Figure 5.17), and decreased the minimum and Chapter 6 - 182 - quartile distribution of snow-covered slope gradient (Figure 5.19). However, no trends were observable for any potential effect that latitude may have on snow cover. Latitudinal breakdowns of snow covered area (Figure 5.14), snow cover change (Figure 5.16), altitude (Figure 5.17), aspect (Figure 5.18), slope gradient (Figure 5.19), snowmelt (Figure 5.32) and westward distance of snow cover from the escarpment (Figure 5.20) all indicated that meso-scale topography had a dominant effect, with the latitudes hosting the highest altitudes and the most favourable topography for snow capture permitting larger accumulations of snow cover, irrespective of latitude. 6.3.6 Topography Large areas of the High Drakensberg, particularly the Amphitheatre to midway between Champagne Castle and Cathedral Peak, the Mafadi Summit area, the Thabana-Ntlenyana area and the Leqooa area favour high proportions of snow cover (Figure 5.6). These regions represent the most significant areas of high altitude in the study area, all of which have large areas above 3200m ASL and high mean altitudes (Figure 4.1). Results confirm that snow cover is controlled by various topographic factors that favour cooler conditions and reduced insolation, such as south-facing aspects, slope gradients steeper than 4? and high altitudes. An analysis of snow cover change at specific study areas, in context with the local topography, confirms many of the above results. Areas such as the northwest part of the study area, the Mafadi Summit region, the Thabana-Ntlenyana region and Leqooa range, all show large areas of residual snow and snowmelt between 03 and 19 August 1990 (Figure 5.33). These regions are all noted as having large areas at higher altitudes than the surrounding Drakensberg and Lesotho Highlands, a factor that seems to have determined the spatial distribution of snow cover on the 03 August 1990. Topography has also had a marked effect on new snow cover identified on the 19 August data image (Figure 5.34). Change analysis confirms that the new snow is confined to escarpment and near-escarpment regions of the northern part of the study area. The distribution of new snow in this region is not limited to the highest elevations, but is instead evenly distributed on all the south-facing slopes of the mountains. This snowfall, 6 days prior to the image, appears to have been the result of Chapter 6 - 183 - orographic uplift of moist air over the High Drakensberg due to onshore flow (Appendix 3, [4]). In the northwestern part of the study area, residual snow between 03 and 19 August is distributed in patches on the south-facing, moderate gradient slopes of mountain ridges (Figure 5.35). Areas of snow cover on flat and north-facing slopes between the patches are covered in snow on the 03 August, but have undergone considerable melting (80%) by the 19 August. This region of north and northwest- facing snow cover is recorded in the data on flatter slope gradients of 0? to 15? (Figure 5.11). The new snow cover in the adjacent and overlapping area of the Amphitheatre is already restricted to moderate gradient south-facing slopes by the 19 August (Figure 5.36). It is observed that new snow has in some places extended the size of residual snow patches, particularly on the upslope side of the pre-existing patches, where the slope gradient is flatter and snow is more exposed to solar radiation. New snow in this escarpment region had thus not undergone the same level of melting by the 19 August, than the original snow cover had by the 03 August. Apart from the primary south-facing aspect of snow cover, the east-west aligned mountain ranges in the Lesotho Highlands have also allowed snow to preferentially survive on southeast-facing as opposed to southwest-facing slopes. The Sani Pass and Leqooa ranges are excellent examples of this preference (Figures 5.37 and 5.38). In both cases, the snow is predominant on the southeast-facing slopes of valleys draining the southern side of the mountain ranges. The Leqooa mountain range on the north side of the Leqooa River demonstrates excellent preservation levels for snow cover, with very little snow melt taking place on the fringes of the large areas of residual snow that survived between 03 and 19 August. The availability of natural topographic depressions in certain mountain ranges in the Lesotho Highlands, such as the Leqooa (Figure 5.38), Sekhokong and Tsatsa-La-Mangaung Ranges (Figure 5.37) appears to allow for a significant ability to capture and maintain snow cover. The higher number of south-facing hollows in these areas as opposed to high altitude regions further north with fewer such topographical phenomena (such as the northwestern part of the study area, the Thabana-Ntlenyana and Mafadi Summit region) has allowed for a higher number of long-lasting snow patches to be recorded. Chapter 6 - 184 - Low-resolution imagery confirmed the preference of snow cover at the highest elevations in the High Drakensberg and Lesotho Highlands. The recurrence of snow cover over 41 snow cover images (Figure 5.40) corresponds favourably with snow cover in the higher resolution images of 03 and 19 August 1990 (Figures 5.33 and 5.34). More regular snow cover is once again recorded over the northwestern part of the study area, the Mafadi Summit area, the Thabana-Ntlenyana area and the Leqooa Range, all of which receive up to 28 incidences of recorded snow cover between 1989 and 2004. The relationship between such snow cover and altitude is analysed through a rationing process (Figure 5.41) and a correlation between altitude and the number of snow recordings occurring in each pixel (r=0.34) (Figure 5.43), confirming that areas with a higher altitude record more snow cover. 6.3.7 Seasonal trends Snow cover analysis in relation to the month and season of satellite image capture shows a preference for snow cover in the winter months of June, July and August, where a higher number of snow cover incidences (12 in August) and a greater spatial extent in snow cover are noted (Figures 5.44 and 5.45). This is in accordance with the lower mean temperatures over the winter months, when the mean monthly air temperature is approximately 0?C, and increased occurrences of cold front events to approximately once a week (Killick, 1963; Preston-Whyte et al., 1991; Sene et al., 1998; Grab and Simpson, 2000). Snow distribution appears to be uniform across all months, recurring most frequently in areas with higher elevations such as the Leqooa, Thabana-Ntlenyana, Mafadi Summit region and the northwestern regions of the study area (Figure 5.44). The occurrence of late seasonal snowfalls in September and October appear to show the greatest inconsistencies, where snow cover is generally low in sections north of Mafadi Summit and even lower northwards of Champagne Castle (Figure 5.45). This may be the result of the contraction of the circumpolar vortex (Budin, 1985; Preston-Whyte and Tyson, 1997), weakening westerly waves, mid-latitude cyclones and associated cold fronts over the subcontinent, which may not penetrate as far north during spring. Chapter 6 - 185 - The long-term seasonal trends of snow cover in the High Drakensberg are not conclusive. The study area appears to have high levels of snow cover for large durations in some years, most notably in 1996 and 2002, when a minimum of 8 snowfall events is recorded (Figure 5.39 and Appendix 1), but these findings are on the assumption that all instances of snow cover have been recorded, which cannot be confirmed. Extensive snow cover may occur between the dates of data capture, yet are not recorded on the next satellite overpass, either due to rapid ablation or due to cloud cover. The monitoring of snow cover in areas such as the High Drakensberg for global climate change (Ellis and Paul, 2001) or the effect of ENSO conditions on snowfall (Brown, 1998) may not be viable using Landsat imagery due to the low frequency of snow cover data and poor temporal resolution. Similarly, SWE estimates (Cline et al., 1998; Molotch et al., 2001) for the run-off modelling in the Lesotho Highlands are not viable using the current data. 6.4 CLIMATIC CONSIDERATIONS IN SNOW COVER DISTRIBUTION The determination of snowfall events resulting in snow cover on satellite data images was easily noted for medium and heavy snowfall events, as these were significant enough to be reported by the South African Weather Bureau in their daily reports (SAWB, 1989-2004). Lower magnitude snowfall events were not mentioned, thus requiring some deductions to be made as to the likely cause of snowfall. In most instances, these were easily determined as only one significant snow forming system moved through the region in the days prior to the data capture. Occasionally, multiple systems moved through the region, thus making the determination of the snowfall events more complex. Consideration had to be made for the possibility that more than one snowfall event led to the snow accumulations recorded in the data. Frequent discontinuities in weather recordings at the Underberg weather station have limited observations in a few instances (Appendix 1), but the trends of temperature fluctuations in the remaining records and in those from Bethlehem have enabled extrapolations to be made on weather conditions. Five different types of snow forming weather conditions were identified (broad cold fronts, low pressure systems, cut-off lows, low pressures in the interior, Chapter 6 - 186 - and onshore flow) that produce precipitation in the High Drakensberg following a reclassification of precipitation producing synoptic events over KwaZulu-Natal (Table 4.6). Broad cold fronts are the cause of the majority of snowfall events, covering the largest area in many instances, with the highest frequency of snow cover on the higher elevations in the study area: the Leqooa Range, Thabana-Ntlenyana area, Mafadi Summit area, and the northwestern part of the High Drakensberg (Figure 5.46). Only cut-off lows deposit snow in quantities as high as (or exceeding) those resulting from cold fronts, but these are less frequent than cold front events. Whereas broad cold fronts deposit snow preferentially on all four major high lying regions, cut-off low events appear to only favour three, with the northwestern region of the study area receiving significantly less snow cover. The cause of this discrepancy is probably a result of its westward displacement away from the escarpment edge. Snowfall does not appear to occur frequently in the interior of Lesotho during such events due to the orographic blocking caused by the High Drakensberg, with a high frequency of snow cover observed in a narrow band alongside the escarpment instead (Figure 5.46). By comparison, broad cold front events move across the entire Lesotho Highlands and High Drakensberg from the southwest, and are not substantially affected by the orographic barrier offered by the Great Escarpment. Several observations can be made regarding weather conditions prior to and after snowfall events, and their subsequent impact on snow cover distribution based on the available images (Figure 5.39) and weather data (Table 5.5, Appendixes 1 and 2). Although climatic conditions between Underberg and Bethlehem are similar for the majority of broad cold front events, they often differ significantly under other synoptic conditions. The location of the two weather stations, to the east and the northwest of the study area respectively, is beneficial for understanding the spatial variability of weather conditions during synoptic events that result in different snow cover distributions. Low-pressure systems over the interior appear to bring larger snowfalls to the northwestern parts of the study area (Figure 5.46), which is reflected in the weather data (e.g. Appendix 1 ? 03 August 1990 and 08 August 1992) where Bethlehem receives 5.4mm and 34.2mm, yet Underberg remains dry. A similar trend is also observed when onshore flow conditions produce orographically induced snowfall along regions adjacent to the escarpment in the northern High Drakensberg (Figure 5.46) (e.g. Appendix 1 ? 19 August 1990). Underberg receives 4mm of Chapter 6 - 187 - rainfall, whilst Bethlehem remains dry due to its location above and to the west of the main escarpment. No clear trend in snow cover distribution is observed for coastal low events, during which snow cover usually occurs at all high lying locations, partially due to their infrequent nature, with only 4 records between 1989 and 2004. 6.5 SNOW COVER DISTRIBUTION AND GEOMORPHOLOGICAL PHENOMENA The overlay of geomorphological data mapped during ground-truthing and from the existing literature with snow cover imagery allows for detailed analysis of the spatial relationship between certain geomorphic features and snow cover. The identification of geomorphic features from aerial photographic data was unsuccessful due to the small scale of geomorphic features in the region of the photo data set. Features such as large debris ridges were not identified in the photographs, even at their 1:30 000 scale. However, imagery obtained using the Google Earth programme proved highly successful, with the identification of numerous debris ridges and some block streams possible due to improvements in the resolution of satellite imagery for the region (Appendix 2). Twenty-five potential debris ridges were identified, with 96% of them occurring in hollows on the south-facing slopes of mountain ridges. The largest and most developed ridges were located in the Leqooa region. The Google satellite data showed that the ridges vary in height between an absolute low of 2818m ASL (Appendix 2 ? ridge S) to an absolute high of 3209m ASL (Appendix 2 ? ridge J). The most notable ridges were ridge B and ridge J, with horizontal lengths of 306m and 364m and widths of 40m and 82m respectively. The light snow covers the majority of images in the Google Earth data on protected south-facing slopes, which allows for investigations into the preferential distribution of snow patches in relation to the debris ridges. The heaviest snow cover in most of the relevant hollows occurs directly above the debris ridges supporting a hypothesis that the ridges are related to large accumulations of snow and ice. Ridge J is potentially favourably suited to the reception of snowfall, situated well away from any orographic barriers that may hinder snow accumulation. Other geomorphological phenomena observed through Google Earth included long-lasting snow cover at the base of a basalt terrace and a debris flow that appears to have displaced a stream course (Appendix 2). Chapter 6 - 188 - 6.5.1 Block fields and block streams Block deposits were identified at high altitude sites within regions with large areas at high altitudes exceeding 3000m ASL. These include the Leqooa region, the Thabana-Ntlenyana / Sani Pass region, the Mafadi Summit region and the northwestern part of the study area, where altitudes are all over 3000m ASL (Figures 5.47 and 5.48). However, the occurrence of blockfields is also noted at Giant?s Castle at an altitude above 3100m. Closer analysis of blockfields and blockstreams in the Mafadi Summit and Thabana-Ntlenyana areas show that these features are common to south-facing slopes and the highest summits of the High Drakensberg (Figures 5.54 ? 5.56). Although contemporary winter snow cover frequently occurs adjacent to the block deposits (Figure 5.53), it disappears more rapidly in areas overlaying the block deposits. Block deposits require cold climates and exposed slopes that are generally snow free. This concurs with the literature, which supports the contention that active frost processes required for block formation necessitates snow free slopes (Hastenrath and Wilkinson, 1973; Boelhouwers, 1994; Grab, 1999a, 2000a; Boelhouwers and Meiklejohn, 2002; Boelhouwers et al., 2002). The late-lying snow cover distribution on 03 August 1990 has shown that the locations of mapped block deposits were predominantly free of snow cover. Although block deposits are considered to be relicts from a colder climate, the imagery demonstrates that such locations may be free of late-lying snow cover for long periods of time during cold conditions. The snow patch distribution shows a distinct preference for large accumulations further than 6km away from the escarpment, yet is absent in nearby regions where block deposits predominate. The effect of the Little Berg climate below the escarpment may be an important control, melting snow immediately adjacent to the escarpment at a relatively high rate due to factors such as orographic uplift of warm maritime air from the east. 6.5.2 Patterned ground Sorted and unsorted patterned ground was identified in various high altitude sites, and an analysis of their locations suggests that the larger forms are most common on mountain top locations above 3200m ASL (Figure 5.49) (Grab, 2002a, 2004). There does not appear to be a relationship between snow cover on 03 August Chapter 6 - 189 - 1990 and the location of small sorted patterns less than 20cm in diameter, with the feature locations not indicating any favoured slope aspect. Locations for large-sorted patterned ground also appear free of snow cover (Figures 5.57 and 5.58). The process formation of large sorted stone circles and stripes is frost heaving, which requires the absence of snow cover (Lewis, 1988b; Boelhouwers, 1994). This is confirmed by the absence of late-lying snow cover on most of these features during in this instance. Large sorted patterned ground on Mafadi Summit and Thabana-Ntlenyana with depths in excess of 85cm have shown that they developed during the LGM where the temperatures were lower than ?1.6?C, further suggesting that permafrost was present above 3400m ASL (Grab, 2002). 6.5.3 Stone-banked lobes An analysis of the spatial distribution of stone-banked lobes shows a preference for the same high altitude sites where patterned ground are located, however with a stronger preference for south-facing slopes (Figure 5.50) (Grab, 2004). Snow cover is absent in the majority of these regions on the 03 August 1990 (Figure 5.60). The site locations are thus favourable for gelifluction and frost creep processes, which are important periglacial mass movement processes (Van Steijn et al., 1995; Grab, 2000a). 6.5.4 Solifluction lobes No specific research has been undertaken in the High Drakensberg on solifluction lobes, yet they are noted on south-facing slopes in the Leqooa, Sekhokong and Thabana-Ntlenyana regions (Figure 5.52). Snow cover on the 03 August 1990 overlaps with lobe locations (Figure 5.62), suggesting the possibility that gradual snow melt may help with the formation of such features through the gradual supply of water to the soil. 6.5.5 Thufur Thufur are widely distributed throughout the mid-to-high lying regions of the High Drakensberg (Figure 5.51). They are recorded at altitudes from at least 2700m Chapter 6 - 190 - ASL to over 3400m in shallow valleys near mountain summits (e.g. Mafadi Summit). Past research (Grab, 1994, 1997a, 1998a) and recent surveys have shown that thufur are restricted to the upper drainage basins in the region, with a strong preference next to streams and within valley wetlands (Grab, 1994, 1997a). Snow cover is generally absent at these locations due to the gentle slope gradients and subsequent high insolation levels (Figure 5.61). 6.5.6 Debris ridges Debris ridges on the south-facing slopes of mountain ranges are recorded in the Leqooa, Sekhokong and Sani Pass regions, whilst ridges in cutbacks of the escarpment are located in the Thabana-Ntlenyana and Champagne Castle regions (Figure 5.53). In all instances, ridges are found in hollows that are well protected from insolation. Snow cover is not recorded for the locations in the escarpment cutback, with orographic updrafts from the warmer Little Berg possibly having accelerated snowmelt before the data imagery was recorded. However, the deposits in the south-facing valleys along the mountain ridges show a strong relationship with snow cover, with snow patches shown either as covering the deposits or covering southeast-facing slopes immediately above them (Figures 5.63 and 5.64). Accurate GPS mapping of the debris ridges along the Tsatsa-La-Mangaung and Leqooa Ranges allowed for the precise geo-registration of the debris ridges onto the snow cover data and DEM. The Tsatsa-La-Mangaung debris ridge occupies the base of the south-facing valley, curving around the adjacent southeast-facing slope (Figure 5.65). Snow cover was recorded on the upper elevations of this slope on 03 and 19 August 1990, whilst snow cover was noted to occur preferentially on the southeast face rather than the southwest face of the valley during ground-truthing (Figure 5.66). The ridge has been identified as moraine by Mills and Grab (2005). The situation of a large persistent snow pack forming a small glacier may be envisaged for the southeast-facing slope during a colder climate. Such a glacier would have required sufficient weight to exert a gravitational force upon the existing colluvial material so as to bulldoze the moraine to its present position. The snow and ice pack may subsequently have contributed to the continued accumulation of material through debris avalanches across the snow-covered surface. Chapter 6 - 191 - A pair of debris ridges is located on a large south-facing slope in the Leqooa valley (Figure 5.67). The larger western ridge has a significant snow-covered area on the southeast-facing slope above and to the side of it, whilst the smaller ridge, 300m to the east, has a thin snow covered area occupying its shallow valley as of the 03 August 1990. The upper slope of the of the western debris ridge is superimposed over the mountain bedrock (Figure 5.69), alluding to its depositional origin, with detailed analysis of the morphology and sedimentological structure indicating that the ridges are glacial moraines (Mills and Grab, 2005) (Figure 5.68). Similar conditions to that at the Tsatsa-La-Mangaung deposit could have created large snow and icepacks on south-facing valley slopes, with a large glacial mass pushing debris off the southeast- facing slopes to create the larger ridge. The likely positioning of a palaeo-icepack near the smaller deposit is not apparent. However, the distribution of the late-lying snow cover on 03 and 09 August on south- and southwest-facing slopes in the shallow valley may indicate the former location of an icepack, which is well shaded from afternoon insolation beneath the 3431m high Leqooa peak. All debris ridges identified by ground truthing and Google Earth are found in hollows on the south side of mountain ridges (Appendix 2). The hollows are generally steep sided, with narrow necks at the base. The debris ridges are closely associated with large snow patches on the southeast-facing slopes within the hollows, with significant altitude and area above the ridges for snow accumulation. The preferred location of these aspects in contemporary climates supports proposals that icepacks preferentially developed on these aspects during colder palaeoclimates. In many instances, the ridges are offset to the east of the main axis of the hollow and subsequently the stream bottom, indicating that the glacial mass may have had sufficient force to push the moraine partially up the southwest-facing slope. The spatial distribution of debris ridges indicates a preference for locations in the central and southern parts of the High Drakensberg. No debris ridges were identified by ground truthing or through Google Earth in the northern High Drakensberg Region. The absence of such ridges in the Google Earth data suggests that either the ridges are too small to be identified through this method, or they are non-existent in the northern region. Chapter 6 - 192 - 6.6 SNOW COVER IMPLICATIONS FOR THE HIGH DRAKENSBERG ENVIRONMENT The distribution of snow cover in the High Drakensberg will have major implications for the soil conditions, vegetation, water reservoir capacity, climatology and geomorphology of the region (Thorn, 1978; French, 1996). The existence of snow cover for extended lengths of time following snowfalls will have varying impacts dependent on the depth of the snow cover. Deep layers of snow cover will provide a large frozen water reservoir, with the slow release of snowmelt ensuring a steady supply of water to the environment and the Lesotho Water Highlands Project during the dry winter season. It will also have an insulating effect against ground temperature fluctuations, erosional run-off and frost and thermal weathering, thus restricting geomorphological activity at the ground surface (French, 1996; Palacios and S?nchez- Colomer, 1997). It may also form a protective layer for soil and vegetation against burning and overgrazing, both of which are prevalent in winter (Grobbelaar and Stegmann, 1987; Meakins and Duckett, 1993; Grab and Morris, 1999; Grab and N?sser, 2001). Snow patches on steeper gradients will also act as transportational surfaces to weathered debris. Snowmelt further supplies a significant amount of moisture to the underlying soil over a lengthy period of time, with weathering and erosional processes increasing with the available water during spring (Thorn, 1978). The preferential location of snow on south and southeast-facing slopes implies that vegetation type, soil moisture levels and geomorphological processes will be most important on these slopes, compared to those facing other aspects. The impact of snow may last longer than the date of the last recordable snow cover, thus small patches of snow may continue to survive in suitable topographic positions for several weeks beyond that of the surrounding topography. Snow cover and snowmelt affects geomorphological features through weathering and erosional processes that control their development. The high incidence of snow cover on south-facing slopes is an important consideration in the debate on valley asymmetry (Boelhouwers, 1988, 2003; Meiklejohn, 1992, 1994; Grab et al., 1999) as the snow provides a long-term source of moisture for weathering processes. The effect of snow cover on slope weathering is also a consideration when comparing Chapter 6 - 193 - southeast- to southwest-facing slopes (Figures 4.4, 5.68 and 5.69). In the analysis of snow cover distribution, no evidence was found for the accumulation of snow due to snow blow, despite the strong gradient winds that predominate in the High Drakensberg (Barry, 1992; Freiman et al., 1998). Mountain tops are generally free of snow cover within a few days of snowfall due to the high levels of insolation on the exposed horizontal surface, which would encourage freeze-thaw and thermal stress processes. The high incidence of block deposits, stone-banked lobes and patterned ground on these summit interfluves, associated with severely cold temperatures and large daily fluctuations as a result of the high altitude, is possible due to the lack of snow cover at these locations. The lack of snow cover on mountain tops may also be a consequence of wind deflation (Grab, 2002a). On a larger scale, despite preferential snowfall near the escarpment under certain synoptic conditions, results show an increase in late-lying snow cover away from the escarpment edge, potentially as a result of ablation due to airflow from warmer regions below the escarpment. However, most research on periglacial and glacial geomorphology has focused on the narrow escarpment region, suggesting that studies may also need to be undertaken at high altitude locations further into the Lesotho interior. The macro-scale topography of the Drakensberg, with its southeast to northwest alignment north of Giant?s Castle, also produces a snow shadow in the northern sections when cold fronts with insufficient strength are limited to regions south of this position. The reduction of snowfall in the late snow season (spring) in the northern region will affect the geomorphology, potentially explaining why debris ridge density and occurrence appears to favour southern regions. With warmer air temperatures, snow cover may melt out quicker, limiting the impact on geomorphology. The expansion of the circumpolar vortex during the LGM and its subsequent increase in the frequency and severity of cold fronts and the increase in gradient wind speeds (Budin, 1985; Preston-Whyte and Tyson, 1997), would have promoted a climate more conducive for glaciation. A good understanding of the present distribution of snow cover can assist with modelling colder palaeoenvironments. Contemporary sites of preferred snow cover may indicate locations where snow cover was deeper and more frequent during the past, particularly during the LGM (Evans, 1977; Rapp, 1984). Based on the snow cover analysis, the formation of perennial snow and icepacks will thus favour southeast-facing slopes in suitable topographic Chapter 6 - 194 - locations where snow cover is presently found for extended lengths of time. No snow cover was recorded for cutbacks along the escarpment due to the proposed orographic uplift of warm air, which casts debate on considerations for niche glaciation at these locations (Hall, 1994; Grab, 1996a). Meanwhile, the proposal by Marker (1991) that wind blown snow was deposited in north-facing hollows to form cirque glaciers does not concur with international literature, which indicates preferential cirque formation in poleward-facing hollows (Nelson, 1998). In addition, the current study has recorded a dominance of snow capture on south-facing orientations, which supports the occurrence of debris ridges at the base of southeast-facing slopes in the Leqooa and Tsatsa-La-Mangaung regions, thought to be of glacial origin (Mills and Grab, 2005). The analysis of snow cover distribution in the High Drakensberg and Lesotho Highlands has led to the further identification of ridges and their close relationship to contemporary snow patches (Appendix 2). The High Drakensberg is presently an ecotone, a climatic and ecological boundary that is a strong indicator of changing climatological conditions (Marker, 1998). As such, its vegetation and geomorphological features cannot only provide us with a quick indication of changing contemporary climates, but also permit the establishment of palaeoclimates. The calculated 6?C drop in the MAAT during the LGM (Talma and Vogel, 1992) is considered to be inadequate for glaciation to have occurred in the High Drakensberg, with the dominance of solifluction processes in the region further arguing against widespread glaciation (Grab et al., 1999). The impact of snow cover on geomorphology however, has not been readily considered in the literature. Although precipitation values during the LGM are reported to have dropped by 30% of current values (Partridge, 1997b), the cooler temperatures would have resulted in a significantly higher proportion of precipitation falling as snow. An increased lapse rate for higher altitudes, as well as more frequent temperature depressions as a result of increased frequency and severity of cold fronts (Grab and Simpson, 2000), coupled together with snow?s high albedo and reflectance of energy away from the surface, may have produced very favourable conditions for the development of long-lasting snow patches, particularly within some topographic locations which allow for an annual accumulation of snow. A climate with MAAT reductions greater than 6?C, with snow and ice packs on south-facing slopes, will thus have considerable implications for geomorphic processes. Permafrost may have been Chapter 6 - 195 - possible on the high mountain interfluves, which would have remained snow free due to wind deflation and insolation. Large relict sorted circles support the contention that permafrost may have occurred above 3400m ASL during the Late Quaternary (Grab, 2002a). Similarly, periglacial phenomena that are not yet confirmed for the High Drakensberg, such as protalus ramparts and cryoplanation terraces, need to be considered in respect to larger accumulations of snow during this period. Finally, consideration needs to be given to the meltout of large patches of snow and ice on south-facing slopes towards the end of the LGM, and its effect on mass movement and possible debris flow activity as suggested by Grab (1999a). 6.7 CONCLUSION The High Drakensberg has often been regarded as a periglacial environment, but with considerable criticism for any suggestion that some features may be glacial in origin (Sumner, 1995, 2000, 2004a; Sumner and Meiklejohn, 2000; Boelhouwers and Meiklejohn, 2002). This dissertation has demonstrated the spatial occurrence and potential role of snow cover in the High Drakensberg and Lesotho Highlands, which have previously been underestimated and sometimes disregarded. Through the use of a GIS and a DEM, high- and low-resolution remotely sensed Landsat images have determined the spatial and temporal distribution of late- lying snow cover for numerous dates between 1989 and 2004 for the High Drakensberg and Lesotho Highlands. With the supplementary use of weather data, it has been shown that a minimum of 8 separate snowfalls may occur during some winters (Figure 5.39 and Appendix 1). Ground-truthing was performed to verify the existence and location of specific geomorphological features of periglacial and glacial origin, such as stone-banked lobes, block fields and block streams, solifluction lobes and debris ridges. Snow ablation rates are dependent on the environmental conditions of the area, with the volume of initial snowfall, the topographic location of the snow cover, temperature, wind and insolation being important controls. Snow cover distribution has been analysed in respect of topography, altitude, latitude, aspect, slope gradient and proximity to the escarpment in order to determine any trends. Chapter 6 - 196 - Altitude is a predominant factor in snow cover, with an increase in snow cover occurring at higher altitudes. Snow cover on 03 August 1990 increases from 5% of area between 2950m and 3000m ASL to 23% of the area between 3150m and 3250m ASL. Late-lying snow cover favours gently sloping south- and southeast-facing aspects, with a preference for bearings between 127.5? and 202.5? (Figure 5.9). Southeast-facing slopes have a slightly higher level of snow preservation with a 38% reduction in snow cover between 03 and 19 August 1990, whilst southwest facing aspects had a 48% snow reduction, possibly due to the western aspect being warmer than the eastern aspect in the afternoon. Snow cover on northern aspects undergoes strong ablation due to higher levels of insolation, with approximately 80% of snow cover melting on 262.5? to 52.5? aspects between the 03 and 19 August. Snow also appears preferentially on gently sloping gradients between 4? and 34?. However, the flattest gradients, particularly on mountain summits above 3250m ASL, show a significantly lower level in snow cover due to higher levels of insolation. Snow cover typically occupies less than 8% of area at the highest altitudes above 3400m ASL. Snow cover adjacent to the escarpment may ablate rapidly under particular synoptic conditions, due to the effect of airflow from the warmer KwaZulu-Natal region when warmer easterly winds are orographically lifted over the High Drakensberg. Snow cover towards the interior of Lesotho may be preserved for longer periods of time at high altitudes due to the protection it receives from its continental position. However, snowfalls over these interior locations may not be as high as those adjacent to the escarpment due to topographic controls. The higher altitudes found adjacent to the escarpment and the preference for synoptic events to produce snow nearer to the orographic barrier of the escarpment may initially produce greater snow cover in areas adjacent to the escarpment. Local topography may also have an impact, with south- facing slopes on east-west ridgelines favouring the accumulation and protection of snow cover, particularly in south-facing hollows along these ridges. The absence of a suitable topographic situation in the northern areas of the study area, as well as the lower number of snowfall events in northern sections of the High Drakensberg during spring, owing to topographic snow-shadowing by the Drakensberg and the weakening of cold front systems, may produce a slight preference for late-lying snow cover for regions south of 29? S. However, no other discernable trends for latitude were identified, with the meso-scale topography dominating the locations of preferred snow cover accumulation and preservation. Chapter 6 - 197 - Cold fronts produce the majority of snow cover in the High Drakensberg, accounting for 66% of the recorded snowfall incidences between 1989 and 2004. Only cut-off lows may produce equivalent or greater amounts of snow cover, but only represent 15% of snow producing incidences. Onshore flow and orographic uplift of moist maritime air over the High Drakensberg usually produces snow cover adjacent to the escarpment. Meanwhile, low pressure systems over the interior may occasionally bring snowfall that is limited to the northwest section of the study area. Specific glacial and periglacial geomorphological features were identified and correlated to contemporary snow cover. Block deposits, patterned ground and stone- banked lobes are found in high altitude regions where an absence of snow cover is noted, particularly exposed mountain summits. Thufur are most prolific in high altitude valleys with available ground moisture. No late-lying snow cover is recorded near thufur due to their occurrence on gentle slope gradients. Solifluction lobes and debris ridges are commonly associated with snow patches on south-facing slopes of mountain ranges. High numbers of debris ridges, some of which have been identified as glacial moraine (Mills and Grab, 2005), have been identified through satellite imagery on Google Earth at sites that support the accumulation of snow. The debris ridges show a very strong relationship to snow patches on southeast facing slopes in hollows, particularly ridges in the Leqooa and Tsatsa-La-Mangaung ranges. These features are significant in the consideration of possible plateau, cirque and niche glaciation in the region. Contemporary snow patch distribution, which favours southeast-facing slopes above these ridges, suggests that these slopes were favourable for the accumulation of snow and conversion of snow to ice during the LGM. The data do not support the hypothesis of cirque glaciers in north-facing hollows as a result of wind blown snow (Marker, 1991). The development of niche glaciers in cutbacks along the High Drakensberg escarpment is feasible due to its topographic location, provided that the effect of orographic uplift of warm air from KwaZulu- Natal is curtailed (Grab, 1996a). Snow cover has a significant role in the contemporary periglacial and palaeoglacial environment of the High Drakensberg. An understanding of contemporary snow cover distribution and longevity improves our ability to evaluate Chapter 6 - 198 - both contemporary periglacial and potential palaeo-glacial environments for large parts of the High Drakensberg region. The use of satellite imagery has proven to be a valuable tool in the analysis of snow cover distribution and its association with geomorphological features. Advancements in satellite imagery technology, and particularly the advent of Google Earth and similar satellite map tools on the Internet, is likely to rapidly improve the spatial analysis of periglacial and glacial phenomena within the High Drakensberg and Lesotho Highlands. Further, remote sensing of snow cover in the region may be valuable to other fields of study such as hydrological modelling. In addition, mapping snow cover over Lesotho may assist in disaster risk reduction projects, especially given that lives are frequently lost during the more severe snowfalls. To this end, the current project serves as a foundation for the future projects which will hopefully expand on the current findings with the aid of improved imagery and technology. - 199 - REFERENCES Adams, W. P., Cogley, J. G., Ecclestone, M. A., and Demuth, M. N., 1998: A small glacier as an index of regional mass balance: Baby Glacier, Axel Heiberg Island, 1959-1992, Geografiska Annaler, 80 (A), 37-50. Aizen, V. B., Aizen, E. M., and Melack, J. M., 1997: Snow distribution and melt in central Tien Shan, Susamir Valley, Arctic and Alpine Research, 29 (4), 403- 413. Andr?s, N., Garc?a-Romero, A., Mu?oz, J., and Palacios, D., 2005: Influence of snow distribution in vegetation cover and geomorphological activity: Manzanares Headvalley, Sierra de Guadarrama, Spain, Geophysical Research Abstracts, 7, EGU05-A-08551. Ballantyne, C. K., 1987: Some observations on the morphology and sedimentology of two active protalus ramparts, Lyngen, northern Norway, Arctic and Alpine Research, 19 (2), 167-174. Ballantyne, C. K. and Benn, D. I., 1994: Glaciological constraints on protalus rampart development, Permafrost and Periglacial Processes, 5, 145-154. Ballantyne, C. K., and Kirkbride, M. P., 1986: The characteristics and significance of some lateglacial protalus ramparts in upland Britain, Earth Surface Processes and Landforms, 11, 659-671. Baral, D. J., and Gupta, R. P., 1997: Integration of satellite sensor data with DEM for the study of snow cover distribution and depletion pattern, International Journal of Remote Sensing, 18 (18), 3889-3894. Barry, R. G., and Van Wie, C. C., 1974: Topo- and microclimatology in alpine areas, In: Ives, J., and Barry, R. G., (eds.), Arctic and Alpine Environments, Methuen, London, 73-84. - 200 - Barry, R. G., 1992: Mountain weather and climate (2nd edition), Routledge, London. Barsch, D., 1993: Periglacial geomorphology in the 21st Century, Geomorphology, 7, 141-163. Benn, D. I., and Lehmkuhl, F., 2000: Mass balance and equilibrium-line altitudes of glaciers in high-mountain environments, Quaternary International, 65/66, 15- 29. Berrisford, M. S., 1991: Evidence for enhanced mechanical weathering associated with seasonally late-lying and perennial snow patches, Jotunheimen, Norway, Permafrost and Periglacial Processes, 2 (4), 331-340. Boelhouwers, J. C., 1988: An interpretation of valley asymmetry in the Natal Drakensberg, South Africa, South African Journal of Science, 84, 913-916. Boelhouwers, J. C., 1991a: Present day periglacial activity in the Natal Drakensberg, Southern Africa: a short review, Permafrost and Periglacial Processes, 2, 5- 12. Boelhouwers, J. C., 1991b: Periglacial evidence from the western Cape mountains, South Africa: a progress report, Permafrost and Periglacial Processes, 2, 13- 20. Boelhouwers, J. C., 1994: Periglacial landforms at Giant?s Castle, Natal Drakensberg, South Africa, Permafrost and Periglacial Processes, 5, 129-136. Boelhouwers, J. C., 1995: Present-day soil frost activity at Mt Superior, Hex River mountains, South Africa, Zeitschrift f?r Geomorphologie, 39, 237-248. Boelhouwers, J. C., 1999: Block deposits in southern Africa and their significance to periglacial autochthonous blockfield development, Polar Geography, 23 (1), 12-22. - 201 - Boelhouwers, J. C., 2003: Quaternary slope development in the Lesotho highlands: review and alternative model, South African Journal of Science, 99, 44-46. Boelhouwers, J. C., Holness, S., Meiklejohn, K. I. and Sumner, P. D., 2002: Observations on a blockstream in the vicinity of Sani Pass, Lesotho highlands, southern Africa, Permafrost and Periglacial Processes, 13, 251-257. Boelhouwers, J. C., and Meiklejohn, K. I., 2002: Quaternary periglacial geomorphology of southern Africa: review and synthesis, South African Journal of Science, 98, 47-55. Boelhouwers, J. C., and Sumner, P. D., 2003: The palaeoenvironmental significance of southern African blockfields and blockstreams, In: Phillips, M., Springman, S. M., and Arenson, L. U., (eds.), Permafrost - ICOP 2003, Swets and Zeitlinger, Lissse, 73-78. Borchert, G., and Sanger, H., 1981: Research findings of a Pleistocene glaciation of the Cape mountain-ridge in South Africa, Zeitschrift f?r Geomorphologie N. F., 25 (2), 222-224. Brown, I., and Ward, R., 1996: The influence of topography on snowpatch distribution in southern Iceland: a new hypothesis for glacier formation?, Geografiska Annaler, 78 (A), 197-207. Brown, R., 1998: El Ni?o and North American snow-cover. In: Proceedings, 55th Eastern Snow Conference, Jackson, New Hampshire, June, 165-172. Available at [http://www1.tor.ec.gc.ca/crysys/esc]. Budin, G. R., 1985: Interannual variability of Australian snowfall, Australian Meteorological Magazine, 33 (3), 145-159. - 202 - Butzer, K. W., and Helgren, D. M., 1972: Late Cenozoic evolution of the Cape coast between Knysna and Cape St. Francis, South Africa, Quaternary Research, 2, 143-169. Butzer, K. W., 1973: Pleistocene ?periglacial? phenomena in southern Africa, Boreas, 2, 1-11. Carbutt, C., and Edwards, T. J., 2004: The flora of the Drakensberg Alpine Centre, Edinburgh Journal of Botany, 60 (3) 581-607. Carroll, T. R., 1990: Operational airborne and satellite snow cover products of the National Operational Hydrological Remote Sensing Center. In: Proceedings of the forty-seventh annual Eastern Snow Conference, June 7-8, 1990, Bangor, Maine. Chinn, T, J., 2001: Distribution of perennial snow and ice water resources in New Zealand, Journal of Hydrology (NZ), 40 (2), 139-187. Christiansen, H. H., 1998: Nivation forms and processes in unconsolidated sediments, NE Greenland, Earth Surface Processes and Landforms, 23, 751-760. Cline, D. W., Bales, R. C., Dozier, J., 1998: Estimating the spatial distribution of snow in mountain basins using remote sensing and energy balance modelling, Water Resources Research, 34 (5), 1275-1285. Costin, A. B., and Polach, H. A., 1973: Age and significance of slope deposits, Black Mountain, Canberra, Australian Journal of Soil Research, 11 (1), 13-25. Czudek, T., 1995: Cryoplanation terraces ? a brief review and some remarks, Geografiska Annaler, 77 (A), 95-102. Dardis, G. F., and Granger, J. E., 1986: Contemporary periglacial phenomena in the Natal Drakensberg, South Africa, Palaeoecology of Africa, 17, 89-93. - 203 - Derbyshire, E., 1971: A synoptic approach to the atmospheric circulation of the last glacial maximum in southeastern Australia, Palaeogeography, Palaeoclimatology, Palaeoecology, 10 (2-3), 103-124. Derbyshire, E., and Evans, I. S., 1976: The climatic factor in cirque variation. In: Derbyshire, E., (ed.): Geomorphology and Climate, Wiley, London, 242-251. Derksen, C., and LeDrew, E., 2000: Variability and change in terrestrial snow cover: data acquisition and links to the atmosphere, Progress in Physical Geography, 24 (4), 469-498. Derwent, S., Porter, R., and Sandwith, T., 2001: Maloti-Drakensberg Transfrontier Conservation and Development Programme, Ezemvelo KwaZulu-Natal Wildlife, Pietermaritzburg. Dong, J., Walker, J. P., and Houser, P. R., 2005: Factors affecting remotely sensed snow water equivalent uncertainty, Remote Sensing of Environment, 97, 69 ? 82. Dozier, J. 1989: Spectral signature of alpine snow-cover from Landsat Thematic Mapper, Remote Sensing of Environment, 28, 9-22. Dozier, J., 1991: Remote sensing of alpine snow cover: invisible and near-infrared wavelengths, Proceedings from the Centre for Snow Science at Alta 1991 Symposium, Available at [http://www.avalanche.org/?moonstone/cssa91?.htm] from the Westwide Avalanche Network. Dunlevey, J. N., Ramluckan, V. R., and Mitchell, A. A., 1993: Secondary mineral zonation in the Drakensberg Basalt Formation, South Africa, South African Journal of Geology, 96, 215-220. Dyer, T. G. J., and Marker, M. E., 1979: On some aspects of Lesotho hollows, Zeitschrift f?r Geomorphologie, 23 (3), 256-270. - 204 - Eastman, R. J., 1991: Change and time series analysis techniques: a review, In: Eastman, R. J., and McKendry, J., (eds.), Change and time series analysis techniques, Idrisi Software Manual, Clark Labs, Worcester, Maine. Ellis, A. W., and Paul, J. R., 2001: Temporal patterns of diurnal snowmelt across the northern Great Plains of the United States, Physical Geography, 22, 1, 59-78. Embleton, C., and King, C. A. M., 1968: Glacial and Periglacial Geomorphology, St. Martin?s Press, New York, pp 608. Embleton, C., and King, C. A. M., 1975: Periglacial Geomorphology, Edward Arnold, London. Etienne, S., 1999: The influence of topography on snowpatch distribution in southern Iceland: a new hypothesis for glacier formation?: a comment, Geografiska Annaler, 81 (A), 101-104. Evans, I. S., 1977: World-wide variations in the direction and concentration of cirque and glacier aspects, Geografiska Annaler, 59 (A), 151-175. Evans, I. S., 2006: Local aspect asymmetry of mountain glaciation: a global survey of consistency of favoured directions for glacier numbers and altitudes, Geomorphology, 73, 166-184. Evans, I. S., and Cox, N., 1974: Geomorphometry and the operational definition of cirques, Area, 6 (2), 150-153. Every, H., 2002: Slope deposits as possible evidence for quaternary glaciation in the High Drakensberg, BSc (Hons) Geography paper, School of Geography, Archaeology and Environmental Studies, University of the Witwatersrand. Fitzharris, B. B., Lawson, W., and Owens, I. F., 1999: Research on glaciers and snow in New Zealand, Progress in Physical Geography, 23 (4), 469-500. - 205 - Fitzharris, B. B., and Owens, I. F., 1984: Avalanche tarns, Journal of Glaciology, 30, 308-312. Fitzpatrick, R. W., 1978: Periglacial soils with fossil permafrost horizons in southern Africa, Annals of the Natal Museum, 23, 475-484. French, H. M., 1996: The periglacial environment, 2nd edition, Longman, Harlow, 341pp. Freiman, M. T., D?Abreton, P. C., Piketh, S. J., 1998: Regional airflow over the southern Drakensberg mountains of South Africa, South African Journal of Science, 94, 561-566. Fukai, K., 2003: Investigating mountain permafrost distribution by ground temperature measurements in the Tateyama Mountains, the northern Japanese Alps, central Japan, Zeitschrift f?r Geomorphologie, 130, 179-193. Gao, J., and Liu, Y., 2001: Applications of remote sensing, GIS and GPS in glaciology: a review, Progress in Physical Geography, 25, 4, 520-540. Garc?a-Ruiz, J. M., G?mez-Villar, A., Ortigosa, L., and Mart?-Bono, C., 2000: Morphometry of glacial cirques in the Central Spanish Pyrenees, Geografiska Annaler, 82 A (4), 433-442. Garland, G. G., 1979: Asymmetry in Drakensberg valleys: an alternative point of view, South African Journal of Science, 75, 138-139. Garland, G. G., 1987: Erosion risk from the footpaths and vegetation burning in the central Drakensberg, Natal Town and Regional Planning, 20, Pietermaritzburg, pp 74. Gil?ad, D., and Bonne, J., 1990: The snowmelt of Mt. Hermon and its contribution to the sources of the Jordan River, Journal of Hydrology, 114 (1), 1-15. - 206 - Google, 2006: Google Earth satellite map program and imagery, Available at [http://earth.google.com] from Google Inc., Mountain View, California. Grab, S. W., 1994: Thufur in the Mohlesi Valley, Lesotho, Southern Africa, Permafrost and Periglacial Processes, 5, 111-118. Grab, S. W., 1996a: Debris deposits in the High Drakensberg, South Africa: possible indicators for plateau, niche and cirque glaciation, Zeitschrift f?r Geomorphologie N.F. Supplement, 103, 389-403. Grab, S. W., 1996b: A note on the morphology of miniature sorted stripes at Mafadi Summit, High Drakensberg, The South African Geographical Journal, 78, 59- 63. Grab, S. W., 1997a: Thermal regime for a thufa apex and its adjoining depression, Mashai Valley, Lesotho, Permafrost and Periglacial Processes, 8, 437-445. Grab, S. W., 1997b: Annually reforming miniature sorted patterned ground in the high Drakensberg, Lesotho, southern Africa, Earth Surface Processes and Landforms, 22, 733-745. Grab, S.W., 1997c: Analysis and characteristics of high altitude air temperature data from northern Lesotho: implications for cryogenic occurrences, Geo?koplus, 4, 109-118. Grab, S. W., 1998a: Non-sorted patterned ground in the High Drakensberg, southern Africa: some new data, The Geographical Journal, 164 (1), 19-31. Grab, S. W., 1998b: Periglacial research in Africa: past, present and future, Progress in Physical Geography, 22 (3), 375-384. - 207 - Grab, S. W., 1999a: Block and debris deposits in the High Drakensberg, Lesotho, Southern Africa: implications for high altitude slope processes, Geografiska Annaler, 81 A (1), 1-16. Grab, S. W., 1999b: A pilot study on needle ice induced stream-bank erosion in the Mashai valley, Lesotho Highlands, South African Geographical Journal, 81 (3), 126-134. Grab, S. W., 2000a: Periglacial features. In: Partridge, T. C., and Maud, R. R. (eds.), The Cenozoic of southern Africa, Oxford University Press, Oxford, 207-217. Grab, S. W., 2000b: Stone banked lobes and environmental implications, high Drakensberg, southern Africa, Permafrost and Periglacial Processes, 11, 177- 187. Grab, S. W., 2002a: Characteristics and palaeoenvironmental significance of relict sorted pattered ground, Drakensberg plateau, southern Africa, Quaternary Science Reviews, 21, 1729-1744. Grab, S. W., 2002b: A note on needle-ice mound formation in the high Drakensberg, southern Africa, Permafrost and Periglacial Processes, 13, 315-318. Grab. S. W., 2002c: Turf exfoliation in the high Drakensberg, southern Africa, Geografiska Annaler, 84A (1), 39-50. Grab, S. W., 2002d: Glacial and periglacial phenomena in Ethiopia: a review, Permafrost and Periglacial Processes, 13, 71-76. Grab, S. W., 2004: Thermal regime for a sorted circle and stone-banked lobe, Drakensberg, southern Africa, Zeitschrift f?r Geomorphologie N. F., 48, 4, 501-518. - 208 - Grab, S. W., and Hall, K, 1996: North-facing hollows in the Lesotho / Drakensberg mountains: hypothetical palaeoenvironmental reconstructions? South African Journal of Science, 92, 183-184. Grab, S. W., and Morris, C., 1999: Soil and water resources issues of the eastern alpine belt wetlands in Lesotho. In: Hurni, H., and Ramamonjisoa, J., (eds.), African Mountain Development in a Changing World, African Mountains Association, African Highlands Initiative, and United Nations University, 207- 219. Grab, S. W., Boelhouwers, J. C., Hall, K., Meiklejohn, K. I., and Sumner, P. D. (eds.), 1999: Quaternary Periglacial Phenomena in the Sani Pass Area, Field Guide to the XV INQUA Conference, Durban. Grab, S. W., and N?sser, M., 2001: Towards an integrated research approach for the Drakensberg and Lesotho mountain environments: a case study from the Sani plateau region, South African Geographical Journal, 83 (1), 64-68. Grab, S. W., and Simpson, A., 2000: Climatic and environmental impacts of cold fronts over KwaZulu-Natal and the adjacent interior of southern Africa, South African Journal of Science, 96, 602-608. Grab, S. W., Van Zyl, C., and Mulder, N., 2005: Controls on basalt terrace formation in the eastern Lesotho highlands, Geomorphology, 67, 3-4, 473-485. Graf, W. L., 1976: Cirques as glacier locations, Arctic and Alpine Research, 8 (1), 79- 90. Granger, J. E., and Schulze, R. E., 1977: Incoming solar radiation patterns and vegetation response: examples from the Natal Drakensberg, Vegetatio, 35 (1), 47-54. - 209 - Grobbelaar, J. U., and Stegmann, P., 1987: Limnological characteristics, water quality and conservation measures of a high altitude bog and rivers in the Maluti Mountains, Lesotho, Water SA, 13 (3), 151-158. Grudd, H., 1989: Small glaciers as sensitive indicators of climatic fluctuations, Geografiska Annaler, 72 A (1), 119-123. Hall, D. K., Foster, J. L., Chien, J. Y. L., and Riggs, G. A., 1995a: Determination of actual snow-covered area using Landsat TM and digital elevation model data in Glacier National Park, Montana, Polar Record, 41 (177), 191-198. Hall, D. K., Riggs, G. A., and Salomonson, V. V., 1995b: Development of methods for mapping global snow cover using Moderate Resolution Image Spectroradiometer (MODIS) data, Remote Sensing of Environment, 54, 137- 140. Hall, D. K., Foster, J. L., Verbyla, D. L., Klein, A. G., and Benson, C. S., 1998: Assessment of snow-cover mapping accuracy in a variety of vegetation-cover densities in central Alaska, Remote Sensing of Environment, 66, 129-137. Hall, D. K., Chang, A. T. C., and Foster, J. L., 1999: BOREAS RSS-08 snow maps derived from Landsat TM imagery, Available at [http://www.eosdis.ornl.gov/] from the ORML Distributed Active Archive Center, Oak Ridge National Laboratory, Oak Ridge, Tennessee, U.S.A. Hall, K., 1991: The allocation of freeze-thaw weathering in geocryological studies: a critical comment, South African Geographical Journal, 73, 10-13. Hall, K., 1992: A discussion on the need for greater rigour in southern African cryogenic studies, South African Geographical Journal, 74, 69-71. Hall, K., 1994: Cutback in the Natal Drakensberg escarpment: a hypothesis on their origin, South African Journal of Science, 90, 263-164. - 210 - Hall, K., 1995: Response to: Sumner, P., Cutbacks in the Natal Drakensberg Escarpment: comments on a hypothesis on their origin, South African Journal of Science, 91, 286. Hall, K. and Meiklejohn, K. I., 1997: Some observations regarding protalus ramparts, Permafrost and Periglacial Processes, 8, 245-249. Hanvey, P. M., and Marker, M. E., 1992: Present day periglacial microforms in the Lesotho highlands: implications for present and past climate conditions, Permafrost and Periglacial Processes, 3, 35-361. Harper, G., 1969: Periglacial evidence in South Africa during the Pleistocene epoch, Palaeoecology of Africa, 4, 71-91. Harris, S. A., French, H. M., Heginbottom, J. A., Johnston, G. H., Ladanyi, B., Sego, D. C., and Van Everdingen, R. O., 1988: Glossary of permafrost and related ground-ice terms, Associate Committee on Geotechnical Research, National Research Council of Canada, Ottawa. Hastenrath, S., and Wilkinson, J., 1973: A contribution to the periglacial morphology of Lesotho, southern Africa, Biuletyn Periglacjalny, 22, 157-167. Hecht, J., 1997: Tibetan ice forces climate rethink, New Scientist, 2088, 17. Hennessy, K., Whetton, P., Smith, I., Bathols, J., Hutchinson, M., and Sharples, J., 2003: The impact of climate change on snow conditions in mainland Australia, CSIRO Atmospheric Research, Aspendale, Victoria, Australia. Higgins, C. G., 1982: Grazing-step terracettes and their significance, Zeitschrift f?r Geomorphologie N.F., 26, 459-472. IOL, 2004: Durban school hikers attacked in tents, Available at [http://www.iol.co.za/index.php?set_id=&click_id=79&art_id=vn2004092206 2634190C977392] from Independent Online, South Africa. - 211 - Jomelli, V., and Bertran, P., 2001: Wet snow avalanche deposits in the French Alps: structure and sedimentology, Geografiska Annaler, 83 A (1-2), 15-28. Klimaszewski, M., 1993: Conditions of the Pleistocene glaciation of mountainous regions, Zeitschrift f?r Geomorphologie N.F., 37 (1), 1-18. Killick, D. J. B., 1963: An account of the plant ecology of the Cathedral Peak area of the Natal Drakensberg, Memoirs of the Botanical Survey of South Africa, No. 34. Killick, D. J. B., 1978a: The Afro-alpine region. In: Werger, M. J. (ed.), Biogeography and ecology of Southern Africa, W. Junk, The Hague, 515-560. Killick, D. J. B., 1978b: Notes on the vegetation of the Sani Pass area of the southern Drakensberg, Bothalia, 12 (3), 537-542. Killick, D. J. B., 1978c: Further data on the climate of the alpine vegetation belt of eastern Lesotho, Bothalia, 12, 567-572. Killick, D. J. B., 1990: A field guide to the flora of the Natal Drakensberg, Jonathan Bull and Ad. Donker Publishers, Johannesburg. King, L. C., 1974: Aspects of the High Drakensberg: illustrated by air photographs, South African Geographical Journal, 56 (2), 128-136. King, L. C., and King, L. A., 1959: A reappraisal of the Natal monocline, South African Geographical Journal, 41, 15-30. Klein, A. G., and Isacks, B. L., 1998: Alpine glacial geomorphological studies in the central Andes using Landsat Thematic Mapper images, Glacial geology and geomorphology, rp01/1998. Available at [http://ggg.qub.ac.uk/papers/full/1998/rp011998/rp01.htm]. - 212 - Klein, A. G., and Isacks, B. L., 1999: Spectral mixture analysis of Landsat thematic mapper images applied to the detection of the transient snowline on tropical Andean glaciers, Global and Planetary Change, 22, 139-154. Klug, J. R., De Villiers, J. M., Tainton, N. M., and Matela, L. S., 1989: Final report of the terrain analysis project for the Drakensberg / Maluti Mountain Catchment Conservation Programme, Natal Parks Board, Pietermaritzburg. Lewis, C.A., 1988a: Periglacial features in southern Africa: an assessment, Palaeoecology of Africa, 19, 357-370. Lewis, C. A., 1988b: Periglacial landforms. In: Moon, B. P., and Dardis, G. F., (eds.), The Geomorphology of Southern Africa, Southern, Johannesburg, pp. 103-119. Lewis, C. A., 1994: Protalus ramparts and the altitude of the local equilibrium line during the Last Glacial Stage in Bokspruit, East Cape Drakensberg, South Africa, Geografiska Annaler, 76A, 37-48. Lewis, C. A., and Dardis, G. F., 1985: Periglacial ice-wedge casts and head deposits at Dynevor Park, Barkly Pass area, north-eastern Cape Province, South African Journal of Science, 81, 673-677. Lewis, C. A., and Hanvey, P. M., 1993: The remains of rock glaciers in Bottelnek, East Cape Drakensberg, South Africa, Transactions of the Royal Society of South Africa, 48, 265-289. Lewis, C. A., and Illgner, P. A., 2001: Late Quaternary glaciation in southern Africa: moraine ridges and glacial deposits at Mount Enterprise in the Drakensberg of the Eastern Cape Province, South Africa, Journal of Quaternary Science, 16 (4), 365-374. Liebling, R. S., and Scherp, H. S., 1981: Structure control on cirque-like features in south-east New York State, U.S.A., Journal of Glaciology, 27, 95, 198-200. - 213 - Linton, D. L., 1969: Evidence of Pleistocene cryonival phenomena in South Africa, Palaeoecology of Africa, 5, 71-89. Liston, G. E., 1999: Interrelationships among snow distribution, snowmelt, and snow cover depletion: implications for atmospheric, hydrologic and ecologic modeling, Journal of Applied Meteorology, 38, 1474-1487. Lozinsky, W., 1910: Die periglaziale fazies der mechanischen verwitterung, 11th International Geological Congress, Stockholm, 1039-1053. Lynch, C. D., and Watson, J. P., 1992: The distribution and ecology of Otomys sloggetti (Mammalia: Rodentia) with notes on its taxonomy, Navorsinge van die Nasionale Museum, Bloemfontein, 8, 141-157. Makhoalibe, S., 1999: Management of Water Resources in the Maloti / Drakensberg Mountains of Lesotho, Ambio, 28 (5), 460-461. Marcus, M. G., French, H. M., and St?blein, G., 1992: Recent studies on glacial and periglacial geomorphology ? preface, Zeitschrift f?r Geomorphologie, Suppl Bd 86, v-vii. Marker, M. E., 1991: The evidence for cirque glaciation in Lesotho, Permafrost and Periglacial Processes, 2, 21-30. Marker, M. E., 1994: Sedimentary sequences at Sani Top, Lesotho highlands, southern Africa, The Holocene, 4, 406-412. Marker, M. E., 1995: Late Quaternary environmental implications from sedimentary sequences at two high altitude Lesotho sites, South African Journal of Science, 91, 294-298. Marker, M. E., 1998: Cenozoic climate change in southern Africa: the evidence from geomorphology, 1967-1996, a personal view, South African Geographical Journal, 80 (1), 1-8. - 214 - Marker, M. E., and Whittington G., 1971: Observations on some valley forms and deposits in the Sani Pass area, Lesotho, South African Geographical Journal, 53, 96-99. Meakins, R. H., and Duckett, J. D., 1993: Vanishing bogs of the mountain kingdom, Veld and Flora, 79 (2), 49-51. Meierding, T. C., 1982: Late Pleistocene glacial equilibrium-line altitudes in the Colorado Front Range: a comparison of methods, Quaternary Research, 18, 289-310. Meiklejohn, K. I., 1992: Some aspects of valley asymmetry in the high Drakensberg, South African Geographical Journal, 74, 49-53. Meiklejohn, K. I., 1994: Valley asymmetry on south-eastern Alexander Island, Antarctica, and valley forms in the high Drakensberg, southern Africa, South African Geographical Journal, 76 (2), 68-72. Metternicht, G. I., and Zinek, J. A., 1998: Evaluating the information content of JERS-1 SAR and Landsat TM data for discrimination of soil erosion features, Journal of Photogrammetry and Remote Sensing, 53, 143-153. Mitchell, A. A., Ramluckan, V. R., Dunlevey, J. N., and Eglington, B. M., 1996: The basalt stratigraphy of the Sani Pass, Kwazulu/Natal Drakensberg, South African Journal of Geology, 99 (3), 251-262. Mills, S. C., and Grab, S. W., 2005: Debris ridges along the southern Drakensberg escarpment as evidence for Quaternary glaciation in southern Africa, Quaternary International, 129 (1), 61-73. Molotch, N. P., Painter, T. H., Colee, M. T., Rosenthal, C. W., Dozier, J., and Bales, R. C., 2001: Analysis of the spatial variability of snow cover depletion in an - 215 - alpine watershed, Tokopah Basin, Sierra Nevada, California, U.S.A., Proceedings of the Western Snow Conference, Sun Valley, Idaho, U.S.A. Morris, C. D., Tainton, N. M., and Boleme, S., 1993: Classification of the eastern alpine vegetation of Lesotho, African Journal of Range and Forage Science, 10 (1), 47-53. Mulder, N., and Grab, S. W., 2002: Remote sensing for snow cover analysis along the Drakensberg escarpment, South African Journal of Science, 98, 213-217. Mulders, M. A., 2001: Advances in the application of remote sensing and GIS for surveying mountainous land , International Journal of Applied Earth Observation and Geoinformation, 3 (1), 3-10. NASA, 2006: The Landsat Programme: a Landsat timeline, Available at [http://landsat.gsfc.nasa.gov/about/timeline.html] from the National Aeronautics and Space Administration. Nel, W., and Sumner, P. D., 2005: First rainfall data from the KZN Drakensberg escarpment edge (2002 and 2003), Water SA, 31 (3), 399-402. Nelson, F. E., 1989: Cryoplanation terraces: periglacial cirque analogs, Geografiska Annaler, 71 (A), 31-41. Nelson, F. E., 1998: Cryoplanation terrace orientation in Alaska, Geografiska Annaler, 80 (A), 135-151. N?sser, M., and Grab, S. W., 2002: Land degradation and soil erosion in the eastern highlands of Lesotho, southern Africa, Die Erde, 133, 291-311. Ohmura, A., Wild, M., and Bengtsson, L., 1996: A possible change in mass balance of Greenland and Antarctic ice sheets in the coming century, Journal of Climate, 9 (9), 2124-2135. - 216 - Palacios, D., and S?nchez-Colomer, M. G., 1997: The influence of geomorphologic heritage on present nival erosion: Pe?alara, Spain, Geografiska Annaler, 79 A (1-2), 25-40. Palacios, D., Andr?s, N., and Luengo, E., 2003: Distribution and effectiveness of nivation in Mediterranean mountains: Pe?alara (Spain), Geomorphology, 54, 157-178. Partridge, T. C., 1997a: Evolution of landscapes. In: Cowling, R. M., Richardson, D. M., and Pierce, S. M. (eds.), Vegetation of southern Africa, Cambridge University Press, Cambridge, 5-20. Partridge, T. C., 1997b: Cainozoic environmental change in southern Africa, with special emphasis on the last 200 000 years, Progress in Physical Geography, 21, 3-22. Partridge, T. C., and Maud, R. R., 1987: Geomorphic evolution of southern Africa since the Mesozoic, South African Journal of Geology, 90, 179-208. Peterson, J. A., Chandra, S., and Lundberg, C., 2004: Landforms from the Quaternary glaciation of Papua New Guinea: an overview of ice extent during the Late Glacial Maximum. In: Ehlers, J. and Gibbard, P. L., (eds.), Quaternary Glaciations - Extent and Chronology, Part III: South America, Asia, Africa, Australia, Antarctica. Amsterdam, Elsevier. Petit J. R., Jouzel J., Raynaud D., Barkov N. I., Barnola J. M., Basile I., Bender M., Chappellaz J., Davis J., Delaygue G., Delmotte M., Kotlyakov V. M., Legrand M., Lipenkov V., Lorius C., P?pin L., Ritz C., Saltzman E., Stievenard M., 1999, Climate and Atmospheric History of the Past 420,000 years from the Vostok Ice Core, Antarctica, Nature, 399, 429-436. Pickles, J., 1985: Landscape appreciation and preferences in the Natal Drakensberg, Natal Town and Regional Planning Commission Supplementary Report, 17, University of Natal, Pietermaritzburg. - 217 - Piesbergen, J., Holecz, F., and Haefner, H., 1997: Multi-source snow cover monitoring in eastern Switzerland, Proceedings from the Third ERS-1 Symposium in Florence, Italy, Available at [http://www.geo.unizh.ch/rsl/projects/data_fusion/florence97/] from the Remote Sensing Laboratories, Switzerland. Pinkham, D. M., 2000: The snow and sea ice mapping algorithms, Available at [http://snowmelt.gsfc.nasa.gov/MODIS_Snow/algor.html] from the Hydrological Sciences Branch, NASA / Goddard Space Flight Center, Greenbelt, Maryland, U.S.A. Porter, S. C., 2001: Snowline depression in the tropics during the last glaciation, Quaternary Science Reviews, 20, 1067-1091. Preston-Whyte, R. A., 1971: Diurnal wind variations over the Drakensberg escarpment at Sani Pass, South African Geographical Journal, 53, 89-92. Preston-Whyte, R. A., Diab, R. D., and Washington, R., 1991: Diurnal variation of rainfall events by synoptic type in Natal, South African Geographical Journal, 73, 22-28. Preston-Whyte, R. A., and Tyson, P. D., 1997: The Atmosphere and Weather of Southern Africa, Oxford University Press, Cape Town. Raczkowska, Z., 1990: Observations on nivation and its geomorphological effect in mountains at high latitude (with Mt. Njulla Massif in northern Sweden as example), Pirineos, 136, 19-32. Rae, K., 2001: First Winter Cold: snowfalls in May 2001, Available at [http://www.weathersa.co.za/Pressroom/2001/2001FirstWinterCold.jsp] from the South African Weather Bureau. Rango, - 218 - Rango, A., Gomez-Landesa, E., Bleiweiss, M., 2002: Comparative satellite capabilities for remote sensing of snow cover in the Rio Grande Basin, Proceedings of the Western Snow Conference, 70th Annual Meeting 2002, 21- 26. Rapp, A., 1984: Nivation hollows and glacial cirques in S?der?sen, Scania, south Sweden, Geografiska Annaler, 66A, 11-28. Rosenthal, W., and Dozier, J., 1996: Automated mapping of montane snow cover at subpixel resolution from the Landsat Thematic Mapper, Water Resources Research, 32 (1), 115-130. Sabins, F. F., 1986: Remote Sensing: Principles and Interpretations (2nd edition), W. H. Freeman and Co., New York. San Diego Natural History Museum, 2006: Geologic Timeline, Available at [http://www.sdnhm.org/exhibits/mystery/fg_timeline.html] from the San Diego Natural History Museum. Sancho, L. G., Palacios, D., De Marcos, J., Valladares, F., 2001: Geomorphological significance of lichen colonization in a present snow hollow: Hoya del Cuchillar de las Navajas, Seirra de Gredos (Spain), Catena, 43, 323-340. S?nger, H., 1987: In: Matheis, G., and Schandelmeier, H., (eds.), Current Research in African Earth Sciences, Balkema, Rotterdam, 447. S?nger, H., 1988: Recent periglacial morphodynamics and Pleistocene glaciation of the Western Cape folded belt, South Africa. In: Dardis, G. F., and Moon, B.P., (eds.), Geomorphological Studies in Southern Africa, Balkema, Rotterdam, 383-388. Schaper, J., Martinec, J., and Seidel, K., 1999: Distributed mapping of snow and glaciers for improved runoff modelling, Hydrological Processes, 13, 2023- 2031. - 219 - Schmitz, G., and Rooyani, F., 1987: Lesotho: Geology, Geomorphology, Soils, National University of Lesotho, Roma. Schowengerdt, R. A., 2006: Remote Sensing: Models and Methods for Image Processing, Third Edition, Academic Press, Burlington, MA. Schulze, B. R., 1965: Climate of South Africa, part 8, General Survey. Weather Bureau, Pretoria. Schulze, R. E., 1979: Hydrology and water resources of the Drakensberg, Natal Town and Regional Planning Reports, 42, University of Natal Pietermaritzburg. Schwabe, C. A., 1989: The assessment, planning and management of wetlands in the Maluti / Drakensberg mountain catchments, Institute of Natural Resources Investigation Report, 38, Pietermaritzburg. Schwabe, C. A., 1995: Alpine mires of the eastern highlands of Lesotho. In: Cowan, G. I. (ed.), Wetlands of South Africa, Department of Environmental Affairs and Tourism, Pretoria, 292. Sene, K. J., Jones, D. A., Meigh, J. R., and Farquharson, F. A. K., 1998: Rainfall and flow variations in the Lesotho Highlands, International Journal of Climatology, 18, 329-345. Shakesby, R. A., 1997: Pronival (protalus) ramparts: a review of forms, processes, diagnostic criteria and palaeoenvironmental implications, Progress in Physical Geography, 21, 394-418. Shakesby, R. A., Matthews, J. A., McEwan, L. J., and Berrisford, M. S., 1999: Snow- push processes in pronival (protalus) rampart formation: geomorphological evidence from Sm?rbotn, Romsdalsalpane, southern Norway, Geografiska Annaler, 81 (A), 31-45. - 220 - Singh, A., 1989: Digital change detection techniques using remotely-sensed data, International Journal of Remote Sensing, 10 (6), 989-1003. Slater, A., 1995: Snow in Australia: Trends and Variability, Unpublished undergraduate thesis, Macquarie University, New South Wales, Australia, available at [http://cires.Colorado.edu/~aslater/snowy.html]. Sokol, J., Pultz, T. J., and Walker, A. E., 1999: Passive and active airborne microwave remote sensing of snow cover, Fourth International Airborne Remote Sensing Conference and Exhibition / 21st Canadian Symposium on Remote Sensing, Ottawa, Canada. South African Weather Bureau (SAWB), 1989-2004: Daily weather bulletin for the months of April to October, 1989 to 2004. Sparrow, G. W. A., 1964: Pleistocene periglacial landforms in the Southern Hemisphere, South African Journal of Science, 60, 143-147. Sparrow, G. W. A., 1967a: Pleistocene periglacial topography in southern Africa, Journal of Glaciology, 6, 551-559. Sparrow, G. W. A., 1967b: South African cirques and ar?tes, Journal of Geography, 2, 9-11. Sparrow, G. W. A., 1971: Some Pleistocene studies in southern Africa, Journal for Geography, 3, 809-814. Sparrow, G. W. A., 1974: Non-glacial cirque formation in southern Africa, Boreas, 3, 61-68. Str?mquist, L., 1985: Geomorphic impact of snowmelt on slope erosion and sediment production, Zeitschrift f?r Geomorphologie N.F., 29 (2), 129-138. - 221 - Sumner, P. D., 1995: Cutbacks in the Natal Drakensberg Escarpment: comments on an hypothesis on their origin, South African Journal of Science, 91, 285-286. Sumner, P. D., 2000: Relict sorted ground near Sani Pass ? evidence for Pleistocene deep freeze, Programme and Abstracts, SAAG 2000, 6th Biennial Conference, Hammanskraal, 9-12 July 2000. Sumner, P. D., 2003: A contemporary winter ground thermal profile in the Lesotho highlands and implications for active and relict soil frost phenomena, Earth Surface Processes and Landform, 28, 1451-1458. Sumner, P. D., 2004a: Geomorphic and climatic implications of relict openwork block accumulations near Thabana-Ntlenyana, Lesotho, Geografiska Annaler, 86A, 289-302. Sumner, P. D., 2004b: Relict sorted patterned ground in Lesotho, Permafrost and Periglacial Processes, 15, 89-93. Sumner, P. D., and Meiklejohn, K. I., 2000: Landscape evolution in a changing environment. In: Fox, R., and Rowntree, K. (eds.): The Geography of South Africa in a Changing World, Oxford, Cape Town, 304-325. Sunday Times, 2004: Schoolgirls attacked in the Drakensberg, Available at [http://www.sundaytimes.co.za/zones/sundaytimesnew/topstories/topstories10 95762477.aspx] from the Sunday Times, South Africa. Talma, A. S., and Vogel, J. C., 1992: Late Quaternary paleotemperatures derived from a speleothem from the Cango Caves, Cape Province, South Africa, Quaternary Research, 37, 203-213. Thorn, C. E., 1976: Quantitative evaluation of nivation in the Colorado Front Range, Geological Society of America Bulletin, 87, 1169-1178. - 222 - Thorn, C. E., 1978: The geomorphic role of snow, Annals of the Association of American Geographers, 68 (3), 414-425. Thorn, C. E., 1979: Ground temperatures and surficial transport in colluvium during snowpatch meltout; Colorado Front Range, Arctic and Alpine Research, 11 (1), 41-52. Thorn, C. E., 1992: Periglacial geomorphology: what, where, when?, In: Dixon, J. C., and Abrahams, A. D. (eds.), Periglacial Geomorphology, Wiley, New York, 1-30. Thorn, C. E., and Hall, K., 2002: Nivation and cryoplanation: the case for scrutiny and integration, Progress in Physical Geography, 26 (4), 533-550. Trenhaile, A. S., 1977: Cirque elevation and Pleistocene snowlines, Zeitschrift f?r Geomorphologie N. F., 21 (4), 445-459. Troll, C., 1944: Strukturb?den, solifluktion und frostklimate der Erde, Geologische Rundschau, 34, 545-694. Tyson, P. D., 1999: Atmospheric circulation changes and palaeoclimates of southern Africa, South African Journal of Science, 95, 194-201. Tyson, P. D., Preston-Whyte, R. A., and Schulze, R. E., 1976: The Climate of the Drakensberg, Natal Town and Regional Planning Commission, Pietermaritzburg. Van Rooy, J. L., and Nixon, N., 1990: Mineralogical alteration and durability of Drakensberg basalts, South African Journal of Geology, 93 (5/6), 729-737. Van Rooy, J. L., and Van Schalkwyk, A., 1993: The geology of the Lesotho Highlands Water Project with special reference to durability of construction materials, Journal of African Earth Sciences, 16 (1/2), 181-192. - 223 - Van Steijn, H., Bertran, P., Francou, B., Hetu, B., Texier, J. P., 1995: Models for the genetic and environmental interpretation of stratified slope deposits: review, Permafrost and Periglacial Processes, 6, 125-146. Van Zinderen Bakker, E. M., and Werger, M. G. A., 1974: Environment, vegetation, and phytogeography of the high-altitude bogs of Lesotho, Vegetatio, 29, 37- 49. Vieira, G. T., Mora, C., and Ramos, M., 2003: Ground temperature regimes and geomorphological implications in a Mediterranean mountain (Serra da Estrela, Portugal), Geomorphology, 52 (1-2), 57-72. Vitek, J. D., Giardino, J. R., and Fitzgerald, J. W., 1996: Mapping geomorphology: a journey from paper maps, through computer mapping to GIS and virtual reality, Geomorphology, 16, 233-249. Vogel, J. C., Fuls, A., and Ellis, R. P., 1978: The geographical distribution of Kranz grasses in South Africa, South African Journal of Science, 74, 209-215. Warburton, J. A., and Wetzel, M. A., 1992: Field study of the potential for winter precipitation enhancement in the Australian Snowy Mountains, Atmospheric Research, 28 (3-4), 327-363. Watson, H. K., 1998: Terracettes in the Natal Drakensberg, South Africa. In: Dardis, G. F., and Moon, B. P., (eds.): Geomorphological Studies in Southern Africa, Balkema, Rotterdam. Webster, P. J., and Streten, N. A., 1978: Late Quaternary ice age climates of tropical Australasia, interpretations and reconstructions, Quaternary Research, 10 (3), 279-309. Weinert, H. H., 1961: Climate and weathered Karoo dolerites, Nature, 4786, 32-39. - 224 - Whetton, P. H., Haylock, M. R., and Galloway, R., 1996: Climate change and snow- cover duration in the Australian Alps, Climatic Change, 32 (4), 447-479. William, R. S., and Ferrigno, J. G., 1998: Satellite image atlas of the world ? glaciers of South America, USGA Professional Paper 1386-I. Wilson, P., and Clark, R., 1998: Characteristics and implications of some Loch Lomond Stadial moraine ridges and later landforms, eastern Lake District, northern England, Geological Journal, 33, 73-87. Zietsman, H. L., Vlok, A. C., and Nel, I., 1996: The identification of irrigated land in an intensively cultivated agricultural area in the south-western Cape by means of satellite remote sensing, Report to the Water Research Commission, WRC 440/1/96, Water Research Commission, Pretoria. Appendix 2 - 247 - - APPENDIX 2 GOOGLE EARTH DATA Real colour composite satellite images of the High Drakensberg and Lesotho Highlands were obtained and viewed from various angles through the Google Earth map program (Google, 2005). Various debris ridges were visible on the images, with preferences for hollows on south-facing slopes on mountain ranges. The majority of debris ridges were found below southeast-facing slopes, with late-lying snow cover visible on the slope. The topographic location of the debris ridge, as well as the upper and lower altitudes, horizontal length and maximum width were noted through the use of Google Earth. Other palaeogeomorphological features, such as block streams and late-lying snow cover at the base of south-facing basalt terraces were also observed and their location noted. Appendix 2 - 248 - - Table 1: The upper and lower altitudes, lengths and widths of potential debris ridges in the High Drakensberg. Ridge Region (and part) Upper Altitude (m ASL) Lower Altitude (m ASL) Approximate Length Maximum Width A Leqooa Valley ? northern 3103m 3071m 201m 42m B Leqooa Valley ? northern 3115m 3068m 306m 40m C Leqooa Valley ? northern 3041m 3016m 169m 31m D Leqooa Valley ? northern 3046m 3015m 330m 28m E Leqooa Valley ? northern 2979m 2961m 118m 19m F Leqooa Valley ? northern 3129m 3039m 552m 36m G Leqooa Valley ? western 3168m 3097m 130m 30m H Leqooa Valley ? western 3023m 2973m 220m 23m I Leqooa Valley ? western 3134m 3085m 131m 36m J Leqooa Valley ? western 3209m 3075m 364m 82m K Leqooa Valley ? western 3158m 3009m 678m 24m L Lisaleng ? Seperenkane 2938m 2907m 337m 21m M Lisaleng ? Seperenkane 3097m 2994m 358m 14m N Lisaleng ? Seperenkane 3019m 2988m 98m 14m P Sekhokong ridge 3047m 3087m 58m 14m Q Tsatsa-La-Mangaung (east) 3085m 2988m 389m 40m R Tsatsa-La-Mangaung (west) 2971m 2876m 660m 168m S Nhlangeni Cutback 3046m 2818m 618m 44m T Makheka (Mafadi region) 3100m 3092m 188m 37m U Mafadi Summit region 3080m 3022m 296m 28m V Ndadema Dome 3140m 3019m 246m 23m W Ndadema Dome 3170m 3086m 317m 34m Appendix 2 - 249 - - Figure (1): Overview of the Leqooa region in the southern part of the study area showing locations of debris ridges A to K. Figure (2): Two debris ridges (A and B) in a south-facing hollow in the Leqooa Range. The ridges have been described by Mills and Grab (2005) as moraines. Appendix 2 - 250 - - Figure (3): Leqooa debris ridge (A), showing a displacement of the ridge to the east of the valley axis. Figure (4): Leqooa debris ridge (B). Appendix 2 - 251 - - Figure (5): Leqooa debris ridge (D). Figure (6): Leqooa debris ridge (E). Appendix 2 - 252 - - Figure (7): Leqooa debris ridge (F), a ridge on an infrequent south-west facing slope. Figure (8): Leqooa debris ridge (G) at the base of a niche. Appendix 2 - 253 - - Figure (9): Leqooa debris ridge (H). Unknown ridge-like landforms are located in the upper part of the hollow. Figure (10): Leqooa debris ridges (I) Three parallel ridges are found in a south- facing hollow. Appendix 2 - 254 - - Figure (11): Local topography surrounding Leqooa debris ridge (J). Figure (12): Leqooa debris ridge (J), the longest and widest ridge identified. Appendix 2 - 255 - - Figure (13): Leqooa debris ridge (K). Smaller debris deposits are located higher up the hollow. The ridge slope is west-facing and may be the result of fluvial processes. Figure (14): Lisaleng debris ridge (L), near the Seperenkane ridges (M and N). Appendix 2 - 256 - - Figure (15): Seperenkane debris ridge (M) . Figure (16): Seperenkane debris ridge (N), just east of ridge (M). Appendix 2 - 257 - - Figure (17): Sekhokong debris ridge (P). Figure (18): Eastern Tsatsa-La-Mangaung debris ridge (Q), showing 3 pits (light spots at lower, middle and upper parts of ridge) dug by Mills and Grab (2005). Appendix 2 - 258 - - Figure (19): Tsatsa-La-Mangaung debris ridge (R), further west from Ridge (Q), is a large deposit that appears to have altered the river course in the valley. Figure (20): Block stream to the west of Tsatsa-La-Mangaung region, described by Grab et al. (1999). Appendix 2 - 259 - - Figure (21): Debris ridges in the Nhlangeni cutback described by Grab (1996a). Figure (22): Photo of debris ridges (S) in the Nhlangeni cutback, showing comparative detail to the Google Earth images. Appendix 2 - 260 - - Figure (23): A potential debris ridge (T) in the Makheka region near Mafadi Summit, showing its location across the mouth of a hollow, with the possible alteration of a stream course. Figure (24): Detail of the Makheka debris ridge (T). Appendix 2 - 261 - - Figure (25): A basalt terrace in the Makheka region, showing late-lying snow cover at the base of a basalt terrace. No other snow is recorded near by. Figure (26): Mafadi debris deposit (U). Appendix 2 - 262 - - Figure (27): Ndedema Dome debris deposit (V). Appendix 3 - 263 - - APPENDIX 3 DISCUSSION OF WEATHER DATA IN RELATION TO SNOW COVER A discussion of weather conditions prior to and after snow producing synoptic events and the subsequent data capture of snow cover through Landsat TM and ETM+ sensors (SAWB, 1989-2004). 1) 31 July 1989 A cold front moved over the central parts of South Africa on 18 July 1989, bringing snowfall to the study area in the evening or on the 19 July. No weather data for Underberg are available, however a maximum temperature of 4.3?C for Bethlehem was recorded on 19 July. A milder cold front moved through on 26 July, but probably did not bring snow, implying that 11 days of snow ablation and melt occurred before satellite image data were captured on 31 July. The image shows isolated patches of snow in the four highest regions of the Drakensberg (i.e. the Leqooa range, Thabana-Ntlenyana area, Mafadi Summit area, and the northwest part of the High Drakensberg). 2) 16 August 1989 Snow probably fell on 03 August 1989 in the study area when a broad cold front moved through. Twelve days of warmer conditions left only isolated snow in the Leqooa range, Thabana-Ntlenyana area, Mafadi Summit area, and the northwest part of the High Drakensberg 3) 03 August 1990 Snow probably fell on 28-29 July 1990, as a result of a low pressure situated over the interior. Precipitation (snowfall) associated with northwesterly winds probably resulted in the significant snow cover observed in the northwest part of the study area 4 days later. Bethlehem recorded 4mm of precipitation whilst Underberg remained dry, supporting preferred snowfall in the northwestern parts. The Leqooa, Appendix 3 - 264 - - Thabana-Ntlenyana and Mafadi areas were once again the only other regions where significant areas of snow cover was observed. 4) 19 August 1990 Moist air advecting from the northeast due to a strong high pressure over the southeast coast of South Africa probably led to orographic uplift of moist air, resulting in the snowfall observed on 19 August 1990 in the northwestern part of the study area. The onshore flow conditions, 7 days prior to the data image, did not appear to produce snow elsewhere in the study area. 5) 04 September 1990 Snowfall was noted in weather reports for 29 and 30 August 1990, when a low pressure over the interior allowed for moist onshore airflow towards the High Drakensberg. After 4 days of snow melt and ablation, snow was only recorded on the data images in the Leqooa, Thabana-Ntlenyana and Mafadi regions. Precipitation was noted in Underberg whilst Bethlehem remained dry, supporting a proposition that snowfall was limited predominantly to the southern escarpment region. 6) 22 October 1990 ?Snow showers? were predicted for the High Drakensberg by the SAWB when two cold front systems moved through on 18 and 19 October 1990. Heavy precipitation was recorded in Underberg (26mm and 18mm respectively), whilst none was recorded in Bethlehem. Two days of snow ablation and melt appeared to have had little effect on snow cover by 22 October, when data images show large sections of the High Drakensberg and Lesotho Highlands covered in snow. However, no snow appears north of the Mafadi Summit area, suggesting that the snow bearing cold front may have been pushed eastwards beyond this point due to the weaker westerly wave at this time of year. 7) 19 June 1991 Snow was recorded on 19 June 1991 in the Leqooa, Thabana-Ntlenyana and Mafadi areas. More extensive snow cover was recorded for the northwestern part of the study area. Three cold fronts (7, 14, 18 June) appear to have brought snow. Precipitation fell in Underberg during the first front event, whilst precipitation fell in Appendix 3 - 265 - - Bethlehem during the last front event. The snow cover is probably a result of multiple snow producing events, with the snow at Leqooa, Thabana-Ntlenyana and Mafadi areas probably resulting from the 7 June and / or 14 June cold fronts, whilst the snow in the northwest of the study area is probably a product of the 18 June cold front, hence the continuous and extensive nature of the snow cover in this area. 8) 08 August 1992 The image data of 08 August 1992 show that snow cover is limited to the northwestern part of the study area. The weather data support this, as an upper-air low pressure over the interior brought heavy rain to Bethlehem (32.2mm on 03 August), yet none to Underberg. The 4 days of snowmelt and ablation are noticeable on the northern slopes in the data image. 9) 08 June 1993 Light snow probably fell in the study area on 03 June 1993 as a result of a cold front, some 4 days before the data image was captured. The snow cover area is small (7.9km2), located in the northwestern part of the study area. The size and distribution of the snowfall prior to this period of significant ablation cannot be determined. 10) 24 April 1994 Snow cover is noted on 24 April 1994 in the Leqooa, Thabana-Ntlenyana and Mafadi areas of the High Drakensberg. Two possible origins for the snowfall are found in weather reports. The first is on 21 April, when an upper-air low pressure over the interior of South Africa brought rain to Bethlehem, but very little to Underberg. The second, on 23 April, is a broad cold front. However, the distribution of snow cover along the eastern edge of the escarpment, and the complete lack thereof in the northwestern part of the study area, suggests that the cold front event contributed to the recorded snow. Snow cover appears fairly continuous, particularly near Mafadi Summit, which corresponds to the lack of ablation and snowmelt opportunities. 11) 29 July 1994 A cut-off low with an accompanying cold front on 25 July 1994 resulted in extensive snowfall, observable on the 29 July data image. The cold impact of the Appendix 3 - 266 - - system, which brought significant precipitation to Underberg (33.5mm on 25 July), is reflected by the lack of snow cover depletion after 2 days. 12) 30 August 1994 A cold front moving through 9 days before the 30 August 1994 data image is the probable origin of snow recorded on the south-facing slopes in the southern part of the study area. Temperature depressions seem significant in Bethlehem (a maximum temperature of 11.3?C on 21 August). However a lack of data for Underberg over these dates, particularly the lack of precipitation data, leaves the cause of the snowfall inconclusive. 13) 30 June 1995 A cold front over the south east coast of South Africa on 16 June 1995, linked to an upper air low in the interior of the country, probably resulted in the snow visible on 30 June. Snow cover is limited to the southern portions of the study area, particularly in the Sani Pass, Thabana-Ntlenyana and Giants Castle areas. A smaller area of snow cover is found near Mafadi Summit, but overall, the northern part of the study area is has less snow, suggesting that the front, constrained mainly to below the escarpment, moved eastwards away from South Africa before reaching the northern section. 14) 18 September 1995 There may be two possible origins for snow cover on 18 September 1995 in the Thabana-Ntlenyana region. The first is as a result of onshore flow from the east coast and subsequent orographic uplift of moist air over the escarpment on 13 September, whilst the other is a cold front on 17 September. Both weather conditions produced precipitation at Underberg (7.5mm on 13 September and 10.5mm on 17 and 18 September) and Bethlehem (1.2mm on 13 September and 0.7mm on 16 and 17 September). However, the cold front produced colder conditions (Maximum of 11.5?C) for Underberg on 18 September, whilst the maximum on 13 September during the onshore flow was 18.3?C. This suggests that snowfall was more probable during the cold front event. Appendix 3 - 267 - - 15) 20 October 1995 Snowfall probably occurred on 19 October 1995 as a result of a cold front over the eastern parts, in association with a surface low over the interior of South Africa. Precipitation was recorded at both weather stations, with Underberg receiving 25mm. Snow cover on 20 October is isolated to the usual 4 regions: the Leqooa range, Thabana-Ntlenyana area, Mafadi Summit area, and the northwest part of the High Drakensberg. It is fairly continuous in these regions, thus supporting the hypothesis that snow fell recently on the 19 October. 16) 31 May 1996 The only opportunity for snowfall during the 16 days prior to the 31 May 1996 is from a cold front event on 24 May. Underberg received 0.7mm of precipitation, whilst temperatures were significantly depressed on 25 May (a maximum of 9.0?C). 17) 18 July 1996 The extensive blanket of snow cover on 18 July 1996 was the result of three cold front systems. Between 6 and 8 July a cold front and upper air cut-off low depressed temperatures significantly at Bethlehem (a maximum of 0.9?C on 7 July). Temperatures for Underberg were not available, but extensive snowfall was reported. On 13 July a cold front passed the region, probably bringing more snow, but no precipitation was recorded at either Underberg or Bethlehem. On 16 July a third cold front brought confirmed snowfalls in the mountains, but once again no precipitation at either weather station. Temperatures between these events were depressed and little snowmelt and ablation could have taken place. The resulting snow cover is probably a combination from all three frontal events. 18) 19 August 1996 Snow covering middle and higher elevations in the Lesotho Highlands and High Drakensberg was probably deposited during a cold front event on 18 August 1996, one day before the data image was captured. This corresponds with the continuous nature of the snow in the image, which shows no melt or ablation on the north-facing slopes. Appendix 3 - 268 - - 19) 20 September 1996 Two cold fronts in quick succession on 17 and 18 September 1996 probably resulted in the snow cover visible in the 20 September image in the Leqooa, Thabana- Ntlenyana, Giants Castle and Mafadi areas as well as the northwest part of the study area. Depressed temperatures and a small amount of precipitation were recorded at both Underberg and Bethlehem, supporting evidence for the wide distribution of snow cover. 20) 06 October 1996 Two snowfall events probably contributed to the snow cover visible on the 6 October 1996 image. On both 22 September and 1 October, cold fronts passed over the eastern parts of the country. They appear to have had effects limited to below the escarpment. Although both Underberg and Bethlehem received regular precipitation during these periods, heavy snow cover is limited to the Thabana-Ntlenyana, Giant?s Castle and Mafadi areas. Snowfall thus appears to have been localised, probably resultant from onshore flow and orographic uplift behind the latter cold front. 21) 05 May 1997 A cold front on 23 April 1997, complimented by a low-pressure trough over the interior of the country probably resulted in the widespread snow cover visible on 5 May. Although there are 11 days of cool to mild temperatures between the proposed snowfall and data capture, the weather conditions did not appear to be conducive to further snowfall. A large quantity of snow is visible in the northwest of the study area, corresponding with heavy precipitation recorded in Bethlehem as a result of the trough and cold front. 22) 21 July 1997 Snow cover is evident across the entire study area at higher elevations on the 21 July 1997 image, and is most likely the result of a cold front that moved across South Africa on 17 July, bringing 8.5mm of precipitation to Underberg. 23) 22 August 1997 A low pressure east of South Africa and associated cold front along the coastal areas on 20 August 1997 appears to be the cause of snow cover visible on the 22 Appendix 3 - 269 - - August. The snow cover is widespread, but limited to south-facing slopes, suggesting that the one day of ameliorating weather conditions was sufficient to melt north- facing areas of snow cover. A temperature of 24.4?C was recorded for Underberg on this day, whilst the average temperature for the 16 preceding days was 22.3?C, some of the highest temperatures for all periods near incidents of snow cover in the study (Table 5.4). 24) 25 August 1998 A substantial blanket of snow cover across the entire study area, visible on 25 August 1998 is the result of snowfall on 22 and 23 August when a broad cold front extended across South Africa. Temperatures were significantly depressed, with Underberg recording 5.0?C and 5.5?C respectively. Precipitation was also recorded in Underberg on these days. 25) 12 August 1999 On 12 August 1999, a small patch of snow is observed in the Giant?s Castle area. This is probably a remnant of a cut-off low over the interior of South Africa, 12 days prior to the image on 29 July when snowfall was recorded. 26) 10 May 2000 Snow cover visible on the 10 May 2000 is primarily limited to the escarpment region between the Amphitheatre and Giants Castle, although localised snow cover is also found in the Leqooa and Sani Pass areas. Snowfall was probable on two occasions: on 3 May a cold front passed along the coastal sections of South Africa whilst on 5 May a cut-off low in the central interior advected moist air from the northeast along the coastal areas. Although both conditions probably brought snowfall, the moist onshore flow from the northeast on 5 May probably resulted in the orographic uplift of the moist air over the escarpment, producing the snow cover pattern observed on 10 May. 27) 30 June 2001 Extensive snow cover on 30 June 2001 is almost certainly a result of snowfall on 29 June associated with a cold front over the central and eastern parts of South Africa. Virtually no melting appears to have taken place. Appendix 3 - 270 - - 28) 09 August 2001 Minor snow cover in the Leqooa area on 9 August 2001 is probably a remnant of snowfall during a cold front on 4 August that was spread across the central and eastern parts of South Africa. 29) 10 September 2001 Patches of snow cover observable in the Leqooa, Thabana-Ntlenyana and Mafadi areas on 10 September is probably the result of snowfall 10 days prior. On 30 August a broad cold front with an associated low pressure over the interior of South Africa brought depressed temperatures and precipitation to both Underberg (9.6mm) and Bethlehem (51.6mm). 30) 08 May 2002 Snow cover is recorded on 8 May across the whole of the study area. At lower altitudes, snow is limited to south-facing slopes, but is more extensive at higher altitudes, particularly in the Leqooa, Thabana-Ntlenyana, Mafadi and northwest areas of the region. This snow is the result of a cold front on 6 May, which brought precipitation to both Underberg and Bethlehem, and for which snow is recorded on high-lying grounds. The snowmelt at lower altitudes on north-facing slopes was affected in the 1 day of ameliorating conditions prior to the satellite overpass. 31) 24 May 2002 A broad cold front across South Africa on 22 and 23 May 2002 probably resulted in the snow cover recorded on 24 May. Continuous large areas of snow are recorded in the more highly elevated regions, particularly the Leqooa, Thabana- Ntlenyana, Giants Castle, Mafadi and northwestern areas. No snowmelt appears to have occurred, confirming that ameliorating conditions have not occurred since the cold front. 32) 09 June 2002 Snow cover is observable on the southern slopes of the more highly elevated ranges in the High Drakensberg on 9 June 2002. This snow appears to be the remnants Appendix 3 - 271 - - of snow recorded during a cold front on 30 May that spread across the interior of South Africa and was associated with an upper air trough. 33) 25 June 2002 An upper air cut-off low and associated cold front on 13 June 2002 is recorded and the likely cause for snowfalls in mountainous regions prior to the 25 June data image. The area of snow cover still present shows the severity of the weather system after 10 days of ameliorating conditions. However, two cold spells on 19 June and 23 June, which do not appear to have brought snowfall, probably prevented significant snow melt. 34) 11 July 2002 Snow visible on the satellite image of 11 July 2002 may well be a remnant of snowfall that resulted in the snow cover on the 25 June 2002 image (33). It is uncertain whether a broad cold front across South Africa on 9 July may have contributed further snow as it may have overlapped the earlier snow cover. 35) 27 July 2002 Snow cover on 27 July 2002 image is extensive, covering almost the entire study area in the southern regions, as well as middle and high altitudes in the northern regions. The snow is a result of 6 days of cloudy weather conditions from 15 to 21 July, the consequence of consecutive cold fronts, a cut-off low and resulting onshore winds. 36) 12 August 2002 An upper air cut-off low on 4 August 2002 probably brought snow to most parts of the study area. On 12 August, 8 days later, snow is visible on the south-facing slopes of ridges in all high-lying areas of the High Drakensberg and adjoining Lesotho Highlands. It is likely that snow is still remnant from the 15-21 July snowfalls. 37) 13 September 2002 Isolated patches of snow cover on 13 September 2002 appears to originate from a broad cold front on 10 September, when heavy snowfalls were recorded. Snow cover had thus undergone significant snowmelt during the intermediate 2 days. Appendix 3 - 272 - - 38) 22 June 2004 The only plausible origin of a small snowpatch recorded in the northwest part of the study area on 22 June 2004 is from a cold front on 7 June. No further significant temperature depressions or precipitation were recorded at Underberg or particularly Bethlehem during the preceding weeks. 39) 08 July 2004 Snow cover on 8 July 2004 is located at higher altitudes in the Leqooa, Thabana-Ntlenyana and Mafadi areas. In the northwest part of the study area, snow cover is extensive. Two potential snowfalls on the 27 June and 4 July are noted. On the 27 June a cold front was situated across South Africa with an associated low pressure over the interior, whilst on 4 July the area was exposed to 2 cold fronts in quick succession. The heavier snowfalls in the northwest are probably a result of the lingering low pressure over the interior of South Africa during this time. 40) 09 August 2004 Small, isolated patches of snow on the highest peaks on 9 August 2004 are probably the remnants of a series of cold fronts from 25-28 July and 30 July when snow was recorded in the Drakensberg. 41) 10 September 2004 A cold front over the central and eastern parts of South Africa on 6 September 2004, when snow was recorded, and another cold front on 8 September, are probably the origins of snow cover observed on the 10 September data image. Extensive snow is recorded in the south of the study area, and along the escarpment in the north, suggesting that the coastal part of the 6 September front played a dominant role in depositing the snow in the northern parts. 21mm of precipitation was recorded in Underberg on the day, whilst Bethlehem remained dry.