Wind-bounded grasslands of North America

Abstract

Abstract Climate may separate grasslands from surrounding forests and shrublands, but an alternative option is that grasslands are suitable for trees except in the presence of frequent fires, which are spread great distances by strong winds. To develop a ruleset that demarcated the central North American grasslands, I modeled grasslands using precipitation and its coefficient of variation, ratio of potential evapotranspiration to precipitation, climatic moisture index, July vapor pressure, and wind speed (i.e., a fire indicator) during 5000 (4900–5100) years before present, 1600–1699, 1800–1849, and 1850–1899. Wind speeds of 4.5\xa0m\xa0s−1 at 10\xa0m above ground height clearly bounded grasslands. For all time intervals, the random forests and extreme gradient boosting classifiers had accuracies of 0.96 to 0.97 in classification of reserved samples; wind speed was ranked as the most influential model variable (i.e., value of 100 out of 100 scale, with the next most influential variables at\xa0≤\xa025). The simulated wind patterns matched recent wind models from atmospheric reanalysis and the Global Wind Atlas downscaled from an atmospheric reanalysis. Wind boundaries were stable over time, shifting slightly northward during 11,000\xa0years before present to recent decades. Although wind is related to precipitation, I directly mapped and modeled precipitation and derivatives. Therefore, I propose that wind is an index of fire spread in grassland biomes, where conditions are appropriate for both fire and woody vegetation. More research on wind models is required to confirm these results and determine how winds affect grassland formation and maintenance globally.