Interactions between shallow groundwater and land surface processes play an important role in the ecohydrology of riparian zones. Some recent land surface models (LSMs) incorporate groundwater-land surface interactions using parameterizations at varying levels of detail. In this paper, we examine the sensitivity of land surface evapotranspiration (ET) to water table depth, soil texture, and two commonly used soil hydraulic parameter datasets using four models with varying levels of complexity. The selected models are Hydrus-1D, which solves the pressure-based Richards equation, the Integrated Biosphere Simulator (IBIS), which simulates interactions among multiple soil layers using a (water-content) variant of the Richards equation, and two forms of a steady-state capillary flux model coupled with a single-bucket soil moisture model. These models are first evaluated using field observations of climate, soil moisture, and groundwater levels at a semi-arid site in south-central Nebraska, USA. All four models are found to compare reasonably well with observations, particularly when the effects of groundwater are included. We then examine the sensitivity of modelled ET to water table depth for various model formulations, node spacings, and soil textures (using soil hydraulic parameter values from two different sources, namely Rawls and Clapp-Hornberger). The results indicate a strong influence of soil texture and water table depth on groundwater contributions to ET. Furthermore, differences in texture-specific, class-averaged soil parameters obtained from the two literature sources lead to large differences in the simulated depth and thickness of the "critical zone" (i.e., the zone within which variations in water table depth strongly impact surface ET). Depending on the depth-to-groundwater, this can also lead to large discrepancies in simulated ET (in some cases by more than a factor of two). When the Clapp-Hornberger soil parameter dataset is used, the critical zone becomes significantly deeper, and surface ET rates become much higher, resulting in a stronger influence of deep groundwater on the land surface energy and water balance. In general, we find that the simulated sensitivity of ET to the choice of soil hydraulic parameter dataset is greater than the sensitivity to soil texture defined within each dataset, or even to the choice of model formulation. Thus, our findings underscore the need for future modelling and field-based studies to improve the predictability of groundwater-land surface interactions in numerical models, particularly as it relates to the parameterization of soil hydraulic properties.

Soil moisture partly controls land‐atmosphere mass and energy exchanges and ecohydrological processes in natural and agricultural systems. Thus, many models and remote sensing products continue to improve their spatiotemporal resolution of soil moisture, with some land surface models reaching 1 km resolution. However, the reliability and accuracy of both modeled and remotely sensed soil moisture require comparison with ground measurements at the appropriate spatiotemporal scales. One promising technique is the cosmic ray neutron probe. Here we further assess the suitability of this technique for real‐time monitoring across a large area by combining data from three fixed probes and roving surveys over a 12 km × 12 km area in eastern Nebraska. Regression analyses indicated linear relationships between the fixed probe averages and roving estimates of soil moisture for each grid cell, allowing us to derive an 8 h product at spatial resolutions of 1, 3, and 12 km, with root‐mean‐square error of 3%, 1.8%, and 0.9%.