Mapping soil erosion in a quaternary catchment in Eastern Cape using geographic information system and remote sensing

Abstract

In South Africa, soil erosion is considered as an environmental and social problem with serious financial implications particularly in some rural areas where this geomorphological phenomenon is widespread. An example is the Umzimvubu Local Municipality, where most households are strongly reliant on agriculture for their livelihood. Sustainable agriculture and proper land management in these rural areas require information relevant to the spatial distribution of soil erosion. This study was therefore aimed at generating such information using Landsat8 Operational Land Imager (OLI)-derived vegetation indices (VIs) including the Normalised Difference Vegetation Index (NDVI), Soil Adjusted Vegetation Index (SAVI), as well as Soil and Atmospherically Resistance Vegetation Index (SARVI). Raster calculator in ArcMap10.2 was used to classify soil erosion features based on selected suitable thresholds in each VI. SPOT6/7 (Satellites Pour l’Obsevation de la Terre) multispectral data and Google Earth images were used for ground truth purposes. SAVI achieved the highest overall classification accuracy of 83% and kappa statistics of 64%, followed by NDVI and SARVI with equal overall accuracy of 81% and slightly different kappa statistics of 60% for the former and 59% for the latter. Using these indices, the study successfully mapped the spatial distribution of soil erosion within the study area albeit there were some challenges due to coarser spatial resolution (15mx15m) of Landsat8 image. Due to this setback, image fusion and pan-sharpening of Landsat8 with higher spatial resolution images is strongly suggested as an alternative to improve the Landsat8 spatial resolution.

Keywords: Geographic Information System; Remote Sensing; Soil Erosion; Vegetation Indices