Radiometric Intercalibration of Multisource Nighttime Light Images at High Resolution for Disaster Mapping

Abstract

High-resolution nighttime light (NTL) remote sensing has become increasingly important for detailed and timely disaster assessment. However, the long revisit cycles of high-resolution satellites constrain continuous disaster monitoring, making multisource image integration necessary for enhancing temporal frequency. Radiometric intercalibration is, therefore, required to bridge sensor differences and ensure comparability of NTL images. Previous intercalibration methods rely on the pseudoinvariant pixel (PIP) framework, typically selecting PIPs based on socioeconomic stability assumptions or images acquired on overlapping dates. These methods are inapplicable in disaster scenarios because cross-sensor images are rarely obtained simultaneously and may exhibit substantial NTL changes. In this study, we proposed an automatic radiometric intercalibration framework for high-resolution NTL images, using SDGSAT-1 and Yangwang-1 as a case study. The proposed method automatically identified PIPs from preevent and postevent images with dramatic changes, which were then used to establish a patch-to-pixel transformation from the SDGSAT-1 image to the Yangwang-1 image. The nonlinear relationship of the transformation was learned using random forest (RF), enabling the intercalibration of SDGSAT-1 to Yangwang-1-like images. The method was applied to two disaster events (i.e., the Israel–Gaza war and the Türkiye–Syria earthquake) and was evaluated using reference PIPs as validation data. In these reference pseudoinvariant regions, the accuracy of the method was measured by measuring the consistency between NTL values of Yangwang-1 and Yangwang-1-like (i.e., the intercalibrated SDGSAT-1). The results demonstrated high accuracy of our method, with R2 values of 0.94 and 0.89 for the two events, respectively, which are better than two conventional methods. The proposed framework can be extended to the integration of other multisource NTL images in disaster scenarios, supporting emergency mapping for humanitarian relief.