Potential vegetation maps have been developed for major ecosystems of the central US using a variety of approaches that focus on the relationship between historic vegetation and long-term climate, soils, and topography. In many ecosystem types in this region, the historic vegetation remains primarily as isolated remnants in protected locations (e.g., research sites) within a predominantly managed and urban landscape. However, the maps maintain their importance because they reflect the underlying drivers of the mosaic of natural and managed ecosystems, and their vulnerability to changes in these drivers. The geographic locations expected to be the most sensitive to changes in drivers are the ecotonal boundaries between ecosystem types, yet the location of these boundaries is inconsistent among maps created by different groups. Our objective was to provide a more mechanistic explanation for the location of boundaries between three major grassland types. We integrated agricultural data on production with climatic, edaphic, and management data (1920s to 1940s) from a gradient across the Great Plains. Because this time period included the 1930s drought, we hypothesized that the shortgrass steppe and mixedgrass prairie would shift farther east under long-term drought and the mixedgrass prairie and tallgrass prairie would occur farther west in the wetter pre- and post-drought.
Our results show complex interactions among precipitation, temperature, soil properties, and management practices across this gradient. Climatic drivers tended to be more important to production in the more mesic part of the gradient, and soil properties became more important as precipitation decreased to the west. The abandonment of fields during the drought of the 1930s led to high rates of soil erosion and decreased production more than expected based on weather alone. As expected, there was a shift toward more arid ecosystems towards the east during extended drought. Historic records of increased land abandonment and losses of soil erosion throughout much of this region in the 1930s support this finding. A shift towards more mesic grasslands farther west in the wet 1920s and 1940s also is supported by historic accounts. Our large-scale integration of diverse datasets is able to both confirm historic observations, and to provide a more mechanistic basis for the boundaries between grassland types that can be used in future predictive models.