Célia Lucas

Recent advances in automated snowpack modelling notwithstanding, public avalanche forecasting relies heavily on human estimates of avalanche conditions. Like any human judgments, these estimates differ to some extent from one forecaster to the next. We assess inter-rater reliability (IRR) — the proportion of pairwise agreement among forecasters — both for estimates of factors determining avalanche danger and for the danger level (D) itself. Comparing forecast drafts prepared independently by forecasters for the same forecast task, both in Switzerland (SWI) and in Norway (NOR), we find that IRR is lower for the factors determining avalanche danger (SWI: 0.6 − 0.66, NOR: 0.46 − 0.75) than for the danger level (SWI: 0.87, NOR: 0.76). The EAWS Matrix is a tool for estimating D that is supposed to secure IRR. It is not a part of the daily workflow in Switzerland, and one might wonder whether introducing it there would help. We find that the IRR for D would on the contrary have been much lower were the primary D chosen in the Matrix (a mere 0.54)! Where IRR is low, individual forecasts must often be incorrect, since two different forecasts cannot both be right. We calculate on the basis of the observed IRR that when estimating the factors in avalanche danger the average forecaster is incorrect in 32% (NOR: frequency) and 13% (NOR: avalanche size) of cases. Forecasting D without the Matrix, a forecaster is incorrect 7% of the time (in SWI). With the Matrix, a forecaster is incorrect about D in 13% (NOR) and 27% of cases (SWI). We conclude that while the EAWS workflow does give structure to the forecasting process, using the Matrix can be counterproductive. It can inadvertently compound variability in parameter inputs. The IRR of Swiss D forecasts leaves some room for improvement, but introducing the Matrix into the Swiss workflow cannot be expected to increase it while the IRR of Swiss factor forecasts remains as low as it currently is. We recommend that avalanche warning services assess the IRR of forecasters and consider the effects of low input IRR on the IRR of their Matrix outputs. Moreover, they should consider the effect of unreliability on forecast correctness. Where IRR is low, one intervention to boost forecast correctness is training forecasters with a view to increasing IRR. Another is to aggregate several individual forecasts for the same task, as is standard practice in Switzerland.

In public avalanche forecasts, avalanche danger is commonly summarized using one of five ordinal danger levels. This strong simplification of a complex phenomenon inevitably leads to a loss of information. To compensate for this information loss and to ease interpretation of the forecast, the Swiss avalanche warning service introduced sublevels, that indicate whether the forecast danger is in the lower (-), middle (=) or higher (+) range of the danger level. After a 6-year test period, during which it showed that forecasters can reliably assign sublevels to a danger level, the sublevels were published for the first time in the public avalanche forecast in the winter season of 2022/2023. To evaluate the general understanding and the usefulness of the sublevels, and to understand how they influence people’s backcountry behavior, we performed a user survey towards the end of the season. The survey with 3403 participants showed that participants perceived the sublevels as generally useful (85%) and that they understood them in the intended manner (94%). Sublevels also had a positive influence on the planning of backcountry tours. Faced with a forecast sublevel (+), almost half of the participants claimed they would change their plans to a less exposed tour than they had before when the same danger level was given but without sublevel information. This self-stated influence of the sublevels on the choice of a backcountry tour was especially important for participants with little or no avalanche training, showing that sublevels seem most useful for less experienced forecast users, who may find it difficult to extract similar information from the forecast text description.

When riding out of bounds or touring in the backcountry, many recreationists and professionals plan their trip with the aim of finding a tour at a personally acceptable level of risk. Particularly during the planning phase, information from the public avalanche forecast is considered. In Switzerland, the avalanche forecast summarizes avalanche conditions with a danger level (DL) and the aspects and elevations (AE) where the DL applies. Further, it informs about the relevant avalanche problems, including a text description. Previous studies have shown that most of this information is related to avalanche risk. However, it is poorly understood which information users consider helpful for trip planning and whether these (often) self-stated claims relate to what is remembered at the beginning of a trip. We therefore conducted two independent surveys. The online survey, with 3400 participants, showed that all information provided in the forecast is perceived as valuable for planning. The DL was considered the most important, followed by the avalanche problem(s). In comparison, the AE were considered less important. This ranking of the users is not in line with recent findings regarding avalanche risk. For the second survey, we went outdoors, either in a popular area for out of bounds skiing or at trail heads of ski tours and asked the users directly. Compared to the responses from the online survey, a different picture emerged: While the DL was well-remembered (90 % correct), only around half of the participants (44 %) could correctly name the avalanche problems. In an application exercise, a clear minority (24 %) could say whether slopes of a particular aspect and elevation were assessed as critical in the forecast. This apparent discrepancy between the self-stated use of forecast information during trip planning and the retention of the information at the start of a trip comes partly as a surprise. Especially the avalanche problems, intended to emphasize and focus on the cause of instability resulting in impact-oriented travel advice, are remembered poorly. The findings are supported by other recent surveys in Europe and show that the information from the forecast is not effectively retained, even at the start of a trip. It highlights the need to rethink and improve the communication of avalanche danger, including the level of detail. While providing detailed information may be helpful for experts, it may be overwhelming for less experienced users. Our study aims to contribute to the ongoing discussion on how to best communicate avalanche danger to reduce risks for recreational skiers.