Content uploaded by Laura M.D. Maguire
Author content
All content in this area was uploaded by Laura M.D. Maguire on Oct 19, 2018
Content may be subject to copyright.
SENSEMAKING IN THE SNOW: EXAMINING THE COGNITIVE WORK OF AVA-
LANCHE FORECASTING IN A CANADIAN SKI OPERATION.
Laura Maguire*, Jesse Percival2
1 The Ohio State University, Columbus, OH, USA
2 Mt. Washington Alpine Resort, Courtenay, BC, Canada
ABSTRACT: The cognitive work of making sense of risk in avalanche forecasting is an under ex-
plored area in the field. This study examines the formal descriptions of how work is conducted in a
Canadian ski operation and the ‘as practiced' cognitive strategies employed by expert practitioners to
successfully manage avalanche hazards in practice within a complex and changing mountain environ-
ment. The three key findings were: 1) Much of the cognitive work required for forecasting is hidden in
the explicit protocols; 2) The cognitive effort needed to manage avalanche risk is a near continuous
activity in season and; 3) It is an inherently distributed cognitive task across both individuals, teams and
the broader industry. These findings have important implications for the design of work systems and
management of professional avalanche forecasting activities in ski resort settings.
KEYWORDS: cognition, human factors, forecasting, expertise, work;
1.0 INTRODUCTION
This study explores how a forecasting team at a
sub-alpine coastal ski resort uses a variety of
tools, protocols, expert practices and distributed
cognitive capacities to manage the avalanche
safety program. This paper describes the opera-
tional aspects of managing snow safety within the
resort and analyze the artifacts that support the
scope of practice. It then examines three promi-
nent themes of successful forecasting activity that
emerged from the research: 1) Existing protocols
hide much of the essential cognitive effort re-
quired to perform these tasks; 2) It is a near con-
tinuous activity of updating mental models and; 3)
It is a distributed cognitive activity. Finally, we dis-
cuss what these insights mean for understanding
and supporting the cognitive work inherent in av-
alanche forecasting.
Since the late 1800’s the body of knowledge sur-
rounding avalanche forecasting and techniques
for successful practice has grown (Perla & Marti-
nelli, 1976; LaChapellle, 1980; Schweitzer &
Fohn,1996; McClung, 2002a/b; White, 2002; Ad-
ams, 2005; Holler, 2012). However, how experts
put that knowledge into practice and what consti-
tutes the conditions of practice has been less
thoroughly explored. The purpose of this study is
to explore expert judgement in the real world.
Hutchins (1995) describes this as operating in the
‘natural laboratory’.
2.0 METHODS
Branlat et al (2009) point out that “direct observa-
tion is one of the most used and valuable meth-
odological approaches in the analysis of opera-
tors’ activity; however, for both practical and the-
oretical reasons, many environments present im-
portant challenges to the actual implementation
of this family of techniques (Guérin et al., 2006;
Crandall, Klein and Hoffman, 2006)” (p.15).
With this in mind, ethnographic methods were
chosen to provide the tools to explore the “messy
details” (Nemeth et al, 2016) of real work while
minimizing the risk of oversimplification. The
study used mixed methods including artifact anal-
ysis (Hutchins, 1995) and semi-structured inter-
view techniques (McIntosh & Morse, 2015).
Artifact analysis was conducted to develop an un-
derstanding of how the tools used for avalanche
forecasting in a ski operation – technological and
analog- both reflect and shape the conditions of
work. Rigorous review of established procedures
and industry guidelines helped form hypothesis
about the constraints in the domain and the de-
gree of interaction between roles.
The initial part of the interview used protocol anal-
ysis as a knowledge elicitation technique focusing
on how the forecaster on duty role completed the
Avalanche Forecasting Worksheet. A semi-struc-
tured approach was taken to explore individual
performance variation. McIntosh & Morse (2015)
note its utility “when there is sufficient objective
knowledge about an experience or phenomenon,
but the subjective knowledge is lacking”(pg 1).
Finally, cases were elicited where forecasters ex-
perienced surprise or an unexpected result during
their duty. Results from these cases were used
* Corresponding author address:
Laura MD Maguire, The Ohio State University
Cognitive Systems Engineering Lab
Columbus, OH 43202;
email: maguire.81@osu.edu
to provide converging evidence of cognitive work
in practice.
3.0 FINDINGS
The resort’s highest peak is at 1588m with the
lowest lift starting at 1083m with avalanche terrain
at tree line and below. The majority of avalanche
terrain is on aspects ranging from NE to NW with
a very small portion facing SE to SW. Most ava-
lanche paths have multiple start zones with few
paths and a vertical drop of 100-200m. The North
and East aspects face the Strait of Georgia with
SW/SE prevailing winds, so the majority of load-
ing patterns are typically on NW aspects. Located
in a Maritime climate, the predominant challenges
are freezing levels and new snow instabilities. Alt-
hough being in the rain shadow of Strathcona
Park the resort receives an average of 2200mm
of precipitation per year.
Due to the maritime climate and deep average
deep snow depth, the control consists of mainly
new snow instabilities. Avalanche control con-
sists of ski cutting, cornice cutting, hand charging
and the very infrequent use of an avalauncher.
The winter of 2007-08 was an anomalous low
snow winter with a deep basal instability. A boot
and ski compaction program was implemented to
successfully remedy the problem and is still in-
cluded in the Avalanche Safety Plan (ASP) today.
Avalanche access control defines areas outlined
both visually and with a written description. The
zonal map identifies and communicates the ava-
lanche danger and risk to all members of resort
staff that may be exposed within the operational
work areas, such as lifts or grooming areas. The
map is also used to indicate closures and is up-
dated prior to shifts beginning and at days end.
The mapping is used to communicate avalanche
risk information during meetings so teams can
plan any required deviation from certain tasks or
terrain until the risk level has been reduced and
the terrain is re-opened.
The snow safety team is a part of the 30-person
ski patrol and includes 3 Canadian Avalanche As-
sociation (CAA) Professional members with Level
2 certifications who rotate through the Forecaster
on Duty (FOD) role, 6 active members with Level
1 certification (3 members currently injured or
pregnant and unable to perform avalanche work)
and 2-4 technicians in training. Two of the snow
safety team retain year round employment with
the resort. Turnover averages 20-25% annually.
One Level 2 member is also the mountain’s Patrol
Director/Director of Avalanche program. All ava-
lanche control workers have a combination of pro-
vincially regulated tickets and in-house training
for handling charges. The Director applies annu-
ally for a nationally regulated license for explosive
storage through the Natural Resource Canada
explosive regulatory division.
3.1 Start Up operations
While recruitment is a continual process of identi-
fying high potential external and internal (patrol
team) candidates and developing internal snow
safety team members, hiring is generally finalized
August-October for the upcoming season. Early
season (before the resort opens) consists of train-
ing, scheduling, renewing regulatory require-
ments for magazine licensing, ordering and re-
ceiving explosives, cleaning and organizing gear
and facilities, conducting pre-season checklists
as per the Avalanche Safety Plan (ASP) and, as
the snowpack develops, marking hazards, in-
stalling protective equipment and performing
manual compaction activity.
3.2 Daily operations
The mountain is located approximately 30
minutes from the closest town and the majority of
workers commute daily. As forecasting activities
take place earlier than most other operations on
the mountain, the forecaster on duty will drive up
often carpooling with the FOD from the previous
day. As per procedures, the FOD begins at
05h30 with standard weather observations at the
base weather station, then updates public daily
reports (snowphone and website) and begins col-
lecting data needed to formulate a control plan as
specified by the ASP.
The origin of the control work in the mid-1980’s
when worker injuries from ski cutting initiated the
need for explosives control. In 2005, additional
terrain expansion at the resort exposed guests
and workers to overhead hazards and increased
avalanche risk. In 2007, the ASP was formalized
into a newly formed snow safety program which
specified qualifications for forecasters, require-
ments for data gathering and hazard evaluation
record keeping and participation in the industry
information sharing program. On-going changes
to the plan included qualifications for avalanche
specialists, new and revised explosive usage pro-
tocols, training, safe work procedures, communi-
cation flows across the team and mountain oper-
ations, known avalanche terrain grouped into
zones and provides templates for the systematic
collection of data and assessment of risk. Chang-
ing regulatory requirements for occupational
health & safety and explosives handling and stor-
age, as well as updates to industry guidelines are
integrated to maintain currency with legal require-
ments and best practice.
3.3 Artifacts of forecasting work
The Avalanche Forecast Worksheet (AFW) is a key
component of daily operational safety manage-
ment. In preparation for completing the worksheet
the forecaster consults Avalanche Forecasting work-
sheets and other internal records from the preceding
days - including Weather Observation sheets, Snow
Profile worksheets and shot sheets prepared by
other members of the team. After reviewing internal
records, the forecaster consults a wide variety of
weather tools: the mountain’s weather telemetry
feed, North American Mesoscale forecast system
(NAM); Global Forecast System (GFS);Global Envi-
ronmental Multiscale - Local Assessment Model
(GEMLAM); Mountain Weather Forecast and Coastal
hydrology forecasts. The weather tools are used to
“stick your nose in there and see how much snow is
on the dancefloor” as one participant puts it, to as-
sess current status of the weather and produce the
“nowcast” and, importantly, expected changes to var-
ious meteorological factors that influence snow sta-
bility to produce the day’s forecast and “futurecast”
for the evening and the following day’s activity.
Additionally, avalanche specific resources such as
the InfoxEx - the industry’s subscription based ser-
vice for technical snow, weather, avalanche and ter-
rain information - and daily blogs and bulletins issued
by weather experts and avalanche forecasters in the
region are also referenced for additional input. InfoEx
submissions are done daily throughout the season by
qualified avalanche professionals in Western Can-
ada across a variety of commercial operations and
public and private transportation and infrastructure
entities.
3.3 Cognitive work and coordination
Using these resources, the forecaster shapes an as-
sessment of the current avalanche risk and antici-
pates how the hazard may shift throughout the
day. This assessment is captured in the Avalanche
Forecast worksheet and forms the basis for the con-
trol plan which informs the coordination needed be-
tween the team leaders and avalanche technologists.
The FOD communicates the closures with other
mountain operations departments, then this infor-
mation is scribed on a map and utilized for their morn-
ing and evening safety meetings. During travel to the
upper part of the mountain (either by lift or snowmo-
bile), the FOD is continually gathering cues to inform
their assessment. The upper mountain weather sta-
tion is assessed and revisions to the plan are made
in real time as the team prepares the charges in the
top hut. The teams ski the pre-defined routes, per-
forming control measures (manual or explosive) and
reporting back the results. Repeated cycling through
the terrain throughout the day by the FOD’s and the
technician team helps maintain an accurate repre-
sentation of the changing risks. In recounting recent
critical incidents, the study participants described
events where the nature of the incident, high work-
load or staff shortages limited their capacity to be
moving around the mountain “feeling the snow under
[their] ski’s”.
4.0 DISCUSSION
Three predominant themes emerged from the analy-
sis of successful forecasting.
1. Much of the cognitive work required for forecast-
ing is hidden in the explicit protocols.
The formal representations of what constitutes “good
practice” in forecasting is a small fraction of the actual
strategies experts employ to conduct their work.
2. The cognitive effort required to manage ava-
lanche risk is a near continuous activity in sea-
son.
The nature of the cognitive work in forecasting ap-
pears to require on-going calibration. Disruptions to
this calibration process have adverse effects on per-
formance.
3. It is an inherently distributed cognitive task
across both individuals, teams and the broader
industry.
Successful forecasting requires a team of distributed
practitioners. It’s largely apparent how ‘local’ team
members within the ski resort operation are needed
but, the resources and insights produced by others
within the industry can also be considered a neces-
sary part of the cognitive network required for fore-
casting.
These themes are highly interconnected and best de-
scribed through examples from the findings.
4.1 Preparations for forecasting
Formally, a forecaster’s day begins at 530am – the
protocols as written suggest a worker could arrive on-
site and produce a control plan- but each forecaster
interviewed detailed extensive preparations not for-
mally captured. One forecaster describes installing
a rain gauge on their deck at sea level so they can
begin formulating an expectation of accumulation fur-
ther up on the mountain before their first cup of cof-
fee. Similarly, interviewees described hearing winds
picking up, experiencing temperature fluctuations or
noting lower than expected skier volumes that pro-
voked thoughts of “what do I imagine is happening
up there?” and shaping expectations. In addition,
most reported they reviewed the online resources or
called the FOD for an update on the evening prior to
returning to work to get a sense of how things
changed – particularly when new storm instabilities
or temperature fluctuations may have occurred. The
team often carpools to the mountain and spends the
half hour discussing what activity, if any, was noted,
control measures used, volume of skier traffic in the
avalanche terrain and. Visual cues such as snow
depth and consistency, snow loading on trees and
wind loading on the road during the drive up are also
informative. This suggests formulating the day’s
forecast begins well in advance of the written proce-
dures and that a forecaster arrives for duty with a hy-
pothesis of how recent and overnight weather is go-
ing to affect their avalanche terrain and the additional
activities are a critical component of successful “on
duty” work.
This example succinctly reinforces all three points. It
is evidence of “off books” activity that is a common
practice (and likely necessary) practice not explicitly
noted in formal work protocols and demonstrates a
need for on-going calibration of how conditions are
changing. It is especially interesting that forecasters
sought to do this during periods where uncertainty
and complexity increased. It is well documented that
forecasting takes place under time pressure and by
seeking out data that can help them anticipate condi-
tions in advance, the FOD relieves themselves of
some of this pressure by building slack into the sys-
tem to lessen the cognitive demands required once
they ‘officially’ clock in.
4.2 Disruption, adaptation and surprise
A second example is drawn from the critical incidents
described by study participants. An FOD recalled an
incident in which they were surprised by an unex-
pected in-bounds release. The plan for the day an-
ticipated potential instabilities, control work had gone
as intended and the usual expectation that continual
monitoring of temperature changes would allow for
closures to be made should the risk change materi-
ally during operations. However, during the day one
of the snow safety team members had a personal
emergency and left early which meant the team was
running a person short. Concurrently, a first aid
emergency tied up members who might otherwise be
skiing the avalanche terrain to monitor changes. This
left the FOD ‘in the bump’ on the backside of the
mountain for longer than the usual rotation. His nor-
mal practice of rotating out to be in the terrain was
interrupted. The temperature fluctuated and a skier
triggered release occurred in one of the avalanche
zones. This example is informative in two ways. The
first is that it is reflective of what “normal work” is –
constantly adjusting to workload demands, to ex-
pected and actual availability of resources, adapting
practices to respond to conditions and balancing
tradeoffs that are an inevitable part of technical work.
This variability is an expected part of forecasting, the
OGRS note that “the type of operation and availability
of observers might necessitate different frequencies
and times” (pg. 15) for observations. This hesitation
to be wholly prescriptive noted here and discussed
earlier means there is a substantial ‘grey area’ of
forecasting work that is left to professional discretion
and worthy of investigation into how avalanche pro-
fessionals bring their knowledge to bear in context.
4.3 Mental models
The second way this example is informative is it pro-
vides supporting evidence that practitioners in the
field construct mental models (Adams, 2005) and
continually work to update them. “Mental models are
the mechanisms whereby - humans are able to gen-
erate descriptions of system purpose and form, ex-
planations of system functioning and observed sys-
tem states, and predictions of future system states”
(Rouse & Morris,1985, p. 351). It is used for the re-
trieval of technical knowledge and the ability to flexi-
bly apply that knowledge to the situations in which a
practitioner faces.
Because conditions are constantly changing, mental
models will become dated and stale unless continu-
ally updated. In this example, the model became in-
sufficient after only a few hours of being removed. In
the previous example, when returning to work after
days off the forecaster is aware their model is stale
and seeks information to recalibrate. LaChapelle
(1980) describes a “...prevalent and strong reluc-
tance of working forecasters to experience an inter-
ruption in their winter routine…” (pg. 78) and stressed
the role of redundancy and iteration that supports this
interpretation.
4.4 Distributed cognitive efforts
Notable as well, is the role of a distributed network in
maintaining that mental model – a diverse range of
perspectives informed by different experiences,
knowledge sets and mindsets is needed for any one
practitioner to remain on top of changing conditions.
The schedule for FOD’s is designed to provide an
overlap day to accommodate the need for distributed
cognition. This is an explicit recognition of both the
importance of ensuring currency of the mental model
and of the interactions between practitioners. The
verbal exchanges allow for knowledge transfer, an
opportunity to draw attention to particular details and
shared insights.
Spatial and temporal constraints also drive the need
for distributed cognitive efforts. Once a forecast has
been developed it needs to be continuously cali-
brated in the field. Large terrain to cover coupled with
constraints on moving avalanche technicians from
the bottom of a slope to the top before lifts open can
introduce spatial challenges. The dark winter season,
where daylight comes late, coupled with a regulatory
need for the visibility afforded by dawn to be able to
see results from deploying explosives or ski cutting
creates a temporal tension as the team strives to
complete control work before the hill opens at 9am.
In this manner, the Forecaster On Duty relies on the
team of technicians to ski the routes efficiently, gath-
ering as much real-time data as possible and relaying
that back alongside an assessment. Without ‘local
knowledge’ from each of the identified avalanche ter-
rain zones, the FOD’s mental model is guaranteed to
be only a partial representation of the conditions.
This same need is on-going as technicians repeat-
edly revisit the zones throughout the day, reporting
back to the FOD to aid in maintaining the currency of
their mental model of the avalanche risk. As noted in
the earlier example, disruption to this continual cali-
bration (either through direct sampling or integrating
and interpreting feedback from others) resulted in a
“situational surprise” (Wears & Webb, 2011) where
the FOD was caught off guard by the changing con-
ditions. Some interpretations would simply label this
as ‘human error’ but this example provides evidence
that the cognitive work of forecasting requires contin-
ual feedback from the environment and disrupting
this important flow of information can result in com-
promised expert judgement.
5.0 CONCLUSION
There were 3 significant findings from this study of a
coastal Canadian ski operation. 1) much of the cog-
nitive work required for forecasting is hidden in the
explicit protocols; 2) It is a near continuous activity in
season; 3) It is an inherently distributed cognitive
task across both individuals, teams and the broader
industry. These results broaden the discussion
about the sources of error in avalanche forecasting
by providing a more nuanced understanding of how
forecasters use a variety of strategies and methods
to assess the risks in ski operations. Connecting
these expert techniques to the system of work sur-
rounding them (such as team size, qualifications and
distribution and technological resources) generates
interesting insights into their role and utility in cogni-
tive work. Future research into the variances in cog-
nitive strategies across different professional settings
(mechanized skiing, transportation and industrial set-
tings, backcountry guiding) will likely be valuable to
provide context specific data that could be used to
minimize accidents and increase organizational resil-
ience in the face of an incident. In addition, compar-
ing and contrasting expert vs recreational cognitive
strategies is likely to produce findings beneficial to
refining public safety prevention efforts.
Cognitive systems engineering offers new tools and
insights to the avalanche and mountain safety com-
munities. Making explicit the cognitive work, formally
recognizing the ways in which continual updating oc-
curs and highlighting the interactions between the
distributed cognitive capacities is important for the
domain as these tacit strategies can aid in: develop-
ing training and accelerating learning in new fore-
casters and guides; protecting or generating funding
for critical resources within the distributed network
such as bulletins, blogs, forecasting sites; developing
new forms of coordination between local teams and
regional networks; and enhancing existing or devel-
oping new forms of decision support tools to aid in
forecasting efforts.
ACKNOWLEDGEMENT
Laura would like to thank her co-author and the fore-
casting team for their enthusiasm and thoughtful re-
flections and EAB for comments.
Jesse would like to thank the Canadian Avalanche
Association, Mount Washington resort, Allan Dennis,
Kevin Fogolin, and Jason Chrysafidis for the contin-
ual support and mentorship.
REFERENCES
Adams, L. (2005). A systems approach to human factors and ex-
pert decision-making within Canadian Avalanche Phenom-
ena. MALT Thesis. Royal Roads University, Victoria, BC, 284.
Crandall, B., Klein, G., Klein, G. A., & Hoffman, R. R. (2006). Work-
ing minds: A practitioner's guide to cognitive task analysis. Mit
Press.
Höller, P. (2012). About the Practical Knowledge to Understand
Snow Avalanches–A Chronology. In Natural Disasters.
InTech.
Hutchins, E. (1995). Cognition in the Wild. MIT press.
LaChapelle, E. R. (1980). The fundamental processes in conven-
tional avalanche forecasting. J. Glaciol., 26(94), 75–84.
McCammon, I. (2009). Human factors in avalanche accidents:
Evolution and interventions. In International Snow Science
Workshop (Vol. 27, pp. 644-648).
McClung, D. M. (2002a). The elements of applied avalanche fore-
casting, Part I: The human issues. Natural Hazards, 26(2),
111-129.
McClung, D. M. (2002). The elements of applied avalanche fore-
casting, Part II: the physical issues and the rules of applied
avalanche forecasting. Natural Hazards, 26(2), 131-146.
McIntosh, M. J., & Morse, J. M. (2015). Situating and constructing
diversity in semi-structured interviews. Global qualitative nurs-
ing research, 2, 2333393615597674.
Nemeth, C. P., Cook, R. I., & Woods, D. D. (2004). The messy
details: insights from the study of technical work in
healthcare. IEEE Transactions on Systems Man and Cyber-
netics- Part A Systems and Humans, 34(6), 689-692.
Perla, R. I., & Martinelli Jr, M. (1976). Avalanche handbook. Ava-
lanche handbook., (489).
Rouse, W. B., & Morris, N. M. (1986). On looking into the black
box: Prospects and limits in the search for mental mod-
els. Psychological bulletin, 100(3), 349.
Schweizer, J. and Fohn, P. M. B.. 1996. Avalanche forecasting-
an expert system approach. J. Glaciol., 42(141), 318–332.
Storm, I., (2010) The Canadian Avalanche Centre’s Long-Range
Forecasting Programme, International Snow Science Pro-
ceedings.
Wears, R. L., & Webb, L. K. (2014). Fundamental on situational
surprise: A case study with implications for resilience. Resili-
ence engineering in practice, 2, 33-46.
White, B., (2002) Development of Avalanche Safety and Control
Programs in the Canadian Rocky Mountain National Parks - A
Historical Perspective. International Snow Science Proceed-
ings.
Wears, R. L., & Webb, L. K. (2014). Fundamental on situational
surprise: A case study with implications for resilience. Resili-
ence engineering in practice, 2, 33-46.
