PrognoSF

Challenge

The combination of climate change, increasing urbanisation and the ongoing sealing of land is leading to more frequent flooding caused by heavy rainfall. Although early warning systems have long been a topic of research, the accuracy of forecasts remains poor, particularly for local events. PrognoSF tackled this issue by integrating SINFONY precipitation forecasts, provided by the German Weather Service (DWD) in 2022, with local weather data, thereby enhancing the accuracy rate by 30% and extending the warning time.

The aim was to develop an AI forecasting module for heavy rain that provides a local, up-to-date risk classification of flood areas. Another objective was to expand the forecasting module with vulnerability maps to analyse the risk to particularly vulnerable infrastructure.

The newly developed methods were trialled in the test areas of Steinheim and Lemgo. Both cities integrated them into their smart city platforms.

Motivation

Extreme weather events are becoming more frequent as a result of climate change. Soils are particularly unable to absorb as much moisture during heavy rainfall events after prolonged periods of drought, thereby increasing the risk of flooding. This effect is exacerbated further by growing urbanisation and the ongoing sealing of land surfaces. The catastrophic flooding in the Ahr Valley (Rhineland-Palatinate) and North Rhine-Westphalia serves as a frightening reminder of the devastating impact that heavy rainfall and flash floods can have on people's lives.

Solution

Our concept for achieving the objectives is as follows:

1. Use radio-based, energy-autonomous sensor measuring stations to acquire high-resolution data in risk areas. 
2. Fusion of multiple current and archived sensor data, as well as satellite and interpolated precipitation data.
3. Use of the new, innovative, high-resolution ensemble precipitation forecast product (SINFONY) from the DWD for predicting heavy rainfall events. 
4. Improving forecasts using AI and high-resolution sensor data collection.  
5. Realistic risk assessment (impact assessment) using dynamic vulnerability maps that consider short-term changes in soil conditions caused by factors such as soil sealing, drought or frost, and vegetation. Automatic generation and regular updating of vulnerability maps based on satellite data and soil moisture modelling. 
6. Dynamic situation map visualisation with components for evacuation planning. 
7. Forecast results and vulnerability maps are made available to decision-makers via a map-based platform.
8. Evacuation planning is integrated into the map-based platform.
9. Standard protocols (e.g. CAP) are used to create warning messages and ensure interoperability with existing crisis management systems.