Accuracy Assessment and Map Comparisons for Monitoring Urban Expansion

The availability of global high-resolution built-up area datasets provides researchers and policy makers a tool for monitoring progress on a number of sustainable development targets. These datasets are made possible by the application of increasingly sophisticated computer methods that eliminate the need for human intervention in classifying remotely sensed earth imagery. The European Commission’s Global Human Settlement Layer (GHSL) is one such dataset with historical layers corresponding to the epochs of 2014, 2000, 1990, and 1975. We assess the accuracy of the GHSL landcover classification for the circa 2014 period, as well as the accuracy of the Atlas of Urban Expansion land cover classifications for the circa 2014 period, using reference map data that was manually digitized from high resolution satellite imagery in 200 global cities. We apply the urban extent methodology developed in Atlas of Urban Expansion to create GHSL urban extents and compare them to Atlas of Urban Expansion extents at the 1990, 2000, and 2014 time periods. The overall accuracies of the two datasets are essentially the same, never more than one percentage point apart. Urban extents created with GHSL data were smaller than Atlas extents in 2014, but larger in 2000 and 1990. Discrepancies between the datasets at the 1990 period may require additional investigation to help establish the historical trend.