State-space models increase sensitivity in tree growth responses to environmental predictors

Recent studies have sought to establish a link between local population dynamics and species distribution. One way is to first understand how tree demography responds to the environment, then to calculate the growth rates of local populations as a function of demography, and finally to assess where these growth rates are positive, which determines the niche and thus the potential distribution. However, most studies found limited concurrence between population growth rates and species occurrences. This might be due to extreme events and demographic stochasticity. Tree demography is complicated to estimate at large scale due to high variability in environmental conditions and coarse time resolution of the National Forest Inventories data. Therefore, tree demography is often estimated using averaged environmental conditions, which erase partially the environmental variability. Our goal is to develop a model for tree demography that makes use of annual environmental data rather than averaged conditions. We started with a so called state-space model for tree radial growth, that we fit on Forest Inventories data from west Europe. We then compared this model to the more classic approach of fitting averaged radial growth to averaged environment. We found that often the state-space approach detects more significant responses to environment and that it performs better than the classic approach at predicting growth on an indepedent tree-ring data set. Therefore, state-space models explain better growth variability than the classic approach.