School of Geography, Archaeology and Environmental Studies (ETDs)
Permanent URI for this communityhttps://hdl.handle.net/10539/38007
Browse
Search Results
Item A Geospatial Approach to Mapping Jacaranda Tree Distribution in Johannesburg, South Africa(University of the Witwatersrand, Johannesburg, 2023-11) Reddy, Rohini Chelsea; Fitchett, JenniferAccurate mapping of the spatial distribution of invasive species is vital for the implementation of effective monitoring and management strategies. In countries where resources are scarce and costly, citizen science provides a cost-effective and accurate alternative for large-scale data collection. Citizen’s familiarity with their environment contributes to aspects such as accurate identification of features on the landscape. Advances in a geographic information system (GIS) together with open-sourced photography from Google Street View, provide accurate methods for in-field and remote validation of citizen science data for invasive mapping and assists with the creation and compilation of maps to visualize the spatial distribution of invasive plants upon the landscape. In this study, the first spatial distribution maps for invasive tree species, Jacaranda mimosofolia (common name: Jacaranda), are created for the City of Johannesburg (CoJ). Jacaranda trees are well-known by citizens in the CoJ for their district purple flowers which blanket the landscape during springtime. A combination approach using citizen science, GIS, and Google Street View for data collection, analysis, and creation of the first spatial distribution map of exact location and prevalence of Jacaranda trees within certain suburbs of the CoJ, is produced. A total of 8,931 ground-truthing geopoints together with extensive Google Street View validation for Jacaranda tree presence, formed the basis of accurate spatial distribution maps. The first research question of this study focused on the spatial distribution of Jacaranda trees in the CoJ and was answered as a total of 54 suburbs were confirmed as having a large presence of Jacaranda trees in the CoJ. Citizen science data collected a total of 488 geotags for possible Jacaranda tree presence in the CoJ, over a 75-day online survey collection period. Although citizen science data provided a lower spatial resolution compared to successful fieldwork and Google Street View approaches, citizen science data provided very high accuracy for the identification and geolocation of Jacaranda tree presence in the CoJ which answers the second research question based on the effectiveness of the geospatial approach towards citizen science, ground-truthing and Google Street View as data collection methods. Since the accuracy of citizen science resulted in 66% of collected geotags within the categories of ‘very high’, ‘high’ and ‘moderate’ accuracy ranges of between <7-24m from a confirmed Jacaranda tree, together with the accuracy of 8,931 in-field collected geolocation of Jacaranda trees and Google Street View’s accuracy and capability of collecting street view imagery – it is concluded that the combined approach of ground-truthing, citizen science and Google Street View contribute not only to effective data collection, but also towards the successful mapping of Jacaranda tree presence in the CoJ.Item Monitoring and evaluating urban land use land cover change using machine learning classification techniques: a case study of Polokwane municipality(University of the Witwatersrand, Johannesburg, 2023) Funani, Tshivhase; Mhangara, PaidaRemote sensing is one of the tools which is very important to produce Land use and land cover maps through the process of image classification. Image classification requires quality multispectral imagery and secondary data, a precise classification technique, and user experience skill. Remote sensing and GIS were used to identify and map land-use/land-cover in the study region. Big Data issues arise when classifying a huge number of satellite images and features, which is a very intensive process. This study primarily uses GEE to evaluate the two classifiers, Support Vector Machine, and gradient boosting, using multi-temporal Landsat-8 images, and to assess their performance while accounting for the impact of data dimension, sample size, and quality. Land use/Landcover (LULC) classification, accuracy assessment, and landscape metrics comprise this study. Gradient Tree Boost and SVM algorithms were used in 2008, 2013, 2017, and 2022. Google Earth Engine was used for supervised classification. The results of change detection showed that urbanization has occurred and most of the encroachments were on agricultural land. In this study, XG boost, and support vector machine (SVM)) were used and compared for image classification to oversight spatio-temporal land use changes in Polokwane Municipality. The Google Earth Engine has been utilized to pre-process the Landsat imagery, and then upload it for classification. Each classification method was evaluated using field observations and high-resolution Google Earth imagery. LULC changes were assessed, utilizing Geographic Information System (GIS) techniques, as well as the dynamics of change in LULCC were analysed using landscape matrix analysis over the last 15 years in four different periods: 2008–2013, 2018 and 2022. The results showed that XGBoost performed better than SVM both in overall accuracies and Kappa statistics as well as F-scores and the ratio of Z-score. The overall accuracy of gradient boosting in 2008 was 0.82, while SVM showed results of 0.82 overall accuracy and kappa statistics of 0.69. The average F-score for SVM in 2008 was from 0.58- 1.00, in 2013 an average of 0.86-0.97, and in 2022 it was 0.76. Z values were not statistically significant as all values were below the z score of 1.96. The ratios for the two classifiers were also taken to know which classifier performs the best. The results showed 212:212 which indicates that during 2008 SVM and XG boost performed the same way as they classified the same number of cases. During 2013 the ratio was 345:312 which shows that XGBoost performed better than SVM. The results of 2017 show 374:316 which shows that XGBoost performed better than SVM. Lastly, in 2022 the ratio was 298:277 which shows that XGBoost performed better than SVM. Overall zscores result show that XGBoost performs better than SVM. Overall, this study offers useful insight into LULC changes that might aid shareholders and decision makers in making informed decisions about controlling land use changes and urban growthItem A GIS framework for the integrated conceptualisation, analysis and visualisation of Gauteng's complex historic and contemporary post-mining urban landscape(University of the Witwatersrand, Johannesburg, 2023) Khanyile, Samkelisiwe; Esterhuysen, Amanda; Kelso, ClareThis research study applies assemblage theory as a philosophical lens. It proposes a framework for integrating contemporary and historical landscape characteristics of post-mining and urban landscapes for an integrated conceptualisation, mapping, and analysis of Gauteng, South Africa. The study utilises a mixed methods approach, incorporating spatial and non-spatial (literature and survey) data of varying formats to identify landscape characteristics. Additionally, it applies three multicriteria decision analysis (MCDA) and GIS mapping techniques, employing a simplified rationale to keep its complexity low. Descriptive and inferential statistics were used to analyse the quantitative data, while the qualitative data was analysed using a thematic analysis. The literature and survey analysis findings were used to inform the development of a framework demonstrating the integration of Gauteng's post-mining and urban landscape characteristics using a fuzzy overlay, weighted overlay and random forest classification, along with an accuracy assessment of the mapped results. Based on the proposed framework, the mapped results' performance was evaluated through four methods: confusion error matrix, cross-evaluation, areal coverage comparison, and an image differencing assessment. The literature and survey analysis findings, used to inform the framework, reveal that the two landscapes consist of an assemblage of characteristics and highlight differences in the characterisation of post- mining and urban landscapes. Distinctions were also apparent between literature-derived characteristics and those identified from local experts. The local expert-derived characteristics demonstrate context- specific characteristics of Gauteng's post-mining and urban landscape. At the same time, those based on the literature emphasise a more distinct and separate portrayal of post-mining and urban landscape characteristics (pages 115-116). The characteristics identified from local experts were less conservative (pages 117-118). They included urban-related characteristics in the description of post-mining landscapes and mining-related characteristics in the description of urban landscapes, presenting some similarities in the characterisation of these two landscapes in Gauteng. Moreover, the findings from local experts also revealed that literature and other written or mapped work informed most definitions of post-mining and urban landscapes. The framework for integrating landscape characteristics (pages 121-123) was spatially represented through the three mapping methods, visually demonstrating several findings providing insight into the Gauteng landscape's uniqueness. First, it demonstrates that the differences in the characterisation of these landscapes also impact how they are spatially represented. The maps of post-mining and urban landscape characteristics based on the literature presented a similar pattern to the traditional mapping of mining and urban landscapes in Gauteng. These mapping techniques show the highest values across the mining belt and at the province's core. These findings highlight the influence of literature on the representation of these two landscapes, which is consistent with local experts' reports. In all three mapping methods, the maps generated from local expert characterisations of post-mining and urban landscapes presented a larger spatial footprint than those based on literature-derived characteristics. This distinction was attributed to incorporating additional post-mining and urban landscape characteristics in the maps based on expert input and applying the three mapping techniques - using representation methods not commonly used in mapping these landscapes. Second, the integrated maps of post-mining and urban landscape characteristics suggested a variance in the presence of post-mining and urban landscape characteristics across the province in the maps generated using fuzzy and weighted overlay techniques. This indicates that some parts of the province have a higher or lower presence of post-mining or urban characteristics (pages 125-132). These findings were visible in the maps generated from literature and local experts, indicating the diversity of both landscapes and the co-existence of post-mining and urban landscape characteristics in the local expert maps. This implies an intricate relationship between these landscapes, challenging the idea of them being strictly separate, as indicated in maps presenting characteristics identified from the literature. Furthermore, a closer inspection of the areas showing the intersection between post-mining and urban landscape characteristics also points towards the porosity of boundaries of these two landscapes and alevel of spatial overlap, organisation and arrangement, which are prevalent at varying levels (pages 164- 168). Third, the maps generated using literature-derived characteristics achieved higher accuracy scores, attributed to using reference data traditionally used to map the two landscapes under investigation. This reference data only comprised classes that characterised the physical mining and urban classes, consistent with those identified in the literature. Consequently, it lacked additional factors characterising the post-mining and urban landscape identified from local experts. The fuzzy overlay maps informed by literature demonstrated an accuracy exceeding 70% for post-mining and urban landscape characteristics. In comparison, those reported by local experts scored 64. The weighted overlay and random forest classification resulted in accuracy rates exceeding 50% for post-mining landscape characteristics maps, regardless of whether literature or expert-derived characteristics were used. Additionally, urban landscape characteristics maps achieved an accuracy of over 76%, regardless of the characteristics used to inform the mapping. These findings were attributed to the different mapping techniques employed, with fuzzy and weighted overlay using a gradual range scale, while random forest classification employed a binary scale. This highlights how different mapping methods affect the representation of space. Additionally, it demonstrates the versatility of these mapping techniques in mapping complex spaces such as post-mining and urban landscapes. In this study, the fuzzy overlay accuracies exceeded 60% for all maps and emerged as the most suitable choice for integrating landscape characteristics due to its ability to represent blurred and porous boundaries between Gauteng's post- mining and urban landscapes. In conclusion, the study challenges the notion of post-mining and urban landscapes as distinct landscapes, emphasising the importance of considering the varying levels of spatial intersection between these two landscapes. With the proposed framework and the alternative representation of these landscapes, including contextual information, this research provides insights into new conceptualisations of urban, post-mining landscapes and mineralised urbanisations as assemblages of different landscapes and characteristics with porous boundaries. This enables a better understanding of Gauteng's post-mining and urban landscapes, which could benefit the representation, communication and management of these landscapes. Recognising the potential applications and limitations of frameworks such as the one developed for this study, the high-level recommendation arising from this study suggests a need for ongoing research into the contextual representation of landscapes and their characteristics. This can be achieved by incorporating input from communities, conducting research on quantifying intangible landscape characteristics and developing tools that facilitate the automation and alignment of such data with development plans.Item Exploring Spatio-Temporal Climate Dynamics over Central Southern Africa: A Cross Border Analysis(University of the Witwatersrand, Johannesburg, 2023-07) Welff, Megan; Fitchett, Jennifer; Esterhuysen, AmandaUnderstanding the diverse nature of climate dynamics in southern Africa is imperative in the face of climate change. Ground-based meteorological stations provide high-resolution climate data that can be used to investigate and analyse climate in detail. However, southern African countries monitor and manage meteorological stations independently which presents various challenges when attempting cross-border studies. While there are many meteorological-station-based climate studies conducted for South Africa or Botswana, there are few that combine meteorological datasets from both these countries to investigate climate dynamics across political boundaries. In this study, meteorological data from Botswana Meteorological Services and the South African Weather Service spanning 1912-2019 is pre-processed, cleaned and combined to produce a cross-border dataset. A total of 44 stations covers the Gauteng and North West provinces in South Africa and the southern, Kweneng, Kgatleng, South-east and Kgalagadi districts of Botswana. The combined cross-border dataset includes the average monthly summer, winter and annual rainfall (RS, RW and RA respectively) and the average monthly minimum and maximum summer, winter and annual temperatures (TSmin, TSmax, TWmin, TWmax, TAmin and TAmax respectively). From the linear regression analysis, an overall increasing trend for temperature is identified barring two stations (TSmin and TAmin for Mahalapye Met station, and TWmin for Vaalharts). Additionally, for rainfall there is a significant decreasing trend identified. Lastly, the spatial variability of the region is determined using an Inverse Distance Weighted interpolation in the GIS Software, ArcMap, to interpolate between stations. From this a west to east reduction in rainfall and a north-western to south-eastern decreasing temperature gradient is identified across the study region.