Predicting landslides better: What a look at hazard maps from the Swabian Alb teaches us

Landslide Öschinger Landhaussiedlung
A combination of geoinformation systems with statistics of past landslides can locally narrow down the probability of future landslides and make forecasts more reliable. This is emphasized by researchers from the TU Freiberg, the KIT and the University of Würzburg in a recent study on hazard maps.

The destructive power of landslides and hillslope debris flows that can be caused by heavy rainfall was clearly demonstrated in the recent flood disasters in several regions of Germany and Europe. Against the backdrop of the debate about assessing predictions, the researchers point out that existing methods for determining vulnerability - especially to landslides - should be improved.

"If we combine existing geodata with landslide events in a specific region, new correlations emerge that make it possible to identify realistic local hazard areas for mass movements after heavy rainfall," says Christoph Butscher, professor of engineering geology at TU Bergakademie Freiberg. To create hazard maps, factors such as topography and soil type are considered statistically. This involves determining how often a landslide occurred within a certain factor class, for example, at 30° slope inclination or with mudstones in the subsoil. The more often a landslide occurred in a factor class, the more hazard points are assigned to the class by a so-called index value. This illustrates the hazard for landslides in graded categories (for example, low, moderate, high, very high hazard) on high spatial resolution maps. Typically, this is represented by a "traffic light" signature - green for low hazard (low index value), through yellow and orange to red for very high hazard (high index value).

Increasing the reliability of hazard maps

Example of a hazard map. For landslides, hazard maps incorporate the relevant geospatial data according to the current state of scientific knowledge and already have a high degree of accuracy. Statistical methods allow the validity of hazard maps to be verified. Nevertheless, according to the research team, the reliability of landslide hazard maps can be further optimized through retrospective evaluation. The authors of the study agree that these findings should be better communicated to authorities and political decision-makers, thus increasing confidence in the hazard maps.

Sustained heavy precipitation in 2013 in the Swabian Alb: Two out of four predictions for landslides were correct

Together with scientists from KIT and the University of Würzburg, engineering geologist Prof. Christoph Butscher retrospectively examined the accuracy with which statistical hazard maps predict landslide events. "Specifically, we compared the predictions of several hazard maps in the Swabian Alb region with the occurrence of landslides after the persistent heavy rainfall of 2013," explains Prof. Christoph Butscher. The event is considered one of the most momentous landslides in Baden-Württemberg. The hazard maps were already created before the landslides.

Four hazard maps were prepared by different authors for the area around Mössingen-Öschingen; the heavy rain triggered a total of five landslides in 2013. "Such an accumulation does not occur too often for a statistical study," Paul Fleuchaus, research associate at the KIT Institute for Applied Geosciences, explains the reason for the study. "So, looking into the past is worthwhile to clarify the validity and reliability of the maps and, if necessary, to improve them with new findings."

Transferring the findings to the prediction of future landslides, the authors of the study come to this conclusion: Only half of the predictions examined were able to correctly locate the hazard. A review of the hazard maps using statistical methods is therefore necessary to validate the accuracy of the predictions. However, in case of doubt, the maps should be additionally verified by field studies, i.e., by investigating the concrete geological conditions.

Original publication: Fleuchaus, P., Blum, P., Wilde, M., Terhorst, B., Butscher, C. (2021): Retrospective evaluation of landslide susceptibility maps and review of validation practice. Environmental Earth Sciences 80, 485. https://doi.org/10.1007/s12665-021-09770-9.

Fragen beantwortet / Contact: 
Prof. Christoph Butscher, christoph.butscher@ifgt.tu-freiberg.de, Phone: +49 3731 39-3163