"Open Avalanche Project" which provides forecast utilizing machine learning to alleviate damage caused by avalanches

Once the snow season approaches the end, the more avalanches, once they occur, cause a lot of damage, such as the death of a caught person, damage to buildings and other property. In order to alleviate this damage, a project aiming at high precision and high-precision forecasting by incorporating machine learning of computersOpen Avalanche Project"Is starting.

Home Page - Open Avalanche Project
https://openavalancheproject.org/

This project aims to provide avalanche forecast for a wide range and high precision to alleviate various damage on avalanche, and it is still in the stage of development at the time of article creation. The area covered is the area of ​​the west coast of the United States, and points which are color-coded according to the degree of danger are displayed on the map.


When you move the mouse pointer to each point, the contents are displayed. In this map, four of "5" from "Low" to "High" are shown according to the danger.


The details of color coding are as follows. "There is no risk of both natural occurrence and artificial avalanche"Low, "There is no danger of spontaneous occurrence / possibility of artificial occurrence"Moderate, "There is a possibility of spontaneous occurrence · An artificial occurrence can occur"Consderable, "Spontaneous generation can happen · artificial occurrence can occur very much"HighAlthough it was displayed in the map above, there is also the highest risk of "occurring both naturally occurring and artificially occur without mistake"ExtremeYou will find that there is.


This forecast has been updated once a day and you can also refer to past data. The data on March 6, 2018 looked something like this, almost everything was "Low" ....


The forecast of February 17 of the same year shows that the danger level was "High" in many areas like this.


Scott Chamberlin, who started this project, used to have developed an application "Powder Alarm" that predicts powder snow falls. This application is a forecast service provided by the Soil Protection Agency of the US Department of AgricultureSNOTELThe thing to sense beforehand that new snow falls by utilizing the data of, and set the alarm earlier than usual so that you can get out to the slope more quickly than others.


In developing this application, Mr. Chamberlin met with Mr. Bob who will become a partner later, and is a site of avalanche forecasting site that combines smartphone technology and snow scienceULLR LabsWas launched. And this time, Mr. Chamberlin has launched the Open Avalanche Project which realizes more accurate forecast service by machine learning of past weather data and data recording occurrence of avalanches.


In carrying out machine learning, training has been carried out to increase accuracy by inputting past snowfall data, avalanche forecast data and the like into a computer for learning, predicting it and comparing it with actual avalanche occurrence data And that. Therefore, at the time of article writing, it is not a stage where it can be used as a perfect forecast but also a page of "Experimental - Not For Hazard Evaluation!" (Experimental application - not for disaster evaluation!) Is displayed.


The prediction model is 6% 8