THE AVALX PUBLIC AVALANCHE FORECASTING SYSTEM Grant Statham 1 *, Scott Campbell 2, Karl Klassen 3 1 Parks Canada Agency, Banff, AB 2 Parks Canada Agency, Gatineau, QC 3 Canadian Avalanche Centre, Revelstoke, BC ABSTRACT In November 2011, Parks Canada, the Canadian Avalanche Centre and Alberta Parks Kananaskis Country launched a new avalanche bulletin system named AvalX, which provides a standardized forecasting method and bulletin layout among the majority of bulletins in Canada. AvalX marks a departure from traditional text heavy public bulletins to a public interface of graphics and short, focused areas of text. The structure of the message is built around specific avalanche problems, their location in the terrain, how likely they might be triggered, and how big the resulting avalanches could be. AvalX challenges conventional thinking on providing avalanche information, and is strongly influenced by a communication theme that less is more. AvalX bulletins put the principles of public communication on an equal footing with technical analysis, incorporate recently developed work in avalanche hazard assessment (Statham et al., 2010a), and offer a case-study of one path to realizing a unified standard among different agencies and levels of government. This presentation describes the design and output of the AvalX software, and shares lessons learned from effecting change in a public environment. 1. INTRODUCTION In 2010, Parks Canada (PC) set out to build a new public avalanche forecasting system in order to 1) implement the Conceptual Model of Avalanche Hazard (Statham et al., 2010a) into the daily workflow of avalanche forecasters, and 2) switch the predominantly text based avalanche bulletins into a more graphical format, similar to developments in other regions Canada, USA and Europe. The project was a partnership between Parks Canada, the Canadian Avalanche Centre (CAC) and Alberta Parks Kananaskis Country (APKC). PC funded and led the development of the system, which was done in-house with a team of avalanche forecasters and software developers. The CAC and APKC contributed expertise towards the design and implementation of the system, and the collective goal was to implement the same system across all three jurisdictions. * Corresponding author address: Grant Statham, Parks Canada, Box 900, Banff, Alberta, Canada, T1L 1K2; Phone: 403 762 1568; Email: grant.statham@pc.gc.ca Specifically, project objectives were: Implement a structured methodology which walks avalanche forecasters through their workflow; Implement a common look and feel among most avalanche bulletins in Canada; Separate out the technical analysis methods of forecasters from public communication strategies; Structure the bulletin around the concept of avalanche problems, and reduce the volume of text; Focus on four key message areas for each avalanche problem (what the problem is; where it is located in the terrain; how likely it could be triggered; and how large the avalanche might be); Design graphics to communicate the main messages, supported by short text statements; Provide more concise and focused advice on terrain and travel; Enable bulletins to be accessed in HTML and CAAML formats, and on mobile devices; and 165
Enable forecasters to work remotely with an automatic network sync upon return to the office. The system was named AvalX, and was launched in November, 2011. System upgrades occurred during the summer of 2012, and a subsequent launch of AvalX 1.2 is planned for November, 2012. 2. DESIGN METHODOLOGY AvalX was built using the Agile method, a group of software development methods based on iterative and incremental development, where requirements and solutions evolve through collaboration between self-organizing, cross-functional teams. The Agile method promotes constant, ongoing collaboration around frequently updated versions of working software, and relies on the development of detailed mockups for project planning. AvalX began with a significant advantage. The Conceptual Model of Avalanche Hazard (Statham et al., 2010a) provides a detailed workflow in order to structure an avalanche hazard analysis, which is a perfect baseline for software design. Additionally, in 2008 Haegeli developed an online application to encourage avalanche professionals to use these same concepts in the context of their operations. Haegeli s online assessment wizard provided the initial model for a software concept, and AvalX was built upon that. The blueprint for AvalX itself was designed by a collaborative team of avalanche forecasters and software developers. Together this group studied the workflow of public avalanche forecasting and developed a set of usage scenarios, or user stories, that describe the initial requirements of the system. An example of a user story is the following: As a forecaster, I need to record the avalanche problem likelihood and size. I want to be able to see a graphic representation of the problems on an x/y chart, possibly with an overlay of each problem on a single chart. In total, AvalX is built upon 124 user stories, most of which evolved, changed or emerged during the life cycle of the project. Combined with detailed design mockups, these user stories formed both the blueprint for AvalX, and the project deliverables. Once the first working version of the software was available, the iterative cycle of testing and revision was continuous until launch. The project was considered delivered when the complete set of user stories had been fulfilled to the satisfaction of the team. 3. AVALANCHE PROBLEMS In 2004, Rogers Atkins proposed that probability and consequence be assessed separately for different types of avalanches (e.g. wind slab, persistent slab, etc). The premise of Atkins idea was that risk management and terrain choices are based directly upon what type of avalanche presents the greatest risk, and that each avalanche type comes with its own unique strategies for choosing terrain and managing risk. Based on this, the Conceptual Model of Avalanche Hazard (Statham et al., 2010a) was built upon the idea of avalanche types, and definitions were developed for eight different types of avalanche character. These definitions were the basis for the CAC s introduction of Avalanche Problems into Canadian avalanche bulletins (Klassen, 2010). Today, Avalanche Problems, and the information related to them, form the backbone of the avalanche information produced by AvalX. This represents a fundamental shift in the format and focus of public avalanche information, much of which was previously aimed at describing the details of weather, snowpack and avalanches in a less structured, text heavy manner. Structuring avalanche bulletins around specific avalanche problems, or scenarios to be most concerned with, allows those making decisions to focus their attention on the one or two scenarios that present the greatest risk (Jamieson et al., 2010). 166
4. FORECASTING WORKFLOW The AvalX system walks forecasters through a three-stage workflow: Nowcast, Forecast and Review. The Nowcast stage of preparing an avalanche bulletin is based on an analysis of current conditions, and examines field observations at the time of publication, with no future outlook. The Forecast stage examines the weather forecast and its effect on avalanche problems and danger ratings into the future, as well as providing basic terrain and travel advice. The Review stage is the final step, and involves a careful study of the information prior to publication. The following steps describe the workflow specifically, and in order: 4.1 (Step 1) Nowcast Field Observations The forecaster begins their analysis by entering detailed weather observations (T max, T min, HS, HN24, wind speed and direction), a snowpack summary and a summary of avalanche activity. The snowpack and avalanche summaries are text based, and limited to 350 characters per field. 4.2 (Step 2) Nowcast Weak Layer Tracking Next the forecaster identifies weak layers in the snowpack and tracks their characteristics over time. The inputs are: Layer name, burial date, status, grain type, layer thickness and range of depth. The status field classifies a weak layer as Developing, Active, Dormant or Inactive. 4.3 (Step 3) Nowcast Avalanche Problems Next the forecaster identifies and prioritizes the avalanche problems to date, and chooses from the following eight types of avalanche problems: loose dry, loose wet, wind slab, storm slab, persistent slab, deep persistent slab, wet slab and cornice. For each problem, the forecaster specifies the aspect, elevation, sensitivity to triggering and spatial distribution, ahead of plotting the avalanche hazard on an x/y graph (Statham et al., 2010a). As shown in figure 1, plotting avalanche hazard involves specifying a range of values for Likelihood of Triggering on the Y-axis, against a range of values for Potential Avalanche Size on the X-axis (Statham, 2008). The position, shape and size of the resulting ellipse indicates the level of avalanche hazard, and its associated uncertainty. Haegeli (2012) proposes to statistically link these ellipses with avalanche danger ratings. Finally, the forecaster provides supporting text to accompany each avalanche problem, with a limit of 250 characters for any information going public. Figure 1. Forecaster s analysis of an avalanche problem (text not shown) 167
4.4 (Step 4) Nowcast Danger Ratings The forecaster completes the nowcast by assigning an avalanche danger rating for each of the alpine, treeline and below treeline zones. The definitions for each danger level include specific criteria for avalanche likelihood and size (Statham et al., 2010b). Therefore when determining avalanche danger ratings, forecasters match their likelihood and size (avalanche hazard) plots to the corresponding danger rating definition. Haegeli (2012) offers a first quantitative link between these hazard plots, and the resulting danger rating. Accompanying each rating, the forecaster specifies the percentage contribution of each avalanche problem to the overall danger rating. This is another technique for highlighting the most important problems in each elevation band. 4.5 (Step 5) Forecast Weather Following the nowcast, the forecaster next determines the future avalanche conditions that are likely to occur within the temporal scale of the forecast. This process begins with a thorough analysis of weather patterns and their associated impacts. The forecaster then prepares a weather forecast synopsis limited to 350 characters, and enters a range of values for freezing levels, temperatures, precipitation, and wind. 4.6 (Step 6) Forecast Avalanche Problems The avalanche problems that were identified in the nowcast are now brought forward and re-assessed as a forecast. In light of the impacts of future weather, modifications can be made to existing problems or new problems can be created that will correspond with the temporal scale of the forecast. A maximum of five avalanche problems can be recorded, with the top three (or fewer) being published. The forecaster then selects Terrain and Travel advice in relation to each problem. These are oneline statements, each to a maximum of 100 characters, intended to clearly highlight the most important terrain and travel advice. A CAC review of historic bulletins indicated that although key terrain and travel advice was usually provided, it was often hard to discern within the larger body of text (Klassen, per comm.). These shorter, bullet style key messages aim to provide more obvious and easy to find terrain and travel advice. Forecasters have the option of selecting predetermined messages from a searchable menu, modifying this message, or creating their own unique message which becomes part of the searchable database. 4.7 (Step 7) Forecast Danger Ratings In this step, the forecaster determines a three-day forecast of avalanche danger ratings for each of the alpine, treeline and below treeline zones (9 ratings in total). No contribution value is assigned to the forecast danger ratings, but the forecaster selects a confidence rating of good, fair or poor that can be supported with a comment. 4.8 (Step 8) Review and Publish This is the final stage before publication, where the forecaster can view the product as the public will see it, and make modifications where necessary. Additionally, a 250 character Headline can be added, publication timing adjusted and Special Avalanche Warnings enabled. Translation to French occurs immediately upon publication of the English bulletin. 5. PUBLIC COMMUNICATION AvalX is founded upon the principle that while avalanche forecasters need a platform for technical analysis, the public needs access to information that is simple, to the point, and communicates well to a variety of audiences. These two purposes are often at odds. 168
Figure 2. A persistent slab avalanche problem; the public view of the same problem illustrated in figure 1. In 2004, Parks Canada s Avalanche Terrain Exposure Scale (Statham et al., 2006) introduced the concept of avalanche forecasters using different models for technical analysis and public communication. The Avalanche Danger Scale revision (Statham et al., 2010b) employed the same principle, designing a Conceptual Model of Avalanche Hazard for analysis, and linking the model output to the Danger Scale for communication. AvalX continues this trend with a platform that facilitates a technical analysis by forecasters, yet publishes a different output format, designed and tested to maximize comprehension. Specifically, this is the substitution of graphics for words combined with shorter, more focused use of text. Figure 2 shows the AvalX output of an avalanche problem where Persistent Slabs are forecast for all elevations and aspects, and it s likely that humans could trigger large avalanches (between D2-3). The supporting text indicates the layer name, grain type, range of depth and test results. Each of the graphics shown in figure 2 (Elevation, Aspect, Likelihood and Size) was tested independently for comprehension using an online survey in June 2011. 793 respondents were presented with various scenarios and asked to make choices matching textual statements with their graphical equivalent. An average of 1963 responses were received for each graphic, with an average of 97% returning the correct match. 6. DISTRIBUTION Parks Canada makes AvalX bulletins available to the public in two languages (English and French), and in two formats (HTML and CAAML). They are viewable on any modern device from a smartphone, to a tablet or desktop computer, and the bulletins will detect the type of device being used in order to size the screen accordingly. The HTML version is an easy to navigate website (avalanche.pc.gc.ca), presented in a mapping format with danger icons overlain on the forecast regions. After getting a regional overview, users can drill down to reach the bulletin itself, with links to 18 remote weather stations and avalanche terrain ratings. The CAAML link is available below the main bulletin, and provides an XML formatted avalanche bulletin for use by other software developers or systems that wish to use the information contained within AvalX. This facilitates Parks Canada s avalanche information being displayed in other formats or applications. Twitter is used for instant notification, and each time an avalanche bulletin is published, a tweet is sent stating the region, the danger rating and providing a link to the actual bulletin. Parks Canada has licensed the AvalX software to the CAC for no cost, and the CAC has configured 169
the software on their server in such a way that other agencies (such as APKC) can login and use AvalX to produce their own bulletins. This way, all Canadian agencies involved in public forecasting can use AvalX at no cost, and the public benefits from the long sought after unified format. This is a good example of leveraging public funds with the flexibility of a not-for-profit organization. 7. PUBLIC REACTION AvalX was launched on November 9, 2011. Within several weeks, feedback from experienced, local backcountry users began in earnest. As expected, the common theme was the loss of technical discussion within the bulletin, replaced with graphics and short statements. Many users assumed that the graphics were intended for less experienced people, and expressed a degree of outrage upon realizing that the text was no longer available. Interestingly, what became apparent was people s reliance upon avalanche bulletins as their main source of snow/ski conditions as much as avalanche information. When the CAC launched AvalX, they added a second page where they continued to offer more detailed discussion on conditions. Their preservation of a discussion of snowpack, avalanches and weather was lauded by experienced users accustomed to receiving this style of information. For 2012, PC will be making modifications to include the same format of information as the CAC. Feedback from less experienced users and from avalanche course providers was that the new format was much easier to understand, and students very quickly came to focus on the most important problems they should concentrate on a given day. 8. CONCLUSION AvalX, and the underlying code that powers it, consolidates and makes operational a number of important developments that have occurred over the last five years in avalanche forecasting and risk communication. In summary these are: Avalanche Problems as a core structure; Risk based, systematic forecasting methods; Different models for technical analysis and public communication; Graphical representation and less text; Mobile technology for forecasters and public; and CAAML, to enable the easy sharing of data. AvalX bulletins, with their reduced text, prominent graphics and focus on the likelihood and size of specific avalanche problems, represent one of the biggest structural changes to public avalanche bulletins in over 30 years. It was no surprise that experienced users reacted instantly to this change, without fully taking the time to understand the new structure. What also became apparent was people s pride in their ability to interpret technical information regardless of their ability to incorporate it into their decision making. Many people are simply interested in the science of the snowpack, and after 30 years of teaching people to study these complexities, it s natural that this is what people have come to know and expect. However, information on avalanche conditions does not need to be overly complex and difficult to interpret. Public information on risk should be simple, and get right to the point. While the complexities of backcountry travel and predicting avalanches are indeed real, they should lie in the mountains, choosing terrain and managing risk rather than in reading about it beforehand. In the famous words of Leonardo Da Vinci, Simplicity is the ultimate sophistication. 9. ACKNOWLDEGEMENTS AvalX was a Parks Canada led project, and the authors extend their sincere thanks to the other members of the working group, all of whom made substantial contributions: Joel Beaulieu, Max Darrah, Chris Gladwin, Jeff Goodrich, Chris Healy, 170
Steve Holeczi, Jake Hunt, Rob Sherwood and Brad White from Parks Canada; Ilya Storm and Kristina Welch from the CAC; Mike Koppang from Alberta Parks Kananaskis Country; Pascal Haegeli from Avisualanche Consulting; Jenn Becker from Hemlock Hills; and Gilles Chateau and Gavan Howe from ebranders. 10. REFERENCES Atkins, R. 2004. An avalanche characterization checklist for backcountry travel decisions. Proceedings of the 2004 International Snow Science Workshop in Jackson Hole, Wyoming, USA, 462-468. Jamieson, B., Schweizer, J., Statham, G., Haegeli, P. 2010. Which obs for which avalanche type? Proceedings of the 2010 International Snow Science Workshop in Squaw Valley, California, USA. Klassen, K., 2010. The Avalanche Hazard Assessment Web-tool A structured approach to public avalanche forecasting. Presentation at the spring conference of the Canadian Avalanche Association, 6-7 May, 2010. Haegeli, P. 2008. Avalanche hazard assessment website [available online at http://avalanchehazard.avisulanche.ca Haegeli, P., Falk, M., Klassen, K. 2012. Linking avalanche problems to avalanche danger a first statistical examination of the conceptual model of avalanche hazard. Proceedings of the 2012 International Snow Science Workshop in Anchorage, Alaska, USA. Statham, G., 2008. Avalanche hazard, danger and risk a practical explanation. Proceedings of the 2008 International Snow Science Workshop in Whistler, British Columbia, Canada. Statham, G., Haegeli, P., Birkeland, K., Greene, E., Israelson, C., Tremper, B., Stethem, C., McMahon, B., White, B., Kelly, J. 2010a. A conceptual model of avalanche hazard. Proceedings of the 2010 International Snow Science Workshop in Squaw Valley, California, USA. Statham, G., Haegeli, P., Birkeland, K., Greene, E., Israelson, C., Tremper, B., Stethem, C., McMahon, B., White, B., Kelly, J. 2010b. The North American public avalanche danger scale. Proceedings of the 2010 International Snow Science Workshop in Squaw Valley, California, USA. Statham, G., McMahon, B., Tomm, I. 2006. The Avalanche Terrain Exposure Scale. Proceedings of the 2006 International Snow Science Workshop in Telluride, Colorado, USA 171