By Torben Haugaard Jensen, AAU Communication and Public Affairs 

At least 135 dead and several cities destroyed. Such were the consequences when the Ahr river in southwest Germany burst its banks in the summer of 2021.

Several days of rain had turned the normally harmless stream into a deadly tidal wave that crushed houses and bridges in its path.

Climate change means that we must get used to more frequent and intense natural disasters like the flooding of the Ahr. But can we be better at predicting disasters?

Maike Schumacher, Assistant Professor in the Department of Sustainability and Planning, Aalborg University thinks we can. Her research area is geodesy, which is about measuring changes in the Earth's shape, rotation and gravitational field.

With a new grant of DKK seven million from VILLUM FONDEN, she will develop a method that will provide more reliable predictions of floods and droughts by combining satellite data and advanced simulation models using data assimilation and artificial intelligence (AI).

"The goal is to predict natural disasters with greater certainty. This will give policymakers and people in risk areas better opportunities to react before the danger arises," says Maike Schumacher. 

The solution is AI and data assimilation

Over the past few decades, satellites have made it possible to observe changes in the Earth's gravitational field. The observations can be used to study how water resources are distributed across the globe.

There are also hydrological models that can make simulations of how the Earth's water resources develop over time.

But we need a method of combining satellite data and models with mathematical methods and knowledge of hydrological processes in order to improve predictions of the timing and extent of floods and droughts –and assess how reliable the predictions are, says Maike Schumacher.

She will develop this method through a new combination of data assimilation and AI technologies such as machine learning and deep learning.

"We are not sufficiently prepared for when and with what force disasters will hit us. But by integrating machine learning and deep learning into a data assimilation system that combines satellite observations and simulation data, we can obtain new knowledge about how floods and droughts occur. This will give us more reliable predictions so that we can mitigate the consequences in enough time," says Maike Schumacher.

Translated by LeeAnn Iovanni, AAU Communication and Public Affairs.