Climate change is altering the frequency and intensity of physical hazards. Quantifying these risks is a challenging task given uncertainties in future projections and the impacts on multi-sector systems. Here, we examine the combined role of climate and hydrologic uncertainties in shaping future projections of agriculture-relevant soil moisture extremes, focusing on the central and eastern U.S. given its global relevance in maize and soybean production. We encode a simple soil moisture submodule of the global Water Balance Model (WBM) in a differentiable programming framework to facilitate fast runtimes and an efficient calibration. We explore uncertainty in the model parameters by calibrating against different observational datasets as well as using several error metric functions. The resulting parameter ensemble is then convolved with a set of downscaled and bias-corrected climate projections to produce a large ensemble of future soil moisture scenarios. We find that accounting for soil parameter uncertainties can induce meaningful increases in the severity of future soil moisture extremes, with the choice of calibration dataset playing a significant role. Our results highlight the importance of considering combined hydrologic and climate uncertainties when constructing projections of decision-relevant hydroclimatic outcomes.