Extreme rain and cloudbursts are becoming increasingly frequent in Denmark. During heavy rain falls, the water level rises rapidly in rivers and watercourses, which can lead to flooding in the fields, destroy crops, cause environmental problems, and threaten biodiversity.
To meet the challenges, DTU researchers are now using artificial intelligence (AI). With an advanced machine learning model trained on water level data from the Danish Environmental Portal’s measuring stations and historical weather data from the Danish Meteorological Institute (DMI), they can accurately predict water levels in watercourses up to two weeks ahead.
In this way, the municipalities will have a strong tool for climate adaptation. This allows for timely action, cost reductions, and protection of vulnerable ecosystems.
“Looking at Denmark as a whole, it’s precisely the watercourses that will lead the rainwater away from the fields. In the cities, water is typically led through underground pipes, but outside the cities, the watercourses are essential. If we don’t manage the task, we’ll experience greater and greater flooding of land areas. This can seriously affect both agriculture and the environment,” says Birger Andersen, Project Manager and Professor at DTU Engineering Technology.
Høje-Taastrup to go first
Initially, the researchers will be developing a data analysis model that can predict the water level in the watercourses in Høje-Taastrup Municipality. The project examines whether the model can be used to optimize the crop cutting–i.e. the cropping of the plants in the watercourses.
Crop cutting is a normal way of maintaining Danish watercourses to ensure that drainage water from the fields flows freely and prevent flooding. But crop cutting also has a negative effect on biodiversity because important habitats providing shelter and food for fish, small animals, and insects disappear.
The goal for Høje-Taastrup Municipality is therefore to develop a model that ensures that the crop cutting taking place neither increases the flood risk of agricultural areas nor dries out the watercourse during periods of less rain. Together with DTU, the municipality has developed a web application that shows the water level forecasts for the municipality’s watercourses, in which a total of three water level sensors are located. The preliminary results look promising, says Magnus Kramshøj from the environmental department of Høje-Taastrup Municipality:
“The water forecasts are far more accurate than we expected. This underlines the value of intelligently exploiting existing data. We are now working to make the model even better in extreme weather situations by integrating weather forecasts so that expected rainfall is included in the calculations. This gives more accurate forecasts—even when the risk of flooding is greatest. This will make it possible to assess the effect of reduced crop cutting and plan emergency measures to avoid flooding. This gives us a strong decision-making tool in terms of nature considerations and risk management.”
Scope of tool to be expanded
With its broad applicability, the online tool has, according to Birger Andersen, the potential to have widespread use—both in Denmark and abroad. He points out that the next step will be to use data for long-term forecasts:
“We need more control of the watercourses and be able to predict water levels several years ahead. As we get more data, we can build models that show inter-year differences and predict long-term water levels. It will be interesting because we can use the forecasts to make decisions: Should we widen the watercourses? Should we establish new watercourses? Or should we change the way we use the land around them?”
