Data understanding for flash flood prediction in urban areas

Abstract

Flash flood has become one of the major disastrous events, especially in urban areas in Malaysia. It has become more prominent to city dwellers, causing massive loss of infrastructures, damage to people, and disruption in business and daily activities. Population growth and rapid development of urban areas have worsened the situation even more. Since the era of Big Data, the possibility to analyse complex data coming from heterogeneous sources, which can be used to predict flash flood, has given a different perspective and hope for finding innovative ways to reduce the impact of flood, especially in urban areas. The purpose of this study is to understand data needed to produce predictive visual analytics for flash flood forecasting using Cross-Industry Standard Process for Data Mining (CRISP-DM) Methodology. Focusing on understanding the flash flood data, this paper intends to characterize data pertaining to disaster management and identify the right data that can facilitate more accurate decision making by stakeholders. Literature review was done to determine which data are needed in the Malaysian urban setting. The research found the critical factors for determining flash flood occurrence in Malaysia are unique due to the tropical climate and urbanization. Therefore, it is important to understand and characterize these factors for more effective and accurate data collection and predictive analytics later. Based on the findings, the most significant factors identified for flash flood prediction are rainfall, urbanization, and fluvial flood which eventually lead to blocked drainage. Details of data under these categories will be analysed as part of data understanding of flash flood occurrence. This study intends to uncover the potential of using Predictive Visual Analytics in flood forecasting and also to discuss how prediction can bring values to the Malaysian environment and create a sustainable ecosystem.