Intelligent early warning system protects against urban flash floods

Extreme rainfall in a very short space of time is always a challenge for urban drainage. And they are occurring more and more frequently as a direct result of climate change. Urban flash floods are particularly problematic and dangerous. These can occur when sewer networks are overloaded and are no longer able to cope with the enormous volumes of water.

The trick is in the detail: if rivers and streams burst their banks, the immediate catchment areas of the watercourses are affected - which is often bad enough. Urban flash floods, however, can theoretically occur anywhere and, above all, simultaneously in different places. This is a safety risk, especially in cities and densely populated regions and particularly where there are many low points at risk of flooding, such as underpasses or subway stations.

According to Bernd Bäumler, member of the management board at fuseki, innovative and “reliable solutions” are therefore needed for dealing with and forecasting extreme weather events. The problem with current methods: They are too imprecise, give warnings over too large an area and therefore sometimes unnecessarily. And: they process the data too slowly and, above all, they generally only deliver static scenarios as a result. “With KIWaSuS, we are taking a different approach. We bring together all relevant data in real time on a single platform. This data is examined and evaluated via an analytics component and then output as a user interface. This means that the data streams are visualized, i.e. made comprehensible, and then show the user where problems are likely to occur.”

In other words: KIWaSuS is a real-time warning and real-time management system for urban flash floods using AI and machine learning methods. Data from the German Weather Service (DWD) and the sewer network control system are primarily evaluated. Special rain sensors developed by the Ruhr West University of Applied Sciences also measure the values on the roads. “So we use various data sources and combine the results. However, the rain data from the DWD is the most important for us; the other values help us to consolidate and supplement the database.”

The result: dynamic digital maps that show the location and extent of flooding in the event of heavy rain at an early stage and with localized detail. Up to 60 minutes warning time is realistic, says Bäumler. Time for the fire department and technical relief organization to warn the population, clear areas and take countermeasures. KIWaSuS works much more reliably and locally more accurately than static models.

The pilot ran for three years in Gelsenkirchen. During this time, KIWaSuS was taught, trained and collected a lot of data. “We are currently reviewing the outcome of the pilot phase in order to develop KIWaSuS into an independent product.” Talks are already underway with the city of Schwerte to install the system there. There are also plans for a research project to follow on from KIWaSuS in order to further optimize the process.

Basically, Bäumler summarizes, “the Ruhr area is a very important region for the water management sector. We have important contacts and cooperation partners here”. The latter include the Ruhr West University of Applied Sciences and the University of Duisburg-Essen: “Both institutions provide us with a great deal of support for our research projects, and they are also home to important young scientists that we need in the region. At the same time, the Ruhr metropolis is also “exactly the right region” for a company like fuseki in terms of end customers: “There are large companies and associations from the water management sector here in the Ruhr area that need the AI solutions we develop. In terms of networking, everything really fits here for us.”

Text: Editorial office Schacht11  
Pictures: fuseki