It is well documented that the South African palaeoenvironmental record is relatively limited in terms of both quantity and quality. This is mainly due to the region's highly seasonal rainfall regimes and generally arid to semi-arid environments which are not conducive to the preservation of sedimentary sequences and associated proxy records. The climate along the southern Cape coast is influenced by both tropical and temperate climate systems, and the region hosts highly diverse vegetation including fynbos and thicket elements and includes the Knysna Afrotemperate Region – the most extensive forest complex in southern Africa. The mechanisms controlling these tropical and temperate systems have responded to changing global boundary conditions and these changes have significantly impacted the regional vegetation mosaic. This ephemeral nature of the region's climate and vegetation suggests it is particularly sensitive to climate change, making it an ideal area to evaluate changes in these systems and how they interact over time. For this study, four sets of records were produced from three wetlands along the southern Cape coast. The Eilandvlei palynological and microcharcoal records span the last ~3000 years with the pollen and microcharcoal records from adjacent Bo Langvlei covering the last ~1300 years. The most outstanding feature in these records are the time periods covering the Medieval Climate Anomaly (MCA; c. AD 950 – 1250) and the Little Ice Age (LIA; c. AD 1300 – 1850). The evidence indicates that conditions in the region during the MCA chronozone were relatively dry and perhaps slightly cooler than present. The most durable phase of forest expansion, and likely more humid conditions, occurred during the transition between the MCA and core cooling of the LIA with the LIA clearly identified as a period of cool, dry conditions between c. AD 1600 to c. AD 1850. In addition, the Eilandvlei pollen record demonstrates the effects of external physiographic dynamics on pollen accumulation and deposition within the lake basin. A complementary set of geochemical and sedimentological records have been generated for Bo Langvlei incorporating the last ~4200 years. These records suggest that the late Holocene evolutionary history of Bo Langvlei comprised of three phases: a marine/lagoonal phase extending until c. 1270 cal yr BP, a short transitional phase between c. 1270 and 1200 cal yr BP, and IV the more recent lacustrine phase. A ~650 year pollen and microcharcoal record have been obtained from nearby Vankervelsvlei. Although discontinuous, this is the first palynological record from this unique waterbody covering this period. In terms of climate, the mechanisms driving the observed changes in the records taken as a whole appear to relate to changes in temperature and dynamics in the influence of tropical systems, perhaps transmitted at least in part via the Agulhas Current and the development of localised precipitation systems. The findings further reinforce the proposed importance of summer rainfall in regulating moisture availability along the south coast of South Africa. The records also highlight the significant impacts of fluctuating sea levels and changes in dune morphology in shaping the embayment, and more recently, the effects of accelerated anthropogenic activities in the area

Contaminated surface water runoff from inadequate drainage and sanitation systems in informal settlements threaten the quality of available freshwater and can negatively impact both human and environmental health. Biofiltration systems (biofilters) provide water pollution controls without inputs of additional energy and chemicals, placing them in the overall context of the need for affordable and sustainable stormwater infrastructure in informal settlements. In addition, cleaned waters from biofilters may be suitable for some reuse applications if they are well-designed and maintained. However, most research is conducted in developed countries where heavy metals are the main surface water pollutant. Consequently, little is known about the extent to which biofilters can be used to meet the water quality targets in conditions likely to be found in informal settlements. In addition, no attempts have been made to recover or reuse the surface water runoff from informal settlements, despite its high nutrient loadings. This study analyses the extent to which biofilters can be used to clean and reuse contaminated surface water runoff from informal settlements. The objectives are threefold: (i) to analyse the performance of two field-scale biofiltration cells (one vegetated and one non-vegetated) that are batch-fed with surface water runoff from an upstream informal settlement; (ii) to determine the effects of varying operating, design and environmental parameters on the performance of the cells; and (iii) to develop a model which predicts the outflow pollutant concentrations under varying conditions. Both cells effectively reduced ammonia (NH3), Total Phosphate (TP) and Escherichia coli (E. coli) concentrations, but leached nitrate (NO3 - ) and nitrite (NO2 - ). The treated waters were suitable for irrigational reuse, however, additional disinfection was required to reduce faecal contamination in some cases. Correlation analyses showed that inflow water quality significantly influenced cell performance, with the vegetated cell outperforming the non-vegetated cell under higher inflow pollutant concentrations. Multiple regression models also investigated several parameters influencing outflow NH3 and showed that inflow pH, temperature and NH3 concentration can be used to determine the outflow NH3 concentration of the cells. These models are important for predicting cell performance and thus can be used to improve the design and/or operation of the cells for varying inflow water quality conditions.

Comparing how wetlands are simulated in different hydrological modelling tools is needed to identify their suitability in different contexts. A simulated wetland will result in predictions of streamflow regulation, e.g., storing flood water and reducing high flows and releasing water in drier periods, which may or may not be realistic for a given area. Evaluating wetland models is critical for navigating the different types of physical wetlands with variable influences on streamflow, and the different simulated wetlands conceived in the plethora of modelling tools (i.e. software) available for use. A recent study found that sometimes wetlands are excluded from hydrological models used to inform water resource decisions. When wetlands are included in a hydrological model, few studies identify process similarities between the actual and modelled wetland or the realism of the modelled impacts of the wetland on streamflow before applying the model's output to water resource decisions. This research aims to identify and evaluate wetland characteristics, processes and impacts on catchment streamflow in different modelling tools and models (i.e. setups in a tool). Evaluating wetland models supports wetland-inclusive modelling and ensures that a wetland model is hydrologically sound and suitable. An unchannelled valley-bottom wetland located in the upper Kromme catchment, Eastern Cape, South Africa, was used. Wetland models were compared as independent units conceptually and as functional units within the catchment by modelling. First, using qualitative analysis, a conceptual assessment of wetland model structures in ACRU, WRSM-Pitman, MIKE SHE coupled with Hydro River and SWAT were considered in the context of the case study wetland. Second, using quantitative analysis, model outputs from wetland models in ACRU and WRSM-Pitman were assessed for model performance, behaviour and streamflow regulation during droughts and floods. The predicted impact of the wetland on catchment hydrology was determined from scenarios with and without a wetland and modelled wetland storage fluxes over the whole simulation period, four severe floods and three drought periods. The results from the qualitative and quantitative comparisons suggest that similarities between the physical and simulated wetland improves the likelihood of model suitability, good model performance and streamflow regulation predictions. Additionally, models setup for the same wetland with the same input data simulated potentially acceptable but different streamflow totals: for an observed total of 9.13 Mm3 ; WRSM-Pitman's comprehensive wetland simulated 10.64 Mm3 ; and from ACRU's riparian zone and wetland HRU's simulated 11.31 Mm3 and 8.89 Mm3 , respectively. Modelled actual evapotranspiration was underestimated by the riparian zone wetland (946.08 mm), overestimated in the comprehensive wetland model (2 054.80 mm) and moderately similar in the wetland HRU when compared with remotely-sensed data (1 520.30 mm). During extreme events, all models simulated flood attenuation while drought responses were variable (two wetland models predicted streamflow attenuation). By implication, the results suggest that good model performance does not guarantee the simulation of expected streamflow regulation roles recorded in literature. Furthermore, variable water yields and wetland impacts from the models demonstrated the possibility for different modelling efforts to result in different water supply, use and conservation measures. The study highlights the importance of contextualising model output for catchments with wetlands before applying the simulations to impact assessments or future climate scenarios.

Drylands farming areas are subject to land cover or land use change, climate change, and a variety of land management practices which may contribute to dust emissions and soil loss, a common symptom of land degradation. The goal of this study was to establish how the shift in land cover or land use change affects the occurrence of dust and determine the climate constraints and mitigations to crop production with focus on the dustiest parts of South Africa in the Free State Moreover, the study wished to establish land management strategies suitable for enhancing the storage of soil organic carbon and associated ecosystem services. To investigate how the shift in land cover or land use affects the occurrence of dust from 2006- 2016, the study firstly identified the spatial and temporal changes of land cover (LC) using the 2000 and 2013-2014 land cover maps derived from Agricultural Research Council. The results showed that grassland is still dominating in 2013-2014 but has decreased by 22 % from 2000. Cultivated land on the other hand has increased by 26 % from 2000 to 2013-2014. Some of the changes seen in recent data might also be the result of the associated data products. The change is however continuous and significant as major LC changes occurred over long-time scales. Cultivated land and grassland are dominant land cover and cultivated land is substantially being used to meet the economic demand (international maize price). Moderate Resolution Imaging Spectroradiometer Enhanced Vegetation Index (MODIS EVI) was used to assess vegetation cover for the site of the start of dust. MODIS EVI indicated that vegetation plays a significant role during the occurrence of atmospheric dust. Sparse vegetation (EVI ranging from 0 – 0.18) is directly proportional to dust production. Dust events were generally observed in cultivated land especially in 2015 and 2016. Moreover, findings showed that vegetation is not solely a factor that affects dust production; rainfall, temperature and wind speed also contribute. Crop and land cover data obtained from Department of Agriculture, Forestry and Fisheries (DAFF) identified the dominating crop and land cover around each dust source as fallow land, maize fields, and pasture. For instance, dust source August 29th 2006 had 11533 ha of fallow land and 10795 ha of maize, and dust source September 15th 2016 had 7737 ha of fallow land and 4599 ha of pasture which suggest that dust does not exclusively depend on the amount of vegetation but also on the type of vegetation. To achieve the second aim, the study used data from Climate Hazard Infrared Precipitation with Stations (CHIRPS), Agricultural Research Council (ARC) and Standardised Precipitation 2 Evapotranspiration Index (SPEI) to determine the variation of rainfall in Free State. SPEI data showed that drought is becoming prominent and severe in Free State (2005, 2012, 2014 and 2016). ARC data showed a slight increase (357-408 mm/year) from two stations and a slight decrease from the other two stations in rainfall from 2005- 2016 and CHIRPS data showed a slight decrease (458-404 mm/year) from all the stations during the same period. Both CHIRPS and ARC data showed that there is a seasonal shift on the onset of rainfall; rainfall has shifted to late summer by approximately 62 days on average from 1981-2019 in Free State. Rainfall in Free State is becoming late (e.g., in 2011 rainfall adequate to make the soil moist was received in November) and number of rain events are decreasing (e.g., September 1981-2019 had 7 rainfall events greater or equal to 15 mm/day, 19 events in October, 52 events in November and 33 in December in Bultfontein weather station). This study used temperature data from ARC weather stations to determine how temperature and frost variation affect crop production in Free State. Findings indicated that temperatures have increased by ~ 1- 2 ○C in Free State and frost has decreased, and its onset has shifted by ~ 11 days. Frost occurred in Julian day of the year (DOY) 144 in 2005 and in 2016 there was no frost for the whole year in Bultfontein. The onset of frost is shifting to May and resulting to frost period decreased over the years. Decrease in frost is an advantage for crops that are planted in late summer, however increased temperatures are likely to result in crop failure, especially if rainfall is also low. Furthermore, the relationship between wind speed and dust occurrence was determined using the ARC and ERA5 datasets. The results showed that some years were characterised by high winds and some by low winds but generally there is no constant trend. The ARC maximum wind speed was 10 - 15 m/s and ERA5 maximum wind speed was 10 m/s. Both datasets showed that high wind speed occurs from August- November and many dust events are observed in these months. Additionally, high winds are mostly observed to occur on surfaces with sparse vegetation and in the absence of wind breaks which further promotes an increase in dust. Thus, it is important for farmers to leave crop residues after harvesting and practice moderate grazing. To achieve the third aim, the study further established land management strategies suitable for enhancing the storage of soil organic carbon. Soils sampled in 2019 from different sites in Free State were used to determine the amount of carbon and nitrogen. It is apparent that grassland can store more carbon especially in the Lixisols which appeared to have stored an average carbon of 1.23 g and maximum of 1.86 g from 0-2 cm to 20-30 cm depths. Moreover, cultivated land can store less carbon especially in the Arenosols which stored an average of 0.18 g and 3 maximum of 0.22 g from 0-2 cm to 30-45 cm depths. Findings suggest that grassland can store more carbon than cultivated land due to minimal disturbance of the land when compared to cultivated sites. Moreover, the study demonstrated that soil texture and grain size distribution play an additional role in carbon storage with higher loads linked to higher silt content (Lixisols) and less carbon being stored in soils with high sand content (Arenosols). More carbon was observed to be stored in soils with moderate sand content (Lixisols = 40 % of sand) and high silt content (54 %) and less carbon was stored in soils with high sand content (Arenosols = 80 % of sand) and moderate silt content (18 %). More carbon was shown to be stored in the top layers of the soil 5-10 cm and was observed to decrease with increasing depth. DAYCENT carbon model was further used to simulate soil carbon in different land management scenarios to come up with suitable land management options for storing more carbon in future. Results illustrated that native grassland can substantially store more carbon than cultivated land, however moderate grazing should be applied. This would allow for a reduction of future dust emissions and increase ecosystem services in the cultivated lands of the Free State. Overall, the study demonstrated that combining the investigation of LCLU, climate change and land management was important to better understand land degradation and identified possible mitigation measures.

Contemporary biodiversity conservation is ‘wickedly complex'. This complexity stems from the need to address the diverse objectives of protecting biodiversity and enhancing social wellbeing. However, centralized and exclusionary conservation approaches are often ill-suited to tackling these coupled objectives. Consequently, increasingly calls have been made for the development of more holistic, participatory, nuanced and context-specific conservation governance approaches. Community-based conservation – which seeks to include local communities and their knowledge and priorities in conservation governance – offers a viable though context-specific alternative. However, thus far communitybased conservation initiatives have produced mixed results, largely due to a lack of understanding of how to effectively initiate, implement and manage such ‘wickedly complex' conservation initiatives. South Africa possesses enabling legislation for community-based conservation, but to date there has been no implementation of legally recognized communityconserved areas in the coastal zone. Accordingly, this research is guided by a desire to better understand this ‘policy-praxis disjuncture', and explores what factors, conditions and processes are required to enable South Africa to embrace a more community-orientated approach to conservation. It is proposed that greater understanding and potentially success can be gained by viewing communitybased conservation including, the initiation, implementation and governance of community-conserved areas, as a ‘change process'. Drawing on Commons Theory, Governance Theory, and the Theory of Change approach, a framework was developed to guide the exploration of the factors, conditions and processes that enable the shift to a community-based mode of conservation governance. Case study investigations were conducted in two established regional community coastal conservation cases, and one South African ‘case-in-progress'. Based on the findings of these cases, and the perceptions of South African conservation actors, this dissertation offers insights for tackling South Africa's policy-praxis disjuncture by developing a South African Empirical Community-Based Conservation Theory of Change Pathway. By exploring the initiation, implementation and governance of community-based conservation initiatives as a change process, this dissertation provides a framework for designing a process to facilitate and implement community-based conservation where contextually appropriate. More specifically, it emphasizes the need to develop a context-appropriate, strategic, systematic and iterative set of actions, with clearly articulated assumptions, which strive to address present or potential issues, to support the change to community-based governance. Consequently, this dissertation provides a framework for understanding how a shift to a community-based mode of conservation governance takes place, and offers a South African specific design pathway, with potential application by diverse conservation actors in other countries.