Abstract Japan has been experiencing depopulation and an ageing population for decades. Particularly, the eastern (Pacific) coast of the Tōhoku region in northeastern Japan was devastated by the tsunami of the Great East Japan Earthquake in 2011. To assess the socioeconomic resilience of local communities in the area, we conducted a comparative time-series analysis of the estimated wealth index using a deep learning model, which used nightlight imagery as input. We examined 12 coastal municipalities in Iwate Prefecture in the east that were devastated by the 2011 tsunami. Additionally, we selected nine coastal municipalities in Akita Prefecture in the west to serve as a control group. The results indicated the estimated wealth index gradually recovered in the East Coast municipalities from 2016 to 2018, owing to the positive impact of the restoration projects. This recovery occurred two to four years later than the government's observations based on the industrial statistics. The estimated index was highly correlated with the general indicators of socioeconomic activity and was cross-validated with several local episodes. These findings prove that nightlight-based remote sensing is applicable not only for visualising population density but also for recovery from great disasters and the effects of depopulation. This study offers the insight that macroscopic estimation should be cross-checked using microscale factual data to close interpretative gaps.
Kondo et al. (Mon,) studied this question.