Toward modeling Pacific Northwest tree species using CARAIB dynamic vegetation model

While uncertainties remain regarding projected temperature and precipitation changes, climate warming is already affecting Pacific Northwest (PNW) ecosystems, notably forest species composition. In this research, we use the process-based dynamic vegetation model CARAIB DVM (Dury et al., iForest - Biogeosciences and Forestry, 4:82-99, 2011) to simulate current distribution of common tree species in PNW ecosystems. This DVM includes the influence of soil water content, atmospheric CO2 concentration and disturbances like fires, all essential to consider for reliably predicting present and future plant species distributions. Classically, dynamic vegetation models represent vegetation at the scale of plant functional types (PFTs). However, since they have a narrower bioclimatic spectrum, individual species are probably more vulnerable to climate change than PFTs.

Here, we first perform simulations with the CARAIB global vegetation classification based on 26 Plant Functional Types (3 herbaceous, 8 shrubby and 15 arboreal PFTs). Then, we apply the vegetation model at the species level in order to analyse the response of a selected set of plant species to current climate. Representing the European vegetation at the scale of individual species has been successfully performed with CARAIB. The simulated individual species are differentiated by their proper climatic requirements and tolerances. Concerning physiological and structural parameters, species share the traits of the respective PFT, but we progressively improve their characterization by the use of global or local trait databases (e.g., TRY database). The model is driven with climatic observation data over the period 1951-2012 across the PNW region at different spatial resolutions. We test the model’s ability to reproduce the present spatial and temporal variations of carbon stocks and fluxes as well as the observed species and biome distributions over the PNW. We then assess model predictions using a variety of available datasets, including eddy covariance and satellite observations.