ALBERT

Description: Watershed-scale hydrological and biogeochemical models are usually discretized at resolutions coarser than where significant heterogeneities in topographic, subsurface abiotic and biotic, and surface vegetation exist. Here we report on a method to use fine-resolution (220 m gridcells) hydrological model predictions to build reduced order models of the statistical properties of near-surface soil moisture at coarse-resolution (25 times coarser; ~7 km). We applied a watershed-scale hydrological model (PAWS+CLM) that has been previously tested in several watersheds and developed simple, relatively accurate (R2 ~ 0.7–0.8) reduced order models for the relationship between mean and higher-order moments of near-surface soil moisture during the non-frozen periods over five years. When applied to transient predictions, soil moisture variance and skewness were relatively accurately predicted (R2 ~ 0.7–0.8), while the kurtosis was less accurately predicted (R2 ~ 0.5). We tested sixteen system attributes hypothesized to explain the negative relationship between soil moisture mean and variance toward the wetter end of the distribution and found that, in the model, 59% of the variance of this relationship can be explained by the elevation gradient convolved with mean evapotranspiration. We did not find significant relationships between the time rate of change of soil moisture variance and covariances between mean moisture and evapotranspiration, drainage, or soil properties, as has been reported in other modeling studies. As seen in previous observational studies, the predicted soil moisture skewness was predominantly positive and negative in drier and wetter regions, respectively. In individual coarse-resolution gridcells, the transition between positive and negative skewness occurred at a mean soil moisture of ~0.25–0.3. The type of numerical modeling experiments presented here can improve understanding of the causes of soil moisture heterogeneity across scales, and inform the types of observations required to more accurately represent unresolved spatial heterogeneity in regional and global hydrological models.