The relationship between climate variation and tree growth encoded in annual growth rings has long been exploited by dendroclimatologists to reconstruct past climates. More recently, statistical climate-growth relationships are being used for ecological forecasting of future tree growth.
Here, we forecast future growth of one of North America’s most important tree species, Pseudotsuga menziesii (Douglas-fir), using the largest tree-ring dataset compiled for this species to date, comprising >29,000 time series from >2,600 locations from the northern distribution limit in British Columbia to the southern edge in southern Mexico. Given this exceptionally large geographic distribution, spanning >30° of latitude, our analysis explicitly accounts for how climate sensitivities vary continuously across environmental gradients. We analyzed climate-growth relationships using a generalized linear mixed effects (LME) model that simultaneously analyzes the influences on tree growth of three groups of predictors: (1) tree size and age (circumventing the need for detrending of raw ring widths), (2) annually varying climate variables, and (3) 30-year climate normals. Our analysis included “representative” samples of increment cores collected within the national forest inventories (NFI) of Mexico and the U.S. Downscaled CMIP5 climate projections were used to forecast future growth under a warming climate.
Results/Conclusions
Output from the LME model yields good model calibration/verification statistics throughout the range of Douglas-fir. The percent of year-to-year variance explained in regional chronologies during the 20th century varies from 40% in the Pacific Northwest to almost 80% in the southwestern United States. Patterns of absolute predicted radial growth increment match well that of independently assessed productivity estimates (i.e., site index) of USFS Forest Inventory & Analysis program. While the results mirror previously reported projections of large-scale growth declines by the end of the 21st century, the projected growth declines are notably smaller, particularly when based solely on the representative NFI samples. We even identify regions where Douglas-fir growth might increase under projected climate change.
This work highlights the value of tree-ring data for ecological forecasting of future tree growth, the importance of using a representative and densely replicated tree-ring dataset and analyses that capture continuous variation in climate sensitivities.