A Large Avalanche in the Ministry of Transportation and Infrastructure’s Jack MacDonald Path
By Mark Grist and the MOTI Columbia Avalanche Program Team
Note: This article was initially published in The Avalanche Journal, Volume 131, Spring 2023
THE LEAD UP
The winter of 2021-22 got off to a good start in the Selkirk Mountains. There was abundant snow throughout November, especially towards
the end of the month when an atmospheric river (AR) flowed over the region. The snowpack at the highway (weather station 1 in Fig. 1) doubled from Nov. 25–28, reaching a healthy 113cm. The weather station on Mount Fidelity in Glacier National Park (GNP), located at 1,905 m elevation, (weather station 5 in Fig. 1) recorded 72 cm of snow (73.1 mm of precipitation) over the last three days of the month, and the height of the snowpack increased to 252 cm from 206 cm. But, as the river warmed up and snow turned to rain, everything changed.
At the highway, temperatures peaked at +5.5 C, while 83.1 mm of rain fell over 97 hours, wiping out half the snowpack. Notably, the average rate of snowpack decline was quite steady at 7 cm/12 hrs over an 84-hour period. At treeline, 34.8 mm of rain fell over 25 hours and temperatures peaked at +2.1 C; the snowpack decreased by 21 cm.
The rain-soaked layer was eventually buried either Dec. 1 or 2, depending on elevation and location, and the resultant hard crust was found as high as 2,200–2,300 m throughout the Columbia Mountains. In our paths, the crust was especially prominent below 1,700–1,900 m. In character, it was more like an ice formation rain crust with a surface glaze than the typical melt-freeze crust; and similar to the November rain crust of 2002, which became very problematic in the tragic early months of 2003.
Due to the substantial November snowpack, the crust was located above most terrain roughness features. We found it ranged from 70–160 cm above ground (the average was 140–150 cm) in our various study plots. Interestingly, strong winds during the rain event (Round Hill, at ridge crest, reported sustained southwest winds of 53–74 km/hr, gusting 105 km/hr) drove heat and rain into the snowpack, resulting in significantly thicker crusts on southerly aspects. The crust was 15 cm thick on a southeast aspect at 2,050 m near Lookout Col, while only two centimetres thick on a north aspect at the same elevation (Kate Ryan, personal communication).
Not surprisingly, we tracked this layer closely over the following weeks. A snow profile conducted on Dec. 9 at Corbin Low revealed the crust was 15 cm thick and located 116–131 cm up from the ground. In another profile in Helen’s (path 46.4), located near Corbin High, we found the crust was 12 cm thick, with a very thin layer of one-to-two-millimetre facets above it (Fig. 2). We had no significant test results at this time, largely due to a lack of slab properties above the crust.
Profiles by GNP field teams between Dec. 5 and Jan. 21 revealed the crust was 10–20 cm thick at treeline elevation. Only one notable result (a hard result on a deep tap test) came back from 18 GNP and MoTI snowpack tests, while several tests on this layer came back with no result.
As time went by, there was a slight trend towards a thicker layer of facets and slightly larger faceted grains above the crust, especially at lower elevations (Fig. 4). We performed control work in early January to test the reactivity of deeper layers and for snowpack reduction. On Jan. 10, control results from 19 shots (13 kg each) in the large paths, including two shots in path 47.8, Jack MacDonald, were limited to size two avalanches.
The first indication the early-December layer had become active was on InfoEx on Jan. 14. Our nearest-neighbour evidence came on Jan. 17 from GNP, when they reported a natural size four persistent slab avalanche with a two-metre crown on a southwest aspect. Over the next few days, several surprisingly large natural and controlled results, with up to one kilometre propagation, appeared on InfoEx. The layer was reactive in propagation saw tests, as demonstrated in a Jan. 19 video recorded by Avalanche Canada near our Corbin Pass Low weather station (vimeo.com/671324845). There was a lot of uncertainty regarding this layer, but it was on our radar for rapid loading and rapid warming events. Given the evidence, it was obvious the potential for large avalanches was there, and we wondered which of our paths might produce similarly large and surprising results. We noted many atypical fracture lines with wide propagations during this period.
On Jan. 21, a GNP snow profile conducted at 1,905 m showed no results on the Dec. 1 layer. It was up 140 cm from the ground and down 189 cm from the surface. Over a metre of high-density slab lay above the crust, with measured densities ranging from 215–390 kg/m3. Table 2 outlines the load on the Dec. 1 crust when we performed control work on Jan 22. Precipitation amounts increased with elevation; however, the snow-to-liquid-water ratios remained similar between low and mid elevations. Note that one millimetre of precipitation adds one kilogram of load per horizontal square metre of surface area.
THE EVENT
We finally had a weather window for a helicopter control mission on Jan. 22. Deploying 13 kg charges, we produced mostly size 2.5–3 avalanches in our larger paths near GNP, but nothing ran over the Lanark or Twins snow sheds and the closest stopped about 100 m from the highway. The first five shots placed in Jack MacDonald resulted in four size three and one size 2.5 avalanches, which stopped 100–800 m from the highway.
The penultimate shot was placed as high as possible under the communications shell at the summit (Fig. 3). The resulting avalanche propagated significantly across the path, while wrapping around to produce a size three in Helen’s and a size 3.5 to the north that ran into the Tangier River. The connected crowns had wrapped almost 220 degrees around the mountain! Longtime staffers in the Columbia Avalanche Program commented how they were surprised to see such a large result and were not expecting something that big to pull out of the path.
The comment in our SAWS database for this avalanche read:
“Pulled entire slope propagating ~500 m down ridge to east, also sympathetic release N asp into Tangier R (sz 3.5) and SW asp into Helens (sz 3). Jack Macdonald terminated beyond historical runouts with ~25 m deep in river and ~5 m deep (~100 m wide) on CP tracks, breaking mature timber and communication line. ~300 m width at terminus with a lot of mature timber from air blast and impact forces.”
FIG. 6: THE AFTERMATH IN JACK MACDONALD. THE LEFT IMAGE SHOWS SOME OF THE 3.1 HECTARES OF FOREST DESTROYED BY THE AVALANCHE. THE RED ARROW POINTS TOWARDS ANOTHER 2.3 HECTARES DESTROYED IN THE LOWER TRACK AS THE AVALANCHE SUPERELEVATED. FOR SCALE, THE SNOW SHED IS 141M LONG. THE RIGHT IMAGE SHOWS MACHINERY CLEARING OVER FIVE METRES OF SNOW FROM THE CP RAILWAY LINE ACROSS THE RIVER. DEBRIS IN THE ILLECILLEWAET RIVER WAS ESTIMATED AT 25–30 M DEEP.
THE AFTERMATH, PART 1: CLEANUP AND FRACTURE LINE PROFILE
The avalanche overran the west end of the snow shed and put an average of three metres of snow on 80 m of highway, and one metre of snow on the road inside the shed. Deposit removal took almost 2.5 hours, with two large loaders working continuously.
We completed a fracture line profile on Jan. 25 in Helen’s (Fig. 4). The Dec. 1 crust was clearly visible and was 10 cm thick. Interestingly, a five-centimetre-thick layer of facets was observed above the crust. The key result on the facet/crust layer was a propagation saw test that ran to the end of the column after 58 cm (PST 58/160 END).
THE AFTERMATH, PART 2: THE WET SLAB CYCLE OF MARCH 29
After the cycle of large avalanches had abated, we wondered if and when a spring wet slab cycle might occur. In a study of 11 winters with notable deep persistent slab avalanche cycles in the Washington Cascades between 1989- 90 and 2020-21, at least four of them had significant deep wet slabs release on the same crust/facet layer around 100–129 days after crust formation (Primomo 2022). Conventional wisdom says watch out for wet slabs after three nights with no refreeze, so we were surprised when a significant cycle occurred on March 29—119 days after crust formation—as the Fidelity weather station had recorded above zero temperatures for only four hours when the cycle began, and the maximum temperature was only 4 C.
More recent research (Levy et. al 2022) suggests incoming radiation, rather than temperature, is a stronger predictor of wet slab activity. This was particularly relevant in our case as the paths that ran naturally faced due south, where the thickest and smoothest crusts were found. Lower elevation paths (especially below 1,500 m) were particularly active.
THE AFTERMATH, PART 3: DEBRIS REMOVAL
By late August, the remaining avalanche debris above the shed was still 10–20 m deep, and it was abundantly obvious it would not melt before winter. This had the potential to compromise drainage through the three-metre culvert running below the shed and threaten the shed structure itself. The ministry contracted three excavators and three rock trucks to clear the snow, rocks, and trees (including some merchantable timber). They worked for 31 days, removing about 45,000 m3 of debris. Special care had to be taken while excavating as a large snow cave had formed above the creek (Fig. 8).
DISCUSSION AND HISTORICAL PERSPECTIVE
This was the largest avalanche in 34 years for this path, when helicopter control on Feb. 17, 1988, produced what was recorded as a size 4.5 avalanche that ran a similar distance and left an average depth of six metres snow (max 10.5 m) along 41 m of road beyond the west end of the shed. The average deposit measurements for this avalanche in the ministry’s SAWS database were listed as 1,000 m length x 150 m width x 10 m depth, giving a volume of 1.5 x 106 m3, which is an order of magnitude greater than the lower end range for a size five avalanche given in Jamieson et.al (2014). The slab was 2.5 m thick and released at ground level, suggesting a rain crust may not have been a factor. Interestingly, 40 minutes later that day, control work in Helen’s produced a sympathetic size four in Jack MacDonald that hit the railroad tracks for a second time!
Fitzharris scoured the old CPR avalanche records from 1909 to 1976 and determined the return period of significant avalanches to the railway to be every 7.5 years for Jack MacDonald. Measurements on Google Earth revealed the distance from the edge of the snowshed to the railway tracks to be 230 m; therefore, any recorded toe mass distance (TMD) greater than 230 m in our SAWS database was taken to be a similarly significant event. There have been six such events since MoTI records began in 1977, giving a return period of 7.5 years!
Fitzharris (1981) also noted that with artillery control, the frequency and magnitude of small and medium avalanches had been altered since 1965, but the effect on very large avalanches was unclear. Our data suggests both the frequency and magnitude of very large avalanches is not different; however, the trigger has changed from 100% natural before 1977 to 17% natural (one of six events) and 83% artificial (five of six events) in the years since. Artillery control began in 1962 with the opening of the Trans-Canada Highway (Schleiss 1990). Previously, avalanche defence of the railway was limited to snow sheds and the Connaught tunnel.
Chris Argue of Dynamic Avalanche Consulting produced a return period graph from MoTI data (Fig. 9) that pegged this as a 1:23-year event. For the large paths in the Selkirks with discontinuous start zones, Fitzharris (1981) used a best-fit frequency distribution to determine the mass of a 30-year avalanche is 5% of Mo (the limit avalanche, or the largest avalanche a path can produce in a 30-year winter). Our independently calculated mass (2.4 x 108 kg) compares favourably with this value (1.45 x 108 kg) for Jack MacDonald, adding further strength to this being approximately a 1:30 year event.
Fitzharris (1981) reported the largest avalanche in Jack MacDonald occurred on Feb. 11, 1937, with a mass of 3.06 x 108 kg (which is 11% of Mo). He used National Research Council (NRC) values for deposit density (320–380 kg/m3), which are low for a size five avalanche [see Table 2 in Jamieson et. al (2014)]. Using a mid-range NRC value (350 kg/m3) for the Jan. 22, 2022, avalanche would yield a mass of 0.91 x 108 kg, roughly one-third the size of the 1937 avalanche. Interestingly, we dated the rings of a mature tree deposited on the snow shed at 84 years old, which means it started growing in 1938, the year after the 1937 avalanche! The tree’s original location is unknown, but it likely came from the 2.3 hectares of forest that was cleared when the avalanche super-elevated in the lower track (see red arrow in Fig. 6).
A few words about size: The black rectangles in Fig. 10 show the range of values for the Jan. 22, 2022, avalanche compared to established criteria for avalanche size classification. A continued log scale is assumed for deposit volume (i.e., size five volume is 1 x 106 m3). MoTI video of the event shows trees being destroyed in the lower start zone, thus, calculating areas where forest was destroyed (or could have been destroyed if trees were present in the track or runout) gives an area of approximately 60 hectares, which places this event squarely in the size five category.
The future impact of climate change on avalanches is a topic of increasing interest. Hendrikx et. al (2022) found the frequency of avalanches to the highway in GNP will likely decrease by the 2090s; however, this key finding may not tell the whole story. Bellaire et. al (2016) found an increase in the frequency of melt-freeze crusts at the Mt. Fidelity study plot in GNP in November or December, especially since With a higher probability of extreme meteorological events under climate change, Hendrikx et. al (2022) note there is a continued potential for extreme avalanche events. The combination of more sliding layers embedded in the early season snowpack and (warm) rapid loading events potentially explains why extreme avalanche events may not diminish over time.
Thus, the setup for a pattern of large destructive avalanches would take on the form of early season AR events occurring when a significant snowpack has already accumulated, and subsequent (warm) rapid loading events overload crust/facet combinations. Typically, snowpack tests have focused on start zones, looking at initiation and propagation propensity. If thick glaze crusts exist at lower elevations, investigations at the track level might be useful to get a handle on whether the crust could ‘turbocharge’ any large avalanches. In other words, we might do well to consider initiation, propagation, AND acceleration when forecasting larger avalanches.
ACKNOWLEDGEMENTS
Special thanks to Bruce Jamieson for article review and insightful discussions about large avalanches. Thanks to Catherine Brown at Parks Canada for GNP weather station data and snow profile information. Thanks also to Paul Harwood, Johann Slam, Wren McElroy, and Kate Ryan for article review and comments.
REFERENCES
Fitzharris, B.B. 1981 “Frequency and climatology of major avalanches at Rogers Pass, 1909 to 1977”. National Research Council of Canada. Division of Building Research.
Schleiss, V.G.. 1990 “Rogers Pass Snow Avalanche Control – A Summary” Canadian Parks Service, Revelstoke, B.C. 22 pages.
Jamieson, B., Beglinger, R., and Wilson, D., 2014. “Case study of a large snow avalanche in the Selkirk Mountains and reflections on the Canadian size classification.” Geohazards 6 – 6th Canadian GeoHazards Conference, Kingston, ON, Canada.
Primomo, M., 2022 “A Pattern of Deep Persistent Slabs in the Washington Cascades.” The Avalanche Review. 40 (3) pp 29-31.
Levy, B., Tupy, A. and Simenhois, R., 2022 “Forecasting Wet Slab Avalanches: Start Zone Characteristics and Weather Variables.” The Avalanche Review. 40 (4) pp 18-20
Hendrikx, J., Jones, A., Argue, C., Buhler, R., Jamieson, B. and Goodrich, J. 2022 “The Potential Impacts of Climate Change on Snow Avalanche Hazards on the Trans Canada Highway in Glacier National Park.” PIARC, Calgary. 16 pages.
Bellaire, S., Jamieson, B., Thumlert, S., Goodrich, J., Statham, G., 2016. Analysis of long-term weather, snow and avalanche data at Glacier National Park, B.C., Canada. Cold Regions Science and Technology, 121, 118-125
