Does flood early warning performance affect flood damage? Evidence from the 2018 Japan Floods

Flood early warning systems are vital for mitigating flood damage, yet limitations in forecasting technologies lead to false alarms and missed events. Repeated occurrences of these issues may cause people to hesitate to take appropriate action (e.g., evacuating or moving assets to safer places) during subsequent warnings, potentially exacerbating flood damage, including both human and economic losses. However, the impact of warning performance on flood damage in Japan has not been examined in the context of 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 outcome variables were (1) fatalities, (2) injuries, (3) economic losses to general assets, and (4) economic losses to crops during the floods.

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. These findings provide valuable insights for improving warning systems and guiding future research.

This presentation is based on our recent publication in Journal of the Meteorological Society of Japan. Ser. II (DOI: 10.2151/jmsj.2025-025).