9.2 Global Built-Up Lands

Land cover classification on a global scale is not an easy task. Currently urban areas are represented world-wide by only one single category, which essentially says that all the urban lands around the world look the same. Overall the planet's urban lands are also under-classified, since given most classification schemes built up land is usually the area left-over after all other categories have been classified.

Two sources were used to create this dataset: the DMSP/OLS Nighttime Lights and the IGBP land cover characterization data set.

Initially the two source data sets were re-gridded to 5 minute resolution. In the newly produced nighttime lights data, each cell represented the average nighttime light value of 4 cells, while each cell in the new land cover data represented the number of cells out of 100 which were categorized in the original IGBP data as “Urban or Built-up land” (Figure 1).

A linear regression was developed for all non-zero values to establish the relation between the built-up area density, as the dependent variable, and nighttime lights, as the independent variable. The statistics were then applied to the nighttime data set to develop modeled built-up area density. To this data set, we added the positive difference to the fractional re-gridded IGBP data set.

Thus, the resultant data contains a combination of modeled built-up areas (base on nighttime lights) and observed built-up areas (based on IGBP land cover data).