Investigating soil controls on soil moisture spatial variability: Numerical simulations and field observations

Abstract

Due to its complex interactions with various processes and factors, soil moisture exhibits significant spatial variability across different spatial scales. In this study, a modeling approach and field observations were used to examine the soil control on the relationship between mean (θ¯) and standard deviation (σθ) of soil moisture content. For the numerical experiments, a 1-D vadose zone model along with van Genuchten parameters generated by pedotransfer functions was used for simulating soil moisture dynamics under different climate and surface conditions. To force the model, hydrometeorological and physiological data that spanned over three years from five research sites within the continental US were used. The modeling results showed that under bare surface conditions, different forms of the θ¯–σθ relationship as observed in experimental studies were produced. For finer soils, a positive θ¯–σθ relationship gradually changed to an upward convex and a negative one from arid to humid conditions; whereas, a positive relationship existed for coarser soils, regardless of climatic conditions. The maximum σθ for finer soils was larger under semiarid conditions than under arid and humid conditions, while the maximum σθ for coarser soils increased with increasing precipitation. Moreover, vegetation tended to reduce θ¯ and σθ, and thus affected the θ¯–σθ relationship. A sensitivity analysis was also conducted to examine the controls of different van Genuchten parameters on the θ¯–σθ relationship under bare surface conditions. It was found that the residual soil moisture content mainly affected σθ under dry conditions, while the saturated soil moisture content and the saturated hydraulic conductivity largely controlled σθ under wet conditions. Importantly, the upward convex θ¯–σθ relationship was mostly caused by the shape factor n that accounts for pore size distribution. Finally, measured soil moisture data from a semiarid region were retrieved from the Automated Weather Data Network. The observed moisture data showed that based on soil texture, a positive θ¯–σθ relationship existed for sandy soils, while an upward convex one was observed for silty soils. The difference in the observed θ¯–sigmaθ relationship can be attributed to the differences in water holding capacities between sand and silt, which is consistent with the modeling results. The field data also revealed that increasing spatial variability in soil texture led to increased variability in soil moisture (e.g., the maximum σθ). Therefore, the effect of soil texture for verifying remotely sensed soil moisture products should be considered.