Header

UZH-Logo

Maintenance Infos

Refined dry-snow avalanche danger ratings in regional avalanche forecasts: Consistent? And better than random?


Techel, Frank; Pielmeier, Christine; Winkler, Kurt (2020). Refined dry-snow avalanche danger ratings in regional avalanche forecasts: Consistent? And better than random? Cold Regions Science and Technology, 180:103162.

Abstract

In public avalanche forecasts, avalanche danger is summarized using a five-level ordinal danger scale. However, in Switzerland - but also in other countries - on about 75% of the forecasting days, only two of the five danger levels are actually used, indicating a lack of refinement in the forecast danger level. A refined classification requires the forecasters to assess the avalanche danger in greater detail than the established danger levels. This leads to the fundamental question, whether a reasonable accuracy and consistency of refined danger ratings can be achieved at all. We address this question relying on a data set from Switzerland, where forecasters of the national avalanche warning service have refined the forecast danger level using three sub-levels (minus, neutral, plus) during four forecasting seasons. These sub-levels, which describe where within a danger level the danger was estimated, were not provided to the public. With the goal to assess whether the forecast sub-levels were better than a random assignment of sub-levels, we compared these forecasts with local nowcast estimates of avalanche danger, for days when two observers reported such an estimate (N= 1146), as ground truth. The agreement between the forecast regional danger level and the local danger level estimate was 81%, with a distinct over-forecast bias in cases when forecast and nowcast disagreed. This tendency towards over-forecasting also showed in a spatial and temporal context. Furthermore, some anomalies in the use of the sub-levels were noted, particularly for sub-level plus in combination with danger level 2-Moderate. Despite these anomalies, the forecast sub-levels were clearly better than a randomly assigned sub-level, resulting in a lower misclassification cost. Furthermore, in case of over-forecasting, the forecast sub-level was in 70% of the cases the sub-level closest to the local estimate, and thus the difference between forecast and nowcast danger level was likely less than one “full” danger level. This indicates that forecasters can often forecast avalanche danger at greater detail than the established danger levels, provided that relevant and reliable data is available in sufficient spatial and temporal density, and that the warning regions, the smallest spatial units used in the forecast are sufficiently small. Therefore, we argue, such refinements of the danger level should be made whenever possible, last but not least for an improved internal assessment of avalanche danger.

Abstract

In public avalanche forecasts, avalanche danger is summarized using a five-level ordinal danger scale. However, in Switzerland - but also in other countries - on about 75% of the forecasting days, only two of the five danger levels are actually used, indicating a lack of refinement in the forecast danger level. A refined classification requires the forecasters to assess the avalanche danger in greater detail than the established danger levels. This leads to the fundamental question, whether a reasonable accuracy and consistency of refined danger ratings can be achieved at all. We address this question relying on a data set from Switzerland, where forecasters of the national avalanche warning service have refined the forecast danger level using three sub-levels (minus, neutral, plus) during four forecasting seasons. These sub-levels, which describe where within a danger level the danger was estimated, were not provided to the public. With the goal to assess whether the forecast sub-levels were better than a random assignment of sub-levels, we compared these forecasts with local nowcast estimates of avalanche danger, for days when two observers reported such an estimate (N= 1146), as ground truth. The agreement between the forecast regional danger level and the local danger level estimate was 81%, with a distinct over-forecast bias in cases when forecast and nowcast disagreed. This tendency towards over-forecasting also showed in a spatial and temporal context. Furthermore, some anomalies in the use of the sub-levels were noted, particularly for sub-level plus in combination with danger level 2-Moderate. Despite these anomalies, the forecast sub-levels were clearly better than a randomly assigned sub-level, resulting in a lower misclassification cost. Furthermore, in case of over-forecasting, the forecast sub-level was in 70% of the cases the sub-level closest to the local estimate, and thus the difference between forecast and nowcast danger level was likely less than one “full” danger level. This indicates that forecasters can often forecast avalanche danger at greater detail than the established danger levels, provided that relevant and reliable data is available in sufficient spatial and temporal density, and that the warning regions, the smallest spatial units used in the forecast are sufficiently small. Therefore, we argue, such refinements of the danger level should be made whenever possible, last but not least for an improved internal assessment of avalanche danger.

Statistics

Citations

Altmetrics

Downloads

2 downloads since deposited on 30 Sep 2020
2 downloads since 12 months
Detailed statistics

Additional indexing

Item Type:Journal Article, refereed, original work
Communities & Collections:07 Faculty of Science > Institute of Geography
Dewey Decimal Classification:910 Geography & travel
Uncontrolled Keywords:Geotechnical Engineering and Engineering Geology, General Earth and Planetary Sciences
Language:English
Date:1 December 2020
Deposited On:30 Sep 2020 10:10
Last Modified:06 Dec 2020 08:47
Publisher:Elsevier
ISSN:0165-232X
OA Status:Closed
Publisher DOI:https://doi.org/10.1016/j.coldregions.2020.103162

Download

Closed Access: Download allowed only for UZH members