Massive forest mortality was observed in California during the most recent drought. Owing to complex interactions of physiological mechanisms under stress, prediction of climate-induced forest mortality using dynamic global vegetation models remains fraught with uncertainty. Given that forest ecosystems approaching mortality tend to exhibit reduction in resilience, we evaluate the time-varying resilience from time series of satellite images to detect early warning signals (EWSs) of mortality. Four metrics of EWSs are used: (1) low greenness, (2) high empirical autocorrelation of greenness, (3) high autocorrelation inferred using a Bayesian dynamic linear model considering the influence of seasonality and climate conditions, and (4) low recovery rate inferred from the drift term in the Langevin equation describing stochastic dynamics. Spatial accuracy and lead-time of these EWSs are evaluated by comparing the EWSs against observed mortality from aerial surveys conducted by the US Forest Service. Our results show that most forested areas in California that underwent mortality exhibit a EWS with a lead time of three months to two years ahead of observed mortality. Notably, EWS is also detected in some areas without mortality, suggesting reduced resilience during drought. Furthermore, the influence of the previous drought (2007-2009) may have propagated into the recent drought (2012-2016) through reduced resilience, hence contributing to the massive forest mortality observed recently. Methodologies developed in this study for detection of EWS will improve the near-term predictability of forest mortality, thus providing crucial information for forest and water resource management.
Liu, Yanlan, Mukesh Kumar, Gabriel George Katul, Amilcare M Porporato (2017): Detection of early warning signals of forest mortality in California. American Geophysical Union 2017 Fall Meeting, New Orleans, Louisiana, 11-15 December 2017.
This Paper/Book acknowledges NSF CZO grant support.