Skip to main content
ARS Home » Research » Publications at this Location » Publication #171088

Title: VARIABLITY OF SURFACE SOIL MOISTURE AT THE WATERSHED SCALE

Author
item Cosh, Michael
item STEDINGER, J
item BRUTSAERT, W

Submitted to: Water Resources Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/15/2004
Publication Date: 12/31/2004
Citation: Cosh, M.H., Stedinger, J.R., Brutsaert, W. 2004. Variability of surface soil moisture at the watershed scale. Water Resources Research. 40, w12513, doi:10.1029/2004WR003487.

Interpretive Summary: The parameters that affect surface soil moisture variability are of keen interest to hydrologic and land surface modelers. Using a study conducted in 1992 in the Little Washita River Watershed, several land surface parameters are analyzed for their impact on variability. A complex regression model is constructed to isolate and quantify soil texture, vegetation density, and time since precipitation effects. Combined, these account for 55% of the total variability of surface soil moisture including their interactions. All other land surface parameters are determined to account for an additional 31% of the variability. The remaining 14% is unexplained by the dataset. This study also puts forth a novel approach to analyzing complex land surface datasets.

Technical Abstract: Data from the Washita '92 experiment illustrate the spatial and temporal variability of surface soil moisture. Statistical moments revealed that there is no significant skewness or kurtosis in the distribution. Over the drying period, the variance decreased with time; however, the coefficient of variation did not significantly change indicating that scaled variability remains roughly constant. Regression analysis showed that soil texture, vegetation density, and time since precipitation accounted for approximately 55% of the variability of surface soil moisture. There are also strong indications of persistence in the spatial structure representing perhaps topography and solar exposure, which was maintained and explained 31% of the total variability. The methodology proposed herein should be generally applicable in the analysis of any spatially variable land surface data set as it evolves with time.