measurements of drifting and blowing snow at iqaluit, nunavut, …€¦ · 2 c e ntr f oe ah o bsv...

ATMOSPHERE-OCEAN 48 (2) 2010, 81–100 doi:10.3137/AO1105.2010Canadian Meteorological and Oceanographic Society

Measurements of Drifting and Blowing Snow at Iqaluit,Nunavut, Canada during the STAR Project

Mark Gordon1,*, Sumita Biswas1, Peter A. Taylor1, John Hanesiak2, Marna Albarran-Melzer1 andShannon Fargey2

1Centre for Research in Earth and Space Science, York UniversityToronto, Ontario

2Centre for Earth Observation Science (CEOS), University of ManitobaWinnipeg, Manitoba

[Original manuscript received 23 April 2009; accepted 18 December 2009]

ABSTrACT A 10 m meteorological tower near Iqaluit Airport was operational from late October 2007 to earlyApril 2008. Measurements included wind speed, temperature, pressure, humidity, visibility, and blowing snownumber flux. Number flux measurements give a frequency of blowing and drifting snow of approximately 10% forthe duration of the study, while meteorological observations from the Iqaluit weather office give a frequency ofapproximately 5%. Winter winds were predominantly from the northwest, and some strong southeasterly windswere also observed, especially in early spring. The average roughness length determined from the variance ofwind speed is z0 = 0.14 mm. Threshold wind speeds for the onset of blowing snow ranged from 7 m s–1 to 12 m s–1, excluding events with falling snow. Measurements of visibility correlate well with the measured numberdensity (R2 = 0.83), assuming a constant particle diameter of d ≈ 100 μm at a height of 2 m. A camera systemwas used during blowing snow events in February to measure the size of blowing snow particles and the massflux of blowing snow. At a height of 0.35 m, the particle size distribution can be approximated by a gamma dis-tribution with shape parameter 4.4 < α < 6.4 and an average particle diameter of 70 < d < 148 μm. The parti-cle size at a height of 0.35 m increases linearly with the 10 m wind speed (R2 = 0.69). Mass flux measurementsdemonstrate a power law relation with height between 0.1 and 0.9 m, with a negative exponent of approximately2.5. Blowing snow density follows a power law relation with height between 0.85 and 1.85 m, with a negativeexponent of approximately 1.3 for friction velocity 0.25 < u* < 0.55 m s–1. In February 2008, a field mill wasinstalled, which measured electric field strengths as high as 26.2 kV m–1 at a height of 0.5 m.

réSuMé [Traduit par la rédaction] Une tour météorologique de 10 m près de l’aéroport d’Iqaluit a été en fonction de la fin d’octobre 2007 jusqu’au début d’avril 2008. Les mesures portaient, entre autres, sur la vitessedu vent, la température, la pression, l’humidité, la visibilité et le flux en nombre de la poudrerie élevée. Lesmesures de flux en nombre donnent une fréquence de poudrerie élevée et basse d’environ 10 % pour la durée del’étude alors que les observations météorologiques provenant du bureau météorologique d’Iqaluit donnent unefréquence d’environ 5 %. Les vents dominants en hiver étaient du nord-ouest et de forts vents du sud-est ont aussiété observés, surtout au début du printemps. La longueur de rugosité moyenne déterminée d’après la variance dela vitesse du vent est z0 = 0,14 mm. Les vitesses de vent seuils pour les événements de poudrerie variaient de 7 m s–1

à 12 m s–1, à l’exclusion des cas où il tombait de la neige. Les mesures de visibilité concordent bien avec la densitéen nombre mesurée (R2 = 0,83), en supposant des particules de diamètre constant d ≈ 100 μm à une hauteur de2 m. Un système à caméra a été utilisé durant les événements de poudrerie élevée pour mesurer la taille des particules de poudrerie et le flux en masse de la poudrerie. À une hauteur de 0,35 m, la distribution de la tailledes particules peut être approximée par une distribution gamma avec un paramètre de forme 4,4 < α < 6,4 et undiamètre moyen des particules de 70 < d < 148 μm. La taille des particules à une hauteur de 0,35 m augmentelinéairement avec la vitesse du vent à 10 m (R2 = 0,69). Les mesures de flux en masse exhibent une relation deloi de puissance avec la hauteur entre 0,1 et 0,9 mètre, avec un exposant négatif d’approximativement 2,5. Ladensité de la poudrerie élevée suit une relation de loi de puissance avec la hauteur entre 0,85 et 1,85 m, avec unexposant négatif d’approximativement 1,3 pour une vitesse de frottement 0,25 < u* < 0,55 m s–1. En février 2008,un moulin à champ a été installé et cet instrument a mesuré des intensités de champ électrique allant jusqu’à26,2 kV m–1 à une hauteur de 0,5 m.

*Corresponding author’s e-mail: [email protected]

1 Introduction

Blowing snow is a frequent weather event in Arctic regions.Hanesiak et al. (2003) found that there were between 500 and

600 hours with blowing snow events per year (6–7%)between 1953 and 2002, measured at 20 weather stations in

82 / Mark Gordon et al.

ATMOSPHERE-OCEAN 48 (2) 2010, 81–100 doi:10.3137/AO1105.2010La Société canadienne de météorologie et d’océanographie

the Canadian Arctic. Hanesiak and Wang (2005) found thefrequency of blowing snow to be as high as 25% in northeastCanada; however they note that the frequency has beendecreasing over the last four or five decades. Déry and Yau(1999) used the European Centre for Medium-range WeatherForecasts’ re-analysis data to infer an average blowing snowfrequency of 6% throughout the year, over the entire northernhemisphere.

Studies of blowing snow using particle counter instrumentshave been undertaken in Antarctica (Budd et al., 1966; Dover,1993; Nishimura and Nemoto, 2005), Japan (Sato et al.,1993), Wyoming, uSA (Schmidt, 1981, 1982), Churchill,Manitoba, Canada (Gordon and Taylor, 2009a), and FranklinBay, Northwest Territories, Canada (Savelyev et al., 2006;Huang et al., 2008; Gordon et al., 2009). Measurementsdemonstrated that the blowing snow size distributions werebest fit by gamma distributions and that particle size decreas-es with height (Budd et al., 1966; Dover, 1993; Schmidt,1981; Gordon and Taylor, 2009a). It has also been shown thatblowing snow number density can be used to determine visi-bility (Savelyev et al., 2006; Huang et al., 2008).

Accurate modelling of blowing snow requires knowledgeof particle sizes and knowledge of distribution of particlenumber and mass with height. Both Schmidt (1981) andDover (1993) found that modelled particle sizes were smallerthan observed. Measured particle sizes are varied for differentlocations and conditions (see Gordon and Taylor, 2009a for asummary). Schmidt (1986) found that the transport rate ofblowing snow was greater over hard snow and ice than oversoft, fresh snow. Hence, measurements may be very specificto location. Although measurements have been made inAntarctica, Japan, Wyoming, Churchill, and over sea ice atFranklin Bay, there is a need for further observations in theArctic, especially on land near populated areas.

Iqaluit, Nunavut (Nu), Canada is significantly affected byblowing snow. reduced visibility due to blowing snow cre-ates dangerous flying conditions and freight is only availableby air transport during the winter months. Accurate predictionof blowing snow events could reduce flight risk. Buildingdesign must also take into consideration the loads of snowdrifts caused by blowing snow (Moore et al., 1994).Improvements in blowing snow models could improve theprediction of these loads. Blowing snow transport to and fromthe nearby sea ice and sublimation of blowing snow may sig-nificantly affect the ice surface energy balance (Bintanja andVan Den Broeke, 1995; Bintanja, 2001). Hence, better knowl-edge of blowing snow at Iqaluit could also lead to more accu-rate prediction of the annual sea-ice melt.

The Storm Studies in the Arctic (STAr) project, describedin Hanesiak et al. (2010), took place in Iqaluit, Nu, Canadaand the surrounding area in the fall and winter of 2007–08.During that study a meteorological tower and various instru-ments were installed near the Iqaluit airport to study blowingsnow. In addition to these automated instruments, measure-ments were taken by observers during the month of February2008. This paper presents these measurements, providing an

overview of blowing snow characteristics at Iqaluit, includingthreshold wind speeds, blowing snow particle size, mass den-sity profiles, visibility reduction, and the generation of anelectric field.

2 Background

Assuming neutral stratification, the wind speed, u, willchange with height, z, as

(1)

where κ = 0.4, z0 is the roughness length, and u* is the fric-tion velocity. The friction velocity is used to represent theshear stress as where t is the surface stress andρa is the density of air. During blowing snow, saltating parti-cles will absorb momentum from the boundary layer andtransfer it to the surface. This results in an increased rough-ness length as the amount of blowing snow increases. On theritscherflya plateau in Antarctica, Bintanja (2001) measuredroughness lengths ranging from z0 = 0.1 mm at u* ≈ 0.25 m s–1

to z0 ≈ 1.8 mm at u* ≈ 0.8 m s–1.The motion of blowing snow particles is generally separat-

ed into two regimes: the saltation layer and the suspensionlayer (e.g., Pomeroy and Gray, 1990). In the saltation layer,particles are ejected from the surface and follow a parabolictrajectory under the influence of gravity. Some of these parti-cles are carried aloft by turbulent eddies into the suspensionlayer. Although there is no clearly defined boundary betweenthe two layers, the transition between the two regimes is at aheight of order 100 mm (Sato et al., 2001; Gordon et al.,2009). In the suspension layer, assuming there is no net influxof particles from the surface and ignoring sublimation, a bal-ance between upward transport by turbulent diffusion anddownward settling of particles for a given radius gives ablowing snow mass density of

(2)

where ρs,r is a reference blowing snow density for a givenradius at height zr, and γ = ω/κu*, where κ = 0.4 and ω is thesettling velocity for a given radius.

During blowing snow, temperature gradients in ice parti-cles produce an electric field. Schmidt et al. (1999) andGordon and Taylor (2009b) demonstrated that the strength ofthe field decreases with height. The results of Schmidt et al.(1998) demonstrate that the field can produce acceleration inparticles comparable to the acceleration due to gravity.Hence, the electric field may significantly affect the transportof blowing snow and the mass distribution given by Eq. (2).Gordon and Taylor (2009b) show that the field strength at agiven height correlates well with the wind speed. They pro-pose a model for the electric field strength in the suspensionlayer which gives

u zu z

z( ) =

* ln ,

κ 0

u a* / ,= τ ρ

ρ ρs s rr

z zz

z( ) = ( )

−

, ,

γ

Measurements of Blowing Snow in Iqaluit during the STAR Project / 83


(3)

where —qp (C kg–1) is the average particle charge density, andεo is the permittivity of free space. Assuming a uniform par-ticle size and charge density, this gives an electric fieldstrength which scales with the number density at a givenheight. However, measurements of the field strength andnumber density at a height of 0.5 m over sea ice at FranklinBay show a poor correlation between field strength and num-ber density (Gordon and Taylor, 2009b).

Particle size distribution has previously been measured byphysically capturing particles on Formvar or oil-coated slides(Budd et al., 1966; Dover, 1993; Schmidt, 1981), or inferringthe sizes from the disruption of a light-sensor system(Nishimura and Nemoto, 2005; Sato et al., 1993; Schmidt,1982; Gordon and Taylor, 2009a). These studies have foundthat the size distribution is best represented by a gamma dis-tribution of the form

(4)

where ND is the total number density, –d is the mean particle

diameter, Γ is the gamma function, and α is the shape para-meter. Measurements of particle size show a large variation inthe value of α, ranging from 0.6 (Dover, 1993) to as high as16.1 (Budd, 1966). Average particle diameters range from 36 <

–d < 144 μm (Nishimura and Nemoto, 2005; Dover, 1993)

at heights between 1 m and 10 m, to 60 < –d < 172 μm (Dover,

1993; Gordon and Taylor, 2009a) at heights below 0.15 m.Blowing snow is often associated with a reduction in visi-

bility, as measured by the Meteorological Optical range(MOr). Huang et al. (2008) found that, for the gamma distri-bution of Eq. (4), MOr relates to the blowing snow particlenumber density and particle size (measured at the sameheight) as

(5)

where N is the particle number density and da is the cross-sec-tional-area weighted mean particle diameter. If the particlesize distribution follows Eq. (4), then

(6)

Over a sea-ice surface in Franklin Bay, Northwest Territories,Huang et al. (2008) found

—da ≈ 100 μm with MOr and num-

ber density measured at a height of 1.5 m.

3 Methodology

The Iqaluit airport (63.756°N, 68.556°W, 34 m above mean

sea level (AMSL)) is at the base of a valley aligned in the northwest-southeast direction. In winter, there is often achanneling effect through the valley and the wind is primari-ly from the northwest (Nawri and Stewart, 2006). A towerwas installed on 29 September 2007 at the Iqaluit WeatherOffice, just west of Iqaluit Airport (Fig. 1). The area sur-rounding the tower is primarily grass covered marsh. Theview to the northwest of the tower is shown in Fig. 2. Thetower consisted of a main 10 m mast, a 3 m stand-off mastlocated 1.5 m south of the main mast, and a smaller 1 m shortmast located 1 m northeast of the stand-off mast. The towermeasured wind speed at heights of 1, 2, 3, and 10 m (U1, U2, U3, and U10, respectively). Note that all heightsgiven here are from the ground, and do not include the heightof the snow pack. On the 10 m mast, a propeller anemometermeasured wind speed and direction at the 10 m height, and acup anemometer was mounted at a height of 3 m. Cupanemometers were mounted at 1 m and 2 m heights on thestand-off mast. Air temperature, Ta, pressure, p, and relativehumidity, Rh, were measured on the main mast at a height of1.5 m. Snow temperature, Ts, was measured just below thesnow surface at the main mast. The temperature difference,ΔT, between the heights of 1.5 m and 9.5 m was also mea-sured. A logger recorded data every 2 s, and output averagesand standard deviations every 10 min. Data were recorded upto 10 April 2008. Meteorological observations from theIqaluit Weather Office (Environment Canada Climate ID:2402590) were also used for comparison to measurements.

Two visibility sensors (manufactured by Sentry) weremounted on separate posts approximately 5 m to the west ofthe main 10 m mast. Both sensors were mounted at a heightof 2 m. The sensor emits a narrow beam of 880 nm wave-length light, some of which is forward scattered into a narrowadmittance angle detector. The output depends on the amountof light forward scattered from particles in the volume of airbetween the emitter and the detector. The sensor output has amaximum MOr of 16 km. The volume of air between theemitter and the detector is situated beside the housing for thesensor electronics. Previous studies (e.g., Huang et al., 2008)have shown that this housing can obstruct blowing snow fromsome directions, resulting in an inaccurate visibility measure-ment. One sensor (the rightmost in Fig. 2) is oriented to faceapproximately northeast so that the housing will obstructblowing snow from approximately 210°. This sensor wasinstalled on 29 September 2007. The other sensor (the left-most in Fig. 2) is oriented to face approximately northwest sothat the housing will obstruct blowing snow from approxi-mately 118°. This sensor was installed on 15 February 2008.

Due to a programming error, the wind speeds at heights of1, 2, and 3 m and the northeast facing visibility sensor wererecorded as instantaneous measurements every 10 min, asopposed to 10 min averages. This error was fixed on22 February after which all values were recorded as 10 minaverages.

Blowing snow particle counters, based on the design ofBrown and Pomeroy (1989), were mounted on the stand-off

E zz q zs p

o

( ) = −( )

−( )3

1

ρ

ε γ,

f dN d d d

d

D( ) =−( )

( ) ( )

−α

α

α

α α

1 exp /

/

,

Γ

MOR =7 84

2

.,

N daπ

d da =+

αα11 2/

.



mast at heights of 0.5, 1, and 2 m. Snow depth beneath thesensors increased to approximately 0.15 m by 9 February2008. Due to a malfunction, the 2 m counter became inoper-able after 16 March 2008. The particle counters are discussedin greater detail in Savelyev et al. (2006). As particles passbetween the light emitting diode and the detector, which isplaced behind a small pinhole (200 μm diameter), the voltageoutput of the detector is reduced. When the voltage is reducedbelow a set threshold, a pulse count is generated. The particlecounts are output to a data logger every 2 s, giving particlenumber flux, v, which relates to the blowing snow density(Eq. (2)) as

(7)

where ρi is the density of ice and up(z) is the average particlespeed at height z. Calibrations have demonstrated that a min-imum particle diameter between 20 and 50 μm is required toproduce a particle count. However, experiments have shownthat this minimum required diameter can also depend on thewind speed. Hence, the minimum detectable particle diameteris not known precisely.

Blowing snow mass flux was measured throughout themonth of February using snow traps mounted on the structureshown in Fig. 2. The traps were made with a fine mesh mate-rial and were mounted at heights of 0.1, 0.3, 0.5, 0.7, and

0.9 m. Each trap had an opening of 100 mm × 100 mm andwas approximately 0.5 m long. Traps were generally left outfor 30 min after which the snow in each trap was weighed.Due to flow distortion around the traps, since the fine meshreduces the wind speed as it passes through the traps, thesetraps can underestimate the true amount of blowing snow.Abe (1989) (cited in Sato et al., 1993) reports a 75% efficiency for 29 mm diameter bag traps. Font et al. (1998)report efficiencies ranging from 5% to 45% for traps with50 mm × 100 mm openings, which are comparable to snowcollector columns with 14 mm diameter inlets. This is alsoseen in the results of Savelyev et al. (2006), who used thesame traps as were used in this study. Savelyev et al. (2006)found mass fluxes measured using the snow bags werebetween 3.4% and 20% of the mass fluxes determined usingan acoustic sensor (described in Chritin et al. (1999)).

A camera system, described in Gordon and Taylor (2009a)and Gordon (2007), was mounted on the short mast to mea-sure the size and number flux of blowing snow particles. Thecamera system records the shadows of blowing snow particlesas they pass between a halogen lamp and the camera. Becauseeach row of pixels in the image array is exposed sequentially,the system measures the dimension of the particle perpendic-ular to the direction of particle motion, referred to here as theparticle width, w. The system used here differs from the sys-tem described in Gordon and Taylor (2009a), as a lens exten-sion is used to increase the resolution to 2.9 μm per pixel. Amore powerful lamp (50 W) is used to compensate forreduced light at the image array due to the use of the lensextension. Further, an entrance passageway replaces the slotof the previous design. For particles to be recorded by thecamera, they must pass through this 10 mm long passageway,which has a 0.65 mm wide by 3 mm long rectangular open-ing. This effectively eliminates the possibility of particlespassing outside the camera's depth of field, a problem whichis discussed in Gordon and Taylor (2009a). The rectangularentrance also limits the maximum size of blowing snow par-ticles that can be measured. The use of the passageway makesthe measurement of velocity unreliable, because particlevelocity will be reduced due to friction if the particle collideswith the walls of the passageway. The minimum detectableparticle size is related to the recorded particle speed, Up, andexposure time, te = 65 μs, as wmin = 0.1 teUp, where wmin is thedimension of the particle in the direction of motion. Thus foran 8 m s–1 particle speed the minimum detectable width is52 μm. However, particle velocity is generally reduced beforethe size is measured due to friction as it enters the passage-way.

An electric field mill (manufactured by MissionInstruments) was installed at the base of the stand-off mast.The field mill, primarily designed as a lightning warning sys-tem, measures the magnitude, but not the direction, of theelectric field strength. The field mill uses a rotating probe thatfaces downward through a 7 cm diameter opening. The open-ing was mounted at a height of 0.5 m from the snow surface.The accuracy of the field mill is given as ±5% for E > 500 V m–1

Iqaluit

Airport

Site

FrobisherBay

Sylvia Grinnell

River

2 km

N

Fig. 1 A map of the study site (at the Iqaluit Weather Office) and sur-rounding area. Contours are at 20 m intervals.

ρρ π

si

p

zd v z

u z( ) =

( )( )3

6,



and the response time is approximately 1 s. Field calibrationtests throughout the study confirmed this accuracy. However,due to a logger error, measurements less than 2.9 kV m–1

were not recorded.

4 Results and discussion

Wind chart diagrams are shown for each month of the studyin Fig. 3. Throughout the winter (December to February),when temperatures are lowest, the wind direction is primarilyfrom the northwest. Later, in March and April, the wind canalso be from the southeast. This is consistent with the resultsof Nawri and Stewart (2006), who note that northwesterly off-shore winds are associated with lower temperatures in winter.They also note that east-northeasterly winds are warmer thanthose from the northwest. This is also seen here in the monthof November, when the average temperature during prevail-

ing northwesterly winds (for U10 > 4 m s–1) is –12.4°C andthe average temperature during prevailing east-northeasterlywinds (for U10 > 4 m s–1) is –7.0°C.

Measurements from day of year (DOY) 341 to 347, (7 to 13December 2007), are shown in greater detail in Fig. 4.Between approximately 10:00 uTC on DOY 341 and14:00 uTC on DOY 343, the wind carried relatively warm airfrom the southeast. There were approximately six hours ofcontinuous blowing snow from 09:00 uTC, DOY 342, asdemonstrated by the 2 m blowing snow number density andreduced visibility. During blowing snow, the relative humid-ity was generally higher, with values near 90% with respectto water, which is near saturation with respect to ice. Afterapproximately 20:00 uTC on DOY 343, there was a relativelystronger and colder wind from the northwest. Here there is amuch longer blowing snow episode, lasting nearly 24 h.

Visibility Sensors

Wind Speed

Temperature,Humidity

Wind Speed

Particle Counters

Snow Traps

Electric Field MillBlowing Snow Camera

Temperature

Fig. 2 A photo of the tower facing northwest, taken 21 February 2008.



During both episodes there is a reduction in visibility due tofalling snow. Snowfall was recorded on DOY 341 andDOY 344 giving 0.5 mm snow water equivalent (SWE) and2 mm SWE, respectively. Hence, the blown snow in bothevents is from new, freshly fallen snow.

a Threshold Wind SpeedNumber flux measurements at a height of 1 m are comparedto the 10 m wind speed measurements in Figs 5a to 5d for four

blowing snow episodes. Although there were more than 30blowing snow episodes during the course of the study, thesefour are chosen as a sample. Visual inspection of the figuresgives a threshold wind speed for the initiation of blowingsnow between 7 and 10 m s–1. Previous studies (e.g., Mann etal., 2000) demonstrated a hysteresis effect, with the value ofthe threshold wind speed showing dependence on the historyof the blowing snow event. Once saltation begins, and the

Apr.Ta = - 12.7oC

Nov.Ta = - 10.9oC

Dec.Ta = - 23.8oC

Jan.Ta = - 28.5oC

Feb.Ta = - 29.6oC

Mar.Ta = - 23.1oC

0

45

90

135

180

225

270

315

0% 2% 4% 6%

0

45

90

135

180

225

270

315

0% 4% 8% 12%

0

45

90

135

180

225

270

315

0% 4% 8%

0

45

90

135

180

225

270

315

0% 8% 16%

0

45

90

135

180

225

270

315

0% 2% 4% 6%

0

45

90

135

180

225

270

315

0% 2% 4% 6% 8%

U10 (m s-1)>0 - 4>4 - 8>8 - 12>12

Fig. 3 Wind chart diagrams for the months of November 2007 to March and the first 10 days of April 2008. The average temperature, Ta, is shown for eachmonth.



bonds of particles at the surface are broken, saltating particlesthen further erode the surface, providing more loose particlesand lowering the threshold wind speed required to sustainblowing snow. In Figs 5a to 5d, each 10 min measurementmade before the peak value of v is reached is marked as +,while each measurement made after the peak value of v ismarked as ×. In Fig. 5a, which occurred on DOY 322, 2007,a hysteresis effect is seen, with wind speeds near 10 m s–1

required to initiate blowing snow. After approximately 5 h of

continuous blowing snow, the same number fluxes (2 × 106 <v < 6 × 106 m–2 s–1) are maintained by lower wind speeds.However, this effect is not seen in other blowing snowepisodes, including the three shown in Figs 5b to 5d. In theseepisodes very little difference is seen between the start andthe end of the episode. This difference is likely due to rela-tively high temperatures (–1°C) and humidities (99% withrespect to ice) which occurred soon after a snowfall ended atapproximately 17:00 uTC on DOY 321. This would result in

341 342 343 344 345 346 347DOY 2007

60708090

100110

Rh (%

)

0

90

180

270

360

Ψ (°

)

0

4

8

12

16

U (m

s-1 )

0

4

8

12

16

MOR

(km)

70

80

90

100

110

Rhice (%)

-24

-20

-16

-12

-8

Ta ( oC)

00.20.40.60.81

10 -6 N(m ) -3

98.698.89999.299.499.699.8

pa (kPa)

Fig. 4 Measurements from DOY 341 to 348, 2007 (times are uTC). Black lines show (from top left) MOr, 10 m wind speed, wind direction, and relativehumidity (with respect to water). Grey lines show (from top right) particle number density at 2 m, air pressure, air temperature, and relative humidity(with respect to ice).



strong bonding of snow particles at the surface, requiring ahigher wind speed to break the bonds and initiate blowingsnow. In the episodes shown in Figs 5b and 5c, the tempera-tures preceding the blowing snow were in the range of –30°C< Ta < –20°C, while temperatures preceding the episodeshown in Fig. 5d were approximately –11°C.

The probability of blowing snow was calculated for each0.5 m s–1 bin of 10 m wind speed based on the criteria v > vthusing the 10 min averages of wind speed and blowing snownumber flux. Three values of vth (0.125 × 106, 0.25 × 106, 0.5 × 106 m–2 s–1) are used to test the sensitivity of the prob-ability to the choice of threshold. These values correspond toaverage counts of 0.5, 1, and 2 particles per second, respec-tively. The resulting probability distributions are shown inFig. 6. To produce a 50% probability of blowing snow, a windspeed ranging from 8.2 m s–1 to 9.4 m s–1 is required at thislocation. For wind speeds between 11 m s–1 and 14 m s–1,depending on the threshold value chosen, there is no blowingsnow between 10 and 30% of the time.

Since particle density should decrease with height duringblowing snow according to Eq. (2), while particle density dur-ing falling snow or ice crystals should be relatively constantwith height, the criteria of ρs(1 m) > ρs(2 m) is also used with

vth = 0.25 × 106 m–2 s–1. If it is assumed that particle diame-ter is equal at heights of 1 m and 2 m, the particle density canbe calculated from Eq. (7) for z = 1 m and z = 2 m. The result-ing probability is shown in Fig. 6 (grey line). This reduces theprobability of blowing snow slightly for wind speeds below6.5 m s–1 and by approximately 6% above 10.5 m s–1.However, if the particle diameter decreases with height theactual probability of blowing snow could be higher.

Following Mann et al. (2000), the threshold velocity isdefined as the hourly averaged velocity measured when thehourly average particle number flux crosses a threshold value of vth. An hourly average particle number flux of vth = 0.25 × 106 m–2 s–1 is chosen as a threshold level. Hence,if the hourly average v is below vth and the next hourly average is above vth, the second hour is used to determine thethreshold values of wind speed, temperature, and relativehumidity. Although Mann et al. (2000) use a counter at aheight of 2 m, here we use the measurement of the particlecounter mounted at a height of 1 m, since the 2 m counter wasnot operational for the full duration of the project. Tests using both the 1 m and 2 m counters show that the choice ofcounter height has little effect on the calculated thresholdspeeds.

4 8 12 16 20U10 (m s-1)

0

10

20

30

40

50

4 8 12 16 20U10 (m s-1)

0

4

8

12

16

20

v (1

06 m-2

s-1)

4 8 12 16 20

0

4

8

12

16

20

4 8 12 16 20

0

4

8

12

16

20

v (1

06 m-2 s

-1)

(a) (b)

(c) (d)

Fig. 5 Blowing snow number flux, v1, at a height of 1 m compared to the 10 m wind speed, U10 for four blowing snow episodes: a) from DOY 322 to 323,2007; b) DOY 47 to 50, 2008; c) DOY 77, 0:00 to 12:00; and d) DOY 81, 12:00 to DOY 83, 12:00 (all times uTC). The symbol ‘+’ denotes measure-ments leading up to the peak value (‘*’), and ‘×’ denotes measurements after the peak value. Note the different y-axis scale in graph (d).



Environment Canada’s hourly meteorological observationswere used to differentiate between instances of particle fluxmeasurements during ice crystal or falling snow events, andparticle flux measurements when neither of these events wereobserved. Ice fog was not reported throughout the duration ofthe study. The threshold wind speeds are compared to air tem-perature and relative humidity in Fig. 7. The filled circles rep-resent observed blowing snow only, while the open circlesrepresent simultaneous observations of blowing snow withice crystals or falling snow. Also shown for comparison (solidcurved line) is the threshold wind speed determined by Li andPomeroy (1997) based on measurements made in theCanadian Prairies,

(8)

where Ut* = 6.98 m s–1 is the minimum threshold velocity andTa is in °C. The threshold speeds during observed blowingsnow only are generally higher than produced by Eq. (8),while the threshold speeds during observations of blowingsnow with ice crystals or falling snow range from 4.8 m s–1

to 15.4 m s–1. The two highest threshold wind speeds withU10,th >13 m s–1 occurred on 30 and 31 January. During thisstorm there was a relatively heavy snowfall (8.6 mm SWEover 60 hours), with winds primarily from the southeast. On29 January, the winds were light (U10 < 2 m s–1) and temper-atures were very cold, reaching as low as –39°C.

The average threshold wind speed for all the data is U10,th = 8.6 m s–1, which is greater than produced by Eq. (8)for –4°C < Ta < –5°C. This suggests that the Li and Pomeroy(1997) model may overestimate blowing snow frequency atthis cold, Arctic location. A comparison of threshold speedand relative humidity (with respect to ice) shows no trend inthe data for blowing snow only. However, at relative humidi-

ties near 100% during observed ice crystal or falling snowevents, the threshold wind speed is generally lower. This maybe due to ice crystals which formed previously in calm weath-er and settled on the surface, which are easily picked up fromthe surface by strengthening winds.

Predicted and calculated frequencies of blowing snow arelisted in Table 1. According to Environment Canada’s mete-orological observations, blowing snow was observed for5.1% of the project duration. The hourly average wind speedis greater than produced by Eq. (8) for 15.5% of the projectduration, while it is greater than the average of 8.6 m s–1 for9.4% of the project duration. The blowing snow number fluxrecorded at a height of 1 m is greater than vth for 10.8% of thetime. However, if data recorded during observations of icecrystals or falling snow are removed, v > vth for 6.2% of thetime, which is closer to the frequency of blowing snow deter-mined by the hourly meteorological observations. Again, theparticle density is calculated from Eq. (7) for z = 1 m and z = 2 m, assuming equal particle diameters at heights of 1 mand 2 m. Because particle density should decrease with heightduring blowing snow, while particle density during fallingsnow or ice crystals should be relatively constant with height,the criteria of ρs(1 m) > ρs(2 m) and v(1 m) > vth are used tocalculate the frequency of blowing snow. These criteria arejointly satisfied 10.6% of the time. If the particle diameterdecreases with height this frequency could be higher. Hence,the true frequency of blowing snow is underestimated bymeteorological observations, and observations of ice crystalsand falling snow may also include blowing snow.

b Velocity Profileroughness length and friction velocity were calculated foreach 10 min average using a linear least-squares fit to Eq. 1with ln(z) as a function of U(z). However, this resulted in

0.125×106 m-2s-1

0.25×106 m-2s-1

0.5×106 m-2s-1

0.25×106 m-2s-1 ρs(1m)>ρs(2m)

4 6 8 10 12 14 16 18U10 (m s-1)

0

0.2

0.4

0.6

0.8

1

P (

v >

vth)

0

200

400

600

800

1000

1200

Num

. per bin

Fig. 6 Probability distribution of blowing snow, based on three values of vth (black lines) and vth = 0.25 × 106 m–2 s–1 with ρs(1 m) > ρs(2 m) (grey line). Thedistribution of wind speed is also shown.

U U Tth t a10

20 0033 27 26, * . . ,= + +( )



values of z0 > 100 mm during some blowing snow episodes.Previous measurements of roughness length during blowingsnow over sea and lake ice (Savelyev et al., 2006; Schmidt,1986; Tabler, 1980), frozen plateau (Bintanja, 2001) or shortgrass (Schmidt, 1982), have found z0 to be on the order of1 mm. Although the cup anemometers were re-calibratedafter the study, they could only be calibrated for wind speedsup to 4 m s–1. Further, as the site was unattended for longperiods throughout the study, the cups would fill with blow-ing snow. This results in an underestimation of wind speed

and an overestimation of roughness length. Hence, we con-sider the cup anemometers unreliable for the calculation ofroughness length from the wind profiles. Wieringa (1973)demonstrates that the friction velocity can be reliably esti-mated using the variance of the wind speed (σU) at a givenheight as σU ≈ 2.5 u*. Measurements of U10 and σU10

can thenbe used to determine z0 from Eq. 1 assuming a logarithmicprofile to 10 m. For U10 > 6 m s–1, this gives an averageroughness length of z0 = 0.14 mm.

c Particle Size DistributionsDuring DOY 35 (4 February 2008), 41 (10 February 2008),and from 47 to 49 (16–18 February 2008), the camera systemwas used to measure number density and size of blowingsnow particles. The camera was mounted at a height of585 mm from the snow surface on DOY 35 and at a height of350 mm during DOY 41 and from DOY 47 to 49. The heightof 350 mm from the snow surface was at the same level as thelowest particle counter. On DOY 35, images were recorded ata rate of 1 Hz from 00:00 to 2:20 uTC. On DOY 40, imageswere recorded at a rate of 6 Hz from 18:30 to 20:15 uTC.Environment Canada observed ice crystals for the duration ofboth these acquisitions. From DOY 47 to 49, images wererecorded at a rate of 1 Hz in four separate blocks: from 3:00to 12:00 uTC, DOY 47 (Block A); from 15:20 uTC, DOY 47to 00:20 uTC, DOY 48 (Block B); from 3:00 to 12:00 uTC,DOY 48 (Block C); and from 18:00 uTC, DOY 48 to3:00 uTC, DOY 49 (Block D). The time between these blockswas required to transfer images to an external storage device.The hourly meteorological observations during Block A weremostly cloudy, with snow beginning at 11:00 uTC. Block Bbegan as mostly cloudy, with blowing snow observed after18:00 uTC. This blowing snow continued throughout Block Cand most of Block D, until 00:00 uTC, DOY 49.

The recorded 10 m wind speed during each block can beseen in Fig. 8. Also shown in the figure are the number flux,v, and particle width, wavg, recorded by the camera, deter-mined as 10 min averages to coincide with the 10 m windspeed measurements. During Block A, when wind speed andparticle number flux are relatively low, the average particlewidth ranges from 82 to 132 μm. As wind speed and particleflux increase in Block B, the average particle width alsoincreases to as large as 168 μm. During Block C, with nearlyconstant wind speeds (U10 ≈ 10 m s–1), the average particle

-40 -30 -20 -10 0Ta (oC)

4

6

8

10

12

14

16U

10,th

(m

s-1

)

70 80 90 100Rhi (%)

4

6

8

10

12

14

16

U10

,th (

m s

-1)

Fig. 7 The threshold wind speed, U10,th compared to air temperature, Ta,and relative humidity with respect to ice, Rhi. Filled circles repre-sent blowing snow only, and open circles represent blowing snowduring either observed falling snow or ice crystals. Also shown forcomparison is the parameterized threshold wind speed of Li andPomeroy (1997).

TABLE 1. Frequency of blowing snow throughout the project duration asdetermined by meteorological observations, threshold wind speeds, and par-ticles counters. ‘No FS/IC’ denotes the removal of data during observationsof falling snow or ice crystals.

Criteria Frequency

Meteorological Observations 5.1%Eq. (8) (Li and Pomeroy, 1997) 15.5%U10,th = 8.6 m s–1 (this study) 9.4%v(1 m) > vth 10.8%v(1 m) > vth (No FS/IC) 6.2%v(1 m) > vth and ρs(1 m) > ρs(2 m) 10.6%



width remains relatively constant, ranging from 108 to 141μm. Near the end of Block D, as wind speed and number fluxdecrease, the average particle width also decreases, to as lowas 97 μm. The 10 min average particle width correlates wellwith U10, as shown in Fig. 9. The friction velocity is alsoshown in the figure, calculated using Eq. 1 with z0 = 0.14 mm,which gives u* = 0.036U10. A best fit to the data gives wavg = 8.8U10 + 35 (μm), with a coefficient of determinationR2 = 0.69 (p < 0.001).

The particle size distributions, from all particles recordedon DOY 35 and 41, are fit to Eq. (4). Each least squares fit isdetermined by calculating the residual error for a range ofshape parameters, 0 < α < 20. The residual error is calculatedas ∑ (f(wj) – n(wj))

2, where n(wj) is the number of particles inthe bin corresponding to particle width wj and f(wj) is deter-

mined from Eq. (4) with d = wj. The best fit is determined asthe shape parameter which gives the minimum residual error.The resulting shape parameters and mean widths for DOY 35and 41 are listed in Table 2. A larger value of α signifies anarrower distribution of particle sizes, while a smaller valuesignifies a wider distribution. The shape parameters and meanwidths are also shown for Blocks A to D (DOY 47 to 49) inTable 2. As discussed in Section 3, the minimum detectablediameter is approximately 50 μm for a particle moving at 8 m s–1. As most particles detected by the camera are movingat speeds lower than this, it is assumed that any small, miss-ing particles will not significantly affect the particle size dis-tributions.

A sufficient number of particles were recorded from DOY47 to 49 to determine the best-fit shape parameter for particles

5

10

15

U10

(m

s-1)

0

40

80

v (1

06 m-2

s-1)

50

100

150

200

wav

g (µ

m)

0

2

4

6

8

10

α

A

B

CD

6DOY47

12 18 0DOY 48

0DOY 49

6 12 18

Fig. 8 The 10 m wind speed, U10, number flux, v, recorded by the camera (solid line) and the particle counter (dotted line) at a height of 350 mm, and the par-ticle size, wavg, all determined as 10 min averages. The data acquisitions are identified as Blocks A to D. Also shown is the shape parameter, α, deter-mined from a best fit to Eq. (4) for the particles recorded in each hour (all times uTC).



recorded during each hour. These hourly shape parametersare shown in Fig. 8. The number of particles used for eachhourly size distribution ranged from 900 to over 14,000. Asthe blowing snow episode begins during Block A the shapeparameter is relatively small, approximately α = 4. As windspeeds decrease, the shape parameter increases to approxi-mately α = 7. During Block B, higher wind speeds give 6 < α < 8, while lower wind speeds give 3 < α < 6. The low-est value of α = 3.3 occurs between 20:00 and 21:00 uTC,DOY 47, following the highest wind speeds. It is possible thatthe high wind speeds before 20:00 uTC resulted in the sus-pension of larger particles. As wind speeds diminished, someof these large particles may have remained suspended orloosely bonded at the surface. Hence, although the numberflux is relatively low between 20:00 and 21:00 uTC, the parti-cles could include a relatively high number of large particles,which would result in a wider size distribution and lowershape parameter. For Blocks C and D, with relatively constantwind speeds, the shape parameter is relatively constant with

an average of α = 6.2. For all these data, there is no correla-tion (R2 = 0.001) between hourly averages of 10 m windspeed and the shape parameter for each hour.

Gordon and Taylor (2009a) measured the shape of blowingsnow particles in Churchill, Manitoba. It was found that theequivalent particle diameter, where Ap isthe projected area of the particle, related to the particle widthas w = 1.13 deqv. If we assume that the blowing snow particlesmeasured by the camera system in this study have the samewidth to equivalent diameter ratio, then the widths reportedhere will cause the equivalent diameters to be overestimatedby approximately 13%. Average equivalent diameters aregiven as deqv = wavg/1.13 in Table 2. using this width toequivalent diameter ratio, the best fit shown in Fig. 9 wouldgive deqv = 7.8U10 + 31 (μm) for 6 < U10 < 15 m s–1. The coef-ficient of determination for this fit is R2 = 0.69 (p < 0.001).

d Mass FluxEquation (2) gives an idealized power law relationshipbetween blowing snow density and height for particles of uni-form size and settling velocity, ω. If the particles are uniformin size then this will hold for both mass density and numberdensity. In reality there will be a range of particle sizesinvolved and a range of settling velocities. This situation wasconsidered by Déry et al. (1998). If the particle size distribu-tion is known or assumed at the reference height and Eq. (2)can be used to determine the height variations of mass ornumber density for individual particle size bins then integra-tion over the size range gives the total number or mass densi-ty of the total spectrum of blowing snow particles. Déry et al.(1998) show that, with no sublimation effects, the log-log plotof mass density versus height is curved in a manner corre-sponding to small particles being lifted higher than large onesand a reduction in the overall, effective ω and γ = ω/κu* withheight.

Mass fluxes calculated from the snow traps are shown inFig. 10. The dashed lines are profiles collected over a 30 minperiod, while the solid lines are for longer periods. We can fitthe mass flux profiles by power-law profiles of the form,

(9)

A fit with γμ ≈ 2.5 would work in the lower portion of theheight range while higher up the exponent appears to dropand is somewhat variable between runs.

4 6 8 10 12 14 16U10 (m s-1)

60

80

100

120

140

160

180

wav

g (µ

m)

0.2 0.3 0.4 0.5u* (m s-1)

Fig. 9 The 10 min average particle width compared to U10. A best fit tothe data is also shown.

TABLE 2. The camera height, zc, duration and rate of acquisition, mean 10 m wind speed, U10, relative humidity, Rh, total number of particles, n, average par-ticle width, wavg, shape parameter, α, and equivalent diameter, deqv for DOY 35, 40, and the four blocks during DOY 47 and 48 (2008).

DOY Start Duration zc rate U10 Ta Rh Rhi n wavg α deqvBlock uTC h mm Hz m s–1 °C % % μm μm

35 00:00 2.3 585 1 8.6 –26 72 95 1220 94 5.5 83 40 18:30 1.7 350 6 8.6 –35 64 92 445 119 4.4 105 47(A) 03:00 9 350 1 7.5 –29 69 93 3700 115 4.8 102 47(B) 15:20 9 350 1 11.1 –30 71 97 57800 139 6.4 123 48(C) 03:00 9 350 1 10.8 –29 74 99 42700 124 6.1 110 48(D) 18:00 9 350 1 9.3 –29 73 99 18500 132 6.2 117

d Aeqv p= 4 / ,π

µ µ γµ= ( )−r rz z/ .



The snow trap at a height of 0.3 m was set while the cam-era system was operational for three periods shown inTable 3. The camera measured the snow flux averaged overan area of 0.65 mm × 3 mm at a height of approximately0.35 m, while the snow trap measured the snow flux over anarea of 100 mm × 100 mm with the opening between heightsbeing 0.25 m and 0.35 m. Assuming there is no significanterror in the camera measurements, and ignoring the heightdifference, this gives snow trap efficiencies between 18.5 and51% for the three periods. As discussed in Section 3, Font etal. (1998) report bag trap efficiencies between 5 and 45%. Assnow begins to fill the traps, efficiency will be reduced fur-ther, as is seen in Table 3, where the efficiency decreases withthe duration of the period. Hence, the traps may underesti-mate efficiency by up to a factor of 5 for long-term exposures.The average duration for the deployment was 4 h. The periodswhen the snow traps were exposed for sampling times of30 min are shown in Fig. 10 as dashed lines. Although effi-ciencies may vary for different sampling times, the similarslopes for all periods suggests that the profile shape does notdepend on the duration of exposure.

The reference mass flux at 0.1 m in Fig. 10 is in the range0.001 < μr < 0.1 kg m–2 s–1. Nishimura and Nemoto (2005)measured mass flux at Mizuho station in northeast Antarctica

and found reference mass flux values at a height of 0.1 m inthe range 10–6 < μr < 0.01 kg m–2 s–1 for 0.21 < u* < 0.56 m s–1.Takahashi (1985) also measured mass flux at Mizuho station,finding 0.07 < μr < 0.7 kg m–2 s–1 at a height of 0.1 m fordaily average 10 m wind speeds between 9 and 16 m s–1. Fontet al. (1998) measured mass flux at La Molina ski resort in theeastern Spanish Pyrenees, and found 0.1 m mass flux in therange 0.02 < μr < 0.2 kg m–2 s–1 during 10 m wind speedsbetween 13 and 22 m s–1. Although the trap efficiencies arelikely below 100%, the mass fluxes measured here are gener-ally within the range of previous measurements.

Nishimura and Nemoto (2005) measured blowing snowmass flux between heights of 0.2 m and 10 m. Five profilesare given for friction velocities ranging from u* = 0.21 m s–1

to 0.56 m s–1. The mass fluxes follow a power law profilewith height between 0.02 < z < 0.2 m with a negative expo-nent ranging from γμ = 1.0 (for u* = 0.56 m s–1) to γμ = 4.4(for u* = 0.21 m s–1). Between heights of 1 m and 10 m, themass fluxes also follow a power law profile with a negativeexponent ranging from γμ = 0.6 (u* = 0.56 m s–1) to γμ = 1.3(u* = 0.28 m s–1). There were no measurements betweenheights of 0.2 m and 1 m, however, the value of γμ = 2.5 seenhere is within the values of 1.3 and 4.4 found by Nishimuraand Nemoto (2005) for low wind speeds. The mass flux

0.000001 0.0001 0.001 0.01 0.1Blowing Snow Mass Flux, µ (kg m-2 s-1)

0.1

1

Hei

ght,

z (m

)0.5

0.3

0.7

-2.5

0.2

Fig. 10 Blowing snow mass flux measurements made using snow traps assuming 100% efficiency. Dashed lines represent 30 min sampling times. Solid linesrepresent sampling times of more than 30 min. A power law profile of μ ∝ z–2.5 is also shown for comparison.



profiles measured by Takahashi (1985) generally follow thepower-law profile of Eq. (9) between 0.4 m and 1 m, althoughsome profiles show a relative decrease in the value of γμ

above 0.5 m. Values of γμ between 0.1 m and 1 m range from1.1 (for U10 = 16 m s–1) to 2.7 (for U10 = 9 m s–1). Similarly,mass flux profiles measured by Font et al. (1998) give a neg-ative exponent ranging from 0.6 < γμ < 2.2 for measurementsbetween 0.1 m and 1 m, however, the range of height is tooshort to demonstrate any change in γμ with height. Hence, thelog-log plots of mass density versus height are generallycurved in a manner corresponding to a reduction of the effec-tive ω and γ = ω/κu* with height and the value γμ = 2.5 seenhere is generally within the range of previous studies.

The average number fluxes recorded by particle countersover the same periods are shown in Fig. 11. The heights of thecounters take into account the snow depth of approximately0.15 m. These profiles also approximate power-law profilesof the form

(10)

with γv = 0.75. The longer term averages of more than 30 min(solid lines) show a closer agreement to the power-law profile.

In order to investigate particle size variation with height wecan use Eq. (2) in conjunction with particle size distributionresults from the camera system on DOY 48, Block C. Thisperiod of time is chosen because snow trap measurementswere made during the camera acquisition, and the wind wasrelatively constant (Fig. 8). Equation (2) is solved for eachparticle size of Eq. (4) with

–d = 110 μm and α = 6.1 at a ref-

erence height of z = 0.35 m (Table 2). The reference numberflux at 0.35 m is taken as the average v = 14 × 106 m–2 s–1.The friction velocity is calculated from Eq. (1) using the aver-age wind speed U10 = 10.8 m s–1 with zo = 0.14 mm, whichgives u* = 0.39 m s–1. Following Déry et al. (1998), we esti-mate ω = ωT, where ωT is the terminal velocity of a sphericalparticle obtained from the drag formulae proposed by Carrier(1953). The resulting number flux, mass flux, and mean diam-eter profiles are shown in Fig. 12 (solid lines). The numberflux and mass flux profiles are calculated as v(z) = N(z) up(z)and μ(z) = ρs(z) up(z) assuming that particle speed is equal towind speed from Eq. (1). The number and mass flux valuesdecrease rapidly with height between 0.1 < z < 0.5 m. Abovethis height they generally approximate a power-law profilewith a slight decrease in negative exponent with height. Bothnumber flux and mass fluxes reference values (0.1 m) aremore than two orders of magnitude higher than those shownin Figs 10 and 11.

As discussed in Xiao (2001), the settling velocity of blow-ing snow particles will most likely be dependent on a numberof factors, including turbulence, particle shape, and orienta-tion of the particle (Clift and Gauvin, 1971). Measurements ofparticle shape in Churchill, Manitoba (Gordon and Taylor,2009a) demonstrate that particles are generally non-spherical,which will increase the drag on the particle relative to asphere. To produce profiles which more accurately approxi-mate the profile shapes seen in Figs 10 and 11, Eq. (2) issolved for the same parameters given above with ω = 0.7 ωT.The resulting calculated number flux, mass flux, and meandiameters are shown as dashed lines in Fig. 12. This suggeststhat the particles generally fall 30% more slowly than ideal-ized spheres in still air, which may be due to their non-spher-ical shape or to acceleration due to turbulent eddies.However, since γ = ω/κu*, the friction velocity may also beunderestimated, as u* = 0.55 m s–1 with ω = ωT will give the same results as u* = 0.39 m s–1 with ω = 0.7ωT. The reference mass flux at 0.1 m of μr ≈ 0.5 kg m–1 s–1 is muchhigher than the values seen in Fig. 10, suggesting that thesnow traps may have efficiencies less than 20%. The refer-ence number flux is also higher than the values seen inFig. 11, which may be due, in part, to the camera system mea-suring higher number fluxes than the particle counters, asshown in Fig. 8.

From Fig. 12 and with ω = 0.7ωT, the mean particle diam-eters are d1 = 84 μm at a height of 1 m and d2 = 75 μm at aheight of 2 m. Assuming that the ratio of d2/d1 ≈ 0.9 is con-stant throughout the study, γ (for mass density) can be calcu-lated from Eqs (2) and (7) as

(11)

where subscripts 1 and 2 refer to number fluxes (v) and parti-cle speeds (up) at heights of 1 m and 2 m respectively and zsis the snow height. Schmidt (1982) and Nishimura andNemoto (2005) demonstrated that blowing snow particlevelocity is approximately equal to the wind velocity forheights above 0.1 m. Hence, wind speed measurements at 1 mand 2 m are used for particle speeds up1 and up2 respectively.As discussed in Section 4b, cup anemometer measurementsproved unreliable, so wind (and blowing snow particle)speeds at 1 m and 2 m were calculated using U10 and Eq. (1)with z0 = 0.14 mm. The calculated values of γ for each 10 minaverage with U10 > 6 m s–1 and ρs(1 m) > ρs(2 m) are sortedwith friction velocity, and are shown in Fig. 13. The error barsshow one standard deviation and the histogram shows the

TABLE 3. Particle number fluxes recorded by the snow trap, μtrap at 0.25 < z < 0.35 m and the camera system, μcam, at z = 0.35 m.

Start Finish Duration mtrap μtrap μcam EfficiencyuTC uTC kg kg m–2 s–1 %

DOY 47, 20:20 DOY 47, 20:50 0 h 30 min 0.048 0.00251 0.00496 51DOY 47, 21:10 DOY 48, 00:50 3 h 40 min 0.520 0.00362 0.01450 25 DOY 48, 00:10 DOY 48, 16:00 14 h 50 min 1.340 0.00231 0.01250 18

v v z zr rv= ( )−

/ ,γ

γ = −

−−

ln / ln

d v u

d v u

z

z

p

p

s

s

232 1

131 2

2

1

,



number of observations used in each 0.05 m s–1 bin. Theexponent γ is approximately 1 for u* < 0.25 m s–1 and increases to approximately 1.3 for 0.25 < u* < 0.55 m s–1.This is slightly higher than the relatively constant value γ = 1found by Gordon et al. (2009) over sea ice at Franklin Bayusing particle counters at heights of 0.5 m and 2 m, assuming

a constant particle size with height. The nearly constant valueof γ = ω/κu* for 0.25 < u* < 0.55 m s–1 suggests that ω

increases with u*. This is as expected, because faster windswill support heavier particles, which have higher settlingvelocities.

0.1 1 10Blowing Snow Number Flux, v (106 m-2 s-1)

1H

eigh

t, z

(m)

0.2

2

-0.75

0.3

0.5

Fig. 11 Blowing snow number flux measurements for the same periods as the mass flux measurements of Fig. 10. A power law profile of v∝ z–0.75 is also shownfor comparison.

1 10 100 1000Number Flux, ν (10-6 m-2 s-1)

0.1

1

Heigh

t, z (m

)

0.0001 0.001 0.01 0.1 1 10Mass Flux, μ (kg m-2 s-1)

2

-0.75

-2.5

2

40 80 120 160 200 240Mean Diameter, d (μm)

0.1

1Height, z (m)

Fig. 12 Number and mass fluxes and mean diameter calculated using Eq. (2) with the data from DOY 48, from 03:00 to 12:00 uTC, with ω = ωT (solid lines)and ω = 0.7ωT (dashed lines). Power law profiles with exponents –0.75 and –2.5 are shown for comparison.



e Electric FieldFigure 14 compares the 10 m wind speed, U10, the numberdensity, N0.35 at 0.35 m above the snow surface, and the max-imum and average electric field strength, E0.5 at 0.5 m abovethe snow surface for two blowing snow episodes. The firstepisode started on DOY 47 (16 February 2008) around09:00 uTC and continued for approximately 40 h. The secondepisode started on DOY 81 (20 March 2008) around16:00 uTC and continued for approximately 32 h. The numberdensity is calculated using the lowest (0.35 m) particlecounter, the wind speed at a height of 1 m, and Eq. (1) with z0 = 0.14 mm to give N0.35 = v0.35/(0.88 U1). Note that elec-tric field measurements of 2.9 kV m–1 or less were not record-ed. The 10 m wind speed of 8 m s–1 is shown in the figure asa dashed line, as a rough approximation of the threshold windspeed required for blowing snow.

Figure 15 compares the 10 min average wind speed andnumber density measurements to the maximum measuredelectric field strength for each of the two blowing snowepisodes. Gordon and Taylor (2009b) have demonstrated acorrelation between the 0.5 m average electric field strengthand the 10 m wind speed in measurements made over sea iceat Franklin Bay. They determine best fits for various blowingsnow episodes of

(12)

where Uth is chosen for each episode by visual inspection aseither 6, 7, or 8 m s–1. They determined values of CU rangingfrom 1.5 to 4.6 kV s m–2, with an average of 2.9 kV s m–2.

Fitting the two episodes of Fig. 15 to Eq. (12) gives values ofCU = 2.0 kV s m–2 (DOY 47-48.5) and 1.5 kV s m–2 (DOY81-83.5) with Uth = 8 m s–1. These coefficients are within therange of values found by Gordon and Taylor (2009b). Theyreport 10 min average electric field strengths as high as 32.3 kV m–1, while the maximum 10 min average value seenhere is 14.6 kV m–1, and the maximum instantaneous value is26.2 kV m–1. In southeastern Wyoming, Schmidt et al. (1999)measured a 14 min average electric field strength of 3.6 kV m–1 at a height of 0.37 m, with an average wind speedof 12.5 m s–1, which is lower than the field strengths seen atthis Arctic location.

The average electric field strength at a height of 0.5 m isalso compared to number density at a height of 0.35 m. Linearbest fits of E = CN N give CN = 0.0037 V m2 (DOY 47–48.5)and CN = 0.0019 V m2 (DOY 81–83.5). These values are lessthan the average of CN = 0.044 V m2 found by Gordon andTaylor (2009b), using average values of E. However, the cor-relation of E and N in this study is stronger (R2 = 0.76 and0.85, both p < 0.001) than the correlation seen in the resultsof Gordon and Taylor (2009b) (0.13 < R2 < 0.66). This dif-ference in the value of CN may be due to the different heightsof the number density measurements in the two studies.Assuming a constant particle diameter of 110 μm, a chargedensity q = –25 μC kg–1 (Schmidt et al., 1998), and γ = 1.7(from Section 4d), Eq. (3) gives a ratio of E/N = 0.0039 V m2

at a height of 0.5 m, which is slightly greater than the highestvalue of CN found in this study. However, the model predictsa weak electric field, producing an acceleration of only0.09 m s–2 on a particle with q = –25 μC kg–1 at a height of

0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55Friction velocity, u* (m s-1)

0

0.5

1

1.5

2

2.5

Neg

ativ

e ex

pone

nt, γ

0

300

600

900

1200

1500

Num

ber of observations per 0.05 m s

-1 bin

Fig. 13 Average values of γ (diamonds) for each 0.05 m s–1 bin of friction velocity, u*. Error bars show one standard deviation. The step plot shows the num-ber of observations used of each average.

E C U UU th= −( ) ,



0.1 m, indicating that electrostatic forces are not significant atthis level.

f VisibilityThe MOr, measured at a height of 2 m, is compared to thenumber density calculated as N2 = v2/U2 using the 2 m parti-cle counter and the 2 m cup anemometer for all data with U10 > 6 m s–1 without observed falling snow or ice crystals.The results are shown in Fig. 16. As discussed in Section 3,prior to 22 February, measurements made by the northeastfacing sensor were sampled, not averaged, and are not used inthe analysis. Average measurements made by the northeastfacing sensor after 22 February are shown as + symbols andaverage values of the northwest facing sensor (installed on15 February) are shown as × symbols.

If Eq. (5) is best fit to the data from both sensors, it gives across-sectional-area weighted mean particle diameter of

da = 109 μm. The coefficient of determination is R2 = 0.83 (p < 0.001). This is close to the diameter of da ≈ 100 μmfound by Huang et al. (2008). For MOr > 2 km, the mea-surements of both visibility sensors is generally greater thanthe best fit, suggesting a smaller average diameter at lowernumber densities. Although studies have shown the shapeparameter, α, to vary considerably between different blowingsnow events and different heights (e.g., Budd, 1966; Gordonand Taylor, 2009a), using the range of 4 < α < 8 from thecamera system measurements (at a height of 0.35 m) gives1.06 <

—da /

–d < 1.12. For da = 109 μm, this gives a range of

d ≈ 100 μm at a height of 2 m.

5 Conclusions

Wind measurements show that the blowing snow at thisIqaluit site is primarily associated with strong, cold north-westerly winds, which are aligned with the Sylvia Grinnell

47 48 49 81 82 83Day of Year, 2008

0

5

10

15

20

25

E0.

5 (kV

m-1

)

0

1

2

3

4

5

N0.

35 (

106 m

-3)

0

4

8

12

16

20

U10

(m

s-1

)

Fig. 14 The 10 m wind speed, U10, 0.35 m number density, N0.35, and 0.5 m maximum (grey line) and average (black line) electric field strength, E0.5, for twoblowing snow episodes starting 16 February and 21 March 2008 (all times uTC). Measurements of E0.5 < 2.9 kV m–1 were not recorded.



Valley. This predominant direction is consistent with theresults of Nawri and Stewart (2006). However, southeasterlywinds also generated a number of weaker blowing snowevents throughout the study.

The threshold wind speeds for the initiation of blowingsnow are generally higher than the threshold wind speeds inthe Canadian Prairies found by Li and Pomeroy (1997).Excluding falling snow, threshold wind speeds based onhourly averages range from U10,th = 7 m s–1 to 12 m s–1. Theaverage value of U10,th = 8.6 m s–1 gives a blowing snow fre-quency of 9.4% for the duration of the study, compared to afrequency of 15.5% based on the model of Li and Pomeroy(1997), and a frequency of 5.1% based on meteorologicalobservations. The particle counters give a frequency ofapproximately 10%, suggesting that meteorological observa-tions may underestimate the true frequency of blowing snow.

There has been considerable discussion of the significanceof sublimation during blowing snow events (e.g., Pomeroyand Gray, 1995) and several models of this process exist (seeXiao et al., 2000). While sublimation may be an importantprocess under some circumstances, we generally found thatthe relative humidity with respect to an ice surface was close

to 100% during most of the blowing snow episodes weobserved near Iqaluit (see for example Fig. 4 and Table 2).Although sublimation may be important at the onset of blow-ing snow, we anticipate that once relative humidity nears100%, sublimation should have minimal effect on blowingsnow profiles and size distributions.

Best fits of the size distribution to Eq. (4) give 4.4 < α < 6.4for different blowing snow episodes. During one blowingsnow episode, best fits to Eq. (4) for each hour give an α

ranging from 3 to 8. Generally, higher values of α are seenwith higher wind speeds, and α remains relatively constantwith nearly constant wind speeds. Measurements of particlesize at a height of 0.35 m show a correlation (coefficient ofdetermination R2 = 0.69) between the particle size and wind speed, with the average particle size ranging from deqv = 70 μm at 5 m s–1 to deqv = 148 μm at 15 m s–1.

Measurements with snow traps demonstrate that the massflux profile follows a power-law profile between heights of0.1 m and 0.9 m, with a negative exponent γμ ≈ 2.5. For com-parable heights, this value is within the range of γμ valuesfound by Nishimura and Nemoto (2005) and Takahashi(1985) in Antarctica and by Font et al. (1998) in the Spanish

6 8 10 12 14 16 18U10 (m s-1)

0

4

8

12

16

20

E0.

5 (kV

m-1

)

6 8 10 12 14 16 18U10 (m s-1)

0

4

8

12

16

20

0 2 4 6 8N0.35 (106 m-3)

0

4

8

12

16

20

E0.

5 (kV

m-1

)

0 2 4 6 8N0.35 (106 m-3)

0

4

8

12

16

20

DOY 47.0 - 49.5

DOY81.0 - 83.5

Fig. 15 The average electric field strength compared to the 10 m wind speed (top panels) and 0.35 m number density (bottom panels) for the time periods of16 February (0:00 uTC) to 18 February (12:00 uTC) (left panels) and 21 March (0:00 uTC) to 23 March (12:00 uTC) (right panels). Best fit lines of E = CU (U10–Uth) and E = CN N are also shown, where Uth is visually estimated as 8 m s–1.



Pyrenees. Measurements with particle counters demonstratethat the number flux profile follows a power-law profilebetween the heights of 0.35 m and 1.85 m, with a negativeexponent γv ≈ 0.75. Solving Eq. (2) with a measured size dis-tribution of particles gives profiles similar in shape to themeasured profiles with ω = 0.7 ωT, suggesting that particlesfall 30% more slowly than the terminal velocity of equivalentspheres in still air. using the particle counter measurementswith the modelled diameters, the negative exponent γ of Eq. (2)is calculated for each 10 min average for 0.85 < z < 1.85 m. Forfriction velocities 0.25 < u* < 0.55 m s–1, γ ≈ 1.3, suggestingthat the particle settling velocity increases with wind speed.This is also suggested by the camera results, which demon-strate an increase in particle size, and hence settling velocity,with wind speed.

Measurements of the electric field strength at a height of0.5 m give 10 min average electric fields as high as 14.6 kVm–1, with a maximum instantaneous value of 26.2 kV m–1.This is generally weaker than the field strengths measured byGordon et al. (2009) over sea ice at Franklin Bay (maximum

of 32.3 kV m–1), but greater than the field strength measuredat a comparable height by Schmidt et al. (1999) in southeast-ern Wyoming (3.6 kV m–1). The field strength correlates well(R2 = 0.76 and 0.85) with number flux, which supports themodel of Gordon and Taylor (2009b). The model predicts arelatively weak electric field for heights above 0.1 m; howev-er, the particle motion may be affected in the saltation layer,where the strength of the field is likely stronger.

The MOr can be approximated (coefficient of determina-tion R2 = 0.83) using the number density with Eq. (5) and across-sectional-area weighted mean particle diameter of da = 109 μm. This gives an average particle size of approxi-mately

–d = 100 μm at 2 m, which is generally less than the

average particle size determined by the camera system at0.35 m.

Acknowledgements

Funding for this work was provided by the CanadianFoundation for Climate and Atmospheric Sciences (CFCAS).

0.01 0.1 1N2 (106 m-3)

0.1

1

10

MOR

(km)

Fig. 16 The 10 min average MOr compared to number density at a height of 2 m for blowing snow events without falling snow or ice crystals for measure-ments from the northeast facing sensor (+) and measurements from the northwest facing sensor (×). Also shown is a best fit of Eq. (5), which gives —da = 109 μm.



References

ABE, O. 1989. A collection efficiency of a net-type blowing snow trap meter.report of the National research Centre for Disaster Prevention, 43:247–254. (In Japanese with English Summary). Cited in Sato, T.; K.Kosugi and A. Sato. 2001. Saltation-layer structure of drifting snowobserved in wind tunnel. Ann. Glaciol. 32(1): 203–208.

BINTANJA, r. 2001. Snowdrift sublimation in a katabatic wind region of theAntarctic ice sheet. J. Appl. Meteorol. 40(11): 1952–1966.

BINTANJA, r. and M.r. VAN DEN BrOEKE. 1995. The surface energy balance ofAntarctic snow and blue ice. J. Appl. Meteorol. 34(4): 902–926.

BrOWN, r.D. and J.W. POMErOY. 1989. A blowing snow particle detector.Cold Reg. Sci. Technol. 16: 167–174.

BuDD, W.F. 1966. The drifting of non-uniform snow particles. In: rubin, M.J. (Ed.) Studies in Antarctic Meteorology. Antarctic research Series Vol.9, American Geophysical union, Washington, DC. pp. 59–70.

BuDD, W.F.; W.r.J. DINGLE and u. rADOK. 1966. The Byrd snow drift project:Outline and basic results. In: rubin, M. J. (Ed.), Studies in AntarcticMeteorology. Antarctic research Series Vol. 9, American Geophysicalunion, Washington, DC. pp. 71–134.

CArrIEr, C.F. 1953. On slow viscous flow. Final report contract Nonr-653-00/1, Office of Naval research, Brown university, Providence, r.I.

CHrITIN, V.; r. BOLOGNESI and H. GuBLEr. 1999. FlowCapt: a new acousticsensor to measure snowdrift and wind velocity for avalanche forecasting.Cold Reg. Sci. Technol. 30: 125–133.

CLIFT, r. and W.H. GAuVIN. 1971. Motion of entrained particles in gasstreams. Can. J. Chem. Eng. 49: 439–448.

DérY, S.J.; P.A. TAYLOr and J. XIAO. 1998. The thermodynamic effects of sub-limating, blowing snow in the atmospheric boundary layer. Boundary-Layer Meteorol. 89: 251–283.

DérY, S.J. and M.K. YAu. 1999. A climatology of adverse winter-type weatherevents. J. Geophys. Res. 104(D14): 16657–16672.

DOVEr, S.E. 1993. Numerical modelling of blowing snow. PhD thesis,university of Leeds, Department of Applied Mathematics.

FONT, D.; M. MASES and J.M. VILAPLANA. 1998. Experimental mass-flux mea-surements: a comparison of different gauges with estimated theoreticaldata. Ann. Glaciol. 26: 225–230.

GOrDON, M. 2007. Drifting and blowing snow, measurements and modelling.PhD thesis, York university, Department of Earth and Space Science.

GOrDON, M. and P.A. TAYLOr. 2009a. Measurements of blowing snow, part I:particle shape, size distribution, velocity, and number flux at Churchill,Manitoba, Canada. Cold Reg. Sci. Technol. 55: 63–74.

GOrDON, M. and P.A. TAYLOr. 2009b. The electric field during blowing snowevents. Boundary-Layer Meteorol. 130: 97–115.

GOrDON, M.; S. SAVELYEV and P.A. TAYLOr. 2009. Measurements of blowingsnow, part II: density profile and saltation height at Franklin Bay, NWT,Canada. Cold Reg. Sci. Technol. 55: 75–85.

HANESIAK, J.M.; T. FISICO, C. rINDAHL and E. CArrIErE. 2003. Climatology ofadverse weather events in the Canadian Arctic. Technical report CEOS-tech-2003-2, Geography Department, Faculty of Environment, universityof Manitoba, Winnipeg, MB.

HANESIAK, J.M. and X.L.L. WANG. 2005. Adverse-weather trends in theCanadian Arctic. J. Clim. 18: 3140–3156.

HANESIAK, J.; r. STEWArT, P.A. TAYLOr, K. MOOrE, D. BArBEr and 20 others.2010. Storm studies in the Arctic: the meteorological field project. Bull.Am. Meteorol. Soc. In press, doi: 10.1175/2009BAMS2693.1.

HuANG, Q.; J.M. HANESIAK, S. SAVELYEV, T. PAPAKYrIAKOu and P.A. TAYLOr.

2008. Visibility during blowing snow events over Arctic sea ice. WeatherForecast. 23(4): 741–751.

LI, L. and J.W. POMErOY. 1997. Estimates of threshold wind speeds for blow-ing snow using meteorological data. J. Appl. Meteorol. 36(3): 205–213.

MANN, G.W.; P.S. ANDErSON and S.D. MOBBS. 2000. Profile measurements ofblowing snow at Halley, Antarctica. J. Geophys. Res. 105(D19):24491–24508.

MOOrE, I.; S.D. MOBBS, D.B. INGHAM and J.C. KING. 1994. A numerical modelof blowing snow around an Antarctic building. Ann. Glaciol. 20: 341–346.

NAWrI, N. and r.E. STEWArT. 2006, Climatological features of orographiclow-level jets over Frobisher Bay. ATMOSPHErE-OCEAN, 44(4): 397–413.

NISHIMurA, K. and M. NEMOTO. 2005. Blowing snow at Mizuho station.Antarctica. Phil. Trans. R. Soc. A, 363: 1647–1662

POMErOY, J.W. and D.M. GrAY. 1990. Saltation of snow. Water Resour. Res.26(7): 1583–1594.

POMErOY, J.W. and D.M. GrAY. 1995. Estimating blowing snow transportrates and sublimation rates on the prairies of Western Canada. In:Snowcover: Accumulation, Relocation and Management. Chapter 6.NHrI Science report No 7. Environment Canada Publication En36-513/7E, 144 pp.

SATO, T.; T. KIMurA, T. ISHIMAru and T. MAruYAMA. 1993. Field test of anew snow-particle counter (SPC) system. Ann. Glaciol. 18: 149–154.

SAVELYEV, S.A.; M. GOrDON, J. HANESIAK, T. PAPAKOurIOKI and P.A. TAYLOr.

2006. Blowing snow studies in CASES (Canadian Arctic Shelf ExchangeStudy) 2003–04. Hydrol. Proc. 20(4): 817–827.

SCHMIDT, r.A. 1981. Estimates of threshold windspeed from particle sizes inblowing snow. Cold Reg. Sci. Technol. 4: 187–193.

SCHMIDT, r.A. 1982. Vertical profiles of wind speed, snow concentration, andhumidity in blowing snow. Boundary-Layer Meteorol. 23: 223–246.

SCHMIDT, r.A. 1986. Transport rate of drifting snow and the mean wind speedprofile. Boundary-Layer Meteorol. 34: 213–241.

SCHMIDT, D.S.; J.D. DENT and r.A. SCHMIDT. 1998. Charge-to-mass ratio ofindividual blowing-snow particles. Ann. Glaciol. 26: 207–211.

SCHMIDT, D.S.; r.A. SCHMIDT and J.D. DENT. 1999. Electrostatic force in blow-ing snow. Boundary-Layer Meteorol. 93: 29–45.

TABLEr, r.D. 1980. Self-similarity of wind profiles in blowing snow allowsoutdoor modelling. J. Glaciol. 26(94): 421–434.

TAKAHASHI, S. 1985. Characteristics of drifting snow at Mizuho station,Antarctica. Ann. Glaciol. 6: 71–75.

WIErINGA, J. 1973. Gust factors over open water and built-up country.Boundary-Layer Meteorol. 3: 424–441.

XIAO, J. 2001. A model of transport and sublimation of blowing snow. PhDthesis, CrESS, York university, Toronto, Ontario, Canada.

XIAO, J.; r. BINTANJA, S.J. DérY, G.W. MANN and P.A. TAYLOr. 2000, An inter-comparison among four models of blowing snow in the atmosphericboundary layer. Boundary-Layer Meteorol. 97: 109–135.

measurements of drifting and blowing snow at iqaluit, nunavut, …€¦ · 2 c e ntr f oe ah o bsv...

Documents