Model representation of the coupling between evapotranspiration and soil water content at different depths

Soil water content (θ) influences the climate system by controlling the fraction of incoming solar and longwave energy that is converted into evapotranspiration (ET). Therefore, investigating the coupling strength between θ and ET is important for the study of land surface–atmosphere interactions. Physical models are commonly tasked with representing the coupling between θ and ET; however, few studies have evaluated the accuracy of model-based estimates of θ ∕ ET coupling (especially at multiple soil depths). To address this issue, we use in situ AmeriFlux observations to evaluate θ ∕ ET coupling strength estimates acquired from multiple land surface models (LSMs) and an ET retrieval algorithm – the Global Land Evaporation Amsterdam Model (GLEAM). For maximum robustness, coupling strength is represented using the sampled normalized mutual information (NMI) between θ estimates acquired at various vertical depths and surface evaporation flux expressed as a fraction of potential evapotranspiration (fPET, the ratio of ET to potential ET). Results indicate that LSMs and GLEAM are generally in agreement with AmeriFlux measurements in that surface soil water content (θs) contains slightly more NMI with fPET than vertically integrated soil water content (θv). Overall, LSMs and GLEAM adequately capture variations in NMI between fPET and θ estimates acquired at various vertical depths. However, GLEAM significantly overestimates the NMI between θ and ET, and the relative contribution of θs to total ET. This bias appears attributable to differences in GLEAM's ET estimation scheme relative to the other two LSMs considered here (i.e., the Noah model with multi-parameterization options and the Catchment Land Surface Model, CLSM). These results provide insight into improved LSM model structure and parameter optimization for land surface–atmosphere coupling analyses.