Generation of artificial rainfall events for flood risk assessment in Taiwan

Typhoons frequently cause casualties, property damage and economic losses in Taiwan. Several flood risk models have been developed to assess loss for Taiwan’s property insurance industry. However, these models have ignored the potential for extreme rainfalls, which could lead to underestimation of assessments of potential property loss. This could be a problem if, e.g., only data from the 239 typhoon alert events recorded by Taiwan’s Central Weather Bureau from 1960–2010 are considered in the flood risk models. This work generates a sufficient number of artificial rainfall events, including potentially extreme rainfall events using truncated lognormal distributions and correlation matrices of historic rainfall amounts for the watershed basins in Taiwan. The artificial events basically demonstrate similar probability distributions and spatial distributions of rainfall as historic events. A practical case study is carried out that includes different artificial events. Comparisons of different event cases and their application for flood risk assessment for a property portfolio are made. Higher losses could occur in the artificial cases of extreme rainfall than the losses from historic cases. This outcome should be seen as a practical supplement for flood risk assessments, considering the influence of potentially extreme future rainfall events.