Key points are not available for this paper at this time.
Abstract Moderate-resolution (30-m) national map products have limited capacity to represent fine-scale, heterogeneous urban forms and processes, yet improvements from incorporating higher resolution predictor data remain rare. In this study, we applied random forest models to high-resolution land cover data for 71 U.S. urban areas, moderate-resolution National Land Cover Database (NLCD) Tree Canopy Cover (TCC), and additional explanatory climatic and structural data to develop an enhanced urban TCC dataset for U.S. urban areas. With a coefficient of determination (R 2 ) of 0.747, our model estimated TCC within 3% for 62 urban areas and added 13.4% more city-level TCC on average, compared to the native NLCD TCC product. Cross validations indicated model stability suitable for building a national-scale TCC dataset (median R 2 of 0.752, 0.675, and 0.743 for 1,000-fold cross validation, urban area leave-one-out cross validation, and cross validation by Census block group median year built, respectively). Additionally, our model code can be used to improve moderate-resolution TCC in other parts of the world where high-resolution land cover data have limited spatiotemporal availability.
Corro et al. (Mon,) studied this question.