Skip to content

Cookies 🍪

This site uses cookies that need consent.

Learn more

Zur Powderguide-Startseite Zur Powderguide-Startseite
news

World of Science | How much did it really snow?

"Mega deepe Faceshots Bro!" is not a meteorological metric.

by Lea Hartl 01/16/2017
Ever since we got the SNOWGRID maps on PowderGuide, we've been getting questions about their accuracy: Why is there much more snow here in my favorite forest slope than on the map? SNOWGRID says there's two meters of snow in Ticino, but my buddy says there's nothing at all? Below we provide some insights into the complex world of precipitation measurement, the efforts of ZAMG to record snow as accurately as possible and why it still doesn't always work.

How do you measure how much it has snowed? Apart from measurements based on the skier's height (boot top, knee deep, hip deep, chest deep, etc.) and the number of Instagram hashtags used (#pow #powder #epic #deep #deepestdayever #chestdeep #faceshots, etc.), there are several other established methods of measuring the depth of snow or fresh snow. The simplest and often the best are hand measurements, where someone uses a meter gauge to check how much snow there is. For the total snow depth, there are permanently installed measuring poles with a scale that can be read off.

So-called snow boards are laid out to measure only the amount of fresh snow. These are actually boards that are left to snow. After the snowfall or at certain times, the amount that has accumulated on the board is measured. The board is then cleaned and the process is repeated. In the high mountains in particular, it is not always possible for someone to check and lay out boards everywhere, which is why precipitation and snow depth are also measured at automatic weather stations. There are several variations of heated precipitation gauges that collect and melt snow and weigh the water periodically. This gives the snow water equivalent (SWE) - a variable that, unlike the depth of new snow, does not depend on snow density and can be compared with summer precipitation. If the snow were not melted, the precipitation gauges would be full after the first significant snowfall and the new snow measurement would no longer work.

Automatic snow depth measurement usually works using ultrasonic or laser sensors, which measure the distance between the sensor suspended from a pole and the ground (or snow surface).

For hydrological applications in particular, there are even more specialized devices, such as so-called snow pillows, which determine the SWE of the total snow cover based on its weight. However, these are only found very occasionally at certain stations. In the mountains, there are not automatic weather stations everywhere or even someone on site. Precipitation radar is therefore also used to measure precipitation over large areas. The radar signal shows where it is raining or snowing and the approximate amounts can also be derived - but for this you need a weather station at least somewhere nearby.

Whoever measures, measures crap

In theory, some methods sound quite simple, but in practice it is often difficult: solid precipitation, i.e. primarily snow, does not always just fall into the bucket provided for this purpose (= precipitation gauge), even if it is equipped with a wind trap. This is a major problem, especially in the mountains, and leads to a systematic underestimation of the amount of precipitation at the pluviometers - the clever word for precipitation gauges. The higher the wind speed, the less actually falls into the bucket. The uncertainties for snowfall are up to 80%. In Switzerland and southern Germany, winter precipitation in the mountains is underestimated by up to 50% on average. As this is a known problem, the data is usually corrected by a corresponding factor that depends on the wind speed and air temperature (the colder and lighter the flakes and the stronger the wind, the less in the bucket). However, this does not always work well, as the exact wind speed in complex terrain is not known and so considerable errors can remain despite correction.

Ultrasonic measurements of the total snow depth are relatively reliable, but also not perfect. The distance measurement is temperature-dependent, among other things. Depending on the air temperature, corrections must also be made here. Laser measurements are more accurate and not temperature-dependent. Of course, neither ultrasound nor lasers can prevent the wind from blowing the snow away.

The actually practical radar has a limited field of vision. It only recognizes what a person standing next to it would see. The radar is usually positioned on mountain peaks. However, it cannot see into deep valleys or behind the mountains that block its view. In addition to underestimation due to shadowing effects on the topography (the "radar shadow" is the area behind the mountain that the radar cannot see), false echoes (e.g. prohibited use of reserved frequencies) can also lead to local overestimates.

Measure better: pluSnow

Of course, there are intensive efforts to better record winter precipitation. The underestimation of precipitation by pluviometers continues as an error in many different applications, for example in the INCA model of ZAMG, which SNOWGRID accesses. The pluSnow project, which is being carried out at the Innsbruck Institute for Interdisciplinary Mountain Research (IGF) in cooperation with ZAMG, is currently looking for ways to automatically record new snow depths as accurately as possible in order to improve precipitation measurement. It is hoped that this will lead to better pluviometer corrections. This also requires precise data on the density of the fresh snow, as the corrections require not only the snow depth, but also the water equivalent of the fresh snow. Both could ultimately further improve the accuracy of analysis applications such as INCA and SNOWGRID.

The pluSnow project initially investigated the extent to which pluviometers actually underestimate winter precipitation in the Austrian Alps. Such studies had previously only been carried out for Germany and Switzerland. It turns out that an average error of 20% can also be expected in Austria, whereby the measurements at very high and wind-exposed stations are more affected than those at more sheltered locations, according to project manager Kay Helfricht from the IGF.

If the melted precipitation in the pluviometer is now compared with a very accurate measurement of the height of fresh snow at the same location, the former could be corrected. Accurate new snow depths can be derived relatively well from the changes in total snow depths measured by ultrasound and laser sensors.

Supposing you have now calculated the amount of new snow from a fairly accurate change in total snow depth, then you would only have to convert the centimetres of new snow into millimetres of water equivalent (melted snow in the pluviometer) and the comparison would be complete. But how exactly did the conversion work again? One millimeter of water is one centimeter of snow or something like that?

This "something like that" covers quite a wide range, which can make the difference between champagne powder and nasty cardboard snow. The density of fresh snow is subject to temporal and spatial fluctuations. At pluSnow, extensive data analysis is now being used to try to understand the reasons for these density fluctuations and to develop better conversion formulas. Every skier intuitively assumes that temperature and wind influence the density of fresh snow. Although this dependency is reflected in the data, it is difficult to quantify and varies from station to station. So there is still a lot of research and improvement to be done here.

From point measurements to snow depth maps

All of the difficulties mentioned relate to measuring locations where there are actually measurements of some kind, be it manual measurements, pluviometers, ultrasound or radar. However, anyone who frequently undertakes ski tours will hardly ever come across weather stations, radar dishes or meteorologists digging in the snow. So there is not measurement data from everywhere. The measurement locations are also not evenly distributed and, for logistical reasons, are more often located in the valley than in the high mountains, which are difficult to access. There is also the question of which snow depth measurement is representative for a location - the one on the wind-exposed molehill or the one from the ditch two meters away?

If you want a map, you have to convert the available data into an areal distribution using a model and the algorithms built into it. The more data there is for a particular area, the better it works. The limiting factor in the conversion to the area is the terrain. Due to limited computing power alone, it is impossible to resolve small-scale terrain structures in the model in the same way as we perceive them when skiing.

SNOWGRID has a model grid point every 100m for which a value is calculated. In combination with the high-resolution terrain model that SNOWGRID uses, this is comparatively very accurate. However, if we only ever considered the terrain as the average of a 100m x 100m square when skiing, we would quickly fail on the mountain. So you have to keep the different scales in mind: The model does not see the situation on individual slopes, which is one of the most important things for the skier. It also does not see local wind effects. Snow redistribution due to wind is sometimes more important for skiing than the exact amount of precipitation, but from a meteorological point of view these are two completely different issues. Wind drift is very complex to model and has not yet been implemented in SNOWGRID. So even when converting point measurements to the area, it's not all that simple.

And what about SNOWGRID and the snow on my local mountain?

Precipitation amounts are also the biggest source of uncertainty for SNOWGRID: "On average, there is generally an underestimation of the snow cover in the high Alps. Locally, there can also be significant overestimates, mainly due to the radar component of the model. These are all problems that are largely difficult to get under control," says Marc Olefs from ZAMG. This is to be remedied by subsequent corrections using additional measurement data and the results of pluSnow, i.e. a better correction of the winter precipitation measurement error. However, this requires a very intelligent, fast and fully automatic test routine. According to Olefs, this is exactly what ZAMG is currently working on in order to further improve the SNOWGRID result.

SNOWGRID is not perfect, but given the difficulties that need to be overcome, it is already pretty good. Olefs points out that there are very strong vertical and horizontal gradients in snow depth and snow distribution in some areas. It is therefore important to always compare the snow conditions at an exact location with the corresponding model value at that location, and not just the regional skier's impression.

There are certainly situations in which skiing is hardly possible in a region due to very uneven snow distribution, but at the same time there is already a meter of snow at a nearby weather station. On the other hand, a station location could be completely blown away, but skiing is possible in the surrounding area. This then leads to discrepancies in your own perception and the color in the SNOWGRID map. While cumulative errors are carried over the entire winter in the total snow depth maps, this is not the case with the 24-hour snow difference map. Accordingly, the difference maps are significantly more accurate than those of the total snow depth, apart from areas in which systematic errors occur due to e.g. radar shadowing.

As with all other weather maps, forecasts and station data, you should always keep in mind where the data comes from, how it is generated and what uncertainties and system-related limits there are with SNOWGRID. Unusual values should be questioned and compared with values from the surrounding area (webcams, measuring stations), but you should also question your own perception from time to time. It is also very helpful if you can remember from previous winters roughly how much snow an area needs to be worthwhile for skiing and what is then displayed at the corresponding measuring stations. Members of the PG editorial team claim, for example, that you can only ski properly on the Krippenstein if the station indicates a snow depth of at least 2m. For alpine meadow wading, however, much less is sufficient.

Conclusion: Weather stations and models are somehow only human.

Literature: Helfricht K., Koch R., Hartl L., Olefs M. 2016. Potential and challenges of an extensive operational use of high accuracy optical snow depth sensors to minimize solid precipitation undercatch. Proceedings, International Snow Science Workshop, Breckenridge, Colorado, 2016.

This article has been automatically translated by DeepL with subsequent editing. If you notice any spelling or grammatical errors or if the translation has lost its meaning, please write an e-mail to the editors.

Show original (German)

Related articles

Comments