Improvement of the Flood Early Warning System for Valkenburg along the Geul River

The combination of climate change and increased urbanization has resulted in cities with no historic flooding experience suddenly vulnerable to extreme flood events. Climate change increases the frequency and intensity of rainfalls, whereas urbanization decreases the total porous surface area, resulting in pluvial (rainwater) flooding. One such affected city is Valkenburg, located in South Limburg in the Netherlands. Valkenburg flooded on 14 July 2021 after experiencing an unusually heavy rainfall that deposited 146 mm of rain into the Geul Catchment, causing up to €600 million in damage. In such communities, flood early warning systems (FEWS) are emerging as possible non-structural solutions to minimizing costs of damage and loss of life from flooding. These systems are people-centered, end-to-end networks that predict floods before they are meant to occur to warn people living in vulnerable areas so that they can protect their homes, businesses, and themselves. 

The FEWS network for the Geul uses forecasted precipitation data to predict discharge and water level conditions for the Geul. In the event that the predictions result in an abnormally high water level, warnings can be communicated to the necessary parties and to the population to allow for ample preparation. At the time of the July 2021 flood, the system was offline, with experts familiar with the network believing that it would not have worked even if it was online. This project aimed to analyze the existing FEWS from data collection to communication of warnings to locate existing issues and potential sources of weakness, thereby improving the system to effectively warn for future floods. 

Each step of the FEWS was tested to find and strengthen potential weaknesses. The data inputs were analyzed and compared to the recorded precipitation that occurred in July 2021. Then, this data was inputted into the FEWS prediction models to understand how the system would have calculated the discharge and water level for that event. Both the July 2021 flood event as well as four non-flooding scenarios (summer storm, winter storm, dry season, wet season) were tested. The models were then used to create a flood map, and this flood map was inputted into the Damage and Casualties Model (SSM2017) to estimate how much in damage costs could be saved for the case with FEWS and the case without FEWS. Communication and evacuation were not extensively tested in this research project due to these components being determined by social and political frameworks.

When inputting the precipitation data associated with the July 2021 flood, it was found that the 1D model overestimated the water level to be 76.5 m+NAP, 6.5 m greater than the expected water level. Implementing a 2D grid reduced this value to 70 m+NAP, which matched the expected water level. It was also found that both HBV and SOBEK produce simulation results that consistently do not align with recorded data, suggesting a need to recalibrate the models to better reflect the behavior of the Geul River. Analyzing the recorded discharge and precipitation data found that using forecasted precipitation data gives Valkenburg enough time to communicate warnings and evacuate if necessary. Cost-benefit analysis that compared the economic impact of warning versus not warning revealed that warning and evacuation is more cost-effective than not not warning and evacuating, even in the case of a false alarm. An evacuation in the event of a false alarm can cost 1/10 of the difference in damage costs with and without evacuation. However, false alarms must still be avoided, as they erode trust in the warning system, thereby reducing its effectiveness in possible cost and loss of life reduction. Analysis of the expected costs of damage with and without the warning system revealed that the inclusion of the warning system has the potential to reduce the total expected damage by more than 50%. 

The insights found in this project can be used to improve the FEWS for the Geul. Future research can be done to create a 2D or quasi-2D model that can predict the discharge and water levels of the Geul in a timely manner (no more than one-two hours’ simulation time). The 2D aspect is important to a warning system as the expected amount of water affects how the community prepares for the disaster. This project can contribute not only to the improvement of the FEWS for the Geul but also for the improvement or creation of FEWS for other river catchments in newly vulnerable cities.

The final report can be downloaded here.