L'intelligence artificielle aide les services d'alerte aux avalanches à évaluer la situation
Published: Tuesday, Apr 30th 2024, 10:40
Updated At: Tuesday, Apr 30th 2024, 11:11
Retour au fil d'actualité
Last season, an artificial intelligence system helped to regularly assess the avalanche danger in Switzerland for the first time. After three years of training, the computer's assessments are as good as those of the experts, according to the avalanche warning service.
Sometimes the predictions of the machine-trained algorithm are not good. "But even we are sometimes wrong," avalanche forecaster Frank Techel was quoted as saying in the press release issued by the SLF avalanche research institute on Tuesday. "The good thing is that the models make different mistakes than we do."
When the employees of the SLF avalanche warning service decide which region to assign which danger level, a computer has been providing its own assessment of the situation since this year. The avalanche warning service thus receives a second, independent opinion and can reconsider its current result for the avalanche bulletin in the event of major discrepancies.
Different approaches
The computer algorithm has been trained to assess the danger situation using the same data that avalanche warning systems use. However, the algorithm is based exclusively on simulations of the snowpack. Human avalanche warning systems use additional information such as current observations in the terrain. This data is not available to the computer. However, due to time constraints, humans have to select data, while the computer analyzes the entire volume of data.
SLF researchers have been working on teaching artificial intelligence to assess avalanche danger since 2019. In future, the team wants to better combine human and machine predictions, according to the SLF. This includes, for example, a more intuitive presentation of the results.
©Keystone/SDA
French 
English 
German 
Spanish 
Arabic