Monitoring land degradation in Southern Africa by assessing changes in primary productivity.

摘要

Land degradation is one of the most serious environmental problems of our time. Land degradation describes circumstances of reduced biological productivity. The fundamental goal of this thesis was to develop land degradation monitoring approaches based on remotely sensed estimates of vegetation production, which are capable of distinguishing human impacts from the effects of natural climatic and spatial variability. Communal homelands in South Africa (SA) are widely regarded to be severely degraded and the existence adjacent, non-degraded areas with the same soils and climate, provides a unique opportunity to test regional land degradation monitoring methods. The relationship between 1km AVHRR, growth season sumNDVI and herbaceous biomass measurements (1989-2003) was firstly tested in Kruger National Park, SA. The relationship was moderately strong, but weaker than expected. This was attributed to the fact that the small areas sampled at field sites were not representative of the spatial variability within 1x1km. The sumNDVI adequately estimated inter-annual changes in vegetation production and should therefore be useful for monitoring land degradation. Degraded areas mapped by the National-Land-Cover in north-eastern SA were compared to non-degraded areas in the same land capability units. The sumNDVI of the degraded areas was consistently lower, regardless of large variations in rainfall. However, the ecological stability and resilience of the degraded areas, as measured by the annual deviations from each pixel's mean sumNDVI, were no different to those of non-degraded areas. This suggests that the degraded areas may be in an alternative, but stable ecological state. To monitor human-induced land degradation it is essential to control for the effects of rainfall on vegetation production. Two methods were tested (i) Rain-Use Efficiency (RUE=NPP/Rainfall) and (ii) negative trends in the differences between the observed sumNDVI and the sumNDVI predicted by the rainfall using regressions calculated for each pixel (RESTREND). RUE had a strong negative correlation with rainfall and did not provide a reliable index of degradation. The RESTREND method identified areas in and around the degraded communal lands that exhibit negative trends in production per unit rainfall. This research made a significant contribution to the development of remote sensing based land degradation monitoring methods.