Correcting the nighttime lighting data underestimation effect based on light source detection and luminance reconstruction

Abstract

Nighttime light remote sensing provides a unique perspective on monitoring human activities and socioeconomic conditions. However, early studies on nighttime lighting usually neglected the evaluation of light source detection and luminance conditions in underdeveloped areas. Recent studies have identified a significant nighttime lighting underestimation effect in underdeveloped areas, which makes it difficult to use nighttime lighting as a valid proxy for other variables. To address this problem, this paper proposes a method based on light source detection and luminance reconstruction. Missing luminous pixels will be identified using the rich information provided by daily-scale lighting data and light blooming patterns, and diverse ancillary data will be utilized for the regression reconstruction of luminance values to supplement annual data and to correct the underestimation effect. The effectiveness of this method was verified using data from 7,776 townships in underdeveloped areas in China. The results demonstrated that the proposed method using the corrected NPP/VIIRS data improved light source detection accuracy by 23.23% and 17.71% at the village and township scales, respectively. The correlations between night lighting and proxy variables were also improved. Spearman's rank correlation coefficients between lighting and population, GDP, and electricity consumption at the county scale increased by 0.045, 0.020, and 0.018, respectively. The correlations between lighting and agriculture-related variables also increased to different degrees at both the township and county scales. The proposed method can contribute to improving modeling accuracy and correcting the nighttime lighting underestimation effect in underdeveloped areas. This will provide better support for sustainable development studies.