E2.4 Dynamic vegetation and ELM-FATES Progress

Abstract

The inclusion of dynamic vegetation demography in Earth System Models (ESMs) has been identified as a critical step in moving ESMs towards more realistic representations of plant ecology, and the processes that govern climatically important fluxes of carbon, energy, and water. Successful application of process-based approaches to simulate plant demography and recovery from disturbances at the global scale is a challenging endeavor. To address this challenge, we have integrated demographic processes using the Functionally-Assembled Terrestrial Ecosystem Simulator (FATES) in the newly developed E3SM Land Model (ELM). We then use an ELM-FATES globally gridded simulation to investigate dynamic plant functional type (PFT) distributions and turnover. Initial global simulations successfully include thirteen interacting and competiting PFTs. Global maps of net primary productivity, leaf area index, and total vegetation biomass by ELM-FATES matched patterns and values compared to several big-leaf models and MODIS estimates. We also applied the improved ELM-FATES model at a Central Amazon tropical site and demonstrate improvements in predicted PFT size- and age-structure and mortality rates. The motivation behind the this tropical simulation is that there is large uncertainty whether Amazon forests will be a carbon sink or source over the 21^st century as atmospheric CO₂ change. To investigate which processes contribute to this uncertainty, we compare four terrestrial models differing in vegetation dynamics (demography vs. ‘big-leaf’ vegetation) and biogeochemical cycling, all driven with CO₂ forcing from the preindustrial period to 2100. With a doubling of CO₂ by 2100, three of the four models predicted an appreciable forest biomass sink (0.52 to 1.04 Mg ha^-1 yr^-1), albeit with different trends in biomass accumulation and vegetation turnover rates. The model that represented phosphorus limitation predicted the lowest forest sink relative to initial biomass stocks (+21%), substantially lower than the other big-leaf model, CLM4.5 (+95%), which only represents nitrogen constraints. We conclude that representing local scale variability (e.g., stochastic mortality events, phosphorus limitation, size structure shifts) is important for large-scale predictions and future forest responses to rising CO₂.