Statistical model of the Mekong water level
Prediction of flooding with satellite data

When the river Mekong overflows its banks, millions of people are often affected by flooding. Claudia Klüppelberg, Professor of Mathematical Statistics, has now developed a mathematical model together with her PhD student Sven Buhl and in co-operation with the German Geodetic Research Institute at the TUM, with which the water level can be determined - almost at any point.
The Mekong extends with a total length of 4,300 kilometers through six countries in Southeast Asia. The complex river basin is variably affected in different sections by persistent heavy rainfall or drought. With a new statistical model, the scientists can predict the consequences for the water levels.
Statistical model from satellite data
Mekong river basin: triangles and squares: crossing points of the satellite ground-tracer on the river for different altimeter missions, color-coded according to the length of the time series; white dots: spatially very well distributed single measurements from the SAR altimeter satellite; red diamonds: river gauges. Without satellite data, only a sparse database would be available.
The basis for the statistical model is established using data from satellite measurements. So-called altimeter instruments transmit radar waves to the earth. These signals bounce back from the water surface towards the source. "By measuring the time taken by the radar waves to travel back to the satellite, we can calculate the water levels," explains Florian Seitz, Professor of Geodetic Geodynamics at the TUM.
This is only possible in places where the satellite's track crosses a body of water. Therefore, Klüppelberg and Buhl used observation data from two different satellite missions for the statistical analysis:
- Altimetry satellites on repetitive orbits, which usually pass over the same points on a repeating cycle of 10 to 35 days. As a result, water level data are captured for each of these points at regular intervals.
- SAR altimetry satellites measure points on the river only once - but distributed across the whole river system. In addition, the SAR altimeter measurements are more accurate.
Universal Kriging links different data
Water levels along the Mekong on 13 August 2008, determined with Universal Kriging from satellite data. Clearly recognizable is the exceptionally high level along the main river, which led to severe flooding. The cause was the tropical storm Kammuri, which brought extreme rainfall in the northern area of the river.
The scientists combined these different satellite data on the water level of the Mekong using "Universal Kriging", a special statistical method. "Being able to include these additional, highly accurate measurements with a good spatial distribution in our model greatly improved the quality of the results," says Claudia Klüppelberg.
The new statistical model allows conclusions about how high the water is at other points of the river system. This allows the researchers to calculate the water level for almost all points. In this way, they enable a smooth water supply, hydrological analyses and more safety - for example against dangers due to flooding.
In the future, the new method can be applied to all major river systems, as Florian Seitz explains: "The seasonal fluctuations in water levels on the Mekong, the diverse topography, and the regular flooding enabled us to test many different scenarios"
The researchers' results have been published in the Journal of Geodesy (mediaTUM) and in the Journal of Hydrology. More information can be found in the press release of TUM, River levels tracked from space.