Coastal Inundation Risk Analysis Using Bayesian Network

Abstract

Recently, natural hazards have been more unpredictable with increasing frequency and strength due to climate change. Especially, coastal areas could be more vulnerable in the future because of climate change . In the case of South Korea (hereafter, Korea), the country is surrounded by ocean and there are many large cities along the coastal areas. Thus, a series of hazard prevention plans are necessary in the coastal areas. However, prior to formulating a plan, the first step would be to find risk areas. The local characteristics of coastal areas should also be considered in order to find vulnerable areas. Therefore, the objective of this study is to find vulnerable areas that could be damaged by coastal hazards caused by typhoon and rainfall when considering the local environment. The contextual scope of this study was narrowed down to coastal inundation cause by typhoon and rainfall. The spatial scope was set up as an administrative district located close to the coastline. Physical and socio-economic characteristics were considered to evaluate the risk in coastal areas. Risk analysis was carried out through the combination of possibility of hazard and the level of damages. Risk analysis was implemented by using Bayesian Networks (BNs), which is a stochastic-statistics method. BNs are based on Bayes Rule, which is calculated by using prior probability to estimate posterior probability . In other words, after creating a network from prior information, the posterior information is calculated by using the Bayesian method. Thus, the possibility of coastal inundation caused by typhoon and rainfall was estimated by using the Bayesian method and the level of damage was estimated by coordinating the probability result of inundation with each of 4 socio-economic dimensions, which are human, infrastructure, environment and socio-economic. The level of damage was also estimated by using the Bayesian technique . As a result, the total risk of risk analysis was calculated by summing up the result from 4 dimensions. According to the result of the study, the Songdo area (Incheon), the Baegot development-prearranged area (Siheung) and the lake region (Assan) were shown to be the vulnerable areas. Songdo needs special coastal management in the future since a study showed that Songdo would become a vulnerable area due to sea level rise and other coastal hazards. Although the Baegot development-prearranged area has not been developed yet, a coastal development and hazard plan should be set up for preventing possible natural hazards. The lake region consists of an agricultural area, experiencing frequent flooding. Thus, the lake region must be protected to minimize damage to agriculture due to coastal inundation. This study, however, has a limitation on data since not all of the past 30 years of information could be used as an input data. Also, dividing the coastal inundation events into three categories was random. Although with these limitations, this study has academic and practical significance. First, the study considered physical and socio-economic variables at the same time, which has not been examined in prior studies. Second, the Bayesian Networks (BNs) were used to find the risk areas, which were not usually used in domestic studies. BNs also allow us to consider many variables, reflecting complicated and diverse environment such as coastal area, and to illustrate the statistical analysis into a spatial result. Since the research required both physical and socio-economic characteristics in evaluating risk analysis for coastal cities, BNs seemed to be a suitable method. In conclusion, the use of Bayesian Networks in risk analysis could be applied to manage coastal cities, and as illustrated in this study, the Integrated Coastal Zone Management (ICZM) of Korea.