Drawing an urban boundary is less straightforward than following an administrative border. Satellite night lights and points of interest each show a different part of the picture.
This analysis combines the two to investigate Beijing’s urban extent, using the density of activity alongside the pattern of light rather than relying on either source alone.
Delineating the physical urban extent of Beijing through multi-source data fusion, integrating VIIRS nighttime lights (NTL) with Point of Interest (POI) spatial distributions.
Implemented using Python (ArcPy). POI density is calculated with spatial kernel density estimation at 500m resolution and a 3,000m bandwidth. A geometric mean integrates nighttime light intensity with POI kernel density surfaces to normalize pixel-level variance. Inflection point analysis across the resulting density curve determines the empirical threshold boundary between built-up urban and rural land.
The resulting delineated urban extent for the Beijing metropolitan area is shown below.
a) KDE of POIs in 2018, b) NTL data for 2018, c) Final Urban extent from composition