Spatial variability constraints to modeling soil water at different scales

Abstract

There is increasing interest in modeling soil water content over relatively large areas or scales. In general, the spatial variability of soil water content increases with scale, but it is not known how much or at which scales. High spatial variability constrains soil water models by reducing the accuracy of input parameters, calibration and verification data. It may also require representation of soil water in a spatially distributed manner. Soil water content data were collected at the Reynolds Creek Experimental Watershed at scales ranging from 12 m 2 to 2.3×10 8 m 2 to determine how scale affects spatial variability. We found significant spatial variability at the 12-m 2 scale, which could be described as random in large-scale models. The increase of spatial variability with scale was controlled by deterministic `sources' such as soil series and elevation-induced climatic effects. The satellite-derived, soil-adjusted vegetation index showed that spatial variability at the scale of Reynolds Creek (2.3×10 8 m 2) is not random, and may have abrupt transitions corresponding to soil series. These results suggest a modeling strategy that incorporates soil series characterized by random spatial variability nested within the larger, elevation-induced climatic gradient. The distinctions between soils and elevations is greatest early in the growing season and gradually diminish as the effects of differential precipitation and snowmelt timing are erased by evapotranspiration until late in the summer, when they virtually disappear. These conclusions are landscape-dependent, so that representation of spatial variability should be an explicit part of model development and application.