The spatial modeling of extreme snow is important for adequate risk management in Alpine and high altitude countries. A natural approach to such modeling is through the theory of max-stable processes, an infinite-dimensional extension of multivariate extreme value theory. In this paper we describe the application of such processes in modeling the spatial dependence of extreme snow depth in Switzerland, based on data for the winters 1966--2008 at 101 stations. The models we propose rely on a climate transformation that allows us to account for the presence of climate regions and for directional effects, resulting from synoptic weather patterns. Estimation is performed through pairwise likelihood inference and the models are compared using penalized likelihood criteria. The max-stable models provide a much better fit to the joint behavior of the extremes than do independence or full dependence models.Comment: Published in at http://dx.doi.org/10.1214/11-AOAS464 the Annals of Applied Statistics (http://www.imstat.org/aoas/) by the Institute of Mathematical Statistics (http://www.imstat.org

The mountain snow cover is an important source of water but also leads to natural hazards, such as avalanches and floods. We use data collected during winters 1999/2000 to 2007/2008 by 239 automatic and manual measurement stations in Switzerland to highlight spatial characteristics of extreme snowfall. With the help of extreme value theory based on a “peaks‐over‐threshold” approach and a Poisson point process representation, we analyze spatial patterns and correlation characteristics. Our analyses show that a significant number of stations do not follow the Gumbel distribution. In particular, low altitude stations in the Swiss Plateau are heavy tailed because of rare extraordinary snowfall events. Spatial characteristics of extreme snowfall are compared to those of the mean snowfall. Altitudinal dependence and spatial distribution of mean and extreme snowfall are similar. Both mean snowfall and extreme snowfall show an increase of magnitude between 400 and 2200 m a.s.l. and a constant or slightly decreasing magnitude at higher altitudes. Below 1200 m a.s.l., the increase with altitude is stronger because of the rain‐snow transition. Another finding is that the spatial correlation pattern of extreme snowfall is similar to that of mean snowfall, both of which are determined by the main climatological regions of Switzerland. An analysis based on those stations with a long record shows that extreme snowfall was 10% lower in the nine winters investigated than in the long‐term period, but the main spatial characteristics of the two periods show no change.

Mountain snow cover is an important source of water and essential for winter tourism in Alpine countries. However, large amounts of snow can lead to destructive avalanches, floods, traffic interruptions or even the collapse of buildings. We use annual maximum snow depth and snowfall data from 25 stations (between 200 and 2,500 m) collected during the last 80 winters (1930/31 to 2009/2010) to highlight temporal trends of annual maximum snow depth and 3-day snowfall sum. The generalized extreme value (GEV) distribution with time as a covariate is used to assess such trends. It allows us in particular to infer how return levels and return periods have been modified during the last 80 years. All the stations, even the highest one, show a decrease in extreme snow depth, which is mainly significant at low altitudes (below 800 m). A negative trend is also observed for extreme snowfalls at low and high altitudes but the pattern at mid-altitudes (between 800 and 1,500 m) is less clear. The decreasing trend of extreme snow depth and snowfall at low altitudes seems to be mainly caused by a reduction in the magnitude of the extremes rather than the scale (variability) of the extremes. This may be caused by the observed decrease in the snow/rain ratio due to increasing air temperatures. In contrast, the decreasing trend in extreme snow depth above 1,500 m is caused by a reduction in the scale (variability) of the extremes and not by a reduction in the magnitude of the extremes. However, the decreasing trends are significant for only about half of the stations and can only be seen as an indication that climate change may be already impacting extreme snow depth and extreme snowfall.

Abstract. For adequate risk management in mountainous countries, hazard maps for extreme snow events are needed. This requires the computation of spatial estimates of return levels. In this article we use recent developments in extreme value theory and compare two main approaches for mapping snow depth return levels from in situ measurements. The first one is based on the spatial interpolation of pointwise extremal distributions (the so-called Generalized Extreme Value distribution, GEV henceforth) computed at station locations. The second one is new and based on the direct estimation of a spatially smooth GEV distribution with the joint use of all stations. We compare and validate the different approaches for modeling annual maximum snow depth measured at 100 sites in Switzerland during winters 1965-1966 to 2007-2008. The results show a better performance of the smooth GEV distribution fitting, in particular where the station network is sparser. Smooth return level maps can be computed from the fitted model without any further interpolation. Their regional variability can be revealed by removing the altitudinal dependent covariates in the model. We show how return levels and their regional variability are linked to the main climatological patterns of Switzerland.

The disintegration of the ice shelves along the Antarctic Peninsula have spurred much discussion on the various processes leading to their eventual dramatic collapse, but without a consensus on an atmospheric forcing that could connect these processes. Here, using an atmospheric river detection algorithm along with a regional climate model and satellite observations, we show that the most intense atmospheric rivers induce extremes in temperature, surface melt, sea-ice disintegration, or large swells that destabilize the ice shelves with 40% probability. This was observed during the collapses of the Larsen A and B ice shelves during the summers of 1995 and 2002 respectively. Overall, 60% of calving events from 2000–2020 were triggered by atmospheric rivers. The loss of the buttressing effect from these ice shelves leads to further continental ice loss and subsequent sea-level rise. Under future warming projections, the Larsen C ice shelf will be at-risk from the same processes.