Early warning signals of grassland ecosystem degradation: A case study from the northeast Qinghai-Tibetan Plateau

As a typical vulnerable ecosystem, alpine steppes can be subject to nonlinear accelerations in degradation, once tipping points are reached or exceeded, and thus far more effort and investment will be needed, to reverse the degraded status of such ecosystems, if effective preventive measures or interventions are not adopted sufficiently early. Thus, detecting early warning signals of tipping events is extremely important for ecosystem protection and recovery, but remains highly challenging. For alpine steppes, coverage or biomass loss and species invasion are both key transitions. Using multi-source data, including ground surveys and remote sensing images with different resolutions, we tried to screen and identify potential early-warning signals for alpine grassland degradation events and analyze their efficacy. The results showed that the degradation levels derived from spatial heterogeneity information combined with a vegetation index were reliable, as was verified by ground surveys, and could illustrate the degradation process of a grassland ecosystem. Six common early-warning signals (variance, skewness, and autocorrelation, from both the spatial and temporal domains) were tested according to degradation degree, and spatial autocorrelation and temporal autocorrelation had the best performances, as they could indicate different degradation levels, and the change trends of temporal autocorrelation could provide early-warning functions. A waved change of spatial variance could reflect a vegetation growth dynamic that might indicate species invasion. In this study, the threshold values of spatial and temporal autocorrelation were 0.7 and 0.2, respectively, and from these, the thresholds of vegetation coverage and biomass were then obtained. However, these signals and thresholds had some limitations during application. These findings provide insights into better understanding of the alpine grassland degradation process, and contribute to a better management approach for vulnerable ecosystems using early-warning methods.