Scientists have shown that sea levels have been rising at the rate of about 3.4 mm per year, threatening to make the weather anomalies of today the average of the future.
With flooding events increasing in frequency and extremity, early-detection and warning systems are becoming more and more relevant to coastal populations around the world.
Unfortunately, current methods of detection, such as remote satellite sensing, local sensor networks, witness statements, and insurance reports are too slow and unreliable to be of much use in disaster-prevention.
To this end, Dr Roger Wang and colleagues from the University of Dundee’s School of Science and Engineering in Scotland had combined Twitter data, citizen reports and artificial intelligence algorithms to model flooding as close to real-time as possible.
Researchers used key words used in Twitter messages, natural language processing to gather more information on severity and location, and computer vision to analyse pictures uploaded by users of the crowdsourcing app MyCoast to look for signs of flooding.
“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,” said Wang in a press release.
Precision of the information extracted by the algorithm was validated by comparing it to precipitation data (which correlated with flood-related tweets) and road closure reports (found to match with the data collected from the app).
Wang recognised the need to further develop the computer vision techniques used in the study – currently at roughly 70% accuracy – but claimed that Twitter could be used to gather coarse-grained information on large-scale events, while data provided by “citizen scientists” is potentially useful on the micro level.
“Taken together, these tools can be used to monitor the water penetration of urban flooding over a city. This can be then used to improve forecasting models and early warning systems to help residents and authorities prepare for an upcoming flood.”
The paper was published in the latest edition of the journal Computers & Geosciences.