Time series analysis of VIIRS-DNB nighttime lights imagery for change detection in urban areas: A case study of devastation in Puerto Rico from hurricanes Irma and Maria

Abstract

Brightness of nighttime lights (NTL) collected by the Suomi National Polar-orbiting Partnership (S-NPP) Visible Infrared Imaging Radiometer Suite (VIIRS) is compatible across different times of images thanks to the on-board calibration system. However, the NTL radiance observed by the S-NPP VIIRS shows clear seasonality corresponding to the seasonal changes in the albedo of land surface. Additionally, the existence of many uncertain factors (e.g. complex atmospheric conditions) renders it inappropriate to directly use the NTL radiances to derive changes on the ground. In this study, we adopt a statistical procedure of time series analysis, namely seasonal and trend decomposition using Loess (STL), to model the time series observations of NTL brightness by decomposing the observations into three separable time series components (i.e. trend, seasonality, and remainder). Based on the time series model, forecast can be made for short-term future with confidence measure, and by comparing the model forecast with observed NTL brightness, significant changes can then be detected at pixel levels. We applied this method to the Puerto Rico area to detect and assess the damages caused by Hurricanes Irma and Maria, and to monitor the recovery after the disaster. Our results show that the proposed method successfully captures the changes of NTL brightness due to the damage of the hurricanes and general economic decline. Moreover, we also find that after removing the seasonal and remainder components, the time series of NTL image can more accurately reflect the temporal trends of economic status in Puerto Rico.