Models for Processing Nighttime Light Data and Software Development

Abstract

DMSP-OLS and NPP-VIIRS provide the most widely used nighttime light data (NTL) sources. These datasets cover the time span of 1992-2013 and 2013-present, respectively. However, the differences in sensor calibration and sensitivity to faint lights between the two satellites prohibit the merging of these two datasets into a longer one, which limits their further application in long-term time series analysis (1992 - present). Here, we provide correction models for the two datasets. The correction process includes oversaturation correction, annual/monthly data synthesis, digital number (DN) value intercalibration, and image continuity correction. Moreover, to test the accuracy and robustness of the merged datasets, we compared the dataset with the corrected products of LiNTL and ChenNTL. The overall accuracy of the dataset in this study was 0.90, and the average of the normalized difference index (ANDI) was 0.046. Additionally, we used the corrected nighttime light data to model Gross Domestic Product (GDP), achieving a correlation of r =0.849. The corrected dataset preserves the quality of the original data without excessive modification and demonstrates certain application value in the socioeconomic field. Finally, to facilitate the broad use of the correction models and the integrated nighttime light datasets, we integrated the above procedures into software with an attached preprocessing workflow. This not only fills the gap in nighttime light data processing software but also allows users to move beyond fixed datasets and customize personalized datasets according to their needs. The study also aims to lay a foundation for future cross-disciplinary applications.

Keywords

• Nighttime light data
• DMSP-OLS
• NPP-VIIRS
• calibration
• remote sensing
• software