ATSUSHI SATO assures me that the flakes of snow billowing down from the ceiling of his laboratory are perfectly safe to eat. I鈥檓 a bit sceptical, until he grabs a gloveful of snow and takes a bite. There is such a high level of automation in this walk-in freezer/laboratory, he jokes that he and his staff might just as well be outside in the warm control room drinking tea. 鈥淭he best time to come here is summer,鈥� says Sato 鈥� who is bundled up in a red down jacket to keep out temperatures of around 鈭�10 掳C. 鈥淚t鈥檚 green and hot outside, but snowing in here.鈥�
At this time of year, winter sports enthusiasts are heading for the mountains. But it is also the prime time for avalanches, which kill dozens of people worldwide each year. Simulating mountain conditions in the lab may be our only hope of forecasting avalanche conditions reliably, preventing tragedies like the massive slide in Russia last September that killed nearly 100 people, or the two in British Columbia last season that killed 14 people, one a group of 7 students. Can we learn enough about snow to predict its behaviour on slopes?
If anyone can answer this question, it鈥檚 Sato. He directs the world鈥檚 largest snowmaking facility, called the Cryospheric Environment Simulator (CES) at the Nagaoka Institute of Snow and Ice Studies in Shinjo, Japan. Here, researchers mimic a variety of outdoor conditions to discover their effect on the stability of layers of fallen snow.
Advertisement
Unlike the snow machines that churn out mere grains of ice to top up the pistes at ski resorts, the CES makes low-density crystals that are closer to the true crystalline shape of natural snowflakes than any other manmade snow. About 8 metres above us, near the lab鈥檚 roof, cold air is being sucked toward slowly rotating horizontal cylinders wrapped in nylon-like belts with tiny holes in them, like thick stockings. Air goes straight through the belts into the cylinders, but moisture is caught on the outside as ice crystals. Each crystal grows as it collects more water vapour, until a moving horizontal screen cuts it from the belt and sends it floating down to a 3 by 5 metre catch basin, or table.
On the table are trained a range of instruments 鈥� cameras to watch the movement of the snow, lamps to simulate the effect of sunlight, sensors to detect the formation of ice crystals. 鈥淔or avalanches, the most important factor is weak layers of the snow cover, because it is at this weak layer where a fracture can occur, causing the snow to collapse. As with other materials, the fracture will happen at the weakest point in the layers,鈥� says Sato.
He and the Shinjo group are using their readings to test and refine the leading computer model in avalanche prediction. The 鈥淪nowPack Evolution Model鈥� was designed by Michael Lehning and his colleagues at the Swiss Federal Institute for Snow and Avalanche Research in Davos. 鈥淪nowPack has now been used operationally for four years and has had success in warning against hazardous conditions,鈥� says Lehning. The program takes meteorological inputs such as temperatures, quantities of snowfall and wind speeds, and predicts weakening of the snowpack that could lead to avalanches.
A snowy slope is stable when the bonding force between snow crystals is strong enough to counter the force caused by gravity鈥檚 differential pull on the upper layers of snow relative to the lower levels. Anything that increases this 鈥渟hear stress鈥�, such as the addition of new snow, spells trouble. 鈥淚f we have snowfall on the slope, shear stress will increase with the additional weight of the snow cover itself,鈥� says Sato. This effect is easy for avalanche forecasters to predict by watching the local weather reports. What is much trickier is predicting when the bonding force between snow layers will weaken, allowing the upper layers to detach from those beneath and slide catastrophically down the slope. 鈥淚magine two books stacked atop each other on a table. Apply horizontal force to the top book, and at some point it will move. In other words, the shear strength of the block of books was overcome,鈥� explains Sato. The same goes for snow layers: when the shear strength of any one of the layers is overcome, a fracture will occur.
Avalanche forecasters already know about one dangerous kind of snow, called depth hoar. Sometimes the snowflakes in a layer of snow deep in the snowpack metamorphose into tiny ice crystals that lack the lacy arms of snowflakes. Depth hoar has little bonding power and its formation increases the risk of the layers of snow above it breaking off and starting an avalanche. Many backcountry skiers have learned to distinguish this kind of snow, sometimes with the help of a little index card with pictures of crystals. On the hillside, they dig into the snow to try and identify dangerous layers before beginning to ski.
The formation of depth hoar is linked to a drop in surface temperatures. Vapour from the ground or warmer, deeper snow deposits onto snowflakes, forming large cup-shaped crystals that do not bond together well. When fresh snow falls, that layer is likely to give way. SnowPack can predict this, using data on the meteorological conditions that cause depth hoar, to give predictions of increased risk of an avalanche under those circumstances. The program appears to predict correctly the formation of depth hoar, and the associated avalanches. But there are other avalanches that SnowPack misses, and there are often discrepancies between modelled and actual snow conditions.
Baby it鈥檚 cold outside
Sato recently investigated what might be causing SnowPack鈥檚 mistakes. Using data collected from two wintry outposts in the Arctic Circle, one in Kevo, Finland, the other near Fairbanks, Alaska, Sato鈥檚 team has found that SnowPack does accurately predict the characteristics of the snow layers when dealing with snowfall in cold places, at least.
In cold settings, there is often a significant temperature gradient through the snowpack. The surface temperatures could be between 鈭�20 and 鈭�30 掳C or colder, while temperatures within the snow hover around freezing. Snow is, after all, a good thermal insulator. These conditions often lead to the formation of weak-bonding depth hoar. 鈥淒epth hoar is the most common feature of the snowpack in cold places,鈥� says Sato. SnowPack has so far had significant success in such areas, suggesting that the program is correctly identifying the conditions that cause depth hoar.
But Sato has discovered that for warmer places such as Japan, SnowPack gives poorer results. Wedged between the Sea of Japan and the Pacific, the mountains of Japan can get pounded with snowfall, especially in Sato鈥檚 home prefecture of Niigata. But the altitude is nothing like that in the Alps of Europe or the Rockies in the US, and temperatures are often mild, sometimes only just cold enough for snow to fall. 鈥淚n Japan we get lots of snow, but we have warm surface temperatures, so there is almost no temperature gradient in many places.鈥� Warmer, wetter air usually means wetter snow than the champagne powder that falls in super-cold climates or high elevations, and anyone who has ever tried to shovel a driveway full of wet snow knows it is more like cement than cotton. 鈥淲etness just means there is liquid water in the snow,鈥� says Sato, and this wetness can lead to the formation of a weak layer that is just as dangerous as depth hoar. If the wet snow cools again, a layer of coarse-grain ice crystals known in Japanese as zarame-yuki often develops. Each crystal of zarame-yuki is much bigger than one of depth hoar, but packed together they are equally slippery.
Other crystal types can also lead to hazardous conditions, such as flat crystals that form a slick sliding surface on the snow, or globe-shaped particles that don鈥檛 bond. But it is zarame-yuki that appears to be tripping up SnowPack the most. If the snow lab researchers can find a way to catalogue the conditions that cause zarame-yuki, SnowPack will become a much more valuable forecasting tool.
Sato thinks that the program does not always correctly model heat and moisture transport through the snowpack and the surface temperature of the snow. At the CES, thin wire sensors running through the snow feed temperature and humidity data back to the control room, while cameras eye the simulated slope for 鈥渃reep鈥� 鈥� a process in which accumulated snow changes shape over time, not unlike a glacier, although considerably faster. Putting both together, Sato and others can trace how the formation of different crystal types correlates with slippage of the snowpack, and what external conditions prompt the formation of more dangerous layers within the snow.
Although the lab snow is more realistic than that of many other simulators, it still does not quite match natural snow. The most significant difference is that the flakes only have one or two arms, instead of the six found on a natural snowflake. This is because the CES flakes form on a belt, so only one or two arms have space to grow before they are cut from the belt and fall to the ground. But Sato points out that the snowpack itself contains so many broken and damaged flakes that the immature nature of his flakes does not matter too much.
And for research purposes, the snow lab may well make a better environment than the pistes. The main advantage is its stable conditions. 鈥淎ir temperature, for example, is obviously a key variable for understanding snow conditions within the snowpack,鈥� says Sato. 鈥淚n nature, air temperature is changing all the time, so it鈥檚 difficult to do controlled research. But in the CES, we can set temperature flat for a day, or make it change hourly.鈥� With other variables kept constant, Sato can see how changes in air temperature influence the metamorphosis of snow crystals. He can do the same with humidity, slope and radiation, and measure how those factors affect the stability of the snow.
A key variable to watch that is crucial in the formation of zarame-yuki is the surface temperature of the snow, because zarame-yuki forms in warm conditions. But predicting surface temperature is complicated. Like many aspects of weather, the surface temperature of snow is influenced by so many factors 鈥� wind, elevation, solar radiation 鈥� that are constantly changing. In the CES, Sato uses a thermal imaging camera to study the extent to which solar radiation affects the snow layers, though the source of the 鈥渟olar鈥� radiation in this case is a movable, 2-by-4-metre array of overhead lamps. SnowPack currently only uses a handful of data points from sensors on mountains, so Sato expects the data from the lab to greatly increase the model鈥檚 accuracy.
Another way to beef up SnowPack鈥檚 forecasting abilities may be to use additional computer models to predict heating of the snowpack. The snow lab can then be used to test such models. Ed Adams, professor of civil engineering at Montana State University, is working on a program that predicts snow surface temperatures using a bundle of physical and meteorological inputs such as terrain, elevation, shadows, long and short-wave radiation and convection. Dubbed RadTherm, the program is funded by the state highway department with the aim of predicting the temperatures of roads in winter. Now Adams wants to apply the same model to the snowpack.
The result should be a new ability to predict the temperature gradient in the snowpack. RadTherm will give detailed predictions of surface temperature, enabling SnowPack to calculate the temperature in the snow more accurately.
The aim is one all-purpose model that forecasters can use in any snowy environment to keep people safe. 鈥淲e could never predict the exact time that a particular slope will have an avalanche,鈥� says Sato. 鈥淏ut what we can do is improve our ability to forecast the possibility of an avalanche occurrence.鈥�
