Improved characterization of dryland degradation using trends in vegetation/ rainfall sequential linear regression (SERGS-TREND)

Land degradation in drylands has been investigated extensively over recent decades and several remote sensing based techniques attempt to decouple the human influence from the natural climate variability, but are contested in literature. We introduce a novel approach termed SeRGS-TREND that is designed to monitor land degradation by suppressing the impact from climate variability and highlight vegetation disturbances may it be human or climate-induced. SeRGS-TREND is based on the interpretation of the slope of a linear regression analysis within a sequentially moving window along the temporal axis of the time series of remote sensing data. The use of a moving window increases the probability of a statistically significant linear vegetation-rainfall relationship (VRR), which in turn provides an improved statistical basis for the results produced and thereby confidence in the assessment of degradation. We test and compare SeRGS-TREND and the commonly used RESTREND by simulating different degradation scenarios and find that SeRGS reveals both, more significant and more exact information about degradation events (e.g. starting and end point) while keeping the VRR correlation coefficients high, thus rendering results more reliable.