Estimating soil moisture typically involves calibrating models to sparse networks of in situ sensors, which introduces considerable error in locations where sensors are not available. We address this issue by calibrating parameters of a parsimonious soil moisture model, which requires only antecedent precipitation information, at gauged locations and then extrapolating these values to ungauged locations via a hydroclimatic classification system. Fifteen sites within the Soil Climate Analysis Network (SCAN) containing multiyear time series data for precipitation and soil moisture are used to calibrate the model. By calibrating at 1 of these 15 sites and validating at another, we observe that the best results are obtained where calibration and validation occur within the same hydroclimatic class. Additionally, soil texture data are tested for their importance in improving predictions between calibration and validation sites. Results have the largest errors when calibration-validation pairs differ hydroclimatically and edaphically, improve when one of these two characteristics are aligned, and are strongest when the calibration and validation sites are hydroclimatically and edaphically similar. These findings indicate considerable promise for improving soil moisture estimation in ungauged locations by considering these similarities.
ASJC Scopus subject areas
- Water Science and Technology
- Earth and Planetary Sciences (miscellaneous)