by Colton S. ’22
The Northwest Avalanche Center (NWAC) is a program that posts daily avalanche forecasts subdivided by region and elevation during their operational seasons, usually from October to May. Each post contains a forecast for the next day and an outlook for the day after that. These posts utilize a five-level danger rating system: low, moderate, considerable, high, extreme. However, due to limited funding and the demands of a daily operation cycle, NWAC is focused on producing daily forecasts rather than retrospective analysis on past data. My Lakeside Summer Research Institute project is centered around analyzing trends from NWAC’s past forecasts in hopes of making future forecasting more consistent.
Because NWAC forecasts for each day twice, once two days prior and once the day beforehand, the forecast for a given day may change. For example, on Thursday, the forecast for Saturday may state the danger rating of the Olympics above the treeline to be low, while the same forecast on Friday may state it to be moderate. The forecast shifting like this represents a deficiency in consistency, especially multi-step shifts of more than one danger rating. This deficiency in consistency is often symptomatic of forecaster bias, a misunderstanding of the conditions in the region, or other causes of unreliable forecasts. My work on this project centered around trends in these shifts in the forecast referred to as escalations or de-escalations. In order to facilitate analysis of such trends, I automated a process of producing bar graphs to track the escalations and de-escalations by region, elevation, and a third sub-category: avalanche problem type. NWAC forecasts not only the danger rating but also the type of avalanche most likely to strike certain regions. These types of avalanches are referred to as avalanche problems.
To demonstrate, below are two example graphs. These graphs track all of the escalations and de-escalations, with each column representing a different shift (e.g. low to moderate, considerable to high, etc.). From these graphs we can see, for example, that the most common shift is the de-escalation of considerable to moderate. In addition, we can see that multi-step shifts are far more common in escalations rather than de-escalations. My ultimate goal is to make it easy to view and analyze forecast shifts in this way, which would allow NWAC to improve the consistency and reliability of their forecasts.
Figure 1: This graph shows every type of escalation in the dataset.
Figure 2: This graph shows every type of de-escalation in the dataset.