A thesis submitted to the Faculty of Science, University of the Witwatersrand, Johannesburg, in fulfilment of the requirements for the degree of Doctor of Philosophy. 25 October 2016 in Johannesburg, South Africa. / Ecosystem services, nature’s benefit to people, contribute to human well-being. Extensive reliance on, and unsustainable use of, natural resources is typical of the rural poor in developing countries and can lead to ecosystem degradation, decreased ecosystem service provision, and increased vulnerabilities of rural populations. Most ecosystem services are intangible or difficult to quantify, but fuelwood provisioning can be measured directly and can serve as a proxy for the status of other ecosystem services (e.g. aesthetic and spiritual services, nutrient cycling, carbon sequestration). South African rural communities have a high reliance on fuelwood despite extensive access to electricity. Within this context, live wood harvesting occurring around rural settlements in increasing amounts has been deemed unsustainable. However, the ‘fuelwood crisis’ of the 1970s, and subsequent predictions of woodland collapse through fuelwood supply-demand models, has still not occurred despite substantial population growth in developing countries. Hypothesised reasons for modelled supply-demand mismatches are based on underestimation of fuelwood supply and woodland regeneration, as well as overestimation of fuelwood demand by discounting behavioural adaptability of users. It is likely that the spatial configuration of fuelwood use allows for the co-adaptability of both humans and ecosystems. A lack of understanding of the spatial configuration of these social-ecological dynamics limits our insights into current and future adaptive responses and thus, the degree of sustainability. This thesis aimed to assess woody biomass stocks and vertical structure changes, as a proxy for provisioning ecosystem services, in a spatially and temporally explicit manner, to describe the status and impact of wood extraction in semi-arid, savanna communal lands. Using repeat, airborne light detection and ranging (LiDAR) data from 2008 and 2012, we surveyed three-dimensional woodland structure in Bushbuckridge Municipality communal lands – the grazing and harvesting areas for densely populated rural settlements in former Apartheid ‘homelands’ in South Africa. Woody biomass in 2008 ranged from 9 Mg ha-1 on gabbro geology to 27 Mg ha-1 on granitic geology. Land-use pressure was associated with compensatory regrowth of savanna tree species through post-harvest coppice in the 1-3m height class. Woody biomass increased at all sites, contrary to previous fuelwood models of the area. Change detection in the vertical canopy structure revealed that biomass increases were almost solely attributable to the 1-3m and 3-5m height classes. These changes were exacerbated by wood extraction intensity in the communal lands – the communal land with the highest wood extraction pressure experienced the greatest biomass increases, likely a strong regrowth response to high harvesting levels. Within communal lands, areas closest to roads and settlements experienced substantial biomass increases as a result of shrub level gains. This relationship was mediated by the usage gradient – the greater the wood extraction pressure on the communal land, the larger and more spatially coalesced the ‘hotspots’ of shrub-level increases were in relation to ease of access to the communal land. However, biomass increases are not necessarily indicative of woodland recovery, as shrub-level increases were coupled with losses of trees >3m in height. To explore these tall tree dynamics further, we tracked >450 000 individual tree canopies over two years over contrasting landscapes – a private reserve containing elephants, two communal lands under different wood extraction pressures, and a nature reserve fenced off from both elephants and humans. Humans are considerable drivers of treefall (defined here as a ≥75% reduction in the maximum height of each tree canopy) in communal lands. Humanmediated biennial treefall rates were 2-3.5 fold higher than the background treefall rate of 1.5% treefall ha-1 (in the control site – the reserve containing neither elephants nor humans). Elephant-mediated treefall was five fold higher than the background rate. Rate and spatial patterns of treefall were mediated by geology and surface water provision in the elephantutilised site where relative treefall was higher on nutrient-rich geology, and intense treefall hotspots occurred around permanent water points. Human-mediated rates and spatial patterns of treefall were influenced by settlement and crop-land expansion, as well as ease of access to communal lands. Frequent fires facilitated the persistence of trees >3m in height, but was associated with height loss in trees <3m. The combined loss of large trees and gain in shrubs could result in a structurally simple landscape with reduced functional capacity. Shrub-level increases in the communal lands are likely an interactive combination of newly established woody encroachers and strong coppice regrowth in harvested species. The more intensely used the communal land, the greater the bush thickening and the stronger the relationship between biomass gains and structural changes in the lowest height classes. The exacerbation of bush thickening in natural resource-dependent communities has critical implications for ecosystem service provision. There is potential for coppice regrowth to provide fuelwood to communities using ‘tree thinning’ programmes, but there is a lack of data on the quantity and quality of the regrowth, as well as the sustainability of coppice, the impacts of different harvesting methods, and the potential feedbacks with changing climate and CO2 fertilisation. Woody resource spatial distribution in communal lands is centred around settlement-level wood extraction pressure, as well as natural resource accessibility in the woodlands. In highly utilised areas, woodland regenerative capacity has been underestimated. Additionally, natural resource extraction is still highly localised, even at the communal land scale, with major structural changes occurring around the periphery or close to existing infrastructure. However, it is these underrated coupled adaptive responses in social-ecological systems that explain the failure of fuelwood supply-demand models’ predictive abilities. Nevertheless, loss of large trees in the landscape and the persistence of ‘functionally juvenile’ coppice stands will have implications for seedling production and establishment in the landscape with repercussions for the future population structure and ecosystem service provision. I discuss the implications of increased natural resource reliance in an African development context and the positive feedback between rural poverty and environmental impoverishment. Potential constraints to the data are unpacked, together with opportunities for further research in this area. / LG2017

M.Sc. (Geography) / Please refer to full text to view abstract

Woody vegetation patches play an important role in the savanna ecosystem. They provide habitat for animal species and are important for some ecological processes such as. However, the coexistence of trees and grass depend on several factors such as fire, elephant, rainfall and soil type. In Kruger National Park, factors that affect the survival of woody vegetation patches are elephants and fire. The loss of woody vegetation patches leads to fragmentation of the landscape. The main aim of this study was to determine the relative difference in the composition, complexity and configuration of woody vegetation patches in and out of the enclosures, and to evaluate the changes in number and size, shape and connectivity of woody vegetation patches between basalt and granite soils over a period of time. There was no linear relationship between composition and complexity and mean nearest neighbour with year, whilst edge density appeared to have a linear relationship with year. Based on edge density results, the results suggested that there was increase in landscape fragmentation in Kruger National Park. The increase in woody patch fragmentation can be attributed to the increase in elephant population, as well as fire frequency and intensity. There was no difference in the composition, complexity and configuration between the inside of the enclosure and the outside. The results show that there is difference between the basalt (Makhohlola) and granite (Nwashitshumbe) region. The results indicated that Makhohlola seems to be highly fragmented than the Nwashitshumbe. The basalt region is situated on rich soils and receives more rainfall than the granite region; hence thus plant growth is greater than that on granite region. High plant growth is associated with high fire frequency and intensity, which result high influence on woody vegetation. The basalt region is dominated with palatable vegetation, as result; the basalt region is associated with large number of herbivores. Due, to high number of herbivores, and high frequency and intensity, the basalt region may be highly fragmented. The change in woody vegetation patches over time, on both basalt and granite region was associated with the interaction between the increase in elephant density, and fire frequency and (or) intensity.

A thesis submitted to the Faculty of Science, University of the Witwatersrand, Johannesburg, in fulfilment of the requirements for the degree of Doctor of Philosophy 29 July 2013 in Johannesburg, South Africa / Effective management of protected areas and communal rangelands, which are often juxtaposed in developing countries, is essential to prevent biodiversity decline and ensure a sustainable resource base for rural communities. However, in human-modified landscapes, there are complex interactions between factors that determine woody vegetation structural patterns. While the underlying biophysical template continues to influence vegetation patterns in a predictable manner; the intensity and type of disturbances that are the result of resource extraction, fire and herbivory can have an overriding impact. In order to effectively conserve biodiversity and plan for sustainable resource use, an understanding of land-use and land management is required. A case study of adjacent protected areas (Kruger National Park (KNP), a national protected area and Sabi Sand Wildtuin (SSW), a private game reserve) and communal rangelands (in Bushbuckridge Municipality (BBR) with varying intensities of use) in north-eastern South African savannas was used to study the spatio-temporal patterns of three-dimensional (3D) woody vegetation structure as a result of natural resource management and abiotic drivers. The aim of this PhD thesis is to advance our understanding of the effects of management of natural resources on spatio-temporal patterns of 3D woody vegetation structure across land uses in a heterogeneous semi-arid savanna system. Vegetation structure was measured using small-footprint, discrete-return LiDAR (Light Detection and Ranging) collected by the CAO (Carnegie Airborne Observatory) Alpha System over 35 000 ha across the study area. 3D woody vegetation structure was compared both within land uses (KNP versus northern SSW, and within BBR) and between land-uses (southern SSW versus BBR) to address two objectives, namely 1. Can LiDAR be used as a monitoring tool for management of woody vegetation structure and biodiversity in semi-arid savannas and 2. What is the impact of land use and the corresponding management of resources on woody vegetation structure in semi-arid savannas? Different land-use legacy timelines and current management objectives at sites in KNP and northern SSW has resulted in an average of 2.5 times higher vegetation density <3 m and >6 m in SSW. These differences in vegetation structure are exacerbated by current management practices, with implications for faunal biodiversity conservation across all scales. Not all reserves are equal in their ability to conserve biodiversity and such knowledge should be considered in conservation planning and management. In the communal rangelands, intense fuelwood harvesting has resulted in coppiced trees <3 m in height, and the only trees >5 m are preserved for cultural reasons, producing similar vegetation patterns to Sabi Sand Wildtuin. Disturbance (extraction and grazing) gradients occur with distance from settlements, with utilization intensity affecting vegetation cover within the size class distributions, but not the shape. Gradients diminish under heavier utilization resulting in a more structurally homogenous landscape, which may be used as an early warning sign of woodland degradation. The increase in >3 m tall trees was twice as high in low intensity use CRs adjacent to SSW compared to those in southern SSW from 2008 to 2010, indicating the impacts of treefall from megaherbivores and fire management reducing plant recruitment/regeneration in the protected area. Knowledge from investigation of socio-ecological drivers in the two land-uses were used to construct an ecologically relevant 3D woody vegetation structural classification which can be used by land managers to plan for sustainable resource use and effective conservation of biodiversity. The management of natural resources, including direct use of fuelwood and the management of herbivory and fire affects woody structural dynamics; however, a lack of knowledge exists around the social and ecological context of natural resource management. The use of remote sensing, the knowledge of savanna ecology and an understanding of community-based natural resource management is integrated in this thesis to contribute to the context specific understanding of drivers of woody vegetation structure in two socio-ecological systems (protected areas and communal rangelands) which can be used in sustainable natural resource management plans.

Ph.D., Faculty of Science, University of the Witwatersrand, 1997

A dissertation submitted to the Faculty of Science, University of the Witwatersrand, Johannesburg, in fulfilment of requirements for the degree of Master of Science. 31 October 2014, Johannesburg. / Over one-third of South Africa’s surface area is covered by savanna woodlands. The structure and dynamics of savannas within communal rangelands have not only been formed by environmental determinants (rainfall and soil) but have also been influenced and manipulated by anthropogenic disturbances (fire, herbivory by livestock, harvesting of resources and cultivation). The aim of this study was to determine the individual and interactive influences of rainfall and catenal position on woody vegetation composition and structure in human-impacted woodlands of Bushbuckridge, Mpumalanga Province, from 2011 to 2013. Three zones were selected that differed in mean annual rainfall: (a) wet west (>700mm), (b) mesic (600-700mm), and semi-dry east (<600m), with three villages per zone. For the rangeland of each village, plots were sampled in 2011, 2012 and 2013 to cover the upland and bottomland variations in catenal position. All trees >6m in height, and their individual stems, were counted and measured within a total of 56 circular plots (only 28 in 2011) each with a radius of 50m. Trees <6m, and their stems, were counted and measured in a circular plot with a radius of 6m, nested centrally within each 50m plot. All analyses were undertaken on (a) total trees and stems and (b) recently harvested (within the last 12 month) trees and stems. The density of small trees (<6m in height) was significantly higher than that of large trees (>6m in height) from 2011 to 2013. Trees and stems were more abundant in the smaller height and diameter classes, respectively, indicating stable populations. The densities of stems for large trees did not show any change over time, whereas the densities of small trees decreased from 2011 to 2013. When comparing across time between rainfall zones, the densities were higher in the high rainfall zone than in the low and medium rainfall zones for each survey year. On the other hand, densities were similar between uplands and bottomlands for each survey year. The intensity of harvesting increased for large trees over time (between 0% in 2011, 2.3% in 2012 and 10.6% in 2013), whereas small trees did not show any change over time. The most harvested trees were between 0.6-4m in height and 1.1-10cm in stem diameter. There were however some signs that harvesting in the larger size classes (>6m in height and >20cm diameter) were increasing over time. The highest proportion of trees was harvested in the medium rainfall zone compared to the other two rainfall zones, but there was similar harvesting intensity between uplands and bottomlands. Overall species richness, Shannon and Simpson’s (Diversity), and Evenness at the plot level did not change for either large or small trees from 2011 to 2013. The species richness and Shannon’s diversity was higher in the high rainfall zone than in the other two rainfall zones, whereas there was no difference in species richness, diversity or evenness between catenal positions for either large or small trees over time. Because there are similar patterns between the species accumulation and rarefaction curves for both the large and small trees, species are distributed at random across the plots, and this is consistent for the three survey years. The most abundance large tree species were Sclerocarya birrea > Philenoptera violacea > Pterocarpus angolensis, which were very different from the most abundant small tree species that were dominated by Dichrostachys cinerea, Terminalia sericea, Acacia exuvialis, Strychnos madagascariensis and Combretum hereroense. A greater species richness, diversity and evenness of harvested trees were observed in 2013 compared to 2011 and 2012. Species that were most harvested for large trees comprised Combretum collinum, Acacia gerrardii, T. sericea, Acacia robusta, Combretum zeyheri and S. birrea, whereas harvested small trees comprised D. cinerea, T. sericea, A. exuvialis and C. hereroense. Even though there were no differences in density, structure, species richness, diversity or harvesting intensity, the species composition did however differ between the uplands and bottomlands. The bottomlands had more abundant fine-leaved species (e.g. Acacia spp. and Dichrostachys cinerea) and the uplands had more abundant broad-leaved species (particularly Combretum spp.). There was a greater difference in species composition in the high rainfall zone relative to the low and medium rainfall zones. This difference in species composition was consistent with the findings that the high rainfall zone had higher density, and a taller, single stemmed tree structure, as well as species richness and diversity when compared to the low and medium rainfall zones. Harvesting intensity was higher in the bottomlands than in the uplands and also higher in the low and medium rainfall zones than in the high rainfall zone. Recent harvesting appeared to have had less influence on species composition than catenal position or rainfall zone. The harvesting of these resources has an impact on both human livelihoods and the ecosystem and must therefore occur in a sustainable way. When the rate of wood production is less than or equal to the rate of wood harvesting, harvesting can be defined as sustainable. The rate of fuelwood harvesting is driven by the demand for the resource, which in most cases is driven by local human population size. With the increase in human population size over time, and the scarcity of fuelwood from the surrounding rangelands, the existence of fuelwood markets is fast becoming a part of daily life as it ensures fuelwood for daily usage such as cooking. For this reason, long-term monitoring is needed. Long-term monitoring will not only allow for better future management of natural resources, but it also allows for the communities to get involved in protecting the resources which are so vital to a vast number of people for daily living. Future studies analysing the data from these plots over longer time periods will provide a better understanding of the role that environmental and anthropogenic determinants play in the changes observed over time in the woody vegeta

Expansion of woody vegetation results in the transformation of a grass-dominated ecosystem to a tree-dominated ecosystem causing land degradation in most semi-arid areas. The imbalance in the natural ecosystem between herbaceous plants and woody vegetation poses a threat to the natural environment. Such changes alter the flow, availability and quality of nutrient resources in the biogeochemical cycle. Most of the dominating woody plants are often unpalatable to domestic livestock. Therefore, the objective is to assess the spatial extent of woody vegetation over time. Knowledge of the spatial and temporal characteristics of woody vegetation dynamics will enable the development of management plans. These characteristics can be derived using remote sensing techniques which have become efficient in such studies. This study aimed to characterize woody vegetation dynamics along the route between Alice and King Williams’s town in Eastern Cape Province South Africa using Landsat data. This aim was achieved by focussing on three specific objectives. The first objective was to compare the performance of multispectral data and Normalized Difference Vegetation Index (NDVI) data of Landsat imagery in mapping woody vegetation cover. The second objective was to investigate the effect of the spatial resolution of remotely-sensed data on discrimination of woody vegetation from other land cover types. The third objective characterised woody vegetation dynamics between 1986 and 2013/2014 using the results from the first objective. The study used Landsat imagery acquired in November or February of 1986, 1994/1995, 2002/2003 and 2013/2014. Due to lack of data which covered the study area two separate dates (November and February) where used for the study resulting in naming the study area western and eastern parts. Unsupervised classification was performed on the multispectral, NDVI and pan-sharpened images to generate four generic land cover classes, namely water, bare land, grassland and woodland. Accuracy assessments of the classified images was done using error matrix. The results showed that the classification based on NDVI images yielded a better overall accuracy than the classification based on multispectral images for the western (83 percent and 75 percent, respectively) and eastern (82 percent and 76 percent, respectively) parts of the study area. Similarly, pan-sharpening resulted in better overall classification accuracy than multispectral, but comparable to the classification of the NDVI images for both the western (82 percent) and eastern (83 percent) parts of the study area. Remote sensing is an effective tool in assessing changes in the physical environment. Landsat imagery is suitable in assessing land cover dynamics given the long-term and free availability of the image. In addition, the large spatial coverage it provides, enables Landsat data to be used on studies that have wide spatial coverage. Classification for the purpose of time-series analysis was then performed on the NDVI images of each date (1986, 1994/1995, 2002/2003 and 2013/2014). Both woody vegetation and grassland experienced changes from 1986 to 2013/2014 with grassland occupying (75 percent) compared to woodland (17 percent) in 1986. In the year 2013/14 grassland occupied 32 percent and woodland occupied 51 percent of the study area. The increase in woody vegetation in the study area can be attributed to livestock rearing and migration of people from the rural to urban areas post-Apartheid. The study output will aid in the development of a database on land cover distribution of the area between King William’s town and Alice town, providing useful information to decision-making and further studies on woody vegetation.

Urbanization has increased rapidly throughout the world. The densification of urban areas has greatly reduced the number of natural areas occurring within the urban environment as well as impacting the ecosystem services that these areas provide. Urban greening and sustainable practices have been advocated as a means to once again provide the urban population with ecosystem services. Sacred natural areas that occur in surrounding forest, temple and cemetery sites have been known harbour a variety of biological diversity, as well as provide people with a number of cultural and spiritual benefits. Much of the literature on sacred natural sites comes from rural or eastern countries, leaving a large gap in the knowledge pertaining to information on these sites in both developed nations as well as urban areas. The aim of this study was to determine the abundance and composition of woody species, as well as the spiritual and cultural significance of sacred natural sites in Grahamstown. This study defined a sacred urban area as any form of garden surrounding a church, temple or mosque, as well as cemeteries. It looked at a total of 28 church gardens, one Hindu garden, one Mosque garden and five cemeteries in Grahamstown, South Africa. For each site the area was calculated and a tree and shrub inventory was done for all individuals above 1.5 m tall. Church/cemetery age, denomination and appearance were recorded as well as soil samples collected. An ordination of the data was done to summarize the community data, relating the community variation to environmental gradients. Questionnaires were completed by congregants who attended a religious building with a garden, as well as those that were not surrounded by a garden. These questionnaires were used to determine the cultural, spiritual and aesthetic value of trees and the sacred area, as well as the perceived and felt benefits that these areas provide. Those that were completed by congregants without gardens looked to find out whether or not it was believed that these areas would improve their experiences. There was an average plant density of 106.1 woody plants per hectare, with a total of 139 different species encountered. Of these, 56 percent were exotic species. This is slightly lower than that of studies done elsewhere in the world, but may be due to the omission of non-woody ornamentals and lawn species in this study. Of the top 11 most frequently occurring species, only two were indigenous. There was generally low similarity between plant assemblages found at the different sites. A significantly positive relationship was found between site size and woody plant basal area as well as the total number of woody plants. Site age and religious denomination had little influence on woody plant density, basal area, species richness or woody plant abundance. Congregants stated that a garden surrounding a religious building improved both their spiritual and aesthetic experiences. Stated spiritual and aesthetic experience was significantly influenced by basal area, while abundance significantly influenced stated aesthetic experience. Greenery was therefore important to many of the congregants, however, the species that were present were less influential. A greater sample size from a variety of religions and sacred areas within urban environments throughout the world would prove to be an interesting comparison for future research.

A major challenge in savanna rangeland studies is estimating woody vegetation cover and densities over large areas where field based census alone is impractical. It is therefore crucial that the management and conservation oriented research in savannas identify data sources that provides quick, timely and economical means to obtain information on vegetation cover. Satellite remote sensing can provide such information. Remote sensing investigations, however, require establishing statistical relationships between field and remotely sensed data. Usually regression is the empirical method applied to field and remotely sensed data for the spatial estimation of woody vegetation variables. Geostatistical techniques, which take spatial autocorrelation of variables into consideration, have rarely been used for this purpose. We investigated the possibility of improving woody biomass predictions in tropical savannas using cokriging. Cokriging was used to evaluate the cross-correlated information between SPOT (Satellites Pour l’Observation de la Terre or Earth-observing Satellites)-derived vegetation variables and field sampled woody vegetation percentage canopy cover and density. The main focus was to estimate woody density and map the distribution of woody cover in an African savanna environment. In order to select the best SPOT-derived vegetation variable that best correlate with field sampled woody variables, several spectral vegetation and texture indices were evaluated. Next, variogram models were developed: one for woody canopy cover and density, one for the best SPOT-derived vegetation variable, and a crossvariogram between woody variables and best SPOT-derived data. These variograms were then used in cokriging to estimate woody density and map its spatial distribution. Results obtained indicate that through cokriging, the estimation accuracy can be improved compared to ordinary kriging and stepwise linear regression. Cokriging therefore provided a method to combine field and remotely sensed data to accurately estimate woody cover variables. / Thesis (M.Sc.)-University of KwaZulu-Natal, Pietermaritzburg, 2008.

The historically poorly conserved subtropical transitional thicket (STT) of the Eastern Cape is overutilised by domestic stock and game in the more xeric areas, and has shown no signs of recovery from this grazing pressure. It has been postulated that no regeneration through seeding occurs. This project was undertaken to determine: 1) how much STT has disappeared between 1950 and present, and what the current conservation status is; and 2) whether regeneration of the xeric succulent thicket is taking place through seeding, and if so, where. The study was conducted at the Andries Vosloo Kudu Reserve near Grahamstown. Approximately one-third less STT was mapped in this study, based on 1981 Landsat images, than was mapped in 1950. Approximately 10 % of all remaining STT is conserved. The order Kaffrarian thicket is poorly conserved. Newly germinated seedlings of a wide range of shrub species occurred under the canopies of a wide range of shrubs which served as nurse plants, throughout a gradient of veld condition. Seedlings of Portulacaria afra, the dominant shrub in xeric succulent thicket, were most common. Similarly all saplings recorded in a survey of saplings were associated with bushclumps. One-third of all saplings have the potential to contribute to the spread of bushclumps. Regeneration of xeric succulent thicket through seeding probably does occur, contrary to current ideas. Ptareoxylon obliquum was the most common sapling, despite mature trees now being scarce following earlier heavier utilisation . P. obliquum was also the nurse plant which supported the highest density of newly germinated seedlings. The possible role of P. obliquum in the functioning of xeric succulent thicket is discussed. The confinement of seedlings and saplings to areas under the canopies of trees and shrubs implies that the xeric succulent thicket will not recover rapidly if allowed to rest. Active management techniques will be necessary if rapid recovery is required. Bare areas between bushclumps may no longer be suitable germination habitats because of high Al concentrations. There was no evidence to support the idea that germination and establishment of shrubs in clear areas is linked to episodic climatic events. Dung middens of recently reintroduced black rhinoceros may however aid in germination of seeds and establishment of seedlings under certain climatic conditions. Recommendations for further studies, based on the findings of this project, are made. Possible management techniques aimed at the rapid recovery of this veld are suggested, and management proposals for the Sam Knott Nature Reserve/Andries Vosloo Kudu Reserve complex are made.