A hydrometeorological approach for probabilistic simulation of monthly soil moisture under bare and crop land conditions

This study focuses on the probabilistic estimation of monthly soil moisture variation by considering (a) the influence of hydrometeorological forcing to model the temporal variation and (b) the information of Hydrological Soil Groups (HSGs) and Agro‐Climatic Zones (ACZs) to capture the spatial variation. The innovative contributions of this study are: (i) development of a Combined Hydro‐Meteorological (CHM) index to extract the information of different influencing hydrometeorological variables, (ii) consideration of soil‐hydrologic characteristics (through HSGs) and climate regime‐based zoning for agriculture (through ACZs), and (iii) quantification of uncertainty range of the estimated soil moisture. Usage of Supervised Principal Component Analysis (SPCA) in the development of the CHM index helps to eliminate the “curse of dimensionality,” typically arises in the multivariate analysis. The usage of SPCA also ensures the maximum possible association between the developed CHM index and soil moisture variation. The association between these variables is modeled through their joint distribution which is obtained by using the theory of copula. The proposed approach is also spatially transferable, since the information on HSGs and ACZs is considered. The “leave‐one‐out” cross‐validation (LOO‐CV) approach is adopted for stations belong to a particular HSG to examine the spatial transferability. The simulated soil moisture values are also compared with a few existing soil moisture data sets, derived from different Land Surface Models (LSMs) or retrieved from different satellite‐based missions. The potential of the proposed approach is found to be promising and even applicable to crop land also, though with a lesser degree of efficiency as compared to bare land conditions. Key Points: Hydrometeorological approach is developed to model the soil moisture variation Developed approach is spatially transferable and temporally consistent Statistical range of uncertainty, associated with estimates, is available