The effects of small-scale trait variability on regional-scale ecosystem behavior

Earth has hundreds of thousands of species, each with different responses to climatic, edaphic, and other sources of variability. However, for regional or global studies, it is all but impossible to represent much of this diversity of behavior explicitly, both because it’s computationally infeasible and because not enough information about the landscape is available. Our group has studied how this aggregated behavior affects affect water and carbon fluxes at larger scales, which we have studied for various aspects of stomatal conductance, wildfire fuel moisture content and burned area, and more. We also use remote sensing and model-data fusion (data assimilation) to build better tools for studying this issue. We have mapped ‘effective ecosystem-scale’ ecosystem traits that integrate the aggregate behavior of the small-scale variability and represent differences in behavior at intermediate stand-scales. Most land surface and earth system models largely ignore this variability of behavior, instead parametrizing all vegetation of the same plant functional type similarly. Indeed, decades of modeling effort have focused on building more-detailed representations of ecosystem processes, but much less attention has been paid to correctly assigning ecosystem traits and their spatial variability. Our group has shown that this approach is counter-productive. In fact, simpler models make better predictions than ever-more complicated models, unless there is enough data and a robust approach to parametrize the more complicated models correctly. As an alternative, together with colleagues at JPL, the University of Edinburgh, and several other institutions, we have contributed to the development and use of the CARbon DAta MOdel fraMework (CARDAMOM), a terrestrial carbon cycle model that uses model-data fusion in combination with several remote sensing data streams to determine optimal representations of carbon pools, fluxes, and parameters across the more. Lastly, we are developing and testing alternative approaches that could be used to replace or enhance the decades-old plant functional type paradigm.