Refining Long-Time Series of Urban Built-Up-Area Extraction Based on Night-Time Light—A Case Study of the Dongting Lake Area in China

Abstract

By studying the development law of urbanization, the problems of disorderly expansion and resource wastage in urban built-up areas can be effectively avoided, which is crucial for the long-term sustainable development of cities. This study proposes a high-precision urban built-up-area extraction method for county-level cities for small and medium-sized towns in county-level regions. Our process is based on the Defense Meteorological Satellite/Operational Linescan System (DMSP/OLS) and the NASA/NOAA Visible Infrared Imaging Radiometer Suite (VIIRS), which develops long-term series of coordinated night-time light (NTL) datasets. We then combined this with the Normalized Vegetation Index (NDVI) to calculate the Vegetation-Adjusted NTL Urban Index (VANUI). We combine land use data and a support vector machine (SVM) for semi-supervised classification learning to propose a high-precision urban built-up-area extraction method for county-level cities. We achieved the following results: (1) we fit binary polynomials to the DMSP/OLS and VIIRS NTL datasets based on the correspondence of the mean values to construct a consistent time series of NTL data. (2) Our method effectively improves the accuracy of urban built-up-area extraction, especially for county-level cities, with an overall accuracy of 91.84% and a Kappa coefficient of 0.83. (3) Our method can perform a long-time series of urban built-up-area extraction, and, by studying the spatial and temporal changes in urban built-up areas, it can provide valuable information for sustainable urban development and urban planning.