USP Develops Software to Aid in Identifying Flood Victims

Floods are recurrent and affect a large part of the Brazilian territory. Especially since April 29th, the State of Rio Grande do Sul has been suffering from extreme rains, which is being responsible for one of the biggest climate disasters in the country’s history, which has already affected more than 2 million people. To mitigate the effect of these disasters, USP has developed a series of technologies to combat flooding, including software that works with artificial intelligence that relates directly to social networks and official government data.

“In fact, it is a program that uses artificial intelligence to combine information from tweets with data from sensors installed by the State of São Paulo. We made this correlation and verified that the combination of information from social networks with data from sensors increases the certainty, so to speak, that a flood is about to occur”, says professor Jó Ueyama from the Institute of Mathematical and Computer Sciences ( ICMC) from USP in São Carlos.

Agile response

This technology can generate models of how these events spread across different regions of a State, anticipating recognition. Doctoral student Caetano Ranieri, also from ICMC, adds that, for example, “instead of alerting the population ten minutes earlier, we can alert them an hour earlier”.

According to Thiago Costa, a doctoral student at ICMC, artificial intelligence uses a computational pattern called “multimodal data fusion”. The objective is to identify flood victims from heterogeneous data sources. From this perspective, we have artificial intelligence responsible for identifying whether messages published on the social network Twitter are related to climate information.

For Jó Ueyama, what makes implementing this technology difficult is the need for investment. “It is necessary for public bodies to take the problem of flooding seriously and install this type of equipment, as this can happen in any city. If we had known there would be this flood, we would have prepared. However, no one imagined that, after 80 or 100 years, a flood of this proportion would occur. As a result, there was no investment for this to be installed and operational”, reports the professor.