Bayesian framework to identify the main effects of long term climatic constraints on soil moisture decline in conterminous United States

We used a Bayesian regression framework based on Hamiltonian Monte Carlo simulations to identify the main effects of mean annual soil moisture, temperature, evapotranspiration, and precipitation, on long-term soil moisture decline across conterminous United States based on 36 years of remotely sensed available data. We found that mean soil moisture was a positive control of soil moisture decline in areas with long-term high precipitation but low evapotranspiration. Furthermore, mean soil moisture is a negative control on soil moisture decline in areas with long-term low precipitation and low evapotranspiration. In contrast, mean soil moisture had no effect on soil moisture decline in areas with long-term low precipitation and high evapotranspiration. These results highlight the importance of having accurate spatial soil moisture information to better inform earth system models to predict regional to global water balance and climate trends. These results support the current understanding of the basic physical mechanisms governing the coupling of soil moisture with temperature, precipitation, and evapotranspiration, but bring attention to high spatial heterogeneity in the constraints of soil moisture at the continental scale. The response of soil moisture to climate variability is considered to be one of the largest uncertainties for global land surface models, and resolving high spatial resolution of soil moisture is an ongoing challenge. Independent estimates of high spatial resolution of soil moisture could improve parameterizations of land surface models and cross-validate the current functions that mainly relay on precipitation, aerodynamic representation of the latent and sensible heat fluxes, and land surface cover type.