Extracting urban areas in China using DMSP/OLS nighttime light data integrated with biophysical composition information

DMSP/OLS nighttime light (NTL) image is a widely used data source for urbanization studies. Although OLS NTL data are able to map nighttime luminosity, the identification accuracy of distribution of urban areas (UAD) is limited by the overestimation of the lit areas resulting from the coarse spatial resolution. In view of geographical condition, we integrate NTL with Biophysical Composition Index (BCI) and propose a new spectral index, the BCI Assisted NTL Index (BANI) to capture UAD. Comparisons between BANI approach and NDVI-assisted SVM classification are carried out using UAD extracted from Landsat TM/ETM+ data as reference. Results show that BANI is capable of improving the accuracy of UAD extraction using NTL data. The average overall accuracy (OA) and Kappa coefficient of sample cities increased from 88.53% to 95.10% and from 0.56 to 0.84, respectively. Moreover, with regard to cities with more mixed land covers, the accuracy of extraction results is high and the improvement is obvious. For other cities, the accuracy also increased to varying degrees. Hence, BANI approach could achieve better UAD extraction results compared with NDVI-assisted SVM method, suggesting that the proposed method is a reliable alternative method for a large-scale urbanization study in China’s mainland.