Tree death from drought and heat stress is a critical and uncertain component in forest ecosystem responses to a changing climate. Recent research has illuminated how tree mortality is a complex cascade of changes involving interconnected plant systems over multiple timescales. Explicit consideration of the definitions, dynamics, temporal and biological scales of tree mortality research can guide experimental and modeling approaches. We examine here proposed definitions of tree mortality and their implications for the physiological variables measured in experiments and modeling approaches. We then test the ability of two different modeling approaches with differing degrees of physiological realism to simulate spatial patterns of widespread, drought-induced aspen forest die-off in western Colorado. The first modeling approach involves simulation of meteorological drought and soil moisture during the severe 2000-2003 period in Colorado. The second approach incorporates in a plant water deficit component in line with a water balance calculation for a tree water reserve. We compare these results against satellite estimates in the San Juan National Forest and aerial survey estimates of forest mortality at a statewide scale.
Results/Conclusions
We first propose a definition of tree death based on a failure of whole-organism function and transport that can guide modeling approaches. From our modeling exercises, we find that elevated 2002 summer temperature and low shallow soil moisture were most associated with the spatial patterns of aspen mortality at a statewide scale. Four modeled climate and soil parameters explained around 32% of the spatial variance in watershed-scale mortality. These results agreed with ongoing stable isotope work regarding from what soil layers and precipitation source aspens draw their water. The second modeling approach revealed that topographically mediated controls on climatic water deficit provide a useful way of simulating tree mortality from a water balance approach. We conclude by discussing the relative importance of physiological understanding in hindcasting and forecasting forest mortality at different spatial and temporal scales, as well as key gaps in data needed for testing and evaluating models.