Understanding the drivers of tree mortality is critical for both understanding forest dynamics and predicting forest responses to global change. Indeed, the rate of vegetation change and the potential for threshold responses are driven much more by changes in disturbance and mortality than changes in productivity, yet compared to the sophistication in models of photosynthesis and growth, our ability to predict mortality remains simplistic. Herein we present a hierarchical Bayes analysis of patterns of individual tree mortality for the eastern and central U.S. based on the USFS Forest Inventory and Analysis (FIA) that integrates large number of data sets across many spatial scales. A state-space model is employed to estimate the latent mortality status for each tree between census intervals and thus leverages the spatial asynchrony in the FIA to estimate interannual variability. Covariates range from regional climate gradients, anthropogenic pollutants (acid-rain, N-deposition, ozone), and epidemic pests, to landscape variation in topography, soil moisture, and soils, and finally down to the tree level (species, size, crowding). Finally, we are also assessing the ability of a regional-scale forest ecosystem model, the Ecosystem Demography model (ED) to capture mortality patterns based on individual estimates of carbon balance.

Results/Conclusions

We found that the strongest predictors of tree mortality were regional patterns of precipitation and temperature, though the magnitude and direction of the effects of these varied among functional and taxonomic groups. Size was also a consistently strong predictor. In addition to observing the expected decline in mortality as juveniles mature to adults and the expected increase in mortality as adults begin to senesce, we were surprised to find a consistent mid-life pulse of mortality. We hypothesize that this pulse is related to stand thinning, though further exploration of this pattern is needed. Mortality patterns also varied consistently with shade-tolerance class, with greater shade-tolerance conferring lower overall mortality and later senescence, but preliminary analyses do not show a clear phylogenetic signal. In terms of anthropogenic pollutants, increases in both acid and nitrogen deposition moderately increased mortality, while the effects of ozone were weak but statistically significant. Despite the large sample size, interannual variability was hard to detect because of the long census intervals (average 10yrs). Preliminary analyses suggest that simulated carbon balance correlates with mortality rates but that the ED model's mortality routine underestimates regional patterns associated with climatological gradients.