Paper Title
Distinguishing Between Human-Induced Land Degradation From Effects of Rainfall: Case of The Greater Sekhukhune District Municipality

Abstract
Land degradation (LD) is a global issue that affects the sustainability and livelihoods of approximately 1.5 billion people, especially in arid/semi-arid regions. Assessing and identifying LD, especially driven by anthropogenic activities, is important for proposal of suitable sustainable land management interventions. Therefore, the study aimed to distinguish anthropogenic LD from rainfall effects in The Greater Sekhukhune District Municipality from 1990 to 2019. Vegetation production, thus Normalized Difference Vegetation Index (NDVI) from Advanced Very High-Resolution Radiometer (AVHRR) was used as an indicator for LD. Rainfall data from Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) has been widely proven to be highly correlated with vegetation productivity, hence was a climatic factor for assessment of vegetation dynamics. Linear regression was performed between NDVI and rainfall. Human-induced LD was then distinguished from that of rainfall using residual trend (RESTREND) method and Mann-Kendall trend test. Spatial RESTREND revealed that 11.59% of the district is degrading due to human impacts such as overgrazing and lack of rangeland management while 41.41% is due to rainfall impacts such as severe droughts in 1992, 2002-2004 and 2015. Additionally, climate variability affected vegetation and contributed to soil erosion and gully formations. Increase in vegetation biomass (53.83%) in other areas was noted to be result of bush encroachment (sign of LD) caused by human activities i.e., overgrazing and abandoned agricultural fields. These findings are crucial as they provide spatial information on rainfall or human-induced LD useful for policy formulation and designing LD mitigation measures in semi-arid regions. Keyword - Land Degradation, NDVI, Rainfall, Mann-Kendall trend, Land Use and Land Cover Change, Restrend.