[A-HW20] Integrated Watershed Management under the Future Extreme Disaster

[AHW20-02] Using the Bayesian theorem to assess flood damage in industrial areas in Taiwan

*Yuan-Yuan Tien¹, Tsun-Hua Yang¹ (1.National Yang Ming Chiao Tung University)

Keywords:Multivariable, Flood damage, Bayesian Theorem, Stage-damage curve

As global climate change intensifies, the frequency of flood-related disasters is increasing. In Taiwan, hydrometeorological disasters, including typhoons and floods, are predominant and account for a significant portion of overall calamities. However, there is still a lack of efficient approaches to assess the economic damages of floods, especially in industrial areas in Taiwan. Therefore, this study proposed three Bayesian theorem-based models, which are (1) stage-damage function (SDF), (2) multivariable Bayesian regression (BR) approach, and (3) Bayesian networks (BN) to estimate flood damages. The first approach is an univariable approach, exclusively assuming that flood damage is a function of water depth. The second approach is to establish a multivariable probabilistic flood loss model that considers different influencing factors. Given observation data, the parameters in the probabilistic model were estimated using the Bayesian theorem-based regression approach. The last model is to establish a BN-based flood damage model in which the influencing factors and their conditional dependencies are represented in a directed acyclic graph (DAG) structure. From this network, the probability distributions of the flood loss can be predicted based on knowledge about the influencing factors. Using data from the HOWAS21 database, maintained by the Helmholtz Centre Potsdam GFZ German Research Centre for Geosciences, all three models and associated parameters are carefully calibrated and verified. Subsequently, the ambit of this inquiry extends to industrial locales within Taiwan under the presumption of exhaustive data coverage of all pertinent influencing factors whilst disregarding latent variables. Ultimately, flood risk is quantified in terms of monetary values, aiming to establish a quick flood loss estimation model pertinent to industrial areas in Taiwan.