Climate variability and forest change

Background/Question/Methods

Forests are constantly undergoing change whether by ongoing succession, human activity, or natural disturbances including fire, pests, and severe weather. Future climate forecasts indicate higher global average annual temperature but also increased climate variability. While increased temperature could lead to geographic shifts in ecological niches and increased energy availability for weather-driven disturbance events (among other effects), it is unknown the role increased climate variability will play in driving forest change. Our objective was to examine the role of climate variability in tree mortality and regeneration. Using the 4-km PRISM and 32-km NARR climate datasets, inter-annual variability metrics were derived for Forest Inventory and Analysis (FIA) plot locations across the eastern United States. Tree mortality and seedling abundance data were obtained from FIA plots that were visited twice in the years from 2000 to 2010. A two-stage analysis first used logistic regression to relate predictors to the presence/absence of mortality or seedling abundance change and subsequently applied a mixed effects model, conditional on presence, to relate predictors to the magnitude of observed changes.

Results/Conclusions

Models that excluded climate information altogether did not perform as well as those that included annual mean values or inter-annual standard deviation of precipitation or temperature.  Models that included both mean and variability climate information outperformed all other models using either the PRISM or NARR data.  Model performance was diagnosed using a variety of measures, such as residual deviance, Tjur's R2 (coefficient of determination), a Hosmer-Lemeshow test, and an information criterion (AIC). Results indicate probability of tree mortality is positively related to inter-annual standard deviation in temperature or precipitation, implying the possibility that mortality will increase with increasing inter-annual climate variability.  Model refinements include  identification of potential variable interactions and improvement of overall model fit.  In addition to inter-annual climate variability, metrics that incorporate intra-annual or inter-seasonal variations may be important predictors of mortality and seedling abundance.