Assessing urban flood disaster risk using Bayesian network model and GIS applications

Abstract

With global climate change, cities face the challenge of increasing flood disaster caused by heavy rainfall, and the prediction and assessment of flood disaster risk is a crucial step towards risk mitigation and adaptation planning. In this study, a method combining Bayesian network (BN) model and geographic information system (GIS), which can capture the potential relationships between factors impacting flood disaster and has capacity of quantifying uncertainty and utilizing both data and knowledgebased sources, was proposed to assess flood disaster risk. The proposed methodology was applied in a case study to assess flood disaster risk and to diagnose the reason for flood disaster in Zhengzhou City, and the results were validated by comparing with actual situation. The results show that that the relative error of very-low, low, moderate, high and very-high risk predicted by the proposed model is 12.57%, 13.21%, 2.23%, 19.63% and 21.65%, respectively, which demonstrates the discriminative power of the established model. Based on the spatial distribution of different risk levels, it can be recognized that the flood disaster risk in Zhengzhou City is decreasing from the middle to the surroundings. The results provide some basis for the field control and management of urban flood disaster.

Key Disaster-Causing Factors Chains on Urban Flood Risk Based on Bayesian Network

“…Therefore, these risk results provide baseline information that needs to be considered before formulating management plans and strategies for management, prevention, and reduction of flood disasters. In order to verify the model evaluation results, the high-risk map of the predicted high risk and very high region extracted from Figure 4b is compared with the actual inundation range in the Figure 2 [34]. The matching rate between them is calculated by superposition in ArcGIS, and the matching degree between them is 92.04% (Figure 6b), which conforms basically in reasonable agreement with the actual inundation data.…”

“…Ten factors are selected to evaluate urban flood inundation risk and determine inundation-prone areas in the study area according to previous studies [1,2,11,14,[32][33][34]41,44], the actual situation of the study area, the availability of data, as well as experts' knowledge and experience. The hazard factor is annual rainfall; the vulnerability factors are elevation, slope, soil water retention (SWR), river density, distance to river; the capacity factors are pipe density, road density, population density, and per unit GDP (Figure 2).…”

“…The factors' data collected from different sources can be summarized into three categories: (1) raster data, Digital Elevation Model (DEM), soil, land cover; (2) vector data, the boundary of the study area, river network, road network, and pipe network; Obtaining the historical inundation area is the key to the evaluation of the flood inundation risk in BN to explain the correlation between flood inundation risk and influencing factors [25,34]. In this study, the 100-year return period inundation area is derived from the report of the World Bank Loan Project, as the maximum inundation range of the study area.…”

“…Wu et a. [34] use BN model to assess the flood risk in Zhengzhou, China. In addition to urban flood risk assessment, BN is widely used in river basin flood risk assessment [35], coastal hazard assessment [36], flood detection [3], and early warning for urban flood [37].…”

A probabilistic approach to estimating residential losses from different flood types

“…The flood damage model utilizes a class of graphical, probabilistic models known as Bayesian networks (BNs). In recent years, they have been increasingly used for flood risk modelling applications (Paprotny and Morales-Nápoles 2017;Beuzen et al 2018;Couasnon et al 2018;Jäger et al 2018;Wu et al 2019). Still, BN-based flood damage models have been few created from flood loss data for Germany (Schröter et al 2014;Vogel et al 2018;Paprotny et al 2020b) and the Netherlands (Wagenaar et al 2017(Wagenaar et al , 2018.…”