An easily implemented method to estimate impervious surface area on a large scale from MODIS time-series and improved DMSP-OLS nighttime light data

It is important for researchers and policy-makers to frequently update the amount and spatial distribution of impervious surface area (ISA) on earth, because the level of imperviousness not only indicates urbanization, but is also a key indicator of ecological conditions. In this study, we developed an easily implemented method for estimating the ISA percentage (ISA%) from vegetation index data obtained from a moderate resolution imaging spectroradiometer (MODIS) and nighttime light data obtained from the Defense Meteorological Satellite Program’s Operational Line-scan System (DMSP-OLS). The proposed method consists of four major steps. First, a non-vegetation fraction map was generated from 16-day composited time-series MODIS normalized difference vegetation index data using the temporal mixture analysis method. Second, the enhanced-vegetation-index-adjusted nighttime light index (EANTLI) was used to overcome the saturation problem and blooming effects in the original DMSP-OLS data. Third, the relationship between ISA% and EANTLI was derived based on a statistical analysis of the non-vegetation fraction image and the EANTLI image to obtain a preliminary ISA% map. Finally, the final ISA% map was obtained by selecting smaller values from the preliminary ISA% map and non-vegetation fraction map for each pixel. The validation results showed that the developed method has promising accuracy for estimating the ISA% in our study area (mainly consisting of four Southeast Asian countries: Thailand, Laos, Cambodia, and Vietnam), with a root mean square error value of 0.111, a systematic error value of 0.061, and a determination coefficient of 0.87. Another important finding is that there are two relationships between ISA% and improved nighttime light (i.e., EANTLI): the natural logarithmic function is suitable for ISA% values between 0% and 50%, and the quadratic polynomial function should be used for ISA% values larger than 50%. The developed method has high potential for application to the generation of global ISA% maps with frequent updates due to its easy implementation and the ready availability of input data.