A meteorological and blowing snow dataset ( 2000-2016 ) from a high-altitude alpine site ( Col du Lac Blanc , France , 2 720 m a . s . l . )

A meteorological and blowing snow dataset issued from the high-altitude experimental site of Col du Lac Blanc (2720 m altitude, Grandes Rousses mountain range, France) is presented and detailed in this paper. Emphasis is placed on data 15 relevant to the observations and modelling of wind-induced snow transport in alpine terrain. This process strongly influences the spatial distribution of snow cover in mountainous terrain with consequences for snowpack, hydrological and avalanche hazard forecasting. In-situ data consist of wind (speed and direction), snow depth, and air temperature measurements (recorded at four automatic weather stations), a database of blowing snow occurrence and measurements of blowing snow fluxes obtained from a vertical profile of Snow Particle Counters. Observations data span the period from December 1 st to 20 March 31 st for each winter season from 2000-2001 until 2015-2016. The time resolution varies from 15 min. at the beginning of the period to 10 min. for the last years. Atmospheric data from a local meteorological reanalysis (SAFRAN) are also provided from 1 August 2000 to 1 August 2016. A Digital Elevation Model (DEM) of the study area (1,5 km2) at 20-cm resolution is also provided in RGF 93 Lambert 93 coordinates This dataset has been used in the past to develop and evaluate physical parameterizations and numerical models of blowing and drifting snow in alpine terrain. Col du Lac Blanc is also a 25 target site to evaluate meteorological and climate models in alpine terrain. It belongs to the Cryobs-Clim observatory (the CRYosphere, an OBServatory of the CLIMate) which is a part of the national research infrastructure OZCAR (Critical Zone Observatories – Application and Research) (Gaillardet et al., 2018). The data are placed on the repository of the OSUG datacenter doi:10.17178/CRYOBSCLIM.CLB.all . Earth Syst. Sci. Data Discuss., https://doi.org/10.5194/essd-2018-74 O pe n A cc es s Earth System Science Data D icu ssio n s Manuscript under review for journal Earth Syst. Sci. Data Discussion started: 25 June 2018 c © Author(s) 2018. CC BY 4.0 License.


Introduction
Wind-induced snow transport strongly influences the temporal and spatial distribution of the snow cover in mountainous areas (e.g.Mott et al, 2010, Vionnet et al. 2014).It occurs throughout the winter in a succession of blowing snow events with and without concurrent snowfall.The redistribution of snow through saltation and turbulent suspension results from complex interactions between the local topography, the near-surface meteorological conditions and the surface of the snowpack (e.g.Pomeroy andGray, 1995, Naaim-Bouvet et al., 2010).This spatial variability has consequences on the snowpack stability and influences the danger of avalanches as cornices and wind slabs are formed during blowing snow events.It has also hydrological consequences since the melt response of alpine catchment depends on the snow spatial distribution at peak accumulation (Egli et al., 2012, Revuelto et al., 2016).Observations of blowing snow and associated meteorological and snowpack parameters are therefore crucial to better understand the complex snowpack/atmosphere interactions during blowing snow events and to develop and evaluate numerical models used in support of avalanche hazard and hydrological forecasting in alpine terrain.
Since the beginning of the 90', the Snow Research Centre (Météo-France -CNRS) and the ETNA unit (IRSTEA, Univ. Grenoble Alpes) have joined their efforts to investigate in-situ the effects of wind on snowpack evolution, during or after snowfall.A high-altitude experimental site has been set up at the Col du Lac Blanc, a north-south oriented pass, located at 2720 m altitude (45.13°N, 6.12°E) in the Grandes Rousses mountain range, France.Recent studies have focused on fine scale processes during blowing snow events (Naaim Bouvet et al., 2010, 2011, 2013;Nishimura et al., 2014;Schön et al., 2015), intercompraison of blowing snow sensors (Cierco et al., 2007;Trouvilliez et al., 2015) and the development and evaluation of blowing snow models (Durand et al., 2005;Vionnet et al., 2013Vionnet et al., , 2014Vionnet et al., , 2017Vionnet et al., , 2018)).In this paper, we present a unique meteorological and blowing snow dataset for each winter of the period 2000-2016.Meteorological data are available from 4 automatic weather stations (AWS) surrounding the experimental site.Blowing snow data stem from two sources: (i) a database of blowing snow occurrence and (ii) blowing snow fluxes derived from Snow Particles Counters over the last 6 years.The paper is organized as follows.Section 2 describes the experimental site and the sensors used at each automatic weather station.The methods applied to derive blowing snow data are also mentioned.Then, Section 3 presents an overview of the meteorological and blowing snow dataset over the last seasons.Finally, Section 4 details the data availability.

Site description
The Col du Lac Blanc (CLB) experimental site, located at 2720 m altitude in the Grandes Rousses mountain range (45.13°N, 6.12°E, Fig. 1), has been operated by Météo France and Irstea with punctual collaboration with other academic partners, since 1988.This experimental site can be assimilated to a natural wind tunnel due to its orientation and the specific configuration of the surrounding summits.Indeed, the Grandes Rousses range on the eastern side and the "Dôme des Petites Rousses" summit on the western side channel the atmospheric flow according to a North-South axis (Fig 1).This characteristic of the site is particularly useful for studies on the effects of wind on snow redistribution.Snow is typically present on the ground around the site (Fig. 2) from late October to early June with a strong inter-annual variability.The underlying ground is covered by bare and rocky soil, typical of high altitude alpine regions.Patches of low alpine grass are also present around the site.In wintertime, the site is characterized by a strong spatial variability of the snowpack due to intense snow redistribution during blowing snow events.In particular, two 10-m slope breaks on the northern and southern side of the pass accumulate large amount of snow during these events depending on the main wind direction (Vionnet et al., 2014, Schön et al, 2015, 2018).
The CLB experimental site consists of four automatic stations located around the pass (Fig. 1 and 2).The exact position and elevation of the stations is given in Table 1.AWS Lac Blanc, Col and Muzelle are located at approximately the same elevation next to the pass where the atmospheric flow is strongly channeled by the surrounding topography.AWS Dome lies at higher elevation on the top of the Dôme des Petites Rousses.At this station, atmospheric conditions are less influenced by the topography of the pass and closer to the synoptic conditions.Two wooden shelters are also installed at the pass (Fig. 2).
They host the data acquisition system, the equipment storage and the living facilities.They are located on the eastern side of the pass, aside from the main wind direction, so that they have a minimum impact on measured snow and meteorological conditions at AWS Lac Blanc, Col and Muzelle.The meteorological, snow depth and blowing snow data collected at these stations are presented in the next two sections.

Meteorological and snow depth data
Table 2 provides an overview of the meteorological and snow parameters measured around CLB, with the corresponding instrument type and height.Each AWS located around CLB measures wind speed and direction at a time step of 15 min or 10 min depending on the station and/or the period.Additionally, measurements of air temperature are available at three AWS.Finally, three AWS situated around the pass measure snow depth using ultrasonic sensors (at AWS Lac Blanc and Muzelle) and laser sensor (at AWS Col).These data have undergone a careful manual quality check.They are available between 1 st December and 31 st March of each winter.This period has been selected since it corresponds to the main period during which most of blowing snow events occur at CLB (Vionnet et al., 2013)

Wind speed and direction
Wind speed and direction are measured using non-heated anemometers (Young) at AWS Lac Blanc, Muzelle and Dome.The starting threshold of the wind velocity for the Young sensor is 1 m.s -1 .The maximum, minimum and mean wind velocity of the time step are recorded.At AWS Col, wind speed and direction are measured using a heated ultrasonic anemometer from Metek Corporation (USA1).For high drifting snow fluxes, particles hitting the transmission/reception cell can disturb the measurement process The recorded wind direction is the most frequent for the time step.Due to snow accumulation at the bottom of the stations, the measurement height is changing during the course of the winter.Table 2 gives the height of the wind sensors over snow-free ground for each station.Snow depth measurements at AWS Lac Blanc, Muzelle and Col (see Sect. 2.2.3) can be used to retrieve the height of the wind sensor above the snow surface as in Vionnet et al. (2013).

Air temperature
Air temperature is measured with PT100 wires at AWS Lac Blanc, Muzelle and Dome.The sensors are placed in ventilated shelter and the uncertainty in the measurements lies within 0.1 K. Due to snow accumulation at the bottom of the stations, the measurement height is changing during the course of the winter.Table 2 gives the height of the temperature sensors over snow-free ground for each station.

Snow depth
Snow depth is measured using ultra-sound depth sensors at AWS Lac Blanc and Muzelle.The correction of the impact of air temperature on the velocity of sound in the atmosphere is carried out using the air temperature measurement previously described.The data are further manually corrected to remove outliers in the dataset, most often occurring during snowfall.
Ultra-sound depth sensors provide measurements accurate within 1 cm for a surface area of a few cm 2 on the ground.The overall accuracy of the automated snow depth record is thus on the order of 1 cm.
Snow depth is measured using a laser sensor (SHM30) at AWS Col.The sensor uses an optoelectronic distance measurement principle to achieve a specified measurement uncertainty of better than 5 mm.Divergence of the SHM30 laser beam amounts to 0.6 mrad, which implies that the beam diameter is up 11 mm in size at the measurement point.This sensor is mainly used to accurately determine the position of Snow Particle Counters above the snow layer during a blowing snow event.Moreover, this technology has the advantage of being less disturbed by blowing snow particles than the sonic sensor but its power consumption is higher.

Atmospheric parameters from a meteorological reanalysis
Atmospheric data from the SAFRAN meteorological analysis system (Durand et al. 1993)  Blanc that cannot be precisely measured at the site during snowfall under windy conditions due to strong gauge undercatch.
Therefore, SAFRAN precipitation is considered as the reference precipitation dataset at Col du Lac Blanc.On the other hand, SAFRAN tends to underestimate wind speed at CLB due to the influence of the surrounding topography which is not included in the conceptual representation of the topography in SAFRAN.It is recommended to replace SAFRAN wind speed and direction by the observations collected at CLB when running a land surface scheme at CLB in wintertime as described in Vionnet et al. (2013).

Blowing snow data
Blowing snow data stem from two sources: These two datasets are described below as well as comparison of the occurrence of blowing snow with the two methods.

Database of blowing occurrence
A database of blowing snow events at CLB was established for each winter of the period 2000-2016.It consists in an extension of the database presented in Vionnet et al. (2013).A blowing snow event is defined as a time-period when snow on the ground was transported in the atmosphere in saltation and in turbulent suspension.Such event may occur with concurrent snowfall.Therefore, the database makes the distinction between blowing snow events with and without concurrent snowfall.The identification of a blowing snow event in the database relied on an empirical method that required the combination of several datasets (wind, precipitation and snow depth) and their detailed analysis by an expert.This method is extensively described in Vionnet et al (2013) and its main characteristics are summarized here.
Periods of concurrent snowfall and ground snow transport were identified first.They are defined as periods with precipitating snow (total snowfall from SAFRAN reanalysis greater than 5 mm SWE over the period) with a 5 m-wind speed above the snow surface, U 5 , at AWS Lac Blanc higher than U 5t = 6 m.s -1 .This threshold wind speed followed the observations of Sato et al ( 2008) collected during their wind-tunnel experiments on the processes of fracture and accumulation of snowflakes.U 5 is obtained from wind speed and snow depth measured at AWS Lac Blanc using a standard log-law for the vertical profile of wind speed near the surface, as in Vionnet et al. (2013).Periods of ground snow transport were then identified at an hourly time step from an analysis of the recordings from the snow depth sensor.This indirect method was developed and tested over fifteen years of observations at Col du Lac Blanc and has been previously used in Earth Syst.Sci.Data Discuss., https://doi.org/10.5194/essd-2018-74 Open

Data from Snow Particle Counter
The Snow Particle Counter (SPC-S7, Niigata Electric) is an optical device (Sugiuria et al., 1998)  Where q d is the horizontal snow mass flux for the diameter d, n d is the number of drifting snow particles, S the sample area, t the sample period, and ρ p the density of the drifting snow particles (917 kg m −3 ).
Depending on the winter season, up to four Snow Particle Counters SPC-S7s were installed on a mast (Bellot et al., 2016).
The SPC-S7s could be raised manually when the snow depth increased and risked burying the sensors.Horizontal snow flux is highly dependent on height above the snow surface.Because of snowfall and blowing snow events the elevation of SPC above the snow varied substantially during the winter season preventing any direct comparison over time being made.That's why in the present database, the snow fluxes near the surface (that means the snow flux provided by the SPC closest to the snow surface) and the corresponding height of the sensor are provided.When available, snow fluxes at a higher position are used to standardize the horizontal snow flux and to estimate the mean horizontal snow flux at 1 m above the snow surface and vertically integrated over 1 m (between 0,2 and 1,2 m over the snow surface).The computation of these fluxes is described below.
According to the diffusion theory of snowdrift, it is possible to approximate averaged drift density [kg m −3 ] as a function of height and wind velocity (Radok, 1977, Gordon et al. 2009).If the average wind profile is approximated by a power law, the vertical distribution function for the snow flux μ (g.m −2 .s - ) is expressed as follow: where A is a calibration parameter and m the exponent which is independent of z; both are derived by regression from measured data (Trouvilliez et al., 2015, Bellot et al., 2016).From this, the mean horizontal snow flux at 1 m and vertically integrated over 1 m can be estimated.
Nevertheless, this typical profile is not always observed or cannot be determined in situations when: i) only one sensor provides data ii) snow flux at different heights are not significantly different (case of falling snow with low wind,…) iii) data are physically inconsistent (flux at the higher position is much greater than near the surface due for example to icing of the self-steering windvane).
Therefore the SPC data were processed as follows.First a filter was applied to the raw data to supress possible electronic noise: events with a particles flux smaller than 20 particles cm -2 during 10 minutes were discarded.Then, an algorithm was designed and used to categorize each 10-min profile of SPC data set into six different groups (specified in the database):  "Undetermined": only one sensor provides data making it impossible to do the vertical interpolation.
 "Power_law": the vertical distribution function can be expressed as a power law.
 "Inconsistent": Data are physically inconsistent. "No_flux": In this case no sensor of the SPC vertical profile detects more than 20 particles per cm² during ten minutes.
 "Maintenance": During sensor maintenance, data acquisition is stopped.
When possible ("Power law" and "Mean"), mean horizontal snow flux at 1 m and vertically integrated over 1 m (between 0,2 and 1,2 m) are determined.

Comparison of the two methods used for the determination of blowing snow periods
During wintertime, SPC provided a non-zero-value during 50 % of time (and an average quantity of snow transported between 0.2 and 1.2 m per linear meter and per 30 days of 6245 kg during the winter season over the period 2010-2016) whereas blowing snow occurrence reported in the empirical database is 12% during the same period.This call for detailed comments.
Periods without significant blowing snow fluxes should be excluded to take the analysis a few steps further.That's why a filter has been applied over the raw SPC data: events with a particles flux smaller than 120 particles cm -2 min -1 (Naaim-Bouvet et al., 2014) were considered as periods without drifting snow.Then similar data processing was applied with detection threshold parameters of 960 particles.cm - .min - to highlight the global trends.Table 3 compares the occurrences of blowing snow derived from the SPC data with the different threshold to the occurrences reported in the empirical database.influenced by the local topography (Fig. 1).The distribution of blowing snow fluxes (Fig. 5b) is quite consistent with the distribution of wind speed at the site (Fig. 5a).

Figure 1 :
Figure 1: Location of the Col du Lac Blanc experimental site seen at different scales.Map (c) shows the location of the four AWS surrounding the site and described inTable 1.

Figure 2 :
Figure 2: Overview of the experimental site at Col du Lac Blanc (2720 m altitude, Grandes Rousses mountain range, France).Insets show the detailed view of each AWS.AWS Dome lies outside the picture.See text for further details 90 on each AWS.
at the elevation of Col du Lac Blanc in the Grandes Rousses range are provided from August 1 st 2000 to July 31 st 2016.SAFRAN combines meteorological fields from the numerical weather prediction system ARPEGE with neighboring observation to get an estimation of Earth Syst.Sci.Data Discuss., https://doi.org/10.5194/essd-2018forjournal Earth Syst.Sci.Data Discussion started: 25 June 2018 c Author(s) 2018.CC BY 4.0 License.meteorological parameters in the French mountains.It is used operationally in support of avalanche hazard forecasting (Lafaysse et al., 2013).SAFRAN data at CLB are provided to get all the meteorological parameters required to run continuously a land surface model at CLB without the need to restart the system every winter.SAFRAN data includes 2-m wind speed, 2-m air temperature and humidity, incoming longwave and shortwave radiation and snowfall and rainfall amount at an hourly time step.In particular, SAFRAN provides a valuable estimation of the snowfall amount at Col du Lac (i) a database of blowing snow occurrence from winter 2000-2001 to winter 2015-2016 and (ii) blowing snow fluxes derived from Snow Particles Counters from winter 2010-2011 to winter 2015-2016.
review for journal Earth Syst.Sci.Data Discussion started: 25 June 2018 c Author(s) 2018.CC BY 4.0 License.Guyomarc'h and Mérindol (1998) and Vionnet et al (2013) to identify periods of snow transport.Positive values of the difference between the maximum and minimum snow depth recorded over an hour and associated large values of its standard deviation are characteristic of the presence of snow particles between the sensor and the surface of the snowpack.Snow particles in motion above the snowpack surface create indeed interference in the ultrasonic signal.The records from a webcam (installed in 2004) completed the analysis.Such an empirical method provides the start and end dates of each blowing snow event.Only events of duration longer than 4 h were recorded in the database.Overall, the database contains at an hourly time step information on the occurrence of blowing snow and the type of event (with or without snowfall).

Figure 3
Figure 3 shows an overview of the inter-annual variability of blowing snow occurrence reported in the database with a winter average ranging from 6.4% in 2002-2003 to 19.0 % in 2011/2012.Using this method, blowing snow occurred during 11.7 % of time at Col du Lac Blanc over the period 2000-2016.36.7 % of time blowing snow occurred with concurrent snow fall.These estimations are similar to those reported in Vionnet et al. (2013) for the period 2001-2011.

Figure 3 :
Figure 3: Percentage of time when blowing snow events (with and without snowfall) are recorded for each winter (December 1 st to March 31 st ) over the period 2000-2016 at Col du Lac Blanc experimental site.
detecting snow particles between 40 and 500 μm in mean diameter by their shadows on photodiode.The SPC-S7 has a self-steering wind vane and the sampling area, perpendicular to horizontal wind vector, is 50 mm 2 (2 mm × 25 mm).Assuming spherical snow particles, the horizontal snow mass flux can be calculated as follow: Earth Syst.Sci.Data Discuss., https://doi.org/10.5194/essd-2018forjournal Earth Syst.Sci.Data Discussion started: 25 June 2018 c Author(s) 2018.CC BY 4.0 License.
Sci. Data Discuss., https://doi.org/10.5194/essd-2018forjournal Earth Syst.Sci.Data Discussion started: 25 June 2018 c Author(s) 2018.CC BY 4.0 License. "Mean": snow fluxes at different heights are not significantly different, so that the mean horizontal snow flux at 1 m is estimated by the average of the measured flux.

Figure 4
Figure 4 shows an overview of the inter-annual variability of blowing snow occurrence and intensity derived from the SPC data.SPC provided a non-zero value during a percentage of time ranging from 43 % in 2010-2011 to 63 % in 2013/2014 corresponding to an average quantity of snow transported between 0.2 and 1.2 m per linear meter during 30 days ranging from 2547 kg in 2010-2011 to 10331 kg in 2013/2014.SPC provided a non-zero-value during 50% of time and an average quantity of snow transported between 0.2 and 1.2 m per linear meter and per 30 days of 6245 kg during the winter seasonover the period 2010-2016.Such percentage is quite different from those reported in the empirical database (Fig.3) but we have to keep in mind that SPC which detects each particle is able to identify trace precipitation which do not significantly contribute to blowing snow quantities.This point is discussed in details in the next paragraph.

Figure 4 :
Figure 4: Percentage of time when particles are detected for each winter (December 1 st to March 31 st ) over the period 2010-2016 at Col du Lac Blanc experimental site and the corresponding quantity of snow transported between 0.2 and 1.2 m per linear meter.Due to missing or invalid SPC data, the length of the time series varies from a winter season to another winter season preventing to easily study the inter-annual variability of blowing snow intensity.To overcome this, an average quantity of snow transported between 0.2 and 1.2 m per linear meter during 30 days has been calculated.Numbers in brackets indicates the percentage of valid data delivered by SPC during the considered winter season.
Figure5and 6 illustrate the meteorological and blowing snow conditions at the site over the last six snow seasons(2010-  2016).In particular, Figure5shows the strong control exerted by the local topography on the atmospheric flow at the pass.Indeed, the wind field at AWS Lac Blanc (Fig.5a) is characterized by a channelling along a North-South axis and an increase in wind speed compared to AWS Dome (Fig5c), located on the top of the Dôme des Petites Rousses and less Figure6shows an overview of snow depth, maximum wind speed and blowing snow fluxes measured at the site over the last six seasons.A strong inter-annual variability is found in terms of snow depth with a maximum snow depth reaching 4.01 m in 2012/2013 and only 2.84 m for winter 2013/2014.Events with a maximum wind speed above 15 m s -1 are frequently measured at the site.These windy events are generally associated with blowing snow events if erodible snow is present at the snow surface.The intensity of these events varies greatly (note the logarithmic scale on the graphs showing blowing snow fluxes).Additional and more detailed summary plots for each year of the present dataset are provided as Supplement to this article.

Figure 5 :
Figure 5: Overview of the period 2010-2016: (a) wind rose for the AWS Lac Blanc, (b) rose of blowing snow fluxes at AWS Col with snow flux vertically integrated over 1 m between 0.2 and 1.2 m and (c) wind rose for the AWS Dome.Yearly wind roses and roses of blowing snow fluxes are provided as Supplement online material.

Figure 6 :
Figure 6 : Overview of snow depth (AWS Lac Blanc), maximum wind speed (AWS Lac Blanc) and blowing snow fluxes measured by SPC vertically integrated between 0.2 and 1.2 m (AWS Col) from 2010 to 2016.More detailed yearly graphs are provided as Supplement online material for each snow season from 2000-2001 to 2015-2016.330

Table 2 : Overview of the sensors used at each station between 2000 and 2016 at Col du Lac Blanc, France.
*Height above snow-free ground