An article published in the journal “Computers & Geosciences” describes a research into the possibilities of combining artificial intelligence, Twitter and what is called citizen science to create early warning systems for flood-prone communities. A team of researchers from the British University of Dundee led by Roger Wang worked to create solutions that can detect as quickly as possible the first signs of danger to activate countermeasures.
The problem of floods in urban areas has become increasingly severe in recent years, partly due to some difficulties in collecting and processing data on these phenomena. According to Dr. Roger Wang of the University of Dundee’s School of Science and Engineering and his team the application of the most modern technologies could allow the creation of cheap but very effective early warning systems thanks to the collaboration of common people.
The concept of citizen science has been emerging in recent years thanks to the new possibilities of collaboration offered to common people thanks to the many devices connected via the Internet. It’s a concept that doesn’t have a rigorous definition but generically concerns the scientific activities in which common people participate.
This concept is typically linked to that of crowdsourcing, where someone uses the help of a group of people to solve a problem or get something. The development of Internet connections allowed the development of global projects based on crowdsourcing and certain types of scientific research also make use of them.
In the case of the flood problem in urban areas, Roger Wang’s team discovered that crowdsourcing can be a key factor in gathering information to be combined with those made available through traditional monitoring systems. In recent years, more and more people have been writing on social media what’s happening in front of their eyes and even more send photos and videos collected thanks to cameras now part of many mobile devices.
The images, which are phots or videos, can be analyzed by computer vision systems that can recognize certain specific characteristics. In particular, tweets from people who provided real-time flood information on Twitter were used in the research. A system capable of analyzing the language of the tweets’ text part was used to determine severity, location and other information.
These systems can recognize the relevant tweets to provide useful information to monitor these phenomena from the beginning in order to quickly activate countermeasures. Roger Wang’s team will keep on training the system to improve its accuracy, also using photos from MyCoast, a system used by various environmental agencies that takes advantage of citizen science. The ultimate goal is to recognize the first signs of danger to have early warning systems with a very quick response.