Researchers are combining Twitter, citizen science and artificial intelligence (AI) techniques to develop an early-warning system for flood-prone communities in urban areas. In a study, published in the journal Computers & Geosciences, the researchers showed how AI can be used to extract data from Twitter and crowdsourced information from mobile phone apps to build up hyper-resolution monitoring of urban flooding.
“By combining social media, citizen science and artificial intelligence in urban flooding research, we hope to generate accurate predictions and provide warnings days in advance,” said Roger Wang from University of Dundee in Britain. Urban flooding is difficult to monitor due to complexities in data collection and processing.
This prevents detailed risk analysis, flooding control and the validation of numerical models. The research team set about trying to solve this problem by exploring how the latest AI technology can be used to mine social media and apps for the data that users provide.
They found that social media and crowdsourcing can be used to complement datasets based on traditional remote sensing and witness reports. Applying these methods in case studies, they found them to be genuinely informative and that AI can play a key role in future flood warning and monitoring systems.
“The present recording systems — remote satellite sensors, a local sensor network, witness statements and insurance reports — all have their disadvantages. Therefore, we were forced to think outside the box and one of the things that occurred to us was how Twitter users provide real-time commentary on floods,” Wang said.
“A tweet can be very informative in terms of flooding data. Key words were our first filter, then we used natural language processing to find out more about severity, location and other information,” Wang said. The researchers applied computer vision techniques to the data collected from MyCoast, a crowdsourcing app, to automatically identify scenes of flooding from the images that users post.
“We found these big data-based flood monitoring approaches can definitely complement the existing means of data collection and demonstrate great promise for improving monitoring and warnings in future,” Wang said.
Twitter data was streamed over a one-month period in 2015, with the filtering keywords of “flood”, “inundation”, “dam”, “dike”, and “levee”. More than 7,500 tweets were analysed over this time. “We have reached the point of 70 percent accuracy and we are using the thousands of images available on MyCoast to further improve this,” Wang said.