BC Snow Pillows

I’ve updated and improved my code that tracks the state of BC’s snowpack using the automated snow pillow network, and thought I’d give a little bit more detail into what goes into the map and the graphs. There is a great Q+A with Tony Litke, one of the province’s snow specialists, here.

First off, what’s is a snow pillow? Basically, it’s an anti-freeze and water filled bladder that gets installed on a large concrete pad somewhere in the mountains. The hydrostatic pressure of the overlying snowpack pushes the antifreeze/water solution into a standpipe, and the height of that is measured with an automatic sensor. There are also newer versions that are basically giant scales measuring the weight of the snowpack.

Snow pillow (on the ground) and snow depth sensor (mounted off the tower) somewhere in the woods. From Engel et al. (2022; http://dx.doi.org/10.21079/11681/44122)

Second, the snow pillows record snow water equivalence or SWE, and SWE is probably the most important snowpack variable from a hydrology perspective. It’s also the hardest to measure continuously as its a function of both snow depth and snow density, and those snow pillows are tough to install and maintain. The best way to think about SWE is to think about how much water you’d have if you melted the snow completely. One cubic meter of snow, with a density of 200 kg/m3, would turn into 0.2 cubic metres (or 200 litres) of water.

On to the code: the snow pillows in BC update automatically through the magic of satellites and cellular networks, and the province of BC updates a .csv file gives the current year’s SWE values for all the stations in the province. The code goes through each station, and matches the ID with the historical SWE data that pulled together by Vionnet et al. (2021) in their CanSWE dataset. For each day of the year, the quantiles of SWE data are calculated (see below), and then I plot the 25th, 50th, and 75th of SWE on each day of the year along with the current year’s values.

Example of historical SWE data from snow pillow 1B01P (Mount Wells) showing the seasonal patterns of snow accumulation and melt. Each individual SWE year is given as a gray trace, and the percentiles of SWE by day of year are given in blue (25th), orange (50th), and green (75th).

Because I’m also interested in the accumulation and melt rates, I look at the change in SWE from day to day in the historical data and the current year.

Left: Historical SWE (25th and 75th percentiles are dashed blue, median is solid blue) and current year’s SWE (orange points). Right: Daily change in SWE for historical median values (blue) and current year (orange).

As the code goes through each station, it grabs the current day’s SWE, and the median historical SWE for the day of year, and stores that with yet another dataset that contains the latitude, longitude, and elevation of each snow pillow station. From there, I calculate the percent difference from normal, and map it using colours to represent how the snowpack is doing, in real-time (red is way below normal, blue is way above normal). The little histogram in the corner just shows the number of stations with a given percentage of normal, and the average (black line) for the day of analysis.

Map of current “percent of normal” for BC’s snow pillow network, and inset histogram with overall mean (vertical black line).

Sara Darychuk at the ESA!

"Learning from the European Space Agency (ESA) has been a fantastic experience so far. We just finished our third day, which focused on optical remote sensing. My favourite day, however, was the SAR day. I am excited to try some of the new techniques I learned in my own research, such as creating coherence and intensity images from multi-temporal SAR images or performing spectral unmixing on optical data. The instructors so far have been knowledgeable and engaging, and I enjoy the different perspectives they bring from ESA, Industry, and academia. I am with a group of 30 graduate students, with varied backgrounds from chemical engineering to archeology. I have already learned a lot from my peers, especially my roommate who also works with SAR and Google Earth Engine. I am the only Canadian here but the group is very welcoming. Our hotel is in the beautiful val-de-poix, and I was welcomed my first night by snowfall."

Guest post by Sara Darychuk, MOSH Lab PhD Candidate

Short Course: Principles of Hydrology

MOSH Lab member Ali Bishop is in Kananaskis, Alberta for the Principles of Hydrology Short Course run by Dr. John Pomeroy. She sends this update:

All going well in hydrology bootcamp, update below:

Another exciting day at the Kananaskis Field Station. We may have 20 cm of fresh snow above the ground, but today our focus was on below ground processes - specifically groundwater. It was a jam-packed day covering everything you would learn in a 2nd year hydrogeology course. Dr. Ed Cey was an excellent lecturer and communicated these concepts by providing real world examples, such as the 2013 Calgary flood, as well as breaking out his spectacular groundwater simulation fish tank!

New Paper: K Mukherjee et al. on mass balances at Peyto and Place glaciers

First, the citation:

Mukherjee, K., Menounos, B., Shea, J., Mortezapour, M., Ednie, M., & Demuth, M. (2022). Evaluation of surface mass-balance records using geodetic data and physically-based modelling, Place and Peyto glaciers, western Canada. Journal of Glaciology, 1-18. https://dx.doi.org/10.1017/jog.2022.83

Peyto Glacier in 2017 (photo: J. Shea)

Now, the summary:

Glacier mass balance observations are challenging to make, and even more challenging to maintain over long periods on shoestring budgets. The Government of Canada has made continuous observations of glacier mass change at two main sites in western Canada: Peyto Glacier (Canadian Rockies) and Place Glacier (Southern Coast Mountains) since 1965 - some of our co-authors have worked for decades to continue these observations. The records for these sites are invaluable for documenting long-term glacier change in the region, but they are also used to calibrate a wide range of models (mass balance, hydrology, weather/climate).

Our recent paper (led by Dr. Kriti Mukherjee) takes a closer look at the last 40 years of these records, and finds that (1) we can use geodetic mass change information (from air photos and LiDAR) to calibrate a mass balance model but we need to include ice dynamics if we want to model mass balance properly, and (2) there are some potential errors in the reported mass balances at each of the sites. Without the original field notebooks, it is difficult to pinpoint the exact source of the error, but we can flag certain periods in the records where the modelled and observed annual mass balances strongly disagree:

• Place Glacier, 1987 - 1993

• Peyto Glacier, 2001 - 2006

Caveat emptor, as they say!

Glaciological, modelled, and geodetic mass balances at Peyto and Place glaciers.

New Paper! Basin Hypsometry and Snowpack Responses to Climate Change

Intuitively, the shape of a mountain basin - and in particular how much area is occupied at different elevation bands - would be an important factor in how sensitive the basin is to climatic change. Top-heavy basins, with a higher proportion of area at high elevations, would be less sensitive to warming. Bottom-heavy basins, with more area at lower elevations, would probably be more sensitive. Exactly how sensitive, and what role does the geometry play? Well, those are the questions we try to answer in our new paper published in Frontiers last month.

It turns out that mountain basins in western Canada fall into 3 different shape categories when we look at their hypsometry (how much area there is at different elevations in the basin; Figure 1 below). And if we strip away information about slope and aspect, give them all the exact same climate inputs and elevation range, and only keep the differences in hypsometry in play, the basins respond very differently to simple increases in temperature using a robust, physics-based snow melt model. In this paper we use CRHM.

Figure 1: Fifty shades of basin hypsometry, clustered into three main categories: top-heavy (purple), middle (green), and bottom heavy (yellow).

Snow Accumulation

For snow accumulation, we assume a linear gradient of snow accumulation: in general, snow accumulation increases with elevation. As the climate warms, we expect this gradient to change: snow at the highest elevations will remain roughly the same, and accumulations will decrease at lower elevations. We simulate this in the model by shifting the elevation where all precipitation falls as rain upwards in response to warming, and we recalculate the gradient (Figure 1). Bottom-heavy basins store more of their snow at the lower elevations, so this leads to big decreases in the total snow volume: on average the bottom-heavy basins could see a 15% decrease for +2C scenarios, and 35% decrease for +4C scenarios. Top-heavy basins could see decreases of 8% and 23% for the same scenarios.

Figure 2: Snowpack volume responses for the three different hypsometry types and two warming scenarios.

Snow Melt

To model snowmelt, all basins are assumed to have the same climate: we use a synthetic temperature curve, some noise, and standard lapse rates to generate reasonable estimates of observed temperatures in the region. As our model is process-based we also need wind speed, but we simplify again here (because wind is notoriously difficult to measure and model) and use a constant wind speed. Shortwave and longwave radiation are calculated using CRHM routines and assumed values for cloudiness and atmospheric transmissivity. We are working with a real model and throwing some reasonable numbers at it, as we are focused here on the geometry of the basin itself.

One would reasonably expect snow melt to start earlier with warmer temperatures, and our snowmelt model shows that this is true. Though your mileage may vary depending on the basin geometry: bottom-heavy basins show melt onset beginning nearly a month early with +4C of warming. Somewhat counterintuitively, we also show that average melt rates decline under warming scenarios (Figure 3)! While melt starts earlier at the lower elevations of the basins, the duration of the melt season advances less rapidly - resulting in lower average melt rates. This also agrees with the “slower slowmelt in a warming world” hypothesis put forward Musselman et al. (2017): as solar radiation is the main driver of snowmelt, an earlier start to the melt season means lower energy is available for melt.

Figure 3: Average daily snowmelt rates for three different basin geometries (purple = top-heavy, yellow = bottom-heavy) and temperature increases.

While slower melt rates might be seen as a positive from a flood forecasting perspective, changes in the timing and magnitude of snowmelt-driven streamflow and the loss of snowpack volumes will result in substantial changes in local hydrology. And again, the basin hypsometry plays an important role: bottom-heavy basins show the greatest sensitivity to a warming climate, and top-heavy basins are less sensitive. The loss of mid-summer snowmelt will create challenges for ecosystems and water managers alike, and increased monitoring of snowpacks at all elevations, through stations and remote sensing observations, will help us adapt to a future that is already here.

Oh: and where do these bottom-heavy basins tend to be located? On the eastern slopes of the Canadian Rockies. The water towers of the prairies.

Figure 4: Study area map, with the location of top-heavy (purple), intermediate (green), and bottom-heavy (yellow) mountain basins.

If you want to play with the code and data used to generate all the figures in the paper, point thy browser in this direction: https://doi.org/10.20383/102.0318.

JMS \m/

Field Work: Alpine Ground Temperature Observations

[guest post by Kevin Ostapowich, MSc candidate]

Alpine snowpacks are a critical source of streamflow in the mountains. Snowmelt is triggered by energy delivered to the snowpack. Most of that energy comes from the atmosphere: sun, warm air, wind. But energy is also delivered to the base of the snowpack, from the heat stored in the ground. Since we don’t know much about this ground heat component and how it relates to winter snowpack variability, I conducted field work this summer with Dr. Joseph Shea to start an alpine ground temperature measurement program as part of my Master’s research.

On August 9th, 2020 we flew into Conrad Glacier, in the Purcell Mountains of B.C. The weather was exceptional on the morning of the flight and, surprisingly, remained so for the duration of the field work.  We spent the following 7 days camped at approximately 2350 m above sea level beside the glacier. Our goal while in the field was to install two meteorological stations and four measurement transects. At each site on a transect, we installed temperature sensors in the bedrock, at the surface, and 10 cm above the surface to measure temperature gradients between the ground and the snowpack during winter.

After establishing camp on August 9th, we installed the first meteorological station. On August 10th and 11th we installed temperature sensors along Transect 1, up to a small summit.  We then turned our attention to accessing the lower sites: Transect 2 and Meteorological Station 2.  Access was a little challenging – to get over the cliff band we roped up, installed a rappel/belay station, and did some low-level alpine climbing. The goats were not impressed.  

While we were able to safely navigate the route to the lower sites, and return the same day, the trip was long and tough. We made the decision to not risk getting stranded below by rapidly changing mountain weather, and we abandoned the lower study area after a team discussion.   

Our subsequent days (August 13-16) consisted of developing new transects to install the ground temperature sensors along and installing ablation stakes for the Conrad Glacier mass balance program, at the request of Dr. Brian Menounos and Dr. Ben Pelto.

In the end we installed three revised transects.  Our revised Transect 2 ascends the west side of the valley, parallel to Conrad Glacier and terminates near the col at the head of the valley.  Revised Transect 3, aligned perpendicular to glacier flow, begins right at the edge of the glacier and ascends towards the camp location.  Revised Transect 4 consists was placed on Conrad Glacier and the west side of the nunatak.  The second map in the slideshow above shows the final transects.

This trip into the high alpine of the Purcell Mountains was a success for numerous reasons.  Not only did we achieve our goal of installing four ground temperature transects while safely managing the terrain, we also experienced phenomenal weather in the mountains for seven straight days.  Although we did not install the second weather station or the lower elevation transect, the majority of the instrument installation went off without a hitch. 

The biggest success of this endeavor was undoubtedly reigniting my passion for the mountains and the processes that are constantly shaping the planet.  And when I say ‘reigniting’, I really mean throwing a five-gallon pail of gas onto an already raging bonfire.  These trips are one of the reasons why we have both chosen this field of study. Our passion for these locales fuels the ongoing research and contributions to the greater scientific community that ultimately increase our understanding of the natural world around us.

These high alpine basins are incredibly important in contributing to the quality of our daily life and yet are not as well understood as lower elevational watersheds or given the attention they necessarily deserve.  It is my goal and hope that this research, and research like it, will be able to fulfill gaps in our knowledge to better understand the processes that shape our everchanging world.

Sentinel-1: Putting Snow Melt on the Radar

(guest post by S.E. Darychuk) Flooding is a natural hazard that effects thousands of British Columbians - the provincial government spent over $162 million in flood response in 2018 alone. A heavy snowpack often contributes to flooding, as seasonal snow melt drives the flow of the majority of river systems in the province. Due to the heavy influence of snow melt on flow, snowpack monitoring is crucial for water management and flood prediction. Remotely sensed data can be a powerful tool for snow mapping as it is able to capture the entire extent of the snowpack. Recent advances in remote sensing technologies have been described as revolutionary, and researchers have greater access to data, and the ability to process that data, than ever before.

Sentinel-1, a satellite constellation that came into full effect in 2016, is one recent advancement that has offered new opportunities to monitor snow. Sentinel-1 is a radar satellite, an active remote sensing product that uses the return strength of a signal sent from the satellite to measure earth surface properties. As an active product, radar data offers some distinct advantages over traditional optical images, including the ability to detect water within a snowpack and the ability to see through clouds that often obscure the mountains. The presence of liquid water is important because it indicates that the snowpack is ready to melt.

In the animation below, I use radar imagery to detect wet snow for the upper Fraser River watershed. I combined this with maps of total snow cover created from optical imagery, which uses visible light reflected from the surface. This animation uses hundreds of Sentinel-1 radar images and optical satellite images to map the monthly progression of wet and dry snow for the 2019 melt season. Dry snow covers the entire basin in January and February, but wet snow starts to appear at lower elevations in March and April. By May, dry snow is found only at the higher elevations of the basin.

While these animations are still a work in progress, I am particularly interested in how these types of datasets could be used in flood prediction. Recent work with Sentinel-1 images has shown a strong link between radar data and the stages of snow melt. Future research will explore this link in mountain watersheds to see if radar data can be used in B.C. to improve snow melt and flood risk forecasts. \m/

Snow Melt Mania

Its that time of the year again: when mountain snowpacks turn the corner from accumulation to melt!

The early April snow survey and water supply forecast from the British Columbia River Forecast Centre showed snowpacks well above normal in the upper Fraser Basin. And as historical data from real-time monitoring networks show, once snow melt starts it tends to go rapidly.

This post examines the most recent snowpack data for BC and points to regions where the accumulation/melt corner has (or has not been turned). But first: a bit of background.

Background

From a river forecasting perspective, its not the depth of snow that’s important, its the amount of water contained in the snowpack. This depth of water is called the snow water equivalence, or SWE. SWE is a function of depth and snow density - while depth is relatively easy to measure automatically, density is not. So automatic measurements of SWE are made by snow pillows which are essentially giant antifreeze-filled bladders that measure the total weight of the snowpack.

I am using data compiled by the fine folks at the BC government, who make all this data accessible. Current year SWE (hourly): http://env.gov.bc.ca/wsd/data_searches/snow/asws/data/SW.csv… Historical SWE (daily): https://pub.data.gov.bc.ca/datasets/5e7acd31-b242-4f09-8a64-000af872d68f/daily_asp_archive.csv

Upper Fraser Basin

So to the current SWE situtation - snowpacks in the Upper Fraser River basin are definitely above normal. And most aren’t showing signs of turning the corner yet. Here’s the first plot, showing the data for the Yellowhead Station (elevation: 1860 m). The plot on the left shows the current year’s daily SWE observations (orange), along with the historical average for each day of the year (solid blue) and the 25th and 75th percentiles (dashed blue). And for the record I used a smoother on the mean values and percentiles to reduce some noise…

The plot on the right shows the daily change in SWE. Again, orange dots are the current year, and blue is the average change (based on the blue line shown on the left). Positive values show accumulation, and negative values show melt/runoff. Note the steepness of the curve: when mountain snowpacks start to go, they really go. But the Yellowhead station isn’t quite there yet - based on the historical observations though I’d expect it to happen in the next few weeks.

Some more stations below: McBride (1611 m), Barkerville (1520 m) and Revolution Creek (1690 m).

It looks like only the Barkerville station, the lowest elevation station of the four, has turned the corner so far this year. The snowpacks at all sites are way above normal for this time of year: and nearly double at Revolution Creek! With the clear skies and warm weather we’ve had this weekend I would guess we will start to see these stations turn the corner shortly. And while these sites represent a very limited sampling of a large area, its not a surprise that the professional river forecasters are concerned for the upcoming snowmelt flood season.

Future work will use time-series analysis to examine current and historical rates of snowpack depletion and the date of the turning point to see if there are any signs of climate change on these mountain snowpacks.

JMS

\m/

Snow pillow locations in the Upper Fraser River Basin.