Season, wind speed, and seasonal rain are major drivers of a regional aeolian sediment transport model.

Wind erosion and sediment transport continue to increase in many parts of the world, leading to decreased soil quality, accelerated snow-melt, respiratory diseases, and traffic accidents. The processes that control sediment transport are well understood at small scales of mm to m but are less well understood at larger scales of km to hundreds of km. Here we test four approaches aimed at improving the variance explained in sediment transport measured in a network of 52 horizontal sediment flux collecting devices located on the Colorado Plateau, USA. First, switching from a regression tree to random forest statistical analysis increased the variance in sediment transport explained from 58% to 91%. Soil moisture as a single variable explained 52% of variation in sediment flux, but had a negligible effect on a random forest model with season (Winter, Spring, Summer), wind speed, and seasonal total precipitation. Similarly, adding four years of new data to an existing five-year dataset or adding measurements of soil roughness and grazing failed to improve variance explained. By explaining 91% of the variance in sediment transport, our model provides baseline model for understanding sediment transport on the landscape scale. Dust flux networks in new regions would likely need to collect at least 300-500 samples to describe variation in sediment transport values using random forest analyses of the effects of season, wind speed, seasonal rain and vegetation type.