AbstractPotential natural vegetation (PNV) remains an important concept in ecology by providing the basis for comparing current or future conditions and trends in vegetation with historic conditions. However, nonstationary climate and land‐use‐driven deviations from baseline conditions challenge the validity of PNV at local scales or for locations that are especially sensitive to environmental changes, such as ecotonal boundaries between ecosystem types. Our goal was to identify the historic boundaries between major grassland types in the Central Great Plains (CGP) of North America (tallgrass prairie and northern and southern mixed‐grass prairies) by examining the environmental drivers of primary production prior to, during, and after the decades‐long drought in the 1930s. We sought to explore differences among grassland types in resistance and resilience to drought, and to evaluate legacy effects resulting from spatially explicit processes during drought. We used county‐level cropland production data from 1926 to 1948 as a proxy for native grass production data. We regressed production data against a suite of models containing variables associated with climate, soils, land use, and drought indices prior to, during, and after drought. We used a dimension reduction approach to identify the locations of grassland boundaries in multivariate space defined by each model. Our results show that climate and soil factors predicted grass production prior to the drought, yet additional land use and erosion factors were needed to explain production in both drought and post‐drought. Grassland responses were most variable along ecotonal boundaries. Tallgrass prairie exhibited stronger production responses to precipitation and temperature compared with northern and southern mixed‐grass prairies during all periods. Tallgrass prairie counties along the ecotone with mixed‐grass prairies were more negatively impacted by soil loss from abandoned cropland during the drought compared with other tallgrass prairie counties. As the intensity and frequency of drought are predicted to increase, landscape‐scale variables and spatially explicit processes that govern novel production responses will need to be accounted for in explaining heterogeneity in grassland responses.