Abstract Regional and global impact assessment tools are increasingly used to explore and evaluate the impact of climate change and extreme events on crop yield and environmental externalities. However, the large uncertainties associated with the inputs or the parameters in crop models within these tools, limits their predictive ability, exceeding the spatiotemporal variability of observed yields. The objective of this study is to explore and quantify different sources of uncertainties and assumptions made behind initial conditions (IC), soil input, meteorological forcing, management practices and model cultivar parameters by running regional simulations for the time period between 2009 and 2019. Simulations were performed for maize and soybean using the pSIMS platform across the U.S Midwest by incrementally accounting for five sources of uncertainty with a 7 k m × 7 k m resolution using the APSIM and DSSAT crop growth models. First, the relative contribution of different sources of uncertainty was estimated over time and space. Then, a series of nitrate leaching hotpots were identified and a regional maize yield productivity index was estimated by decomposing the uncertainty in the same scenario using a hierarchical Bayesian random-effect model. All factors showed a strong spatial pattern in their contribution to the total uncertainty and their contribution was found to be partially dependent on location. However, across the whole region, it was found that the uncertainty around management is larger than IC, soil and meteorological forcing while showing a strong correlation with each of these factors. Given the high spatial correlation, we hypothesize that constraining soil inputs and management uncertainty could allow for the largest reduction in predictive uncertainty for crop yield. Our results showed vast areas over northern IA, IL and IN with high potential for NO₃ leaching and southern IA, IL and east NE with lower maize productivity index compared to the regional average.