Analysis of Mumbai Floods in recent Years with Crowdsourced Data

TL;DR

This study demonstrates how crowdsourced Twitter data can be effectively used to identify flood hotspots, monitor floods in real-time, and evaluate mitigation measures in Mumbai, addressing data scarcity issues in urban flood analysis.

Contribution

It introduces a novel method of using Twitter data combined with geographic information systems to map floods and assess mitigation impacts in a data-sparse urban environment.

Findings

Twitter data accurately identifies flood hotspots

Flood reporting decreased despite increased rainfall due to mitigation

Crowdsourced data enhances real-time flood monitoring capabilities

Abstract

Mumbai, a densely populated city, experiences frequent extreme rainfall events leading to floods and waterlogging. However, the lack of real-time flood monitoring and detailed past flooding data limits the scientific analysis to extreme rainfall assessment. To address this, we explore the usability of crowdsourced data for identifying flood hotspots and extracting reliable flood information from the past. Through an automated program, we filter and retrieve flood-related data from Twitter, using location information to generate flood maps for past heavy rainfall events. The validity of the retrieved data is confirmed by comparing it with volunteered geographic information (VGI) which is more accurate but less abundant. In the absence of direct flood information, Twitter data is cross-verified with the Height above the Nearest Drainage (HAND) map, which serves as a proxy for elevation. Interestingly, while extreme rainfall events are increasing in frequency, recent Twitter-based information shows a decrease in flood reporting, attributed to effective mitigation measures implemented at various flood hotspots. Local surveys support this finding and highlight measures such as underground storage tanks and pumping stations that have reduced flood severity. Our study demonstrates the value of crowdsourced data in identifying urban flood hotspots and its potential for real-time flood monitoring and forecasting. This approach can be adapted for data-sparse urban regions to generate location-specific warnings, contributing to improved early warnings and mitigating the impact on lives and property.