Item

Abstract

Understanding how resilient rangelands are to climatic disturbances such as drought is of major importance to land managers. The resilience of ecosystems can be reduced by livestock grazing and by environmental conditions. Most studies quantifying resilience are based on model simulations. However, natural time series from satellite data offer the possibility to infer aspects of resilience from real systems. The objective of this study was to investigate two aspects of ecological resilience, namely resistance to climate variability and recovery from drought, by applying a change detection method (Breaks For Additive Seasonal and Trend; BFAST) spatially on a 28-year Landsat NDVI time series in a dry rangeland in southern Cyprus. First, we used the number of breakpoints fitted by the BFAST model as an inverted proxy for long-term vegetation resistance to climate variability (the ability to withstand change during a disturbance reduces the likelihood to trigger a breakpoint in the time series). Second, we used the linear slope of the BFAST model after a known drought as a proxy of the recovery rate of the vegetation. This information was then used to analyse the spatial distribution of the total number of breakpoints and of the NDVI recovery trend in relation to grazing and environmental properties. Our results show that high NDVI and a northern orientation (i.e. favourable environmental conditions) were associated with a highly resilient system, due to high resistance to climate variability and fast recovery after drought. Intermediate conditions were associated with low resistance. Unfavourable conditions and high grazing intensities were associated with an unresponsive ecosystem state characterised by high resistance and slow recovery after a drought event. Low grazing intensities positively affected the NDVI recovery trend, but did not improve resistance. On northern slopes, terrain slope had a positive effect on the NDVI recovery trend, while on southern slopes it had a negative effect. Our satellite-driven approach has a strong potential for resilience monitoring, because it can be applied on broad spatial and temporal scales in areas with low availability of field data. Moreover, it allows to jointly extract two important components of resilience: resistance and recovery rate.