Detecting dryland degradation using Time Series Segmentation and Residual Trend analysis (TSS-RESTREND).

Dryland degradation is an issue of international significance as dryland regions play a substantial role in global food production. Remotely sensed data provide the only long term, large scale record of changes within dryland ecosystems. The Residual Trend, or RESTREND, method is applied to satellite observations to detect dryland degradation. Whilst effective in most cases, it has been shown that the RESTREND method can fail to identify degraded pixels if the relationship between vegetation and precipitation has broken-down as a result of severe or rapid degradation. This paper presents an extended version of the RESTREND methodology that incorporates the Breaks For Additive Seasonal and Trend method to identify step changes in the time series that are related to significant structural changes in the ecosystem, e.g. land use changes. When applied to Australia, this new methodology, termed Time Series Segmentation and Residual Trend analysis (TSS-RESTREND), was able to detect degradation in 5.25% of pixels compared to only 2.0% for RESTREND alone. This modified methodology was then assessed in two regions with known histories of degradation where it was found to accurately capture both the timing and directionality of ecosystem change.

Fig. 5. An example pixel from the Simpson-Strzelecki Dunefield Bioregion with a breakpoint in the VPR. The orange represents the values before the breakpoint and the purple represents the values after the breakpoint. a) NDVImax vs time. The dotted red line indicates the position of the detected breakpoint. b) The change in the VPR before (orange) and after (purple) the breakpoint. The dotted grey line represents the VPR that was fitted to the data by a standard RESTREND and the black bar represents the break height (BH = 0.0442). c) The segmented RESTREND applied using the segmented VPR. The red bar indicates the residual change (rc = 0.0488). The total change is calculated by adding the VPR break height to the residual change (T.C. = 0.093). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)