Why are the PLUMBER Results so Terrible? A Diagnostic Evaluation of the Errors in Land Surface Model Simulations Across FLUXNET Sites

PLUMBER, a recent land-surface model (LSM) benchmarking study, demonstrated that simple data-driven models outperform modern LSMs in partitioning net radiation into turbulent heat fluxes over a wide range of FLUXNET locations. PLUMBER asks the question “What is the performance of LSMs?” In this presentation we summarize diagnostic evaluations of the PLUMBER results and ask “why do simple models outperform LSMs?” Specifically, we use the Structure for Unifying Multiple Modeling Alternatives (SUMMA) to construct LSMs of varying complexity, in order to relate model performance to process representation. In this context we analyze the PLUMBER results to determine when and where the data-driven models outperform LSMs, and understand the impact of site characteristics on models' partitioning of net radiation into turbulent heat fluxes. This presentation will focus discussion on the role of benchmarking in LSM evaluation efforts, along with the need for comprehensive diagnostic evaluation to identify key LSM development priorities.