In June 2009, a storm over the Adriatic - a so-called cut-off low, persistent for several days - carried large amounts of moisture to central Europe. Massive amounts of rainfall along the north eastern flanks of the Alps caused severe flooding, but also the south eastern Alpine forelands were hit by thunderstorms, recurring for several days. A state of emergency was declared in the district of Feldbach as more than 3000 landslides caused severe damage to buildings, streets, and agricultural land (ORF, Die Presse).
Due to the high impact of the event, the Austrian Institute of Military Geoinformation (IMG) and the Geological Survey of Austria (GBA) systematically recorded many of the actual landslides. A key feature of this inventory was the precise dating of several thousand events. Often landslide data are not well dated and can therefore only be linked to climatological, i.e., time-averaged weather data, but this data set allowed for directly studying the influence of the actually observed rainfall during the meteorological event on the landslide occurrence. Around the time of the event, an airborne, LIDAR-based high resolution digital terrain model was generated for the region and used by JOANNEUM RESEARCH to derive land-use and land-cover (LULC) data.
Just as I started my new position at the Wegener Center for Climate and Global Change, Herwig Proske from JOANNEUM RESEARCH approached me. The Feldbach region is a hotspot of climate change - temperatures, intense thunderstorms and drought events have become distinctly more frequent over the last decades. In 2014 another severe landslide event had happened, and Herwig and his colleague Philip Leopold from the Austrian Institute of Technology wanted to know how climate change affects such events.
An obvious research topic was the attribution question: how much has climate change contributed to the observed event? But for decision makers it might even be more relevant to understand how the landslide hazard could further change in a warmer future climate. We decided to also investigate the relative role of climatic and LULC changes. To study these questions, we joined forces with a team of landslide modelers around Alexander Brenning and Helene Petschko at the University of Jena in Germany. Based on the observed geological, LULC, rainfall predictors and landslide data, their PhD student Raphael Knevels developed a statistical landslide model to predict the local probability of landslide occurrence across the Feldbach region (Knevels et al., 2020). Given the important - but often neglected - role of soil moisture as a preconditioning factor, we amended the original model accordingly. This modification was particularly relevant as the Feldbach region has suffered from severe drought over recent years, and a further reduction of soil moisture in a warming climate could counteract an increase in landslide hazard. Because no high resolution spatial field of soil moisture observations existed for the region, we used hindcast simulations conducted by Armin Schaffer, a master student, for a different project.
In the years before joining Wegener Center I had led a project investigating the influence of Ocean warming on a devastating rainfall event in Russia (Meredith et al., 2015), and around the same time a perspective on simulating "Tales of Future Weather" had been published (Hazeleger et al., 2015). I found this novel approach - now known as event storylines (Shepherd et al., 2018) - ideal for our study of landslide hazard: it assesses how a particular real world event would unfold in a warmer climate. The approach directly links to the episodic memory of affected people and thus aids risk communication; it allows to optimally exploit the precisely dated landslide inventory; it enables very high resolution limited-area climate model simulations that realistically represent local extreme rainfall. But when starting the practical implementation of the approach we faced a series of challenges: what is "the meteorological event"? Is it just the rainfall caused by the cut-off low, or also the preconditioning snow and rainfall over previous months? When simulating the event over a limited area, how to define the boundary conditions in a warmer climate? As simulated rainfall fields are typically at least slightly displaced compared to real world rainfall (Maraun et al., 2017), how to use climate model rainfall to make statements about real world future landslides? How to include soil moisture as predictor into the landslide model when the former varies weakly in space, but the predictive power of the landslide model arises from spatial variations? Emanuele Bevacqua, Heimo Truhetz, Aditya Mishra and me developed solutions for these challenges and refined the simulation setup, and Aditya finally conducted the RCM simulations.
Because the event-storyline approach takes an event - as defined by the large-scale circulation - as given, it has been argued that it implicitly assumes that the characteristics of the atmospheric circulation do not change in a warming climate. To address this criticism, we invited Giuseppe Zappa, an atmospheric dynamicist at the Institute of Atmospheric Sciences and Climate, Bologna, Italy. Together with his undergraduate student Bryony Puxley at the University of Reading, he studied future changes in the occurrence of low pressure systems leading to persistent heavy rainfall events in the region. They found that the number of such storms would only moderately decrease in a warmer climate.
We finally simulated the event under various plausible future climate and land-use conditions. Depending on the changes of rainfall and soil moisture, the area affected by a high risk of landslides during a 2009-type event could grow by 45% at 4°C global warming. This also means that the number of landslides occurring during such an event would increase correspondingly. Only if rainfall intensities remain roughly constant and at the same time the soil dries out strongly, also a slight reduction in the affected area is possible. The growth in the affected area could be reduced to less than 10% by limiting global warming according to the Paris agreement. Active land-use management plays a key role: land-use changes towards a mixed-leaf forest would fully compensate for such a limited increase in hazard. As the current spruce forests are expected to suffer strongly in a warmer climate, such a land-use change would be an important co-benefit of climate change adaptation.
The research, funded by the Austrian Climate Research Programme and the Austrian Science Fund FWF, has been accompanied by an arts project led by Heike Marie Krause. An exhibition at the Natural History Museum Graz presents the scientific findings, photographs, and interviews with people who were affected by landslides in 2009.
Maraun, D., Knevels, R., Mishra, A.N. et al. A severe landslide event in the Alpine foreland under possible future climate and land-use changes. Commun Earth Environ 3, 87 (2022). https://doi.org/10.1038/s43247-022-00408-7