How do individual species and Plant Functional Type responses to environmental change differ in Dynamic Vegetation Models? - A forest stand analysis

Originally dedicated to simulate vegetation at the global scale, dynamic (global) vegetation models (D(G)VMs) classify vegetation into Plant Functional Types (PFTs) to represent the largest set of plant species. PFTs are clusters of species that share common morphological and physiological traits. However, PFT classification becomes surely too coarse to reflect the large diversity in plant species responses to climate and environmental changes, a critical point for biodiversity questions. Thus, some efforts focus now on applying DVMs at the species level refining the definition of morphophysiological parameters from initial PFT traits to specific traits collected or found in trait databases. What are the effects of using species-specific parameters? Do dynamic vegetation models better reproduce historical forest growth and mortality observed in monitored stands? How will individual species respond to future climate compared to PFTs? To study these questions, we used two process-based dynamic vegetation models CARAIB (Dury et al., 2011) as well as LPJ-GUESS (Smith et al., 2001), and compared their outputs. CARAIB has been previously adapted to model a set of 40 European tree species, differentiated by their specific traits, proper climatic requirements and tolerances. LPJ-GUESS features a detailed representation of climate sensitive tree species dynamics, resource competition and canopy structure (Hickler et al., 2012). The respective tree species are distinguished by taking differences in phenology, allometry and bioclimatic limits into account.
Model simulations were performed in accordance with the experimental protocol of the COST Action PROFOUND (“Towards robust projections of European forests under climate change”) for several European forest stands selected in the project. We particularly focused on Fagus sylvatica stand in Sorø (Denmark), Picea abies stand in Solling (Germany) and Pinus sylvestris (and Picea abies) stands in Hyytiala (Finland). The experiments include site-specific soil characteristics, management practices (planting, thinning and harvest) and climate conditions. For the historical period (from planting year to 2014), besides local observations, simulations were also run with the original (0.5-degree spatial resolution) and locally bias-corrected (LBC) ISIMIP2B outputs of global climate models for testing the reliance of DVM results to the spatial resolution of climatic inputs. For the future period (2005-2100), vegetation models were driven by the ISIMIP2B climates under different Representative Concentration Pathways. The simulations at the PFT level were performed following the same protocol replacing the locally present species by their representative PFTs. The carbon and water fluxes obtained from the different experiments carried out with the two DVMs were compared with eddy-covariance data from each site.