ShakeMap-based prediction of earthquake-induced mass movements in Switzerland calibrated on historical observations

Abstract

In Switzerland, nearly all historical Mw ~ 6 earthquakes have induced damaging landslides, rockslides and snow avalanches that, in some cases, also resulted in damage to infrastructure and loss of lives. We describe the customisation to Swiss conditions of a globally calibrated statistical approach originally developed to rapidly assess earthquake-induced landslide likelihoods worldwide. The probability of occurrence of such earthquake-induced effects is modelled through a set of geospatial susceptibility proxies and peak ground acceleration. The predictive model is tuned to capture the observations from past events and optimised for near-real-time estimates based on USGS-style ShakeMaps routinely produced by the Swiss Seismological Service. Our emphasis is on the use of high-resolution geospatial datasets along with additional local information on ground failure susceptibility. Even if calibrated on historic events with moderate magnitudes, the methodology presented in this paper yields sensible results also for low-magnitude recent events. The model is integrated in the Swiss ShakeMap framework. This study has a high practical relevance to many Swiss ShakeMap stakeholders, especially those managing lifeline systems, and to other global users interested in conducting a similar customisation for their region of interest.

Appendix: Mass-movement likelihood scenarios for other historical events for which documentation of induced mass movements exists

For the 1755 event of Brig-Naters (Mw 5.7; Fig. 12), the historic reports examined by Gisler et al. (2004a) describe only a minor earth slump in the immediate vicinity of Brig, while a recent study revealed that a large (~ 1.5 × 10⁶ m³) rockslide destroyed parts of the village of Niedergraechen above the Matter valley (Fritsche et al. 2012). When compared with the 1855 event in the Visp region, the amount of observations of induced effects is low in the epicentral region, particularly in the category of mass movements. Maximum mass-movement likelihoods computed within 500 m of the town of Niedergraechen are equal to 0.43 in Fig. 12.

Fig. 12

Mass-movement likelihoods and known historic observations for the 1755 Mw 5.7 Brig-Naters earthquake

The 11 March 1584 Mw 5.9 mainshock marked the onset of an earthquake sequence near Aigle, in south-western Switzerland, associated with induced effects that had strong effects on the contemporary natural and built environment. Our mass-movement likelihood estimates are shown in Fig. 13. Following Fritsche et al. (2012), we highlight in the picture the well-documented rockfall occurred near Tour d’Aı, accurately captured by the predictive model. The rockfall was triggered by the mainshock and the geomaterials rested on a terrace for ~ 3 days before moving again downslope (most likely due to the severe Mw 5.4 aftershock occurred on March 14) almost completely destroying two villages. The mainshock is also associated with a tsunami and a seiche in Lake Geneva, most likely caused by a mass movement in the Rhone delta. (Kremer et al. 2015; Fritsche et al. 2012; Schwarz-Zanetti et al. 2003). The historic reports describe flooded and damaged watersides in Villeneuve, Lausanne and Geneva and the temporary draining of the Rhone river in Geneva. Waterfront collapses were identified also in Montreux and Chillon. The reader is referred to Schwarz-Zanetti et al. (2018) for further details. Figure 14 shows an application of the mass-movement predictive model using slope values based on a high-resolution (2 m, subsequently downsampled to match the resolution of the Swiss DHM25) bathymetry dataset of Lake Geneva. While the results shown in Fig. 14 should not be overemphasised because the predictive model was not derived/calibrated for underwater mass movements and our knowledge about site amplification and dynamic response of underwater deposits is scarce, we limit ourselves here to note that the areas of heightened mass-movement hazard are mainly located nearby the delta region and high mass-moment likelihoods are also found near Montreux and Chillon, consistent with the historical sources.

Fig. 13

Same as Fig. 12 for the 1584 Mw 5.9 Aigle earthquake

Fig. 14

Example of the application of the predictive model (Eqs. 3, 4 and 5 with b₀ = − 5) using slope values computed based on high-resolution bathymetry of Lake Geneva and PGA_SM of the 1584 Aigle Mw 5.9 earthquake, known to have induced a tsunami and a seiche in the lake, most likely triggered by underwater mass movements

The strongest historical event known in central Switzerland occurred on 18 May 1601 near Unterwalden (Schwarz-Zanetti et al. 2003) with an epicentral intensity of degree VIII and an estimated magnitude Mw 5.9. Although the mainshock induced several rockfalls and landslides throughout central Switzerland, only a few mass movements are described and locatable based on the historic reports. Rockfalls and subaquatic landslides (Monecke et al. 2004) caused high waves in Lake Lucerne. Our mass-movement likelihood estimates are shown in Fig. 15, along with the most likely locations of the mass movements that occurred within the lake according to Schwarz-Zanetti et al. (2003) and Gassner-Stamm et al. (2017).

Fig. 15

Same as Fig. 12 for the 1601 Mw 5.9 Unterwalden earthquake