Development and evaluation of a new soil moisture and runoff parameterization for the CABLE LSM including subgrid‐scale processes

New conceptual parameterizations of subgrid‐scale soil moisture, runoff generation, and groundwater are developed and tested in the offline version of the Community Atmosphere and Biosphere Land Exchange (CABLE) model. These developments are designed to address a known bias in CABLE: the excess simulation of mean evapotranspiration over most of the Northern Hemisphere and over most vegetation types at global scales. The inclusion of subgrid‐scale soil moisture parameterizations and runoff generation processes largely mitigates the overestimation of evapotranspiration, and simultaneously improves the simulation of mean runoff. There are also pronounced improvements in total water storage anomalies as compared to various observationally derived estimates. It is shown that surface and subsurface runoff generation resulting from subgrid‐scale soil moisture heterogeneity is necessary to accurately simulate the mean observed runoff and evapotranspiration. Surface runoff generation is largely responsible from reducing the mean evapotranspiration bias north of 30°N. We note that the inability of previous CABLE evaluation studies to demonstrate the need for subgrid‐scale heterogeneity is likely due to only testing at flux tower site scales. Evaluating land models at point scale and catchment scale across a large range of climate and vegetation types is necessary to evaluate the contributions of processes that influence the model at various spatial scales. Flux tower data are therefore a necessary but insufficient constraint on land surface models. Key Points: CABLE overestimates mean ET north of 20° Subgrid moisture and runoff generation developed for CABLE New processes drastically improve simulations of ET, Runoff, and SM