Confronting global land-atmosphere models with coupled process metrics

Metrics of land-atmosphere feedback based on the statistics of measurable quantities have emerged over the last decade as advances have been made in the understanding of processes of land-atmosphere coupling on weather and climate time scales. These metrics are invaluable for understanding nature as well as for the diagnosis and development of numerical models. Indices calculated from environmental states and fluxes from global FluxNET sites are compared to reanalysis products (constrained by data assimilation), offline land model simulations and standard free-running coupled land-atmosphere simulations, each with two different modeling systems. In the light of results over the last dozen years suggesting the potential land surface contribution to predictability and prediction skill, the fidelity of coupled land-atmosphere processes in models should be a priority for improving forecast skill. This metric-based analysis points to where model weaknesses lie in terms of both parameter choices and process representation, and where model development should be focused.