A spatial assessment of urban waterlogging risk based on a Weighted Naïve Bayes classifier

Xianzhe Tang; Yuqin Shu; Yanqing Lian; Yaolong Zhao; Yingchun Fu

doi:10.1016/j.scitotenv.2018.02.172

A spatial assessment of urban waterlogging risk based on a Weighted Naïve Bayes classifier

Sci Total Environ. 2018 Jul 15:630:264-274. doi: 10.1016/j.scitotenv.2018.02.172. Epub 2018 Feb 22.

Authors

Xianzhe Tang¹, Yuqin Shu², Yanqing Lian³, Yaolong Zhao¹, Yingchun Fu¹

Affiliations

¹ School of Geography, South China Normal University, 510631 Guangzhou, PR China.
² School of Geography, South China Normal University, 510631 Guangzhou, PR China. Electronic address: shuyq@m.scnu.edu.cn.
³ School of Geography, South China Normal University, 510631 Guangzhou, PR China; Illinois State Water Survey, Prairie Research Institute, University of Illinois, 61820 Champaign, IL, USA.

PMID: 29477824
DOI: 10.1016/j.scitotenv.2018.02.172

Abstract

Urban waterlogging occurs frequently and often causes considerable damage that seriously affects the natural environment, human life, and the social economy. The spatial evaluation of urban waterlogging risk represents an essential analytic step that can be used to prevent urban waterlogging and minimize related losses. The Weighted Naïve Bayes (WNB) classifier is a powerful method for knowledge discovery and probability inference under conditions of uncertainty; a WNB classifier can be applied to estimate the likelihood of hazards. Six spatial factors were considered to be added to the WNB, which may improve the efficiency in predicting urban waterlogging risk during analysis. As such, a spatial framework integrating WNB with GIS was developed to assess the risk of urban waterlogging using the primary urban area of Guangzhou in China as an example. The results show that 1) the rationality of six spatial factors was determined according to the Conditional Probability Tables and weights; 2) the Most Accurate Sampling Table has objectivity; and 3) the areas with a high likelihood of waterlogging risk were mainly located in the southwestern part of the study area. The northeastern zones are relatively free of waterlogging risk. The results reveal a more accurate spatial pattern of urban waterlogging risk that can be used to identify risk "hot spots". The resulting gridded estimates provide a realistic reference for decision making related to urban waterlogging.

Keywords: GIS; MAST; Spatial framework; Spatial pattern.