Multi-year extreme drought events that affect regional to sub-continental extents have catastrophic impacts on ecological systems. Drought impacts are often studied based on local factors, such as climate and soils, interacting with local ecological processes (e.g., competition between plants). However, spatial processes operating within broad extents can lead to connectivity-mediated feedbacks to the vegetation that overwhelm direct effects of drought and lead to unexpected landscape-scale heterogeneity. One of the most devastating droughts occurred in the Great Plains of North America in the 1930s. This climatic event combined with land management practices (plowing and subsequent abandonment) to result in broad-scale plant mortality, soil erosion, and dust storms. However, not all areas were affected similarly, even across short distances. We hypothesized that spatial discontinuities in impacts were related to: (1) local factors of climate, soil properties, and management and/or (2) connectivity by wind and/or water erosion from nearby locations. We integrated legacy data from the U.S. Department of Agriculture on production, management, and erosion with climatic and edaphic data (1920s to 1940s) for a 100-km gradient from western Iowa to eastern Nebraska. Responses were compared during the drought to responses for the same location in the 1920s (pre-drought) and the 1940s (post-drought). Our results show complex interactions among precipitation, temperature, soil properties, and management practices across this relatively short (100-km) gradient. Climatic drivers tended to be more important to production in the more mesic part of the gradient (IA), and soil properties became more important as precipitation decreased to the west (NE). However, local factors alone were unable to explain lower production in NE than IA for locations with similar properties. Landscape features, such as large alluvial plains that reduced connectivity by wind, were more important than local factors to explain different responses along the gradient during the drought. These findings have important implications for landscape configurations and connectivity in the face of large-scale, extreme events.