Controlling Urban Flash Flooding with Big Data
Vehicle rain sensors are to act as mobile measuring stations to make it easier to predict heavy rain
We are experiencing an increase in local heavy rain events that are difficult to predict in terms of the exact location and time when they will occur. This is why there is a need to improve the accuracy of early warning systems with regard to when and where they will happen. In the mobileVIEW research project, IAV is working with two partners on a new approach. Vehicle sensors act as mobile measuring stations to supply high-resolution geographic data for short-term forecasting. In addition to the water sector, this could also have benefits for fire brigades, disaster response services and the general public.
On June 29, 2017, nearly 200 liters of rain per square meter fell in the Tegel district of Berlin in just 24 hours. That is a huge quantity, compared to the overall annual average of just about 800 liters of rainfall per square meter in Germany. Experts call this a heavy rain event. The German Meteorological Service (DWD) defines heavy rain for our region as when at least five liters of rain per square meter fall in five minutes, or 17 liters of rain per square meter in one hour. Heavy rain events like this will occur with increasing frequency, causing major damage including fatalities and injuries. Timely precise forecasting of flash flooding is therefore increasingly important.
Dense network of measuring stations needed
Unfortunately, that is not possible at present. “Heavy rain has a different genesis and other specific characteristics compared to conventional precipitation”, explains Dr. Matthias Pätsch who deals with water management issues at IAV. “It occurs at very short notice on a local scale so that existing measuring methods are inadequate.” The forecasting of normal precipitation and its consequences is based on widespread measuring stations (usually 20 per 1,000 square kilometers, ideally rated to spatial variability of precipitation, e.g. more measuring stations in hilly regions) together with the DWD rain radar and relevant satellite information. Predicting local heavy rain needs a far higher density of measuring stations, particularly in the urban context.
This is where vehicle rain sensors come in. They are omnipresent nearly on a 24/7 basis and could contribute missing data. While not providing exact quantitative measurements, they do indicate nuances in rain intensity which would be sufficient for detecting heavy rain. By linking the data from existing measuring stations, the rain radar and relevant satellite information with the measurements from mobile rain sensors, cells of precipitation could be detected at very short notice, leaving enough time to take countermeasures. “Regional responsibility for flood control lies with the water boards and the local authorities”, says Pätsch. “Timely information as to where the heavy rain is going to fall would let them respond promptly and take appropriate retention measures.” The fire brigade could also be dispatched to seal off underpasses or roads at risk of flooding.
100 test vehicles as mobile weather stations
IAV is joining forces in the mobileVIEW research project with the Research Institute for the Water Industry (FiW) at RWTH Aachen University of Technology and the EG/LV water board (Wasserverband Emschergenossenschaft/Lippeverband Essen) to turn this vision into reality. “Data boxes will be put in 100 vehicles from the EG/LV fleet, connecting them up to an Internet of Things platform so that the sensor data can be transferred”, explains Pätsch. “They are then combined with the data from the stationary measuring stations, the rain radar and the satellites”. IAV is also responsible for the algorithms used to visualize the heavy rain events. The mobileVIEW project began in October 2017 and runs until September 2020. It is being funded by the Federal Ministry of Transport and Digital Infrastructure with altogether € 2.2 million.”
The evaluation process is not a trivial affair. Not all the data are equally valuable. When a car drives down a narrow road, the measurements supplied by the rain sensors probably differ from measurements for the same quantity of precipitation when travelling through open spaces. The analysis therefore always has to take account of the measurement location. “We have to learn to assess the quality of the data”, says Pätsch. “With artificial intelligence, our system continues learning all the time, even during normal rain”. IAV has great expertise in this area, having gained a wealth of experience with car-2-X and also in using Big Data.
Everyone can contribute to the data pool
The 100 test vehicles should be ready by the end of August 2018, marking the start of an 18- month test phase. Meanwhile the project partners are working at data collection, combination and analysis. One of the outcomes at the end of the project should be an app that makes the evaluations available to interested users. In future, Matthias Pätsch would like to see not just all vehicles contributing to the data pool but also private measuring stations used at home. These would then receive exclusive weather information in return.
In this context, data are seen as a kind of raw material. They are processed, refined and converted into smart data in order to develop their full economic potential. The specific beneficiary of the project is the water industry. Furthermore, business models are being discussed and developed parallel to the research project, both at IAV and also with the other partners.
