Predicting rootzone soil moisture from surface observations in cropland using an exponential filter

Abstract

AbstractRootzone soil moisture provides valuable information to guide in‐season management decisions in rainfed and irrigated agricultural systems. Measuring rootzone soil moisture usually requires the deployment of a vertical array of sensors, which can be costly and labor intensive. In this study, we tested the skill of an exponential filter to estimate rootzone soil moisture conditions from a time series of near‐surface soil moisture observations. Daily soil moisture observations from a sensor at 10‐cm depth were used to predict the soil water content at 30‐, 50‐, and 70‐cm depth, and across the entire soil profile (0‐to‐80‐cm depth) in four agricultural fields under irrigated and rainfed conditions. The characteristic time length was the only fitting parameter of the model, which was optimized for each site based on the Nash–Sutcliffe (NS) score. Across the four sites, the mean NS score at individual soil layers ranged from −0.64 to 0.73, but the NS score at the profile level ranged between 0.29 and 0.84. Predictions of soil water storage at the profile level based on near‐surface soil moisture and the exponential filter resulted in average RMSE of 11 mm across the four locations. Our study shows that in agricultural fields the exponential filter performs better when considering the entire soil profile rather than individual soil layers.