A New Approach to Defining Boundaries for Avalanche Canada’s Public Avalanche Forecasts
By Karl Klassen, former Avalanche Warning Services Manager, Avalanche Canada
Note: This article was originally published in The Avalanche Journal, Volume 131, winter 2022-23
IF YOU’VE BEEN USING Avalanche Canada’s public avalanche forecasts or following our social media channels this season, you already know we’ve made some changes to how we display information to our users. Most notably, our old forecast regions have been replaced with flexible regions that change based on conditions. In this article I review how our forecast regions evolved over the decades, discuss the challenges presented by the traditional approach to regions, provide an overview and some background on where we’re at now, and share a few thoughts on where we might go next.
HOW DID WE GET HERE?
When Avalanche Canada was formed in 2004, we inherited a set of forecast regions from the Canadian Avalanche Association, which had previously established a public forecasting program. The historical rationale for where to set region boundaries was largely based on geography, snowpack, and weather patterns. In some cases, such as in the northwest and South Coast Mountains, main ranges were split on a north-south axis, where western areas were the “wet side” and eastern areas were considered “dry.” Some areas encompassed an entire range (e.g., the Cariboos). In other cases, regions cut through mountain ranges east-west, such as the North and South Columbia, which somewhat arbitrarily used Highway 1 as the dividing line between portions of the Selkirk and Monashee Mountains. We did make some changes over the years and as of the 2021-22 season we had 16 regions in British Columbia, Alberta, Yukon, and Newfoundland & Labrador (Image 1).
However, the constraints of the systems used for forecasting limited our capacity to make signfii cant alterations. A lack of funding
and resources also played a part, limiting our changes to relatively minor and incremental revisions. Over the years, there have been two constants. First, our regions have always been relatively large. At the end of 2021-22, region sizes ranged from about 4,000 km2 to just under 52,000 km2, and our total jurisdiction was nearly 366,000 km2.
Second, forecast region boundaries have been fi xed. That is, they were immutable during the forecasting season and, other than an occasional tweak or two between seasons, they remained the same from one year to the next.
CHALLENGES WITH LARGE, FIXED-BOUNDARY REGIONS
There are many challenges associated with large forecast regions. They present considerable diffi culties with data density, terrain familiarity, and spatial and temporal variability. As public forecasters, communicating that variability to our users is a signifi cant challenge. Forecasters can publish only one set of danger ratings, avalanche problems, and discussions per region. Regional variability adds complexity to the messaging, not to mention it all must t into the limited space of the public forecast template. When your average region size is over 22,000 km2, regional variability occurs fairly regularly. When it does, users who base their decisions primarily on danger ratings may form an inaccurate impression of hazard depending on which part of the region they are travelling in. It also means users looking for more detail are forced to carefully parse information to gain an understanding of conditions in their specifi c area of interest.
A secondary challenge in public forecasting arises when there are large areas of uniform conditions and the weather is relatively static. In this case, forecasters are required to create essentially identical forecasts for multiple regions—a waste of resources and capacity that could be better used on other responsibilities.
WHERE ARE WE NOW—AND WHY?
We recognized that giving forecasters the ability to adjust forecast region boundaries to refllect conditions alleviates these issues. To enable responsive forecast regions, senior forecasters were tasked with developing a set of criteria that could be used to divide our forecast area into smaller subregions. To improve our ability to communicate distinct avalanche conditions, the criteria included traditional factors such as snow climate and weather patterns, but we also incorporated more nuanced factors like backcountry user patterns and common access
routes. For example, to effectively communicate risk, we agreed subregion boundaries should not bisect a high-use area or separate
an access route from a destination even though geoclimatic factors might indicate otherwise.
When developing the new subregions in the summer of 2022, the forecasting team used its experience and knowledge along with data from sources like the Mountain Information Network (which helped us determine where use is concentrated) and our snowpack modelling application (which helped us visualize snowpack patterns). This exercise resulted in 92 subregions across B.C., Alberta, Yukon, and Newfoundland & Labrador (Image 2).
These new subregions were incorporated into an updated version of our forecasting software, AvIDfx 1.2, which was developed in collaboration with the Colorado Avalanche Information Center (CAIC) and launched in the fall of 2022. The new forecasting app has been
adopted by the CAIC, Parks Canada, Kananaskis Country, and Avalanche Quebec. While there are many improvements, the most notable
change is a feature that allows forecasters to build dynamic, responsive forecast regions.
Like always, the forecasting team develops its picture of current conditions and incoming weather using various data and tools such as
InfoEx, the Mountain Information Network, our snowpack modelling app, our weather station and avalanche observation visualisation tools (Image 3), and the Mountain Weather Forecast. And, for the first time this season, forecasters are also using a new geospatial weak layer
visualisation system to support their understanding of snowpack conditions (Image 4).
As in the past, this process results in an analysis of what conditions exist where and how conditions are expected to evolve during the forecast period. This analysis is now overlaid on the 92 new subregions instead of the traditional 16 regions. Most importantly, the new software allows forecasters to aggregate subregions into “clusters” that have similar conditions. These aggregated clusters form the forecast regions shown to the public on our map at avalanche.ca (Image 5).
As a result, when large areas are experiencing similar conditions, forecasters can aggregate large numbers of subregions and create fewer, larger forecast regions than have typically been produced at AvCan. When variability is high, forecasters can aggregate fewer subregions per forecast and you might see more, smaller forecast regions than in the past. And, at times, you may see “spot” forecasts where the data indicates an anomaly within a relatively small area, perhaps encompassed by only one or two of the subregions.
In short, public avalanche forecast regions are now dynamic and responsive to evolving avalanche conditions. Because we expect to revise subregions over time as our knowledge increases, and the internal names used by forecasters are subjective and meaningless to most users,
we do not show subregion boundaries or names to users.
Only forecast region boundaries are shown. Because the regions change based on conditions, they are also not named. We’ve removed the traditional icon with elevation band danger ratings from the map and moved to coloured regions that refl ect the highest local danger rating. This, along with the ability to ‘hover’ over a region and see a truncated description of the danger rating, allows users to better visualise the danger where they plan to travel.
To make it easier for users to determine which forecast region is applicable to them, they can use the new search bar. The user enters a place name–for example, a backcountry skiing, snowmobiling, snowshoeing, or climbing destination–and selects their location from a list. The map then centres on that location and the user can open the applicable forecast with a single click.
Various other information can be added, adjusted, or removed from the map using the lter menu; notably, a colour-blind safe version is available. As in the past, popular features like the Mountain Weather Forecast and Mountain Information Network remain directly accessible by using the links shown on the map.
The changes to how our public avalanche forecasts are displayed refl ect common practices in various risk communication contexts and align us with the approach used by most other major public forecasting agencies worldwide. The location tools, features, and techniques included on our map are ubiquitous in the electronic age and bring us to modern standards. Entry-level users will see features like the ones they use when navigating the non-avalanche world (e.g., searching for and fi nding a location on apps like Google Maps). More experienced users who
use advanced mapping tools like Gaia, CalTopo, or FatMaps will feel comfortable with the new features and functions available on the new forecast map.
This new approach has solved one of our biggest challenges, which was variability in conditions across large regions. Now, our forecasters are no longer required to issue forecasts with complex and potentially mixed messages to account for variability, which results in more effective communication of information to our users. The new system provides a forecast that is more responsive to changing backcountry conditions. It also enables more ef ficient forecast production and more effective forecasts, making it simpler and easier for users to plan safe trips into the backcountry.
WHERE DO WE GO FROM HERE?
We see the changes implemented this season as the starting point for an improved suite of products and services. Moving forward, we will be looking into things like:
- customising the search functionality to include even more features, such as common local names;
- a more detailed default base map;
- offering more base map options, such as satellite imagery, shaded topography, or perhaps even 3D maps; and
- updating the forecast structure, format, and iconography, including adding images and video to forecasts.
The new forecasting software and improved public display of forecasts opens the door to an array of longer-term future possibilities, such as:
- new types of avalanche information that address the needs of specifi c locations and/or users: for example, avalanche safety information products designed for ice climbers in the Rockies or snowshoers in the North Shore Mountains;
- computer-generated forecasts;
- integration of terrain into forecast information; and
- vastly improved decision-making support tools.
The AvCan forecasting and software development team has been working for years to create the systems we use. Those systems have allowed us to make the changes you see today. It’s been a long, hard road, and we’re excited about moving into a new era for public avalanche safety.
ABOUT THE AUTHOR
KARL KLASSEN has made innumerable contributions to public avalanche safety and Canada over his lengthy career. He is an internationally certifi ed mountain guide and has held various roles in the North American avalanche community, most recently serving as Avalanche Canada’s Warning Services Manager, where he played a crucial role in developing the forecasting program and creating new and innovative forecasting software. Additionally, he served as both the executive director and the president of the Association of Canadian Mountain Guides.
