Ecohydrological parameters that describe vegetation controls on soil moisture dynamics are not easy to measure at hydrologically meaningful scales and site-specific values are rarely available. We hypothesize that sufficient information required to determine these ecohydrological parameters is encoded in empirical probability density functions (pdfs) of soil saturation, and that this information can be extracted through inverse modeling of the commonly used stochastic soil water balance. We developed a generalizable Bayesian inference approach to estimate soil saturation thresholds at which plants control soil water losses, based only on soil texture, rainfall and soil moisture data at point, footprint, and satellite scales. The optimal analytical soil saturation pdfs were statistically consistent with empirical pdfs and parameter uncertainties were on average under 10 %. Parameter estimates were most constrained for scales and locations at which soil water dynamics are more sensitive to the fitted ecohydrological parameters of interest. The algorithm convergence was most successful and the best goodness-of-fit statistics were obtained at the satellite scale. Robust and accurate results were obtained with as little as 75 daily observations randomly sampled from the full records, demonstrating the advantage of analyzing soil saturation pdfs instead of time series. A sensitivity analysis showed that estimates of soil saturation thresholds at which plants control soil water losses were not sensitive to soil depth and near-surface observations are valuable to characterize ecohydrological factors driving soil water dynamics at ecosystem scales. This work combined modeling and empirical approaches in ecohydrology and provided a simple framework to obtain analytical descriptions of soil moisture dynamics at a range of spatial scales that are consistent with soil moisture observations.