Satellite based inference for disaster induced power outages using daily nighttime light observations

This study demonstrates that daily nighttime light observations can be transformed into outage-sensitive indicators using a physically motivated attenuation representation that accounts for variations in satellite coverage and background illumination. The proposed representation exhibits a strong empirical relationship with observed outage fractions and provides a basis for inferring outages from satellite-derived observations alone. Because this formulation is independent of hazard-specific variables and detailed power-system information, it can be applied across diverse hazards and geographic regions.

They presented a case study on detecting outages in Humboldt County, California, during Public Safety Power Shutoffs (PSPS) implemented to mitigate wildfire risk. The framework demonstrated the ability to identify broad outage patterns across multiple events and hazards, including the case of Hurricane Fiona (2022) in Puerto Rico. The team also presented the application of their framework to detect outages for Antakya and Defne, the most severely affected districts from the 2023 Turkey earthquakes.