Leading—and misleading—indicators of grassland to shrubland regime shifts

Background/Question/Methods There have been extensive efforts to create theoretically derived leading indicators or “warning signs” that physical, physiological, ecological, and societal systems are approaching a critical threshold beyond which a catastrophic shift will occur. If these indicators prove reliable, it would be possible to foretell impending critical transitions with very little knowledge of systems, providing a powerful tool for avoiding undesirable state transitions. As far as we are aware, however, this body of knowledge has yet to be applied to large terrestrial systems. We used a unique long-term data-set (>30 years), with substantial spatial replication (n >150) to test whether spatial auto-correlation and “flickering” act as leading indicators for a well-documented regime shift of grassland to shrubland. Our approach was hierarchical, first looking at leading indicators of transitions at the landscape-scale (>1 km²) and then focusing on the patch-scale (~1 km²).

Results/Conclusions We found that when applied at landscape scales—the scale typically observed using remote sensing techniques—leading indicators generate false positives and fail to warn of impending transitions. However, when we applied these methods at patch-scales, we were able to identify leading indicators with sufficient warning time to engage in resilience-based management. The decision to focus on certain patches was based on prior knowledge of the system’s hydrology and ecology, allowing us to define the “patch scale”, group patches of similar resilience, and identify leading indicators. We conclude that leading indicator theory has considerable potential for application in terrestrial systems, but will require some a priori understanding of the system for successful implementation.