This page presents the output maps from our paper “Cities as Inverted Watersheds”.
We apply our algorithm to 4.6 million points of interest (POIs) in the continental U.S. and identify consumption centers.
For the parameters of our algorithm, we use a 50 x 50 meter cell partition of the continental U.S. to count POIs and different values for the standard deviations (SD) of Gaussian smoothing. The one with SD = 250 meters (= 5 cells x 50 meter) seems to work best and we use it as the baseline. But here we also provide the ones with SD = 5 and 20 for reference.
We identify all consumption centers with two or more points of interest (POIs). However, to prevent cluttering, the maps show only those consumption centers with 20 or more POIs.
We also identify cities in the continental U.S. by applying the same algorithm but with a larger SD parameter.
One can increase the SD parameter of Gaussian smoothing either by increasing cell size or the number of cells. We use a larger cell size (1 square kilometer) to speed up the process.
We also provide the maps of consumption centers for the three largest cities in Canada.
The numbers of POIs for Canadian consumption centers are not directly comparable to their US counterparts; we apply different criteria for which POIs to include for US and Canada due to data availability.