AI analyses flood risks to road networks

Could the technology help to reduce flood-related car accidents?

Spotted: Reducing emissions is key to stabilising the climate and minimising extreme weather events. But while the world works towards net zero, it’s also essential that we mitigate the effects of natural disasters, including by helping communities respond in a timely and informed manner. To this end, engineers at Rice University are developing an AI solution to assess flood risk on roads in real time.

Although existing data sources – such as traffic cameras and water level sensors – can observe floods as they occur, they aren’t always designed with that end in mind, and rarely work together. To address this issue, Rice’s OpenSafe Fusion framework combines and enhances available data, leveraging AI physics-based modelling together with network and spatial analyses. This means the technology can monitor quickly changing flood conditions in urban areas, which would help local inhabitants to avoid the worst-hit routes and prevent life-threatening road accidents.

Alternative monitoring methods would require structural changes or increasing the number physical sensors, which would be costly and deliverable only over the long term. Noting that urban areas “are replete with sources directly or indirectly observing flooding or road conditions,” Professor Jamie Padgett, Chair of Civil and Environmental Engineering at Rice, together with postdoctoral researcher Pranavesh Panakkal, turned their attention to a more practical solution: combining existing, real-time data sources without the need for significant investment.

Inspired by their collaboration with colleagues at the SSPEED Center at Rice, who have developed nest-generation flood alert systems, the researchers are now planning to extensively test and validate the OpenSafe Fusion framework. One such test, recreating the conditions of Houston’s Hurricane Harvey in 2017, resulted in the model observing as many as 37,000 road links. By examining a diverse array of conditions, the team hopes to better understand the system’s performance and introduce improvements where needed.

Written By: Duncan Whitmore