Predicting extreme snow in the Alps in January 2018

Linus Magnusson
David Richardson


January 2018 saw several episodes of extreme snowfall in the Alps, on both the northern and southern side of the mountain range. In this article we focus on the event that affected the south-western part from 7 to 9 January. Reports on the Web talked about 2 to 3 metres of fresh snow. The ski report for Tignes and Val d'Isère reported 110 to 160 cm of fresh snow in two days. Road links to several villages, such as Bonneval-sur-Arc in France and Zermatt in Switzerland, were cut by avalanches and tourists were stranded in the resorts.

Observed precipitation. Twenty-four-hour precipitation between 8 January 06 UTC and 9 January 06 UTC, according to reports received from regional networks contributing to the high-density observation (HDOBS) project at ECMWF.

Cut-off low. The analysis of 500 hPa geopotential height (contours) and the corresponding anomaly (shading), averaged for 7–9 January, indicate the location of the cut-off low in the western Mediterranean that brought extreme precipitation to parts of the Alps.

The extreme precipitation in the south-western Alps was caused by a cut-off low centred over the western Mediterranean that stayed in a similar position for several days and brought moist air northward on its eastern side. The cut-off low was well predicted and ECMWF’s Extreme Forecast Index (EFI) for total precipitation showed a signal more than a week in advance over the south-western Alps. The median of the ensemble forecast (ENS) starting on 1 January showed precipitation of 40 mm/48 hours in Val d’Isère for 7–8 January, and a risk of up to 100 mm. Between 2 and 3 January, the ensemble forecast became more extreme and the median jumped up to above 80 mm.

For all forecasts issued from 3 January onwards, the high-resolution forecast (HRES) gave higher two-day precipitation for 7–8 January in Val d’Isère than the ensemble median. This suggests a sensitivity to horizontal resolution (about 9 km for HRES and 18 km for ENS), which is expected in steep terrain. From the TIGGE-LAM archive hosted by ECMWF, we have access to eight different limited-area ensembles for evaluation purposes. Comparing the ECMWF global ensemble with the COSMO-LEPS ensemble with 7 km resolution from ARPA-ER SIMC in Italy, we find much higher precipitation accumulations in the limited-area ensemble. This is also true if we compare a short-range forecast from a random ensemble member from COSMO-LEPS with ECMWF HRES.

Accurately predicting snow accumulation in complex terrain is difficult for current global models because of their relatively coarse grid spacing. It is impossible to resolve local differences in precipitation due to individual valleys and mountain peaks and local wind patterns. The lack of resolution for the orography also poses difficulties in predicting the freezing level relative to the terrain and whether the precipitation will fall as rain or snow. For verification, observing precipitation accumulation is very challenging during heavy snowfall. During fast accumulations of snow, rain gauge buckets can easily be filled by snow between being emptied, and automatic stations are known to have large errors during intense snowfall. It is therefore very difficult to judge the performance of the forecasts in terms of total precipitation. What we can say is that the ECMWF forecast gave an early indication of unusually large amounts of precipitation in the Alps. However, with the lower amounts of about 40 cm for Val d’Isère in the earlier forecasts, there could have been an expectation that there would be nice skiing conditions with some fresh snow, while the outcome of probably three times as much or more resulted in closed pistes and tourists being trapped in the resorts.

Extreme Forecast Index (EFI) and Shift of Tails (SOT). EFI (shading) and SOT (contours) for three-day total precipitation from 7 to 9 January in the forecast from 1 January.

Ensemble forecasts for Val d’Isère. Total precipitation for Val d’Isère on 7 and 8 January as predicted by forecasts from different initial times. The box-and-whisker plots show the 1st, 10th, 25th, 75th, 90th and 99th percentile and the black dots show the median.

Short-range HRES and COSMO-LEPS forecasts. HRES precipitation forecast initialised on 8 January at 00 UTC for the period 8 January 06 UTC to 9 January 06 UTC (left) and the same forecast from a member of the COSMO-LEPS ensemble (right).