Flood and Typhoon Damage Estimation in North Korea Using a Principal Component Analysis-Based Regression Model

A statistical model was developed in this study for a quantitative estimation of the damage caused by floods and typhoons in North Korea, and its predictive performance was evaluated. Despite North Korea experiencing recurring damages due to natural disasters such as floods and typhoons, a systematic damage estimation framework is lacking. Therefore, a damage estimation model was developed by treating the entire territory of North Korea as a single analysis unit, and using damage statistics from 1993 to 2018 along with data on precipitation and wind speed from 27 meteorological stations. The principal component regression model showed overall superior performance compared to a generalized linear model, achieving (a) an average NRMSE of 0.1901 and R-squared of 42% for flood damage estimation, and (b) an average NRMSE of 0.3077 and R-squared of 85% for typhoon damage estimation. This quantitative methodology proposed for estimating disaster damage in data-scarce environments can serve as a model and be applied to other countries facing similar disasters and lacking systematic damage estimation frameworks.