Extension and variability of impervious surface area revealed by nighttime light imagery

Abstract

The indiscriminate increase in impervious surface area, mainly associated with anthropogenic activities, is causing important environmental changes. Given the relevance of impervious surface area, this paper introduces an approach designed to map impervious areas. The proposed method was applied to the State of São Paulo, southern Brazil, using nighttime light imagery collected by the Visible Infrared Imaging Radiometer Suite Day–Night Band from the Suomi National Polar-orbiting Partnership satellite. Our hypothesis is that impervious surface area is primarily related to the presence of persistent nighttime light in space and time, and that its quantification basically depends on the brightness pattern emitted by the urban surface. Thus, brightness patterns extracted from monthly composites of nighttime light were associated with impervious surface area established through a random forest regression model. The results for the entire study area (248,000 km²) showed that the method was quite effective in predicting impervious surface area (R² = 0.90, p < 0.001), although, on average, an underestimation deviation (~ 20%) occurred. Focusing on densely urbanized areas, on the other hand, an overestimation error was reported, mainly around downtown areas. Despite this error, graduated values of impervious surface area were found in intra-urban spaces, making it possible to characterize, consistently, differences in the land-use patterns.