Current dynamic global vegetation models (DGVMs), including those incorporated into Earth System Models (ESMs), represent terrestrial vegetation by a small number of plant functional types (PFTs) with fixed properties, including a number of plant traits. Such constant parameter values contradict with the observed trait variation in natural vegetation. In the DGVM JSBACH, which is part of the MPI-ESM, we allow three traits (specific leaf area (SLA), maximum carboxylation rate (Vcmax25) and maximum electron transport rate (Jmax25)) to vary within PFTs via trait-climate relationships. These relationships are based on a large database containing observed trait data (TRY, www.try-db.org), and reflect ecological assembly processes.
Results/Conclusions
The results show that for all three traits the default model strongly deviates from observed natural trait data, with mean differences of 32.3 % for Vcmax25, 26.8 % for Jmax25 and 17.3 % for SLA. Simulating trait variation results in a different dominant vegetation cover in more than 35 % of the terrestrial grid cells. GPP increases up to 50 %. The discrepancy between default trait values and natural trait variation, combined with the substantial changes in simulated vegetation properties, together emphasize the importance of trait variation for global vegetation functioning. Incorporating observational data based on the ecological concepts of environmental filtering will improve the modeling of vegetation behavior in DGVMs and as such will enable more reliable projections in unknown climates.