COS 146-4 - Inclusion of plant functional traits improves representation of an alpine tundra hillslope gradient in a land surface model

Abstract: Earth system models simulate vegetation and ecosystem processes using plant functional types (PFTs), where vegetation is broadly grouped based on similar characteristics and biomes. While these broad groupings are useful for investigating global-scale questions, they are unable to characterize the diversity of plant growth strategies that occur at ecosystem scales, particularly in arctic and alpine tundra systems. For instance, for alpine tundra simulations, the default PFT is an Arctic C3 grass. In this study, we used meteorological and eddy covariance tower data as well as plant trait datasets to run point-scale simulations with the Community Land Model (CLM; the terrestrial component of the Community Earth System Model) using the hillslope hydrology configuration at the Niwot Ridge Long Term Ecological Research site in the Colorado Front Range. We modified plant traits and parameters–including foliar traits, phenological characteristics, and hydraulic traits–to represent alpine growth strategies and to configure dry, moist, and wet alpine tundra communities driven by differences in snow accumulation. After validating our simulations with observations from Niwot Ridge, we compared parameterized results to those using the default PFT and ran sensitivity tests by modifying sets of parameters to represent more resource use conservative or acquisitive communities. We found that using foliar trait data from each plant community significantly improved gross primary productivity and latent heat fluxes compared to overestimates in the default simulation, and the model was particularly sensitive to changes in specific leaf area. In addition, phenology metrics derived from phenocam images showed that dry meadow communities had a longer growing season with earlier green-up and peak greenness in comparison to moist and wet meadows where deeper snow delayed the onset of the growing season. Modifying our model configuration to include these phenological differences improved the timing of the growing season, particularly in the dry meadow where plants greened up too early in the default simulation. These findings highlight the importance of incorporating site-level measurements into land models as a means to characterize vegetation and more accurately forecast climate change impacts on ecosystem function.