Estimating natural disaster loss using improved daily night-time light data

Abstract

Satellite-observed night-time light data has been widely applied to estimate natural disaster loss, which is important for evaluating progress of SDG 11 and 13, but large uncertainty of the time series data hindered the evaluation in high accuracy. This study aims to estimate power outage, a proxy of economic loss, after natural disasters using VIIRS daily night-time light data with improved quality, which is able to capture abrupt change of the light. Firstly, an angular normalization algorithm was developed to generate a stable time series of night-time light data with less fluctuation, and it was validated with different metrics. The algorithm was then applied to analyze power outage after 2017 Maria Hurricane in Puerto Rico, and it was found that the Pearson correlation coefficient between the improved night-time light data and power supply is 0.929 in Puerto Rico, suggesting that the data is accurate to reflect the power supply. For the eight regions in Puerto Rico, they totally lost 78.01% of the power supply after the hurricane, while Caguas, Guayama and Humacao are the regions suffering from the hurricane most severely, losing 94.29%, 92.85% and 88.38% of power supply, respectively, and their recovery speeds are also relatively slow compared to other regions. In addition, we discuss the possibility of using the night-time light indicator for estimating GDP loss of service and industry due to disaster-derived power outage.