Global reconstruction of three decades of fine-grained nighttime light data with analysis of large-scale infrastructure and landmarks

Satellite-collected nighttime light provides a unique perspective on human activities, including urbanization, population growth, and epidemics. However, global, long-term, and fine-grained nighttime light observations are lacking, hindering the analysis and application of continuous light changes in specific facilities over extended periods. To address this gap, we reformulated the super-resolution task – reconstructing low-resolution nighttime light data into high-resolution images – by explicitly incorporating temporal difference information. Rigorous validation across three evaluation years demonstrates that our dataset consistently outperforms existing products at global, national, and city scales. Case studies of high-value infrastructure sites, such as artificial islands and international airports, highlight our dataset’s unprecedented ability to capture both long-term gradual illumination changes associated with development and abrupt intensity oscillations caused by wars. The dataset is available at https://doi.org/10.5281/zenodo.15845676 . • We reformulate DMSP-to-VIIRS super-resolution via annual-difference learning. • The first global 1992–2021 VIIRS-like dataset supports fine-grained NTL analyses. • Unprecedented multi-decadal monitoring of large infrastructure and landmarks.