We apply deep learning to daytime satellite imagery to predict changes in income and population at high spatial resolution in US data. For grid cells with lateral dimensions of 1.2 km and 2.4 km (where the average US county has dimension of 51.9 km), our model predictions achieve R2 values of 0.85 to 0.91 in levels, which far exceed the accuracy of existing models, and 0.32 to 0.46 in decadal changes, which have no counterpart in the literature and are 3–4 times larger than for commonly used nighttime lights. Our network has wide application for analyzing localized shocks.
We propose a methodology for defining urban markets based on builtup landcover classified from daytime satellite imagery. Compared to markets defined using minimum thresholds for nighttime light intensity, daytime imagery identify an order of magnitude more markets, capture more of India’s urban population, are more realistically jagged in shape, and reveal more variation in the spatial distribution of economic activity. We conclude that daytime satellite data are a promising source for the study of urban forms.