NUMERICAL EXPERIMENTS OF WIND TRANSPORT OVER A MOUNTAINOUSINSTRUMENTED SITE AT SMALL, MEDIUM AND LARGE SCALES

Florence Bouvet-Naaim1, Yves Durand2

, Mohamed Naaim1,

Gilbert Guyomarc'h2, Jean-Luc Michaux\ Laurent Merindol2

Cemagref, Snow avalanche engineering and torrent control research unit, Grenoble, FranceMeteo-France CEN (Snow Study Center), Grenoble, France

Abstract: For about ten winter seasons, the CEN and the CEMAGREF, two French laboratories involve(;in snow research, have performed intensive field measurement campaigns concerning various snowweather parameters during blowing snow events. The observation sites are located near the Alpe d'Hu~

ski resort at elevations between 2700 and 2800 m (Col du Lac Blanc).In order to determine the avalanche release hazard, it is necessary to take into account the spatialdistribution of snow transported by wind in avalanche starting zone. So we have to determine the effectorthe wind on the snow cover at the avalanche path scale.Therefore the numerical simulation of such phenomena, in the framework of the operational avalanchehazard forecast, also began some years ago. All numerical experiments, presented here, benefit from aIthese archived data and aim at representing the local wind and their effects on the distribution of snow.The main difference is their working scale which can vary from the current operational massif scale (- 20­30 km) to finer definitions varying between 1 m and 100 m.The originality of this project consists in the coupling of this different numerical simulations, using outputsfrom models at large scale as inputs for models at small scale.First results will be presented.Keywords: drifting snow, wind, experimental site, measurement, numerical model

1. INTRODUCTION

In the seventies, after a dramatic avalancheaccident killing more than thirty people, tworesearch centers were created in France:

the Snow Study Center (from Meteo-France)which has in charge the coordination of theoperational avalanche hazard forecast.The Snow avalanche Engineering Unit (fromdepartment of Research and Agriculture)which has in charge snow avalanche mappingand related engineering.

Initially the research in these two laboratorieswere essentially dedicated to studies of snowpack and avalanches dynamics.

Corresponding author address: FlorenceNaaim-Bouvet, Cemagref, Snow avalancheengineering and torrent control research unit, 2rue de la papeterie, BP 76, F-38402 Saint-Martin­d'Heres ; E-mail: florence.naaim@cemagreUr

302

These two teams progressively turned to blowingsnow study because of the significance of thisphenomenon producing cornices and creatingdangerous avalanche starting zones. During tenyears ago, they performed field measurementcampaigns dealing with blowing snow on acommon site in high mountain area. Therefore,they were able to test and validate their ownmodels. Briefly we can say that Meteo Francedeveloped numerical models at a large scale(massif scale) to forecast drifting snow events andCemagref developed physical and numericalmodels at small scale to prevent snowdriftformation.Meteo-France developed SAFRAN (estimation ofrelevant meteorological parameters for snowmodels including wind speed), CROCUS (snow­pack evolution model) at the massif scale (theworking scale is typically about 500 km2). MeteoFrance subsequently developed PROTEON, anapplication for forecasting blowing snow, using theparameters (snow types and wind speed)calculated from SAFRAN and CROCUS.The Cemagref team developed a numerical modelof wind in complex topography called ARIEl.,

which is dedicated to solve engineering problemsand not to forecast weather. Moreover, theystudied physical modeling of blowing snow inwind-tunnel to know the limits of this approach.Afterwards, they carried out a blowing snownumerical model in the framework of engineering,that means at a small scale (the working scaletypically r~nges betw~en 1 and 1,00 m). The aim ofresearch In the coming years IS the forecastingsnowdrift shape at the avalanche path scale bycoupling these different numerical simulations,using outputs from models at large scale as inputsfor models at small scale.The global approach and the first steps, validatedon a common experimental site, will be thereforepresented in this paper.

2. DESCRIPTION OF THE METHODOLOGY

The following described approach is typically adownscaling problem.The largest scales are treated by the operationalsuite composed of two main models SAFRAN andCROCUS. SAFRAN is an automaticmeteorological analysis which combines andspatializes all the available information, such asobservations or meteorological forecast fields, inorder to provide hourly the relevant parametersaffecting the snow pack evolution. CROCUS, fedby these analyzed data, simulates the differentphysical and mass processes into the snow coverand its complete stratigraphy.PROTEON combines this information into adiagnostic application for snowdrift occurrence(Guyomarc'h,1998).In the framework of those studies, these modelsare complemented by the model SYTRON whichevaluates hourly the amount of wind-transportedsnow. Coupled with SAFRAN, for the windcomputation, and with CROCUS, for the snowprofile, it allows to insert into the automatic suitedifferent effects due to the wind as erosion on thewindward aspect and a correspondingaccumulation on the leeward aspect. 'As the previous models are running at the scale ofthe massif, it was necessary to get a finer scaleH OOm ) approximation of the surface wind fieldtherefore to develop a small and flexible modelcoupled to SAFRAN outputs. The modelSA~VER, integrating the Euler primitive equations~~ ,I~entro~ic vertical surfaces, will be used toOItlahze different applications running at smallscale.SAFRAN and CROCUS have been presented insever~1 publications (Durand, 1993, Brun,1992)and Will not be explained in detail.

303

Large scale(Massif.seale)

INPUT

Small scale(Avalanche pathseale)

Figure 1 : Downscale of blowing snow numericalmodels

The NEMO model is dedicated to thedetermination of snowdrift shape. This numericalmodel is based on an eulerian approach in theframework of fluid-mechanics conservation lawswhere the snow is transported and diffused by theair flow. The saltation layer is considered as aboundary condition. The upstream boundaryconditions are the wind speed and the winddirection (calculated by SAMVER), the amount ofavailable snow (calculated by SYTRON). Theinput data depending on snow types (originatingfrom the modified snowpack model) are thethreshold velocity and the fall velocity. Thethreshold velocity could be calculated thanks toPROTEON but the fall velocity is still an unknownquantity and must be therefore estimated.Moreover the real topography is needed.This general approach corresponds to a short­term plan but is not yet completely achieved.Until now, some key steps, which will be explainedin this paper, have been performed and validatedon a common experimental site.

3. EXPERIMENTAL SITE

The experimental site (Michaux (2000) andGuyomarc'h (2000)) is located on the ski resort ofAlpe d' Huez near Grenoble, France,The large north-south oriented pass (Col du LacBlanc) has been dedicated to the study of blowingsnow in high mountainous regions for about tenyears. Indeed, on this site the high wind speedsand the snow cover are favorable to drifting snow.Several parameters (air temperature, winddirection and speed, snow depth, water equivalentof precipitation) are recorded every 15 minuteswith a scan rate of one second. Moreover, the

Figure 4 :snow pole.

Figure 3 : Network of fixed snow poles.

4. WIND NUMERICALSCALE: SAMVER

SAMVER is also a downscaling tool, aiming atinserting the effects of the orography into a Iscale surface wind field. Initialized by SAF(data provided according to the elevation andorientation), it provides a surface wind field overregular grid up to 100 m. The model is basedthe conservation of the vorticity and ~divergence over different vertical isentro~surfaces. The limitations of such a model arenumerous, but it has the advantage to run qUi.and is well suited to provide immediate informson all the studied areas.

304

speed and the direction of the wind are recordedin three other places located near the pass (seefigure 2). This standard measurement device issupplemented by 6 acoustic snowdrift sensors(Michaux,. 2000), consisting of a miniaturemicrophone located at the base of a 2 -m longaluminium pole. The pole is exposed to the snowparticle flux, and during blowing snow, part of theflux impacts on the pole (see figure 3). The soundproduced by these impacts is recorded as anelectrical signal. Until now, the sensor is notcompletely calibrated: the relation between signaland blowing snow mass flux is not yet given.Nevertheless, it makes it possible to determine ina reliable way the periods of beginning and end ofblowing snow events as well as the thresholdvelocity. Moreover, in order to investigate thethree-dimensional spatial distribution of snowduring the winter and to determine where snowaccumulates and settles for a given meteorologicalevent, we set up a network of metallic snow poleson erosion zones as well as on accumulationzones (see figure 3). A square of 20 fixed polesand a 200 m long profile of fixed poles have beenset up. Moreover, on both slopes of the pass(varying around 15°), two areas of about 300 m2

have been chosen where complementary snowdepth measurements (averaged on 20 points) areperformed regularly. In fact, the teams climbed tothis experimental site at least once a week,depending on snowdrift event forecasts in order tocarry out measurements, which are not recordedby data logger (snow depth, stratigraphic profile,ram test, ...).

Figure 2: One point of measurements (windvelocity, wind direction, air temperature, blowingsnow) at Col du Lac Blanc near the snow polesnetwork

5. DRIFTING SNOW NUMERICAL MODEL AT LARGE and MEDIUM SCALE: SYTRON.

Col du Lac Blanc, Grandes Rousses, 2700 m a.s.l.

300 ...-------------------"1lI-----,

250 --------------------------

E.~ 200 --------------------------

0"U>

~150c-"o •~ 100<:

U>

50

v l.t') CD <D ,..... CX) 0> 0O __ NO (\JONo 00__ T""" C'\I (\J

Dates (mmdd)

'" '" 0'" 0 '"'" ... ...000

-North SODrift

-South SODrift

-North SONoOrift

-South SONoOrift

• North Pole

II South Pole

Figure 5 : SYTRON numerical simulations on the 1998-1999 winter season.

Unlike different previous models, we do not wantto perform, at large scale, a full 3D estimation ofthe different transport fluxes over a real orography.As we have simulated hourly information on snowand meteorological conditions at differentelevations of the massifs, we stay at this scale insimulating the effects of an "imaginary" crestbetween two opposite aspects at every elevationsin steady-state conditions. We shall thus considera 1D channel forced by the normal simulated windwhere we want to estimate the amount of snowwhich is removed from one slope andaccumulated on the opposite one in treating onlytwo fictitious points. The computations are donehourly over long periods (figure 5), without any re­initialization, using as estimated transport velocityon the windward aspect from SAFRAN; theevolution of the snow characteristics is determinedby CROCUS except for the aggregation of theblowing snow. The blown snow is easilyaggregated to the new fallen snow if a snowfallOCcurs during the transport, in a coherent manner.As the computations are done hourly, the differentsnow-packs change simultaneously and allow thedeposited snow (blown or fallen) to be erodedagain in the next hour. The different followingparameterizations have principally been deducedf~om in-situ measurements on the experimentalsite for ten years.A new version of SYTRON, running at mediumscale over realistic orography over some parts ofthe massif, is under development. Unlike the

305

previous large scale 1D version, it is a full 2Dmodel integrating all the different transportphenomena. Coupled to SAMVER it will providethe different input parameters needed by the smallscale models.

6. DRIFTING ·SNOW NUMERICAL MODEL ATSMALL SCALE: NEMO

6. 1 Brief description of the model

In the past years, the Cemagref team studied thephysical modeling of blowing snow in wind-tunnel.But the difficulties lied in the disagreementbetween the scientists concerning the mostappropriate sets of modeling requirements,specially for time similarity criteria (Naaim-Bouvet,1993). Moreover, it was not possible to simulatethe evolution of snow pack. Therefore, in theframework of operational snowdrift hazardforecast, Cemagref turns to numerical modelling.This numerical model is based on a physicalmodel for saltation and turbulent diffusion (Naaim,1998). The saltation layer is described by itsheight, its concentration and two turbulent frictionvelocities, one for the solid phase and one for thegaseous phase. The suspension layer is describedby mass and momentum conservation equations.These equations were formulated both for thesolid phase and the gaseous phase. Theinteraction between these two phases was takeninto account by an equation based on the drag

306

ModfficaUon of NEMO model:The real case, however, is more complex.input parameters not only change during allblowing snow event, but the time of comphas to be short in the framework of opera'forecasting models. That is why we can notthe fully coupled wind and snowdrift model,the case of the simulation presented on figuand some additional assumptions are made.We consider the flow as a fully develoturbulent flow. A logarithmic velocityrepresented by the friction velocity (u.(x=O, twill be introduced at the upstream boundary.outputs of this first calculation on thetopography at t=to are the turbulent friction ve(u.(x,t=to)) near the snow surface,concentration field C(x,t=to) near the snow surfand the mass flux exchange with the snow(M(x,t=to)). So, if we changed the upstream frivelocity at the upstream boundary conditionu.(x=O, t=O) to u·(x=O,t), the resultin~ frivelocity in each cell in contact with the scover, u.(x,t), may be determined by:

U.(x,t}=u.(x,t=O}. (u.(x=O,t)/u.(x=O,t=O}}

6.2 Case of a real blow;ng snow event

The mass exchange flux with the snow cover'function of the concentration and the fri .velocity near the ground. According to (pomeand Gray, 1990) the saltation mass flux is g'by:

Q U'r (2 2)s oc- U* -U*t

U.

where U'! is the threshold friction velocity..In each bottom cell, the reference saltation mflux is:

cex,t=to).u.(x,t=to)oc U't (u;(x,t=to)-u;,)

u.(x,t=to) ,

During the episode, in each bottom cell,saltation mass flux is :

C(x, t). u.(x, t) eX:~ (u? (x, t) - u?, )u.(x,t)

From the ratio between these two formulae,deduce the concentration in each bottom cell:

C(;fl-C(; _ u;(x,t=to) u?(x,t)-u~

x,t;- x,t-toJ 2 2 2u.(X,t) u.(x,t=to)-u./

12108642o-2

Adimensional distance (xIh) from the snow fence

·.··.·o··:I:, :}iJjeighlt;{f:l~;';T;;u~penSiOnlayer " ..,,'':, ,ss,ahiimomentUm:equatioris'orsolict,ancigazeoti$:ptias~s) .. :.• v_'_ . '. -< ".-. - -':~,.,., <' ,._. . . -. '.' ,

")'" '-.:',.' ,./\- ",. '~.-

'0' ;;...••<'J.sauatioO',· "",<,0; '!..y~r,. "\' .

',. '(£implrica".,;.",.fqrintil~.ex;·.'":;,

The sublimation of snow particles is not taken intoaccount because we consider that its influence isinsignificant at this scale. It must be repeated thatthis physical process is taken into consideration atmedium scale.

Figure 6 : Principles of NEMO numerical model

force of a particle in a turbulent flow. Turbulencewas modelled by the k-£ model, in which areduction of the turbulence with the concentrationwas introduced (Chen and Wood. 1985). Theexchange between the saltation layer and thesnow cover was described by an' erosion anddeposition model (figure 6).

6.2 Compar;son between numerical model andexper;ments ;n laboratory

E.£ 4 -r----.----=-----------,'lij 3g 2Q) 1

Figure 7 : Comparison between numerical modeland experimental data obtained in wind-tunnelaround a small-scale snow-fence (u.=0.23 ms-1

,

u.t=0.21 ms-1)

NEMO has been tested by comparing leeward andwindward drift equilibrium obtained in a windtunnel near a small-scale snow fence (Naaim,1998). The numerical simulations were performedwith all input parameters known (fall velocity,threshold shear velocity, shear velocity, mass,concentration) which stayed constant throughoutthe experiment. The obtained results were quitesatisfactory (figure 7).

1-Wind speed- VoltageI

10080

80

60

60

• measured deposit

Distance (m)

40

40

20

-·----·----------1I

~:::::::-~ !

20

-simulated depOSIT

o

,--------------~O,5

0,4

0,3

0,2 ~<0

0,1 g.o 0

-<l,1

i----r----,-----,--~--__+ -<l,2100o

2728

2727

2726E-; 2725u~ 2724< 2723

2722

2721 +-----,----~--~--~--_J

Figure 9 : Comparison between numerical modeland experimental data obtained at "Col du LacBlanc" pass

- simulated final a~itude - initial attitude

• measured final a~itude

.Distance (m)

Figure 10 : Comparison between numerical modeland experimental data obtained at "Col du LacBlanc" pass

By comparing the numerical model results with themeasured data, we can notice that the numericalmodel correctly localizes the snowdriftaccumulation (figure 9) but an error in theaccumulation height can reach 30 % (figure 10).This can be due to the strong hypothesis weadopted or to the empirical formulae we used todetermine the flux. Therefore this work will becontinued using the fully coupled model on thewhole episode and using calibrated data from theacoustic sensor.

25 3000

~20 2500>

!. .§.

" 2000 8,: 15 .flQ.

1500 g..".5 10 :;;: Q.

1000 :;

50

Figure 8 : Wind speed and output voltage from theacoustic sensor during the storm

soo

o ~ 02610119927101199 28101/99 2910119930101199 31101199 01/0219902102199

Date

Comparison between numerical model and fieldexperiments .-

The needed inputs of NEMO are the frictionvelocity, the threshold friction velocity, the fallvelocity, the amount of available snow.We have to keep in mind that NEMO model has touse the output data from SYTRON. This twomodels (SYTRON and NEMO) have beendeveloped in parallel. Until now, the output datafrom SYTRON are not completely available. In afirst step, we have therefore to use measureddata, which will be later forecast by SYTRON, thatis hourly mean velocity, average precipitation rate,duration of blowing snow event, given by theacoustic sensors. The snow type, which will beavailable, allows us to determine the thresholdvelocity using PROTEaN. At the present time wedon't know the relation between fall velocity andsnow types. Therefore the fall velocity isestimated.So, by introducing the in situ measured wind.velocity (figure 8) at the upstream boundary of thesimulation area during the storm, we obtain thefollowing snowdrift shape at the end of the-driftingsnow event (figures 9 and 10).

7. PERSPECTIVES

Two French laboratories specialized in snow andavalanches research have joined their means andtheir models to investigate snowdrift phenomena.First results have been presented and furthercommon research is needed in a near future. Butthis approach only constitutes a first step. Such

307

studies have been carried out with the aim ofimproving avalanche hazard forecast. We havenow to develop and/or improve numerical modelsof avalanches (triggering and flow) and to includethem in the chain of models presented in thispaper. That will be our research issues in thecoming years.

8. ACKNOWLEDGMENTS

First of all, the authors are -§rateful to all of theircolleagues, specially F. Ousset, P.Pugliese and J­M Panel, who participated in the field experiments.Thanks are also attended to SATA (Safety Serviceof Alpe d'Huez Resort) for their help on theground. At last, the teams are grateful to GeneralCouncil of (sere Department for partially fundingthis work.

9. REFERENCES

Brun, E, David, P., Sudul, M., Brunot, G., 1992. Anumerical model to simulate snow-coverstratigraphy for operational avalancheforecasting. J. G/aciol., 38(128), 13-22.

Durand, Y., Brun, E, Merindol, L., Guyomarc'h,G., Martin, E 1993. A meteorologicalestimation of the relevant parameters forsnow models. Ann. Glaciol., 18, 65-71.

Durand Y., Guyomarc'h G., Merindol L., 2000.Numerical experiments of wind transportover a mountainous instrumented site. Part1 : Regional scale. In press. An. Glaciol.

Guyomarc'h G. and L. Merindol. 1998. Validationof an application for forecasting blowingsnow. Ann. Glaciol., 26, 138-143.

Guyomarc'h G., Durand Y., Merindol L., Naaim F.,2000. Climatology of an experimentallocation for studies on snowdriftProceedings of International SnowScience Workshop, October 2000, BigSky, Montana

Michaux J-L, Naaim-Bouvet F, Naaim M.,Guyomarc'h G., 2000. The acousticsnowdrift sensor: Interests, calibrationand results. Proceedings of InternationalSnow Science Workshop, October 2000,Big Sky, Montana

Michaux J-L, Naaim-Bouvet F, Naaim M., 2000.Numerical experiments of wind transportover a mountainous instrumented site. Part1 : Avalanche path scale. In press. An.Glaciol.

Naaim-Bouvet F., 1995. Comparison ofrequirements for modeling snowdrift in the

308

case of outdoor and windexperiments. SUN. Geophys., 16(5-6),727.

Naaim, M., Naaim-Bouvet F., Martinez, H., 1Numerical simulation of drifting snoerosion and deposition models. An. G/a26,191-196.