Where
snow
lingers

Seven years in the Canadian Rockies reveal where to look—and why last year’s date can let you down.

Follow the story

Alberta · observations from 2019–2025

Three-dimensional view of real Marmot Creek terrain coloured by estimated main snow-period end, with later estimates in lighter blue and white.
Marmot Creek · 2024–25
Colour shows seasonal timing, not snow depth. Explore the data below.

The last turns / The first hikes

Where would you look
for the last snow?

The same snow can be the reason for a ski trip or the obstacle on a hike. The approach matters as much as the view from below.

A mountain can look ready for summer from the road while snow still covers the climb above it. For a skier, that could mean a few more turns. For a hiker, it could change the day’s plan. How much can the view from below really tell you?

These maps explore the Canadian Rockies in Alberta, western Canada, using snow history and summer satellite observations from 2019–2025. They help trace where snow persisted and how its timing varied; they do not show present conditions.

A mountain has more than one summer

The first stop is Marmot Creek, in Kananaskis Country, near Mount Allan. Here, the July 2024 satellite picture suggests a difference between slopes that face north and those that face south. One photograph is intriguing. Seven years let us ask whether the pattern repeats.

The comparison follows the same selected patches of ground from 2019 through 2025, all between 2,200 and 2,400 metres. Keeping them within the same height band makes the comparison more useful: we are looking across slopes, rather than comparing a valley with a summit.

In every year, the north-facing group reached the estimated end of its main snow season later. The difference between the groups’ middle dates was 13–20 days. A middle date is simply the point halfway through the patches when their dates are sorted from earliest to latest.

That gives a skier a place to investigate and a hiker a reason to check for lingering snow. These dates describe the main stretch of seasonal snow cover. Small patches can outlast it, and a slope can become too thin or broken up to ski before it ends.

The same slopes, a different June

The direction of the difference held. The calendar moved.

For the north-facing group, the middle date was June 29 in 2022 and May 27 in 2023: 33 days apart. A late-June visit in those two years would fall on very different sides of the estimated end of the main snow season. Remembering a promising date from last year would miss that change.

The paired dates below show both parts of the story: the gap between north and south within each year, and the changing dates across years. Dates move both earlier and later through this short record, so it does not establish a steady trend toward earlier snow loss.

North-facing ground was a consistent clue in this sample. Choosing a day still requires evidence from the season you intend to visit.

Alberta / Marmot Creek–Mount Allan / 2,200–2,400 m

North lasts longer.
The calendar moves.

The estimated end of the main snow season came later on the selected north-facing slopes in all 7 years — by 13–20 days.

33 daysearlier on the north-facing sample in 2023 than 2022.

South-facing middle dateNorth-facing middle date

The highlighted years put the contrast in perspective: the north-facing middle date moved from Jun 29 to May 27. Slope direction alone does not give you a reliable trip date.

Exact dates, selected terrain and uncertainty
Fixed-cohort median end of the main snow season
YearSouthNorthNorth laterInput uncertainty medians, S / N
2019May 22Jun 1120 days7 / 7 days
2020Jun 4Jun 2117 days7 / 10 days
2021Jun 1Jun 1615 days4.5 / 4.5 days
2022Jun 14Jun 2915 days7 / 7 days
2023May 12May 2715 days1 / 7 days
2024Jun 10Jun 2313 days11 / 7 days
2025May 12May 2715 days4.5 / 1 days

The same 1,014 north-facing and 562 south-facing cells are retained across all seven years. These are different locations, not causal controls. Selection requires slopes of at least 5°, seasonal-snow status, a finite seasonal end estimate and input uncertainty no greater than 15 days in every year. Missing and persistent-snow/ice cells are excluded; this is a subset of the terrain.

North means aspect 315°–45°; south means 135°–225°. Each date is a descriptive median, rounded to a day from its annual epoch. The uncertainty column summarizes the input uncertainty values; it is not a confidence interval for the median. Spatially correlated terrain cells are not independent trials. Lines connect the two medians, not uncertainty bounds.

Fixed-cohort dates and method · Source datasets and provenance

End of the main annual snow period, not the last snow patch or last skiable day. Each point is the middle estimate for a selected group of slopes. The underlying dates can be uncertain by up to 15 days; these are historical patterns, not exact deadlines for a trip.

Follow the snow through the approach

A promising slope is only part of an outing. The ground between it and the trailhead determines how much walking comes before the snow—and whether the snowy sections connect.

For that question, the explorer moves to a second place: the Peyto area, near the Wapta Icefield. Mount Allan supplies the repeated north–south comparison; Peyto offers a separate ski-touring example in which to examine a route. The result from one place is not a measurement of the other.

The Wapta is glacier-touring terrain. Parks Canada identifies April and May as the recommended alpine touring season on the icefields and snow peaks along the Continental Divide. Travelling there requires avalanche and glacier-travel skills. Its appearance in a summer explorer is an opportunity to study the changing cover, not a recommendation for a summer ski trip.

Choose Bow Summit · Peyto, then Wapta Icefield Routes. On July 7, 2024, every sampled map square along the displayed segment was labelled snow or ice. By August 1, fewer than half were; the rest were labelled clear land, water or unknown.

The height chart gives that line a side view. Follow it along the route and use the position marker to locate the changing colours on the map. Look for interruptions between snowy sections. A white upper slope can draw the eye while the approach tells a different story. Water labels are another reason to examine this mapped segment carefully before treating it as an itinerary.

Peyto / Follow the approach

The mountain.
The route. The remaining snow.

Start with a route, then watch where snow remains as summer advances. The white patches show where a satellite detected snow or ice on that date. They do not tell us whether there is enough snow to ski.

Historical satellite observations, mapped trail segments and real elevation. Your own GPX stays in this browser.

Select a route and an observation date to investigate snow along the way.

What these observations can—and cannot—tell us

Sentinel-2 classifies snow and ice together. Cloud, shadow, unclassified pixels and missing coverage stay unknown. Vegetation is a land classification, not proof that ground underneath a canopy is snow-free. Water is separate. The 20 m source classification is aligned to the existing 30 m analysis grid; it cannot resolve the width of a trail.

Trail lines are OpenStreetMap features, not verified trip itineraries. Some are partial segments. The elevation model can miss small cliffs and gullies. Historical observations do not establish present-day access, snow depth, avalanche stability or a safe route.

Potential sunlight is a terrain calculation under stated assumptions, not observed weather or a snowmelt forecast. Use official trail conditions and avalanche forecasts for current planning.

© OpenStreetMap contributors / ODbL · Copernicus Sentinel-2 classification · Observation index · Route data and provenance

For a closer look at the history of one place, open the annual explorer below. It follows the area selected on the route map. If you stay at Peyto, you are examining Peyto’s history; the seven-year north–south finding above belongs to Mount Allan.

Explore how the snow season changed in one place

These colours show estimated seasonal end dates, rather than snow visible on a particular day. Some glacier-covered or uncertain places have no comparable date. The selected area follows the route map above.

Optional / Compare seasonal timing

Same slope. Different year.

Would the same trip have met a different snow season in another year? Compare the same ground across seven winters and springs.

Marmot Creek terrain with annual snow-period timing colours. Load the explorer to inspect the source estimates.

Three Alberta windows. Seven snow years.
A landscape you can turn, compare and inspect.

Historical annual estimates, 2018–19 to 2024–25. Not current conditions.

How much of each window qualifies? All seven years · available without JavaScript

Percentage of the 90,000 source cells in each 81 km² window with seasonal status 0, a reported end date, and uncertainty ≤ ±15 days. Excluded cells include perennial snow/ice, other classifications and wider observation gaps. These percentages describe data qualification, not snow-covered area.

Qualified cells / all cells · same rule for every window and year
Study window2018–192019–202020–212021–222022–232023–242024–25
Marmot Creek · Mount Allan61.7%78.9%89.8%75.0%84.5%85.6%90.7%
Bow Summit · Peyto6.8%4.3%14.9%11.2%40.7%32.5%37.4%
Jasper · Marmot Basin71.6%64.4%84.0%64.0%90.7%77.6%94.1%

The glacier-bearing Peyto window has much less qualified seasonal-date coverage. A low percentage is a reason to inspect the classifications, not evidence of early melt.

01 / What is mapped

End of the biggest snow period, not necessarily the last summer snow patch. The source class includes snow and ice.

02 / What is unresolved

A satellite cannot always see through clouds or tree cover. A blank or uncertain date means we do not know, not that the ground was clear.

03 / What to take away

Use past years to understand how much timing can change. Use current reports and avalanche forecasts to decide whether a trip is suitable now.

Data, definitions & downloads

Natural Resources Canada’s annual 30 m snow-dynamics product, seven snow years from 2018–19 through 2024–25. This atlas contains bounded regional extracts, not all of Alberta.

NRCan dataset and documentation · Download the data manifest

Before choosing a day

The maps make the next questions more specific. Where does continuous snow begin? Is the report you found about that approach or a slope much higher up? Was the ground visible in the satellite picture, or hidden by cloud? Unknown areas leave the question open.

For a skier, those questions help separate the appeal of a few turns from the work of reaching them. For a hiker, they identify sections where a recent account could change the plan. Snow depth, surface firmness and stability still need other evidence; the satellite colours cannot supply them.

This collection ends in 2025. Before an outing, look for recent reports from the intended route, check access and weather, and consult the applicable avalanche forecast. Parks Canada’s touring guidance explains the preparation these mountains require.

In this sample, the north-facing group reached the estimated end of its main snow season 13–20 days later than the south-facing group, even as its date shifted by 33 days between two years. That makes direction a useful starting point for deciding where to look. The approach determines what reaching that snow would involve, so the next useful evidence is a recent report covering the ground you would actually travel.

A closer look: four pictures of summer 2024

Supporting views · Summer 2024

Where is the snow?
Four clues, not a forecast.

Start with the question each picture answers. Together they show where snow remained in summer 2024—and how to avoid mistaking a cloudy image for a clear route.

These figures show fixed 2024 observations; the explorer controls do not change them. A compares all three areas; B–D examine Marmot Creek, a separate example from the Peyto ski route. On smaller screens, scroll wide charts sideways.

A

Observation strips

Did the snow go, or did the view disappear?

Look at Peyto on June 22: most of the bar is hatched. The satellite could not see enough to tell us what happened to the snow. A small blue bar on that date does not mean a mostly clear mountain. Compare it with July 7, when much more of the ground was visible.

  • Snow / ice
  • Clear land
  • Water
  • Unknown

Marmot Creek

Bow Summit / Peyto

Jasper / Marmot Basin

Takeaway: check the hatched area before trusting the blue area. A mostly hidden scene is a reason to look for another observation. Each bar shows the whole study window, with its own date; clear land can include trees hiding snow underneath.

Observation counts and original scenes
2024 classifications, by 30 m analysis cell
WindowSceneSnow / iceClear landWaterUnknown
Marmot CreekJun 729,15146,970813,871
Marmot CreekJun 2916,68828,175145,136
Marmot CreekJul 78,28778,478153,220
Marmot CreekJul 2451085,436104,044
Marmot CreekAug 121285,118144,656
Marmot CreekAug 317582,15567,764
Marmot CreekSep 51085,45984,523
Marmot CreekSep 25074,1341415,852
Bow Summit / PeytoJun 779,7826,8667462,606
Bow Summit / PeytoJun 2211,15715115278,540
Bow Summit / PeytoJul 761,92123,8411,5782,660
Bow Summit / PeytoJul 1748,31535,9191,8833,883
Bow Summit / PeytoAug 131,45847,0754,4577,010
Bow Summit / PeytoAug 3123,88748,2135,33612,564
Bow Summit / PeytoSep 515,45556,10710,7787,660
Bow Summit / PeytoSep 308,9422,22791077,921
Jasper / Marmot BasinJun 727,50152,5992259,675
Jasper / Marmot BasinJun 1736,64634,79922518,330
Jasper / Marmot BasinJul 75,72981,4323482,491
Jasper / Marmot BasinJul 1768283,9943424,982
Jasper / Marmot BasinAug 11066,78531222,903
Jasper / Marmot BasinAug 19080,3023039,395
Jasper / Marmot BasinSep 5085,9733913,636
Jasper / Marmot BasinSep 18049,76492739,309
B

Elevation × observation

In early July, the snow was higher up.

Read the July 7 column upwards. Around 2,000–2,200 metres, about 4% of the ground the satellite could classify was snow or ice. At 2,600–2,800 metres, it was 57%. The lower slopes and the high mountain were having very different summers.

0% snow / ice100%

The large number is the snow/ice share of visible land. The small number is the share of the whole band the satellite could not classify. Marmot Creek, summer 2024; slopes of at least 5°. Each row covers 200 metres of height, from its lower bound up to—but not including—the upper one.

Marmot Creek · 2024 · elevation in metres
Elevation
2,800–3,000100.0%0.9% unknown; 754 snow or ice, 0 clear land, 0 water, 7 unknown cells100.0%0.7% unknown; 756 snow or ice, 0 clear land, 0 water, 5 unknown cells95.9%10.8% unknown; 651 snow or ice, 28 clear land, 0 water, 82 unknown cells3.1%31.7% unknown; 16 snow or ice, 504 clear land, 0 water, 241 unknown cells2.3%32.5% unknown; 12 snow or ice, 502 clear land, 0 water, 247 unknown cells3.5%43.2% unknown; 15 snow or ice, 417 clear land, 0 water, 329 unknown cells0.0%38.8% unknown; 0 snow or ice, 466 clear land, 0 water, 295 unknown cells0.0%55.3% unknown; 0 snow or ice, 340 clear land, 0 water, 421 unknown cells
2,600–2,800100.0%1.9% unknown; 2,700 snow or ice, 0 clear land, 0 water, 53 unknown cells100.0%15.0% unknown; 2,338 snow or ice, 1 clear land, 0 water, 414 unknown cells57.0%10.4% unknown; 1,405 snow or ice, 1,062 clear land, 0 water, 286 unknown cells5.6%23.9% unknown; 118 snow or ice, 1,978 clear land, 0 water, 657 unknown cells3.6%24.9% unknown; 75 snow or ice, 1,993 clear land, 0 water, 685 unknown cells3.0%32.4% unknown; 55 snow or ice, 1,805 clear land, 0 water, 893 unknown cells0.5%22.7% unknown; 10 snow or ice, 2,117 clear land, 0 water, 626 unknown cells0.0%48.5% unknown; 0 snow or ice, 1,417 clear land, 0 water, 1,336 unknown cells
2,400–2,60095.5%6.3% unknown; 7,723 snow or ice, 364 clear land, 0 water, 541 unknown cells93.3%20.8% unknown; 6,378 snow or ice, 459 clear land, 0 water, 1,791 unknown cells31.1%8.1% unknown; 2,468 snow or ice, 5,460 clear land, 0 water, 700 unknown cells2.6%14.0% unknown; 190 snow or ice, 7,227 clear land, 0 water, 1,211 unknown cells1.2%14.7% unknown; 88 snow or ice, 7,270 clear land, 0 water, 1,270 unknown cells0.0%21.5% unknown; 0 snow or ice, 6,772 clear land, 0 water, 1,856 unknown cells0.0%15.0% unknown; 0 snow or ice, 7,336 clear land, 0 water, 1,292 unknown cells0.0%31.5% unknown; 0 snow or ice, 5,909 clear land, 1 water, 2,718 unknown cells
2,200–2,40083.5%9.0% unknown; 11,240 snow or ice, 2,225 clear land, 8 water, 1,331 unknown cells60.9%34.1% unknown; 5,936 snow or ice, 3,819 clear land, 0 water, 5,049 unknown cells20.4%4.9% unknown; 2,874 snow or ice, 11,201 clear land, 0 water, 729 unknown cells1.1%6.4% unknown; 150 snow or ice, 13,711 clear land, 0 water, 943 unknown cells0.2%7.6% unknown; 29 snow or ice, 13,656 clear land, 0 water, 1,119 unknown cells0.0%11.8% unknown; 0 snow or ice, 13,051 clear land, 0 water, 1,753 unknown cells0.0%8.0% unknown; 0 snow or ice, 13,620 clear land, 3 water, 1,181 unknown cells0.0%24.8% unknown; 0 snow or ice, 11,117 clear land, 11 water, 3,676 unknown cells
2,000–2,20037.4%26.2% unknown; 4,811 snow or ice, 8,051 clear land, 0 water, 4,574 unknown cells11.3%61.6% unknown; 753 snow or ice, 5,937 clear land, 0 water, 10,746 unknown cells3.5%4.0% unknown; 592 snow or ice, 16,143 clear land, 8 water, 693 unknown cells0.0%2.8% unknown; 8 snow or ice, 16,940 clear land, 6 water, 482 unknown cells0.0%3.3% unknown; 3 snow or ice, 16,850 clear land, 9 water, 574 unknown cells0.0%5.7% unknown; 0 snow or ice, 16,436 clear land, 4 water, 996 unknown cells0.0%3.5% unknown; 0 snow or ice, 16,829 clear land, 4 water, 603 unknown cells0.0%19.2% unknown; 0 snow or ice, 14,095 clear land, 1 water, 3,340 unknown cells
1,800–2,0006.0%25.8% unknown; 844 snow or ice, 13,131 clear land, 0 water, 4,848 unknown cells0.4%56.3% unknown; 31 snow or ice, 8,192 clear land, 1 water, 10,599 unknown cells0.0%1.5% unknown; 2 snow or ice, 18,538 clear land, 0 water, 283 unknown cells0.0%1.1% unknown; 0 snow or ice, 18,610 clear land, 0 water, 213 unknown cells0.0%1.4% unknown; 0 snow or ice, 18,556 clear land, 0 water, 267 unknown cells0.0%4.4% unknown; 0 snow or ice, 17,989 clear land, 1 water, 833 unknown cells0.0%1.8% unknown; 0 snow or ice, 18,481 clear land, 0 water, 342 unknown cells0.0%10.8% unknown; 0 snow or ice, 16,782 clear land, 0 water, 2,041 unknown cells
1,600–1,8002.0%13.5% unknown; 199 snow or ice, 9,613 clear land, 0 water, 1,531 unknown cells0.0%47.4% unknown; 0 snow or ice, 5,969 clear land, 0 water, 5,374 unknown cells0.0%1.3% unknown; 0 snow or ice, 11,200 clear land, 0 water, 143 unknown cells0.0%0.5% unknown; 0 snow or ice, 11,281 clear land, 0 water, 62 unknown cells0.0%1.9% unknown; 0 snow or ice, 11,126 clear land, 0 water, 217 unknown cells0.0%6.1% unknown; 0 snow or ice, 10,646 clear land, 0 water, 697 unknown cells0.0%0.6% unknown; 0 snow or ice, 11,276 clear land, 0 water, 67 unknown cells0.0%12.0% unknown; 0 snow or ice, 9,983 clear land, 0 water, 1,360 unknown cells

Takeaway: a clear valley does not tell you whether the upper route still crosses snow. Follow the full climb. This picture says where snow or ice was detected, not whether it formed a continuous, deep-enough ski run.

C

Slope direction

North held more snow.
On this date.

July 7 · Marmot Creek · 2,200–2,400 m · slopes ≥5°

North-facing cells show more classified snow than south-facing cells in this elevation bandN: 46.8% snow or ice among classified land, 18.7% unknown among all cells. E: 21.1% snow or ice among classified land, 1.5% unknown among all cells. S: 12.2% snow or ice among classified land, 0.0% unknown among all cells. W: 11.4% snow or ice among classified land, 4.0% unknown among all cells0%25%50%75%100%N46.8%E21.1%S12.2%W11.4%Snow / ice among classified land

At roughly the same height, 47% of classified north-facing ground was snow or ice, compared with 12% facing south. North is a useful clue for where to investigate—not a promise of July skiing. About 18.7% of the north-facing ground was hidden or unresolved, versus 0.0% facing south. The seven-year comparison earlier in the article checks whether the broader north/south pattern repeats.

Direction counts and missing observations
DirectionSnow / iceClear landWaterUnknown
N1,0291,1680506
E1,0063,756072
S4293,08200
W4103,1950151
D

Two-date transition

What changed on
the visible ground?

Marmot Creek · June 7 → July 24, 2024

25,712 cells changed from snow or ice to clear landOnly the 73,144 cells classified as land or snow on both dates are shown. 510 from Snow / ice to Snow / ice. 25,712 from Snow / ice to Clear land. 0 from Clear land to Snow / ice. 46,922 from Clear land to Clear land. Ribbon widths are proportional to cell counts.June 7July 24Snow / ice → Snow / ice: 510 cellsSnow / ice → Clear land: 25,712 cellsClear land → Clear land: 46,922 cellsSnow / ice26,222Snow / ice510Clear land46,922Clear land72,63473,144 matched cells · ribbon width = cell count

25,712 cells switched from snow / ice to clear land.

Here we keep only ground that could be classified on both dates, so a cloudier picture cannot explain the difference. Of the 26,222 squares labelled snow or ice on June 7, 25,712 were labelled clear land by July 24—about 98%. Takeaway: the summer change was substantial in this shared view. It does not give the exact day a route became clear. The remaining 16,856 squares were left out because one date was unknown or water.

Transition counts
June 7 → July 24Cells
Snow / ice → Snow / ice510
Snow / ice → Clear land25,712
Clear land → Snow / ice0
Clear land → Clear land46,922

Source: archived Sentinel-2 Level-2A scene classifications, resampled to the shared 30 m analysis grid; elevation and direction from the terrain model. Download the counts and source links. These are historical observations, not present-day trail or avalanche conditions.

Behind the story: the comparison, the maps and AI

The annual estimates come from Natural Resources Canada’s snow-dynamics dataset. They describe the end of the largest snow-covered period, estimated from the gap between satellite observations. They are not exact dates of final melt or last skiable snow.

The Mount Allan comparison retains 1,014 north-facing and 562 south-facing 30-metre cells between 2,200 and 2,400 metres, on slopes of at least 5°. Every retained cell has a seasonal-snow estimate with reported timing uncertainty no greater than ±15 days in all seven years. These are approximately 38% of the north-facing and 16% of the south-facing candidate cells. The selection favours ground with usable records. It does not represent every slope, and matching height does not isolate direction from other terrain differences.

Seven-year calculations. The difference between the rounded north-facing dates in 2022 and 2023 is 33 calendar days. Descriptive middle dates are not confidence intervals, and neighbouring cells are not independent measurements.

The separate summer sequence contains 168 Sentinel-2 observations: eight dates per summer, for seven summers from 2019–2025, in each of three study windows. The fixed worked figures show 2024. Their observation strips include all three windows, including a separate Jasper / Marmot Basin window. Marmot Basin is not Marmot Creek: the elevation, direction and change figures concern Marmot Creek near Mount Allan. These figures do not change when an explorer control changes. Clouds and shadow remain unknown. Clear-land labels can miss snow under trees, and snow/ice labels do not distinguish a skiable surface. Local ground measurements have not validated these classifications.

The Peyto example samples 144 unique map cells along one segment labelled “Wapta Icefield Routes.” On July 7, 2024, all 144 were labelled snow/ice. On August 1, the counts were 62 snow/ice, 46 clear land, 33 water and three unknown. These are sampled cells, not measurements of skiable route length or a survey of the full Wapta. Observation counts.

AI agents helped research the sources, process the data and build the interface. Checks compared exported values with original satellite pixels and tested route coordinates. Those checks establish agreement between the source values, files and maps; they do not establish conditions on the ground. Annual data and methods · Summer observations and sources.

Trying your own route on the map

A GPX file stores a route as GPS points. A Wapta trip report from March 2008 links the party’s recorded tracks. Download the file from the author’s page, choose Bow Summit · Peyto, then Upload your route (.gpx). Only parts of two of its ten tracks cross this small study window and receive snow and height information. The explorer starts in 2024; its observations cover 2019–2025, not the party’s 2008 outing.

The author warns against using the old track for navigation. It is an upload example; glacier terrain changes, and an old track cannot establish today’s route. The built-in mapped lines also include hiking routes; their presence is not a ski recommendation.

End of field noteBack to Field Notes