Does the Performance of a Flood Early Warning System Affect Casualties and Economic Losses? Empirical Analysis Using Open Data from the 2018 Japan Floods

Abstract

Flood early warning systems are crucial for mitigating flood damage; however, limitations in forecasting technology lead to false alarms and missed events in warnings. Repeated occurrences of these issues may cause people to hesitate to take appropriate action during subsequent warnings, potentially exacerbating flood damage. However, the effects of warning performance on flood damage in Japan have not been analyzed for actual flood events. This study empirically examined these effects by applying Bayesian regression analyses to open data on the 2018 Japan Floods in 127 municipalities in four prefectures (i.e., Okayama, Hiroshima, Ehime, and Fukuoka) for which data were available on the real-time flood warning map (Kouzui Kikikuru in Japanese) during the 2018 Japan Floods, which provides limited open data on warning performance. Based on these data, the false alarm ratio (FAR) and missed event ratio (MER) for each municipality before the 2018 Japan Floods were calculated and used as explanatory variables. The (1) fatalities, (2) injuries, (3) economic losses to general assets, and (4) economic losses to crops during the 2018 Japan Floods were used as outcome variables. The results indicate that a higher FAR was associated with an increase in fatalities, injuries, and economic losses to general assets. By contrast, no prominent positive effect of MER was found for any outcome variable. Although our results are fundamental, they provide valuable insights for improving warning systems and guiding future research.