The Contribution of Crowdsourcing to Urban Flood Data Collection, Monitoring, and Communication: Case Study in the City of Ouagadougou

Abstract

According to observations and the nearest future, flooding is among the most frequent climate hazards; the frequency and amplitude are expected to increase according to climate change projections. During such events, Facebook, Twitter, Snapchat, LinkedIn, and other social media are used to air firsthand information. Crowdsourcing uses electronic tools to improve data collection, monitoring, and communications including web scraping to collect data and information during and after disasters to gain situational awareness. It also uses an open data kit (ODK) to collect other in-situ datasets. In this study, we used web scraping and ODK techniques to find the link between the triggers and consequences of urban floods. The web scraping was combined with machine learning algorithms and exploratory analysis tools to collect flood-related messages from the Twitter website for the City of Ouagadougou (Burkina Faso). Using natural language processing techniques, the data set was normalized by tokenization and lemmatization to clean and extract emotions for the textual data of each tweet. Further, we used machine learning approaches such as NLTK and TextBlob to analyze the sentiments and polarity in the data sets and supervised machine learning (logistic regression) for text classification. The accuracy of the data was assessed using the rain gauge dataset and the reports of the disaster management agency. The tweet word cloud generated exhibited the onset of urban flood events, the causes, the effects, and the spatial extent of flood events in Ouagadougou. The text classification yielded a 93% accuracy of the Twitter data sets compared to instrumental measurements and officially reported observations. The ODK questionnaire was deployed among multiple stakeholders across the City of Ouagadougou to identify landfill triggers of floods in urban areas such as both official and unofficial dump sites. Hence, crowdsourcing is now designated as an efficient method for data collection, improving the quality and visualization to contribute to event databases, effect-based monitoring, and communications of urban floods.