Towards improved remote sensing based monitoring of dryland ecosystem functioning using sequential linear regression slopes (SeRGS)

Abstract We present a method for remote sensing based monitoring of changes in dryland ecosystem functioning based on the assumption that an altered vegetation rainfall relationship (VRR) indicates changes in vegetation biophysical processes, potentially leading to changes in ecosystem functioning. We describe the VRR through a linear regression between integrated rainfall and vegetation productivity (using NDVI as a proxy) within a combined spatio-temporal window, sequentially moved over the study area and along the temporal axis of a time series. The trend in the slope values derived from such a sequential linear regression, termed SeRGS, thus represents a measure of change in the VRR. Scenarios of land degradation, defined here as a reduction in biological productivity, which may be caused by either climatic or anthropogenic factors are simulated for the period 1970–2016 from CRU rainfall and modelled NDVI data to test and evaluate the performance of the SeRGS method in detecting degradation, and compare it against the well-known RESTREND method. We found that SeRGS showed (1) overall more pronounced trends and higher significance levels (p ≤ 0.01) in detecting degradation events and (2) an improved statistical basis for the calculation of trends in the VRR (expressed by high coefficients of determination throughout the period of analysis), which was found to increase the validity of the results produced. Through the implementation of the temporal moving window the effect of inter-annual rainfall variability on vegetation productivity was effectively reduced, thereby enabling a more exact and reliable identification of the timing of degradation events (e.g. start, maximum and end of degradation) by using a time series breakpoint analysis (BFAST). Finally, the SeRGS method was applied using real data for Senegal (seasonally integrated MODIS NDVI and CHIRPS rainfall data 2000–2016) and we discuss patterns and trends. This study provides the theoretical basis for an improved assessment of changes in dryland ecosystem functioning, which is of relevance to land degradation monitoring targeting loss of vegetation productivity.