Evaluating Road Lighting Quality Using High-Resolution JL1-3B Nighttime Light Remote Sensing Data: A Case Study in Nanjing, China

Abstract

A good lighting environment for roads at night is essential for traffic safety. Accurate and timely knowledge of road lighting quality is meaningful for the planning and management of urban road lighting systems. Traditional field observations and mobile observations have limitations for road lightning quality evaluation at a large scale. This study explored the potential of 0.92 m resolution JL1-3B nighttime light remote sensing images to evaluate road lighting quality in Nanjing, China. Combined with synchronous field measurements and JL1-3B data, multiple regression and random forest regression with several independent variable combinations were developed and compared to determine the optimal model for surface illuminance estimation. Cross validation results showed that the random forest model with Hue, saturability, ln(Intensity), ln(Red), ln(Green) and ln(Blue) as the input independent variables had the best performance (R2 = 0.75 and RMSE = 9.79 lux). Then, this model was used to map the surface illuminance. The spatial scopes of roads were extracted from Google Earth images, and the illuminance within roads was derived to calculate the average, standard deviation and coefficient of variation to indicate the overall brightness level and brightness uniformity of the roads. This study provides a quantitative and effective reference for road lighting evaluation.