The Plumbing of Land Surface Models: Why are Models Performing so Poorly?

The PALS Land sUrface Model Benchmarking Evaluation pRoject (PLUMBER) provided the first true benchmarking-based model evaluation and intercomparison. One of the key results from PLUMBER indicated that in a very broad, averaged sense, Land Surface Models (LSMs) were performing worse for predictions of sensible heat than simple instantaneous linear regressions on met forcings. The models were also out-performed on average for latent heat by a non-linear empirical regression. We examine this result in more detail and show that the result is not due to problems with a lack of energy conservation in the empirical benchmarks or the Fluxnet data, and that the PLUMBER methodology appears sound. We also show that even if the available energy calculation is correct, there is still a significant problem in the energy partitioning, and that the problematic behaviour appears to be shared across LSMs.