|dc.description.abstract||Globally, soil erosion by water is often reported as the worst form of land degradation owing to its adverse effects, cutting across the ecological and socio-economic spectrum. In general, soil erosion negatively affects the soil fertility, effectively rendering the soil unproductive. This poses a serious threat to food security especially in the developing world including South Africa where about 6 million households derive their income from agriculture, and yet more than 70% of the country’s land is subject to erosion of varying intensities. The Eastern Cape in particular is often considered the most hard-hit province in South Africa due to meteorological and geomorphological factors. It is on this premise the present study is aimed at assessing the spatial and temporal patterns of water-borne erosion in the Umzintlava Catchment, Eastern Cape, using the Revised Universal Soil Loss Equation (RUSLE) model together with geospatial technologies, namely Geographic Information System (GIS) and remote sensing. Specific objectives were to: (1) review recent developments on the use of GIS and remote sensing technologies in assessing and deriving soil erosion factors as represented by RUSLE parameters, (2) assess soil erosion vulnerability of the Umzintlava Catchment using geospatial driven RUSLE model, and (3) assess the impact of landuse/landcover (LULC) change dynamics on soil erosion in the study area during the period 1989-2017.
To gain an understanding of recent developments including related successes and challenges on the use of geospatial technologies in deriving individual RUSLE parameters, extensive literature survey was conducted. An integrative methodology, spatially combining the RUSLE model with Systeme Pour l’Obsevation de la Terre (SPOT7) imagery within a digital GIS environment was used to generate relevant information on erosion vulnerability of the Umzintlava Catchment. The results indicated that the catchment suffered from unprecedented rates of soil loss during the study period recording the mean annual soil loss as high as 11 752 t ha−1yr−1. Topography as represented by the LS-factor was the most sensitive parameter to soil loss occurring in hillslopes, whereas in gully-dominated areas, soil type (K-factor) was the overriding factor. In an attempt to understand the impact of LULC change dynamics on soil erosion in the Umzintlava Catchment from the period 1989-2017 (28 years), multi-temporal Landsat data together with RUSLE was used. A post-classification change detection comparison showed that water bodies, agriculture, and grassland decreased by 0.038%, 1.796%, and 13.417%, respectively, whereas areas covered by forest, badlands, and bare soil and built-up area increased by 3.733%, 1.778%, and 9.741% respectively, during the study period. The mean annual soil loss declined from 1027.36 t ha−1yr−1 in 1989 to 138.71 t ha−1yr−1 in 2017. Though soil loss decreased during the observed period, there were however apparent indications of consistent increase in soil loss intensity (risk), most notably, in the elevated parts of the catchment. The proportion of the catchment area with high (25 – 60 t ha−1yr−1) to extremely high (>150 t ha−1yr−1) soil loss risk increased from 0.006% in 1989 to 0.362% in 2017. Further analysis of soil loss results by different LULC classes revealed that some LULC classes, i.e. bare soil and built-up area, agriculture, grassland, and forest, experienced increased soil loss rates during the 28 years study period. Overall, the study concluded that the methodology integrating the RUSLE model with GIS and remote sensing is not only accurate and time-efficient in identifying erosion prone areas in both spatial and temporal terms, but is also a cost-effective alternative to traditional field-based methods. Although successful, few issues were encountered in this study. The estimated soil loss rates in Chapter 3 are above tolerable limits, whereas in Chapter 4, soil loss rates are within tolerable limits. The discrepancy in these results could be explained by the differences in the spatial resolution of SPOT (5m * 5m) and Landsat (30m * 30m) images used in chapters 3 and 4, respectively. Further research should therefore investigate the impact of spatial resolution on RUSLE-estimated soil loss in which case optical sensors including Landsat, Sentinel, and SPOT images may be compared.||en_US