Аннотация
Decomposing the responses of ecosystem structure and function in drylands to changes in human-environmental forcing is a pressing challenge. Though trend detection studies are extensive, these studies often fail to attribute them to potential spatiotemporal drivers. Most attribution studies use a single empirical model or a causal graph that cannot be generalized or extrapolated to larger scales or account for spatial changes and multiple independent processes. Here, we proposed and tested a multi-stage, multi-model framework that detects vegetation trends and attributes them to ten independent social-environmental system (SES) drivers in Kazakhstan (KZ). The time series segmented residual trend analysis showed that 45.71% of KZ experienced vegetation degradation, with land use change as the predominant contributor (22.54%; 0.54 million km2), followed by climate change and climate variability. Pixel-wise fitted Granger Causality and random forest models revealed that sheep & goat density and snow cover had dominant negative and positive impacts on vegetation in degraded areas, respectively. Overall, we attribute vegetation changes to SES driver impacts for 19.81% of KZ (out of 2.39 million km2). The identified vegetation degradation hotspots from this study will help identify locations where restoration projects could have a greater impact and achieve land degradation neutrality in KZ. A detection, contribution, and attribution framework analysis suggests that almost 46% of the land area in Kazakhstan has experienced vegetation degradation, with grazing and snow cover variability identified as the principal drivers of degradation.