Ensemble Evaluation of Hydrologically Enhanced Noah-LSM: Partitioning of the Water Balance in High-Resolution Simulations over the Little Washita River Experimental Watershed

The ability of two versions of the Noah land surface model (LSM) to simulate the water cycle of the Little Washita River experimental watershed is evaluated. One version that uses the standard hydrological parameterizations of Noah 2.7 (STD) is compared another version that replaces STD's subsurface hydrology with a simple aquifer model and topography-related surface and subsurface runoff parameterizations (GW). Simulations on a distributed grid at fine resolution are compared to the long-term distribution of observed daily-mean runoff, the spatial statistics of observed soil moisture, and locally observed latent heat flux. The evaluation targets the typical behavior of ensembles of models that use realistic, near-optimal sets of parameters important to runoff. STD and GW overestimate the ratio of runoff to evapotranspiration. In the subset of STD and GW runs that best reproduce the timing and the volume of streamflow, the surface-to-subsurface runoff ratio is overestimated and simulated streamflow is much flashier than observations. Both models' soil columns wet and dry too quickly, implying that there are structural shortcomings in the formulation of STD that cannot be overcome by adding GW's increased complexity to the model. In its current formulation, GW extremely underestimates baseflow's contribution to total runoff and requires a shallow water table to function realistically. In the catchment (depth to water table > 10m), GW functions as a simple bucket model. Because model parameters are likely scale and site dependent, the need for even "physically based" models to be extensively calibrated for all domains on which they are applied is underscored.