Global nighttime light dataset from 1992 to 2022 with focus on low-light areas
Hui Tang, Yongde Zhong, Jinyang Deng, Hongling Xia, Juan Wei
Abstract
Traditional nighttime light (NTL) research has largely focused on urban areas, neglecting approximately 80% of Earth's low-light or dark sky regions. This oversight may result in a significant underestimation of light pollution, especially in global protected areas that are often biodiversity hotspots. Our study employs a novel approach combining a residual neural network with a raster function model to tackle key challenges, including NTL restoration in high-latitude regions, long-term data continuity, and gap alignment across different sensor types. For the first time, we enable continuous calibration and temporal extension of global DVNL and DMSP/OLS data. Our dataset outperforms similar products by offering greater explanatory power for economic activities, enhanced temporal stability, and improved spatial distribution accuracy. Furthermore, it exhibits heightened sensitivity to subtle changes in low-light areas across global, national, urban, and protected scales, making it especially valuable for monitoring human activities and assessing environmental impacts in critical regions like World Heritage Sites, Dark Sky Preserves, and national parks.