Delineating urban areas using building density

We develop a new dartboard methodology to delineate urban areas using detailed information about building location, which we implement map of all buildings in France. For each pixel, our approach compares actual density after smoothing counterfactual smoothed computed randomly redistributing buildings. define as any area with statistically significant excess density. Within areas, extensions allow us distinguish ‘core’ pixels and detect centres subcentres. Finally, novel one- two-sided tests that provide statistical basis compare maps different delineations, use assess the robustness document large differences between preferred delineation corresponding official one.