Abstract
Summary 1. We present a simple algorithm for reconstructing spatially explicit historical changes in croplands. We initialize our simulation with a satellite‐derived characterization of present‐day croplands c. 1992. This data set of croplands is then used within a simple model, along with historical cropland inventory data at the national and subnational level, to reconstruct historical crop cover. We present an annual data set of cropland areas in North America between 1850 and 1992, at a spatial resolution of 5 min (≈10 km). 2. The reconstructed changes in North American crop cover are generally consistent with qualitative descriptions of change. Crop cover is initially concentrated in the eastern portions of the continent, and subsequently migrates westward into the Midwestern United States and the Prairie Provinces of Canada. We also see cropland abandonment in the eastern portions of the continent during the 20th century. The simulation, however, fails to characterize adequately the changes in crop cover in Mexico. 3. We also estimate the extent to which the different vegetation types of North America have been cleared for cultivation. We find that savannas/grasslands/steppes and forests/woodlands have undergone the most extensive conversion (1.68 and 1.40 million km 2 cleared, respectively, since 1850). We further discuss the wider implications of such large‐scale changes in land cover.