Freshwater biodiversity is highly threatened, but under-represented in media, and less well known than marine and terrestrial counterparts. The current lack of public awareness of freshwater species and ecosystems may limit freshwater conservation as a popular cause. The manner in which species are portrayed in the popular media can influence awareness and support for conservation. In this digital age, images in particular may play a major role in public support for conservation. In this study, I examined images sourced from the web of ten native South African freshwater fish species and images of a general search for ‘freshwater fish South Africa' to investigate the kinds of images people are exposed to. Results show that only 28% of speciesspecific images showed the fish in a water habitat, with the remaining images showing the fish out of water, often with the presence of humans. I then conducted an online survey of people's perceptions of freshwater fish images. Respondents were shown images of native South African fish in water and out of water. From this survey, 73% of respondents stated they would choose images of fish in water in their natural habitat to communicate the importance of conserving freshwater ecosystems and freshwater fish. Over 95% of respondents, agreed that images play an important role in rallying support for nature conservation in general. When shown images of fish in water, 57% of respondents stated the in-water images evoked a desire to know more about freshwater environments and 49% stated a desire to know more about freshwater fish. The survey also investigated respondent willingness to donate money and time to freshwater conservation and explored whether showing either ‘in-water' or ‘out of water' images beforehand influenced this response. The results suggested that the type of images presented beforehand had little to no effect on willingness to donate, but other factors such as gender, natural sciences background, and history of visiting freshwater environments influenced respondent willingness to donate time and money. Overall this study suggests that future considerations should be given to how freshwater fish species are portrayed in science communication and environmental education through imagery.

Vultures are crucial scavengers, removing carcasses, and maintaining the healthy and clean environments. However, their populations are rapidly declining. Six of eight Kenyan vulture species are currently listed as ‘Endangered' or ‘Critically Endangered' on the IUCN Red List. Poisoning caused by human-wildlife conflicts is a major threat to vultures after belief-based use of their body parts. A total of 1387 interviews were administered within Maasai Mara pastoral community. We used the ‘Unmatched Count Technique' to estimate the prevalence and distribution of poison use, and direct questions to characterize poison types, usage and sources, as well as belief-based use of vulture body parts. We found that 54% of respondents reported Carbofuran poison, whereas nearly half (48%) of respondents pointed to agrovets as a major source of all poisons and popularly (84%) smeared on carcasses. The vast majority (75%) particularly use vulture feathers for arrows. Further, 22% fence off their livestock against predatory wildlife. We further explored how predation protection measures used predict individual poisoning likelihoods. None of the five main predation protection measures significantly influences poison use. Both fencing and the use of lights as predator control measures attenuate the poisoning risk. However, herding indicated weak signals for poison use amongst pastoralists. We proposed that an effective vulture poisoning risk reduction should be multi-faceted and collaborative. Regulating and monitoring of the import, local trading and use of poisonous substances. Building partnerships and engagements for more support for local livelihoods. Lastly, upscaling fencing and expanding the communal conservancies. These strategies would curb retaliatory human-wildlife conflicts and poison use against wildlife and vultures in Southern Kenya.

There is increasing understanding of the role that both ecological and built infrastructure can have in economic growth and development in terms of water supply. However, degradation of ecological infrastructure (EI) is resulting in the loss of valuable ecosystem services that benefit human well-being. Invasive alien plants (IAPs) are degrading catchment areas which negatively impacts delivery of hydrological ecosystem services. Clearing IAPs is considered a catchment conservation intervention that preserves these services. This study used South Africa as a case study to analyse the viability of investing in EI by way of IAP clearing compared to built infrastructure augmentation interventions to secure water supply in the long term. Unit reference values (URVs) were used to compare costeffectiveness between ecological and built interventions for 11 of South Africa's regional water supply systems (WSSs). Built infrastructure URVs were available from government reports, while URVs for EI were calculated by modelling spread of IAPs, calculating the cost to clear them between 2022 and 2050 and the potential amount of water saved in their absence. The results provide quantitative evidence of the cost-effectiveness of investing in EI against built infrastructure options to secure water supply. The potential water to be gained by clearing IAPs from catchment areas of existing bulk water infrastructure was approximately 40% of what would be gained by implementing all built infrastructure interventions by 2050. It is recommended that IAP clearing be pushed ahead of built infrastructure interventions to delay costs associated with further built infrastructure development. Governing institutions, economists and natural resource managers are therefore encouraged to coordinate efforts towards designing EI investment frameworks as a sustainable, resilient approach to securing water supply.

The Cape parrot is the only endemic parrot of South Africa and is currently nationally threatened. One of the biggest threats to the Cape parrot is the past and present degradation of indigenous forest. The Amathole Mistbelt Forest in the Eastern Cape is the primary habitat for Cape parrot and has historically been heavily degraded. In order to conserve the Cape parrot effectively, there is a need to understand the spatial distribution of indigenous forest patches and their quality. There is currently not a sufficiently accurate landcover map available to fulfil this need. Thus, this study uses remotely sensed imagery at a 10 m resolution and random forest classification to (1) produce a land cover map of the indigenous forest in the Amathole region; (2) determine habitat quality of the indigenous forest, and (3) determine whether forest loss, as reported by Global Forest Watch (GFW), reflects the loss of indigenous forest or the clearing of plantations and woody alien invasives. The overall accuracy of the classification was very high at 82%. Cross validated accuracies were all high ranging from 95 – 100%, with water having the highest accuracy and indigenous forest, eucalyptus spp., pine spp., and infrastructure having the lowest accuracies. F1 scores ranged from 0.78 – 1.0, with indigenous forest ranking the second lowest at 0.80 and grassland ranking the second highest at 0.91. Indigenous forest covered 26% of the study area. Black wattle, pine spp. and eucalyptus spp. covered a combined 35% of the study area. The detailed map of indigenous forest shows the extent of its fragmentation and outlines some of the management implications associated with small forest patches. Secondly, habitat quality for Cape parrot is questioned as there is a lack of emergent canopy tree species and 30% of the matrix between forest patches is invaded by invasive alien species. Thus, it is suggested that a strong focus is put into clearing and managing invasive alien species. Lastly, GFW ‘forest cover loss' is shown to be comprised primarily of plantation felling and invasive clearing. It is suggested that there has been little loss of indigenous forest in the last 30 years. Further research will include creating an open and accessible product in the form of a Google Earth Engine App to share with conservation managers in the area.

Competition with non-native invasive species is considered a major threat to many rare native species. As such, invasives removals are a common management strategy. Rare native species that interact uniquely with other organisms in their community (e.g., hemiparasitic plants) may be adversely affected by removing invasives. A management plan for a regionally rare hemiparasitic plant in Massachusetts, Pedicularis lanceolata Michx., identified invasives as a threat, but more quantitative evidence is needed to determine how P. lanceolata's persistence is influenced by its co-occurrence with native or invasive hosts. This research asks how P. lanceolata is affected by growth with native versus invasive hosts. Chapter I describes the species associated with P. lanceolata throughout its range, comparing areas where it is considered common and rare. Relative abundances of natives, non-native invasives, non-native non-invasives, and species with both native and non-native genotypes growing with P. lanceolata did not differ significantly at sites where the species is considered common in the Midwest compared to sites where the species is considered rare in the east. Chapter II outlines greenhouse and field removal experiments in which the types of host plants growing with P. lanceolata were manipulated. In the greenhouse, P. lanceolata growth, survival, and flowering were lower when it was growing with invasive compared to native graminoids. However, differences in P. lanceolata growth and survival when natives versus non-native were removed in the field varied from year to year due to succession of native shrubs at the site during the study. Chapter III asks how the population growth of P. lanceolata differs in uninvaded and invaded patches using an Integral Projection Model to perform population projections, sensitivity and elasticity analyses, and a life table response experiment. The population growth rate of P. lanceolata in uninvaded patches was lower than in invaded patches due to the succession of native shrubs in uninvaded patches. Chapter IV describes a metapopulation model for the invaded population of P. lanceolata in Massachusetts. The quasi-extinction probability was significantly affected by probabilities of dispersal, positive correlations in vital rates between sites, and catastrophes. These data will be used to update the management plan for P. lanceolata.

(1) The distribution and habitat preferences of wild ungulates and cattle were determined from observations and counts in Lochinvar National Park and the Game Management Area (GMA) on the west em part of the park on the Kafue river floodplain of Zambia over five months. Oribi, zebra and lechwe were the most widely distributed wild ungulates using habitat patches inside and outside the park. Although the distribution of buffalo was restricted within the park, the species was found in all the habitat types unlike lechwe and zebra that preferred only the floodplain and termitaria areas. Wildebeest preferred the termitaria and floodplain areas inside the park. The highest lechwe density was in Chunga floodplain with a density of 504.2/km2 followed by Mulindi GMA with 451.9/km2. The lowest lechwe density was in Mulindl termitaria with 0.4/km2. Zebra density was highest in Nampongwe floodplain with 118/km2 while the highest wildebeest density was on Gwisho dambo with 27/km2. Nampongwe floodplain had the highest oribi density with 2.8/km2. There were significant differences in the density of lechwe inside and outside the park in Mulindi. Zebra and oribi densities inside and outside the park in Mulindi were low and no differences could be noted. (2) The biomass production in the different habitats was determined by harvesting the above ground biomass in selected areas in each habitat. The termitaria were the most productive of the three habitat types with an estimated production of 4320kg/ha followed by the woodland with 3200kg/ha and the least production was in the floodplain with 2850kg/ha. Of this total, grasses made up 87.4%, 83. 7% and 72.5% of the production in the floodplain, termitaria and woodland respectively. There were no significant differences in biomass production between different habitat types, however, there were significant differences in biomass production within the same habitat. (3) The species composition of the dominant plants was different in all habitat types. and as a result, the forage availability in the different habitats also differed. The floodplain was dominated by Echinoehloa spp, Panicum repens and Oryza longistaminata whereas the termitaria was dominated by Hyparrhenia rufa, Setaria spp and Echinochloa colonum. The woodland was dominated by Setaria and Digitaria spp. (4) Cover was determined in the different habitat types using the Parker Step Point method. The highest amount of grass cover was in the floodplain with 46.6% followed by termitaria with 45.6% and the woodland with 13.3%. Non grass cover was highest in the termitaria with 15.8°, followed by floodplain with 9.4% and woodland with 0.5%. The woodland had the highest amount of litter with 66.5% followed by the floodplain with 30.7%. The termitaria had the least amount of litter with 24.9%. The highest amount of bare ground was in the woodland with 19.8% followed by termitaria with 13.8% and the floodplain with 13.3%. There were significant differences in the proportion of grass and non-grass cover in all habitat types. There were no significant differences in the proportion of litter and bare ground between the floodplain and tennitaria and between floodplain and woodland however, there were significant differences in the proportion of litter and bare ground between the tennitaria and woodland. (5) A questionnaire survey revealed that game meat and grazing land are the two main resources that are used by people near the park. The main problems faced by people living near the park is destruction of crops by problem animals like baboons and bushpigs and the transmission of diseases like tuberculosis from game to livestock. The survey also revealed that people are forced to travel inside the park because it is the shortest route to various destinations. Others travel inside the park because the main sources of employment in the area are found inside the park. Game seems to be used to the presence of people and will only run away when closely approached or when people are travelling using a vehicle. (6) The results of the study do not support the hypothesis that cattle can be accommodated in Lochinvar National Park with no adverse effect on lechwe and other large herbivores. Cattle and game use the same areas outside the park in Mulindi but the areas are used at different times in the year. (7) Lochinvar National Park needs a clear burning policy to effectively manage the bush encroachment problem and overgrazing in some areas inside the park. The income generated by the park should be ploughed back into the management of the park in order to effectively control poaching, human encroachment into the park and to monitor the activities and movements of people inside the park.

This thesis reports upon the findings of a study into the socioeconomic and ecological drivers of plant resource utilisation, and dynamics of woody plants used by rural people in the Northern Province, South Africa. The key components were to (1) quantify the range of socioeconomic drivers of plant resource utilisation in rural communities, (2) identify plant species highly preferred by people and their uses, (3) quantify the impacts of plant resource harvesting, (4) seek relationships between spatial distribution patterns of plants and productivity, (5) determine the responses of trees to harvesting regimes, and (6) elucidate the impacts of selected management actions on plant resource supply and hence sustainability. Each of these was pursued through a combination of surveys and empirical experimentation. High unemployment rate, low educational levels, large family sizes and most importantly low overall family income are characteristic of this rural community. As a result, many people are still reliant on wild growing plants for their household and economic needs. The three main use categories associated with people in this rural village, in order of their importance, are: construction, food and energy. It is construction and energy uses that are associated with very few species. Harvesting of plant materials for these two use categories is destroying the preferred species in communal land. Two plant species, Colophospermum mopane and Androstachys johnsonii, are the most preferred plant species for construction and energy purposes. In a protected area the C. mopane population is stable, an indication that recruitment balances mortality. The same applies to A johnsonii. However, in communal land the size class frequency distribution of C. mopane varies at three distances from the village, suggesting that communal patterns of C. mopane utilisation are unfavourable to this species. This is probably so because of intense browsing within the village and high levels of harvesting. However, A johnsonii at two harvest zones in communal land show the same trend as in the protected area: the inverse J-shaped curve which is a characteristic feature of a stable population. Selective harvesting therefore, has little impact on recruitment of young A. johnsonii trees. Conflicting results were evident with respect to the role of competition in tree populations. Spatial distribution methods (i.e. nearest-neighbour analysis and departure from randomness approach) from a number of plots suggested that competition was not important at most plots. Yet, the strong negative relationship between loge mean stand basal area and log0 stand density indicated that competition is a significant factor affecting individual tree and stand woody productivity. Most plots from the communal land, when plotted on the same graph of plots from the protected site, lay far below the thinning line, suggesting that harvesting is promoting rapid growth of trees in communal land. The five species harvested responded differently from each other. Full and partial harvesting of C. mopane trees during December at ~30 cm above ground level resulted in 100% survival rate after one year. The equivalent diameter of coppice shoots was significantly positively related to stump size and light availability. The former suggests that harvesting bigger trees will result in rapid diameter growth of coppice shoots while the latter suggests that shoots are suppressed by shade from neighbouring trees. The equivalent shoot diameter was significantly negatively related to sum of neighbour size divided distance ratios, an indication of competition between neighbouring trees. All big trees and 70% of the small trees of A johnsonii species died when the same season, height of cutting and harvesting regime applied to C. mopane was used. Partial harvesting resulted in 60 and 45% survival rates for big and small trees respectively. Coppice shoot production of the surviving stems was also very low compared to C. mopane. A static transition model was developed to simulate requirements for sustainable harvesting of Colophospermum mopane woodlands. The model predicts that a combination of high levels of harvesting and browsing will result in the depletion of the resource base within three decades.

Ant and plant assemblages in patches of Pinus species trees in mountain fynbos and renosterbos (Elytropappus rhinocerotis) shru.bland were investigated in order to detennine whether there were differences in ant and plant species richness, abundance and cover between patches of pines and the surrounding indigenous vegetation. The study was carried out in the Matjiesrivier valley on the southern slopes of the Swartrg mountains in the Oudtshoom district, South Africa. Ants were collected at three sites: one in an older (c. 70 years) pine patch in mesic mountain fynbos at about midslope and two in younger (c. 10 years) pine patches in closed and open remnant renosterbos shrublands respectively on lower slopes. On a per site basis, there were no major differences in ant and plant species richness between the pine and no-pine sites in renosterbos, or (with the exception of plants) the pine and no-pine site in mesic mountain fynbos. A total of 2 539 ants belonging to 37 species, distributed among 15 genera, were collected during the study. Camponotus was most species-rich genus (13 species), followed by Monomorium (7 species), Tetramorium (4 species) and Lepisiota (2 species), with 11 of the 15 genera each represented by a single species. No non-indigenous species were observed during the duration of the study. Camponotus sp.11, Lepisiota sp.1 and Tetramorium quadrispinosum were observed only in the open renosterbos site. Camponotus klugii, Camponotus sp. l, Camponotus sp. 6, Cerapachys sp. l, Meranoplus peringueyi, Monomorium fri.dae, Monomoriu1'Jl sp.3 and Technomyrmex albipes were observed only in the closed renosterbos. Camponotus bayneii, Camponotus sp.5, 7, 8 and 9, Crematogaster peringueyi, Pachycondyla pumicoca, Solenopsis punctaticeps and Tetramorium regulare were recorded only in mesic mountain fynbos The most abundant ant guild was the non-dominant generalist (17 species), particularly Pheidole sp. 1# (48 % of all individual ants) followed by nectar feeders (15 species), dominant generalists (2 species), specialist predators (2 species) and a seed harvester (1 species). In general, the non-dominant generalists preferred no-pine sites to sites invaded by pines. There was very little undergrowth under the pines in mesic mountain fynbos, and Clutia polygonoides, Cymbopogon sp. 1, Erica sp. 1, Metalasia cephalotes, Metalasia muricata, Pelagronium betulinum sp. 1 and Restio sp. 1 were significantly more frequent outside the patch. The two renosterbos sites showed no significant difference in Elytropappus rhinoceroti's cover in and out of pine patches, and only two species at the open rerttsterbos site, Dodonea sp. 1 and Re/hania squarrosa, were more abundant under pines than in the surrounding renosterbos. The most abundant plant life form was the ericoids (37 species) followed by proteiods (11 species), grasses (11 species), forbs (9 species) and succulents (4 species). In general, the five life forms did not show -any exclusive pattern of preference for either the pine or no-pine sites.

As global biodiversity decreases, the importance of protected areas for protecting biodiversity and ecosystem complexity, is rising. Increasing development and land use change means that protected areas must restrict species to their boundaries to avoid human-wildlife conflict. Populations of species therefore no longer disperse, and overpopulation can occur. In southern African savanna systems, large populations of savanna elephant (Loxodonta africana) are transforming woodland and reducing vegetation diversity. While historically large, the elephant population in Gonarezhou National Park in Zimbabwe has grown at >5% annually since 1992, and densities currently exceed 2 elephants km-2 . Over the last 70 years, riverine woodland vegetation has undergone substantial changes. While initially it might be compelling to hold elephants responsible, Gonarezhou National Park has a complex natural and socio-ecological history to consider. Before proclamation in 1974, areas supporting woody riverine vegetation along Gonarezhou National Park's biggest river, the Runde, were widely cultivated. Some riverine vegetation was also cleared in the late 1950s to prevent the spread of tsetse fly (a vector for African Sleeping Sickness). This study makes use of aerial photography and satellite imagery of the Runde River, and its confluence with the Save River, covering the period 1948 to 2018. Using supervised classification techniques, imagery was analysed to identify vegetation types and provide an estimation of riverine woodland cover. Further analyses were conducted to assess characteristics and possible drivers of change. Percentage cover of woody riverine vegetation along the Runde River, covering a total area of 60.2 km2 , varied greatly over time, rising initially from 14% cover in 1948 to 40% cover in 2005 with cessation of cultivation and clearing for tsetse fly after the park was proclaimed. It then decreased by 20% between 2005 and 2018 most likely due to high populations of elephants and severe droughts. Confirmation that elephants have caused a decrease in woody cover of riverine woodland within Gonarezhou National Park was gained from comparing a plot inside the park boundary with an equivalent plot outside the park boundary that has supported a lower density of elephants. Woodland cover in these plots showed a significant reduction inside the park boundary suggesting elephants have driven this decline. The area of riverine woodland which was previously cultivated (3.3 km2 along the Runde River) supported between 1% and 12% less riverine woodland cover than areas which were not cultivated for each year of assessment, but the differences were not significant. The area of riverine woodland cleared for tsetse fly control (0.4 km2 along the Runde River) in the late 1950s supported between 8% and 25% less riverine woodland cover than uncleared areas in each year. Although some recovery is evident, there is a significant long-term influence of tsetse clearing on riverine woodland vegetation cover in Gonarezhou National Park. Elephant impacts are expected to decline with distance from permanent water. However, examination of changes in woody cover along a seasonal river, offering a potential refugium, yielded variable results. By 2018 a decrease in woodland cover adjacent to the permanent water source of the Runde River was apparent, suggesting that riverine woody vegetation near permanent water is most affected by elephant damage. A hypothesis that elephant impact on woodlands is lessened where there is a concentration of alternative food sources, especially hygrophilous grassland and Faidherbia albida pods, was examined. Woody cover along the Runde at its junction with the Save, which has substantial alternative forage, was compared to upstream areas without such alternatives. In 2018 a decrease in woody cover is shown with distance from alternative food sources, suggesting elephants use woodland more intensively when alternative forage is unavailable. The probable influence of drought and flooding events on reducing alternative forage and woody cover, requires further consideration. Woody cover is a crude index of change as it does not account for structure or composition of woodland below the canopy. As a result, if a significant change is detected, we can be quite confident that the change has occurred. However, if there is no significant change, this does not necessarily mean that changes have not taken place, they simply may not be detected by such a crude measure. Analysis of woodland cover serves as a starting point. Time and resource limitations meant that structure and composition could not be considered in this study. However, such measures would increase the accuracy of analysis in future studies. Overall, this study demonstrates the importance of considering all possible influences on vegetation change. Strong evidence was provided that elephants impact upon woody vegetation change over time. However, the longer-term influences of cultivation and tsetse clearing and, availability of refugia and alternative forage cannot be disregarded. An assiduous approach is required lest we falsely attribute blame to elephants alone, the management consequences of which are profound.

Large carnivores face numerous threats, including habitat loss and fragmentation, direct killing, and prey depletion, leading to significant global range and population declines. Despite these threats, leopards (Panthera pardus) persist outside protected areas throughout most of their range, occupying a diverse range of habitat types and land uses, including peri-urban and rural areas. Our understanding of leopard population dynamics in mixed-use landscapes is limited, especially in South Africa, where most research has focused on protected areas. Here I use spatially explicit capture-recapture (SECR) models on camera trap data to estimate how leopard density varies across a mixed-use landscape of protected areas, farmland, and urban areas in the Overberg region of the Western Cape, South Africa. Data were derived from 86 paired camera stations, which collected data for 161 camera trap nights, providing 221 independent leopard captures at 50 camera trap stations. A total of 25 individual leopards were identified, and the best-performing SECR model included the covariate sex on the σ (spatial decay), and a combination of sex, vegetation type and the interaction on λ0 (capture probability), with a density estimate of 0.64 leopards per 100 km2 . Elevation, terrain ruggedness, protected area status and NDVI were all important drivers of leopard density in the region, with leopard density highest on elevated remnants of natural land outside of protected areas. These results are similar to previous research findings in the Western Cape, where high-lying natural vegetation was shown to serve as both a refuge and a corridor for leopard movement in otherwise transformed landscapes. Given the low level of risks to lives and livelihoods posed by leopards in this region, the continued persistence of leopards in this shared landscape is considered high. Education of landowners should still be prioritised to improve tolerance towards leopards in the event of occasional negative impacts (e.g., livestock depredation).