A Detailed Assessment of Snow Accumulation in Katabatic Wind Areas on the Ross Ice Shelf, Antarctica

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 102, NO. D25, PAGES 30,047-30,058, DECEMBER 27, 1997

A detailed assessment of snow accumulation in katabatic wind areas on the Ross Ice Shelf, Antarctica

David A. Braaten Departmentof Physicsand Astronomy,University of Kansas,Lawrence

Abstract. An investigationof time dependentsnow accumulation and erosiondynamics in a wind-sweptenvironment was undertakenat two automaticweather stations sites on the RossIce Shelf betweenJanuary 1994 andNovember 1995 usingnewly developedinstrumentation employinga techniquewhich automatically disperses inert, colored (high albedo) glass microspheresonto the snowsurface at fixed intervalsthroughout the year. The microspheresact as a time markerand tracerto allow the accumulationrate and wind erosionprocesses to be quantifiedwith a high temporalresolution. Snow core and snowpit samplingwas conducted twice duringthe studyperiod to identify microspherehorizons in the annualsnow accumulation profile, allowing the snowaccumulation/erosion events to be reconstructed.The two siteschosen for thisinvestigation have characteristically different mean wind speedsand therefore allow a comparativeexamination on the role of wind on ice sheetgrowth. Mass accumulationrate at the twosites for the 14-dayintegration periods available ranged from 0.0 to >2.0 kg m-2 d -l. The meanmass accumulation rate duringthe studyperiod was greaterat the site with strongerwinds (0.69kg m -2 d -1) than the site with lower mean wind speeds (0.61 kg m-2 d-l); however,the differencebetween the two meansis not statisticallysignificant. Accumulationrates derived from an ultrasonicsnow depth gauge operated at one of the sitesare comparedto the actualtracer-derived accumulationrates and show the limitationsof only having a measureof snow surfaceheight with no instantaneousmeasurements of the snowdensity profile. Snowdepth gauge derived accumulationrates were foundto be greatlyoverestimated during high-accumulation periods and were greatlyunderestimated during low-accumulation periods.

1. Introduction complex and has been examinedexperimentally by Budd et al. [ 1966], Kobayashi [ 1978], Wendlet [1989], and Maeno et ai. Polar ice sheets are an important component of the global [1995] using innovative experimental techniquesto examine climate system; however, their remote and harsh environment various characteristicssuch as mass flux and snow grain size confound and limit experimental attempts to understandtheir as a function of height and wind speed. Naturally, these dynamics and interactions with the atmosphere and oceans. processescan only occur when the wind speed exceeds some Polar ice sheetscontain approximately2% of the Earth's water threshold speed for snow grain movement (saltation), which and therefore negative changesin ice sheet mass balance can usually dependson the condition of the snow surface. Age- potentially have serious implications due to a rise in global hardened snow has a much higher threshold saltation speed sea level, although it is still an open question whether polar than recentlydeposited snow. Bromwich [1988] gives a range ice sheets grow or decay under a warming climate scenario of thresholdspeeds between 7 and 13 m s-l; however,during [Jacobs, 1992]. Polar ice sheets are also valuable archives of this study the saltation thresholdwind speedson the Ross Ice direct and indirect evidence of millennial-scale climatic events Shelf were observed to be between 5 and 6 m s-l. Wind and past atmosphericcomposition of trace gasesand aerosols. movement of surface snow results in a variety of different Proper interpretation of this evidence requires a thorough surface features which have been classified as depositional knowledgeof the dynamical processeswhich govern ice sheet and/or erosional patterns by Kobayashi and Ishida [1979]. growth. They concludethat wind speedsless than 15 m s-l (at 1 m Processes which control local ice sheet changes can be above the surface) producetransverse features such as ripples, categorized in terms of source and loss parameters. Source waves, and barchans, whereas wind speedsgreater than 15 m parameters include episodic snowfall events associatedwith s-1 producelongitudinal features such as dunes and sastrugi. synoptic scale features such as cyclones and fronts The microphysicalprocesses which producethese featuresare [Bromwich, 1988] and hoar frost deposition. Loss parameters complex, and essentially nonlinear, in that the turbulent include sublimation [Fujii and Kusunoki, 1982] and snow airflow distorts the snow surface which in turn distorts the erosion caused by drifting or blowing snow; however, snow flow. External characteristicssuch as amplitude,wavelength, grains removed by erosionmay be replacedby snow which has shape, and movement have been examined by Budd et al. eroded from upwind locations. The processof snow erosionis [1966] and Kobayashiand Ishida [1979]; however,important aspects such as snow grain transport and mixing into the Copyright1997 by the AmericanGeophysical Union. accumulation profile remain largely unknown. Since most Papernumber 97JD02337. precipitation events at the coastal margins of Antarctica are 0148-0227/97/97JD-02337509.00 associatedwith wind speedsgreater than the thresholdspeed

30,047 30,048 BRAATEN: SNOW ACCUMULATION IN KATABATIC WIND AREAS

I I I I I I 166øE 167"E 168øE 169øE 170øE 171 øE

Ross Sea 77.25 ø S-

McMurdo Sound Ross Island 77.50"s-

Ross Ice Shelf

77.75" S- - McMurdo/ Station Ferrell AWS

Willie Field WS 78.00" S-

78.25"

Wind Rose Scale Wind Rose Legend l• • , • I .... I --<5m/s 0 10 20 >=5m/s Percentage of Observations

i i i i i 18o ø

90 øE 90 øW

Figure 1. Location of automatic weather stations where snow accumulation experiments were conducted. Wind roses summarize the winds during the study period, showing the prevailing directions from which the winds were coming.

for blowing snow, sourceand loss parametersare interrelated new automatedsystem to dispersecolored glass microspheres in that layers of accumulatedsnow are generally a mixture of onto the snow surface at preset time intervals to act as a tracer. snow producedby individual stormsand from previousstorms These sites are Willie Field automatic weather station (AWS) which has been eroded and transportedfrom upwind locations (77.85øS, 167.08øE) located approximately 10 km east of [Bromwich, 1988]. McMurdo Station (MCM) and Ferrell AWS (78.02øS, This investigation was conducted to obtain greater 170.80øE)located approximately 100 km eastof MCM (Figure quantitative insight into the time dependentprocesses of ice 1). Evaluation of snow accumulation at each site was sheet snow accumulation in windswept areas. Experiments conductedin two phasesbetween January 1994 and November were carried out at two locations on the Ross Ice Shelf using a 1995. For the Willie Field AWS the two evaluationperiods BRAATEN:SNOW ACCUMULATION IN KATABATIC WIND AREAS 30,049 wereJanuary 18, 1994,to January28, 1995,and February l, corridor from 210 ø with a directional standard deviation of 1995, to November 2, 1995, and for the Ferrell site the 18.5'for wind speeds > 5 m s-l. evaluationperiods were January25, 1994, to January21, At Willie Field only, a CampbellScientific ultrasonic snow 1995, andJanuary 25, 1995,to November17, 1995. depthgauge and CRI0 data loggerwith a datastorage module supplementsthe AWS data,providing hourly measurements of distance to the snow surface with an accuracy of +10 mm. 2. Wind and PrecipitationCharacteristics These data were retrieved once per year. For most days, more 2.1. Automatic Weather Stations than 90% of these hourly measurements are valid. Occasionally, erroneous data were recorded for several The two AWS sitesused in this studyare part of a large array consecutivedays which was most likely due to ice crystal of stationsdeployed on the continent[Holmes et al., 1996]. buildup(hoar frost) on the sensorhousing. In 1994,data were EachAWS provides10 min measurementsof air temperature, missingfrom 23% of the days with the longestconsecutive wind speedand direction,atmospheric pressure, relative period being 30 days, and 22% of the days in 1995 were humidity, and vertical differential temperature(2 m) missingwith the longestconsecutive period being 20 days. transmittedto polar-orbiting satellites, although between Figures 2a and 2b give the daily averagesnow depth gauge 15% and 20% of the measurementsare intermittently lost derived accumulationduring the study period at Willie Field. becausea satelliteis not in positionto receivethe broadcast. These data show both large positive and negative Data are collected through the Argos Data Collection and accumulations on short timescales with a substantial long- LocationSystem (ARGOS) and processed by SpaceScience and term positiveaccumulation. Positivechanges are associated EngineeringCenter, University of Wisconsin-Madison with precipitation, wind blown transportof snow into the [Stearnsand Wentiler, 1988]. Sinceboth AWS sitesare target area, and/or the formationof a surfacefeature on the locatedwhere annual snow accumulationis approximately0.5 target area. Negative changescan resultfrom wind erosion, m yr-l, theheight of thewind speed measurement at both sites sublimation, gravitational forces which compact the snow, decreasedfrom --3 to -2 m during the studyperiod. The two and metamorphicprocesses of the snow crystals[Colbeck, studysites have characteristicallydifferent wind regimes, 1983]. Figure 2a correspondsto the period before the first althoughboth sites experience katabatic winds and are close field evaluation of snow accumulation processes,and Figure enoughto be influencedby the samesynoptic scale storms. 2b representsthe periodbefore the secondfield evaluation. FigureI provideswind roses for eachAWS, summarizingthe wind observationsduring the entire study period. Winds at 2.2. McMurdo Station Precipitation Willie Field AWS are moderated by the proximity to Observations topographicfeatures of RossIsland, White Island and Black Island,with 21.2%of the observationsexceeding 5 m s-I Daily precipitation summariesfrom McMurdo Station (approximatesnow grain saltation threshold) and prevailing derived from individual observations of trained weather directionsfor windswith speeds greater than 5 m s-• ranges observerswere obtained for the entire study period with the between 60 ø and 210 ø. It is not unusual for wind speedsat exception of August 1995 which was not available. Willie Fieldto exceed15 m s-1 duringkatabatic wind outflow Quantifying snowfall accumulation is difficult in the eventsin whichprevailing wind directions are mainlybetween windsweptAntarctic environmentand is usually estimated 160ø and 210 ø. Winds at Ferrell are strongly influenced by basedon intensity of precipitationand duration. The bars in moderate katabatic outflow winds with 44.1% of the Figures 2a and 2b identify when precipitation episodes observationsexceeding 5 m s-I anda dominantwind direction occurred and the estimated accumulation amount. Not included

700

600

500-

400- 300 ...re-

200

100

--I ...... I I I I I I I i__lI I .....I I I I I .•t[I I I I ll..hI I II I II I,.,I I

Figure2a. Theconnected data points show the daily average snow depth at WillieField obtained from an ultrasonicsnow depth gauge from the start of thefield experiment until the first field survey on January28, 1995. The verticalbars along the time axis show precipitation amounts reported at McMurdoStation. The datesgiven on thetime axis are the planned microsphere activations at WillieField. 30,050 BRAATEN:SNOW ACCUMULATION IN KATABATICWIND AREAS

1000

900

800

700

600

500

. 400

300

200

100-

Figure 2b. Sameas Figure l a, exceptfor the secondfield periodbetween February 1, 1995, and November 2, 1995. are the 245 days in which trace snowfall amounts(<1.3 mm) aerosol generator units mounted on aerodynamic masts and a were reported. A total of 55 precipitationepisodes occurring pneumatic system (buried -1.5 m below the snow surface) over one or more days are shown, with 16 episodesoccurring controlled by a microcontroller-based timing system to during periods in which the snow depth gauge data were activate the aerosol generatorsand disperse the microspheres. missing. While most of the positive snow accumulation The MDSs at both sites were programmedto activate every 14 changesindicated by the snow depth gauge were associated days for 10 s, dispersing -75 mL of the inert, colored with a precipitation episode, only 18 of the precipitation microspheresfrom a height of-1.2 m. The microsphereshave episodes were associated with a correspondingpositive a terminalsettling speed of 0.9 m s'l andrapidly fall to the changein snowaccumulation greater than the accuracyof the snow surface. A total of four different colors (pink, green, snow depth gauge. As for the remaining 21 precipitation orange, and yellow) are used by MDS, each dispersedduring events in which the snow depth gauge does not indicate a successive14 day periods; therefore any given microsphere correspondingchange in snow surface height, explanations color is dispersedin a 56 day cycle. At the start of each field for this include the possibility that the observedprecipitation evaluationperiod, yellow flat glassshards (150 to 425 I.tm) are was localized, the precipitation resulted in a local dispersed by hand onto the snow surface to allow an accumulation which was less than the accuracy of the snow unambiguousidentification of the base of the accumulation depth gauge,or that moderateto strongwinds restrictedthe layer. accumulation since the vast majority of these events were Performance of MDS during the first field period was associatedwith wind speeds >5 m s'l. evaluated based on a forensic examination of system components, volume of microspheres remaining in the 3. Detailed Characterization of Snow aerosol generator units and a snow core site survey. For the Accumulation second field period, pressuretransmitters were added to the pneumatic system and system pressure recorded hourly by 3.1. Microsphere Dispersal System Campbell Scientific CR10 data loggers. Ambient temperature To allow dating of snow accumulation horizons and to of the pneumaticsystem was also recordedby the CR10 once assessthe role of wind on accumulationdynamics at the two per day. Pneumatic system pressure provides a log of AWS sites, colored glass microspheres(with a high albedo) microspheredispersal activations by confirming an activation were dispersedonto the snow surface periodically throughout occurredor indicating a systemfailure and allows an accurate the study period as time markerswithin the growing ice sheet, estimateof the volume of microspheresdispersed to be made and as a tracer to quantify snow grain mixing and transport basedon the magnitudeof the systempressure change. For processes. A similar idea was used in earlier studiesof wind- each confirmed activation, the AWS at each site provided wind blown sand surfaces by Hunter [1977] who periodically speedand wind directionmeasurements within a few minutesof sprinkled a small amount of dark magnetite grains on a each MDS activation, allowing an accurate estimation of blowing sand surface to record the internal structure of the wherethe microsphereswere deposited. Figure 3 providessite sediment bed to successive deposition of sand grains. mapsof each MDS site during each deploymentand indicates Automated dispersal of the 120 I.tm diameter glass the direction from which the winds were coming and wind microspheres(density = 2.5 g cmø3) was accomplishedby speed (proportional to the length of the line) during each instrumentation referred to as the microsphere dispersal confirmed activation. During the study period a total of 45 and system (MDS) [Braaten, 1994, 1995] which consistsof three 46 MDS activationswere plannedfor Willie Field and Ferrell, BRAATEN:SNOW ACCUMULATION IN KATABATICWIND AREAS 30,051

First Field Survey- January 1995 Willie Field MDS site Ferrell MDS site o ß o o ß o øe ß e•e •õ øø o øo0 o ß oco ß ßß •e•e•e•e

O ß 0 ß

AWS70m AWS 80m

Second Field Survey - November 1995 Willie Field MDS site Ferrell MDS site

% ßo eo oo oo ß o oo øo o ooøOoø

ß • 0 ßD1 ß ß o ß o

AWS 80m AWS 80m

Leaend

ßo Core Sample withwithoutMicrospheres Microspheres I N -Microsphere Generator Unit I] Snow Pit with Microspheres • I I I I] Snow Pit without Microspheres 0 5 10 15 Wind direction during MDS activation meters •=5ms-1 Figure3. Sitemaps for eachfield sampling campaign at WillieField and Ferrell showing locations of microspheregenerators, snow cores, snow pits, and the wind speed and direction the wind was coming from during confirmedmicrosphere activations. respectively,with an actual percentageof successful able to rebuild and test these regulatorsbefore deployingthem activationsbeing 47% for Willie Field (21 activations)and for the secondfield survey. As can be seen in Figure 4, the 52% for Ferrell (24 activations). This does not include the secondfield surveyhad about the samefailure rate as the first. manualdispersals of glassshards at the beginningof each field evaluationperiod at eachsite. Figure4 showswhen each 3.2. On-Site Snow Sampling successfulactivation occurredand when a planned activation Reconstructionof the time dependentsnow accumulation did not occur due to a malfunction. All system malfunctions processesoccurring during each surveyperiod is basedon wereattributed to the gaspressure regulators of the pneumatic observed vertical and horizontal spatial distribution of the system;however, the MDS pneumaticsystem is designed colored glass microspheres,the dates of successful using two gas pressureregulators such that if one microspheredispersals, and on-sitemeteorological data. The malfunctions,microsphere dispersals will continueat 28 day initial deposition location of microspheresfrom each intervals instead of the normal 14 day intervals. Pressure activationis initially estimatedusing the AWS wind speedand regulationfailures were manifestas eitherreduced pressure direction observationsat the time of activation (Figure 3) and downstreamresulting in a large reductionin gas flow rate a known microspheredistribution under calm conditions duringactivation or high pressuredownstream locking the [Braaten, 1995]. At the conclusion of each field survey solenoidvalves in the closedposition. These problemswere period, snow core sampling and snow pit sampling is foundto occur when pneumaticsystem ambient temperatures conductedat each site with the snow sampleskept frozen and reachedapproximately -30 ø C despitea minimumoperating returnedto a laboratoryin the Crary Scienceand Engineering temperaturerating of-40 ø C for theseregulators. This Center at MCM for microsphereidentification. Laboratory problemwas discovered at theend of thefirst field survey in protocolsfor identificationof microspheresinclude melting January1995, but due to thetight field schedule, we wereonly each sample,filtering all material from the meltwater,and 30,052 BRAATEN:SNOW ACCUMULATION IN KATABATICWIND AREAS

I I I I I I I

S PGOYPGOYP O P O P O P O PS PGOYP O P Ferrell - 00000000000000000000000000 0000000000000000000000

SPGOYPGOY SPGOYP OYP O P O P 0 Willie Field - 000000000000000000000000000 00000000000000000000

, , ! , 1/1/94 4/1/94

Figure4. Microsphereactivation schedule at Willie Fieldand Ferrell for dispersalof yellowglass shards (S), pinkmicrospheres (P), greenmicrospheres (G), orangemicrospheres (O), andyellow microspheres (Y). The solidcircles indicate a successfulsystem activation, and open circles indicate the system did not activate as planned due to a malfunction.

examining the filter media with a low-power dissecting directionswere very similar during dispersals of bothpink and microscope,noting the numberand color of microspheres orangemicrospheres. A possiblecause of thisdiscrepancy is present, if any. Using these procedures,it is routine to the occurrenceof wind erosionsoon after dispersalof the identifya singlemicrosphere in any givensnow sample. orangemicrospheres. To test this hypothesis,average wind speedswere calculated for 24, 48, 72, 96, and 120hour periods 3.3. Snow Cores after eachconfirmed microsphere dispersal, and theseresults Snow cores 0.8 m in length and 0.0045 m2 in cross- were groupedby microspherehorizons positively identified sectionalarea were taken from the studysites with a Sipre by snowpit samplingand those which were not. The green auger. These samples provide an overview of the on-site and yellow microspheredispersal events (two each) were microspherespatial distribution integrated over the entire excludedfrom the analysisbecause these microsphere colors accumulation period and are compared to an expected were identifiedin snowcore analysesbut not in snowpit distributionobtained using wind speedand wind directiondata analyses,and thereforeit is unknownwhich dispersal(s) the from the time of dispersalto identify possiblewind erosion recovered microspheres are associated. The mean of each events. Figure 3 showsthe locationsof thesesite survey groupwas calculatedfor eachof the five time periodsand a t snow cores obtainedduring the study period. Microspheres testwas employed to assesswhether the two groupmeans were were found in 71% of these snow cores, and Table l statisticallydifferent at the 0.05 level of significance.These summarizesthe corescollected at eachsite during each survey resultsare shownin Table 2 and supportthis wind erosion period and the microspheres identified. Also shown in hypothesis.For all averagingperiods the meanwind speeds parentheses are the number of confirmed activations for each associated with the identified horizons are less than the color. In general, with fewer activationsof a given color, a smaller numberof surveycores contained microspheres with that color. This is consistentwith expectationssince the Table 1. Summaryof SipreSnow Cores Obtained and the microspheredeposition area of each aerosolgenerator is Fractionof CoresWhich Contain a ParticularMicrosphere typicallyless than 10 m2 [Braaten,1995] and the snowcore Color surveyscover an area which variesbetween 150 and 400 m2. For two periodswith the sameor nearlythe samenumber of Fractionof CoresWith Microspheres activations(Willie Field, period 2, pink and orange, and Ferrell, period l, pink and orange), large differencesin Survey Cores microspherespatial distribution were found. To explainthese Period Extracted Pink Green Orange Yellow differences, wind characteristics during microsphere activationswere examined. In the first case,five dispersals Willie Field eachof pink and orangemicrospheres at Willie Field during surveyperiod 2 resultedin a largefraction of coreswith pink I 34 0.41 (2) 0.44 (2) 0.15 (2) 0.18(2) microspheres,but a small fraction of cores with orange 2 20 0.70 (5) 0.17 (1) 0.30 (5) 0.05 (2) microspheres(Table l). During the pink microsphere dispersals,wind direction varied by 126ø , whereaswind Ferrell directionduring orange microsphere dispersals were all within a surprisinglynarrow -24ø sectorwhich explainswhy these I 40 0.63 (7) 0.10 (2) 0.10 (6) 0.13(2) microsphereswere found in sucha smallfraction of thesurvey 2 25 0.56 (3) 0.04 (1) 0.16 (2) 0.00(l) cores. In the secondcase, sevendispersals of pink and six dispersalsof orangemicrospheres at Ferrell duringperiod 1 The numberof activationsfor a particularmicrosphere color are resulted in a similar discrepancy;however, this time, wind given in parentheses. BRAATEN: SNOW ACCUMULATION IN KATABATIC WIND AREAS 30,053

Table 2. Group Mean Wind Speedsfor 24, 48, 72, 96, and 120 Hour Periods After Confirmed MicrosphereDispersal

Events During the First Study Period ..

. . at Ferrell. .

Mean Wind Statistically Speed, Mean Wind Different Time After Identified Speed,Missing Means? (0.05 Dispersal, Horizons, Horizons, Level of hours m s-• m s-• Significance)

24 3.38 5.71 No

48 3.82 6.69 No 72 3.52 6.47 Yes •"•?•"'.'•'•"'-'

96 3.86 5.95 Yes

120 4.05 5.69 No

Data were grouped basedon whether microspherehorizons were identified or not during site sampling. A t test at the 0.05 level of significancewas pertbrinedto determinewhether the group meansfor each time period were statisticallydifferent.

missing horizons; however, only the 72 and 96 hour averaging periods have differences which are statistically significant. These results on the fate of newly deposited microspheres also offer insights into the mobility of snow grains after snowfall events, in that high wind speeds (exceeding the saltation threshold) occurring during a period -4 days after new precipitation causes significant downwind transport of snow grains. If wind speeds remain low during Figure 5. Collecting snow samples in cuvette tubes from a snow pit wall using an apparatuswhich ensuresexact vertical this period, the new precipitation has a chance to age alignment l¾om a bubble level on the base. sufficiently and become resistant to wind erosion.

3.4. Snow Pits surface features. High-resolution measurements of snow density were also obtained from the cuvette samples by Snow pit sampling provides high-resolutionintbrmation of measuring the volume of snow and the volume of water in the microsphere horizons, allowing a detailed reconstructionof cuvette after melting. the time dependent snow accumulation to be made. Snow pit Table 3 summarizes the high-resolution snow pit samples locations were chosen based on winds during microsphere obtained at each site giving the number of snow pit profiles, dispersals (Figure 3), prevailing winds (Figure 1), and the the number of cuvette samples, and the fraction of samples results of the snow core survey (Figure 3). Snow samplesare collectedin disposablecuvette sampling tubes (1 cm2 x 4.5 which containedeither microspheresor glass shards. Because of persistent high winds during the second Ferrell site visit, cm) which are pushed into the snow pit wall. A custom built no snow pit profiles were obtained. Instead, 12 additional snow pit sampling apparatus, shown in Figure 5, is used to rapi(•y obtain accurately positioned, high-resolution (1 cm depth) snow pit samples. The apparatus holds five cuvette Table 3. Summaryof High Resolutionof Snow Pit Samples tubes at a time, which are guided into the snow pit wall to and the Percentageof Cuvette SamplesContaining extract snow samples. A bubble level on the apparatusensures Microspheres or Shards exact vertical alignment, and prepositioned sampling locations in both vertical and horizontal axes with known Survey Number Number PercentageWith spacing accurate to -0.05 mm allow an entire 0.5 m snow pit Period Profiles Samples Microspheresor Shards profile to be sampled after making a single sample tube depth measurement. The depths of subsequent sample tubes are Willie Field calculated based on the reference tube depth, the sample tube I 7 440 6. l number (1 through 5) and the apparatus position. Most microspherehorizons sampled were found in only one sample 2 l0 470 6.0 tube indicating that the microspheres stayed within a layer Ferrell less than I cm in depth and therefore did not migrate vertically after deposition to the surface. The vertical mixing of I 7 495 4.4 microsphereswhich did occur was attributed to wind-generated 30,054 BRAATEN:SNOW ACCUMULATION IN KATABATICWIND AREAS

Sipresnow cores were extracted. These cores were cut into 156 the survey period, but if this horizon was not identified at the samples3 to 4 cm in lengthand analyzedfor microspheresand site, the profile with the greatest number of identified snow density. The percentageof core sectionswhich were microspherehorizons was designatedas the referenceprofile. found to have microspheresor shardswas 20.5%. This much No attempt was made to match an entire accumulationprofile largerrecovery fraction compared to the cuvettesamples is due with the reference profile since local accumulation is often to the greatervolume of snowanalyzed (by a factorof-40); influenced by formation and movement of surface features however, the vertical resolution is reduced and the uncertainty resulting in different relative accumulationrates throughout in the vertical position is increased. the study period. A good example of how microspheredata from severalsnow 3.5. Snow Horizon Dating pit profiles can be normalized, and how variation in A "snap shot" of snow accumulation variability was accumulation rates during the year can be identified at two determinedfor each studyperiod at each site basedon the depth nearby locations (-4 m apart), is given by Figure 6 which of the glassshards scattered over the dispersalarea at the start shows the snow density profiles, locations of visible ice of the period, several snow stakes positioned around the layers (dashed lines), locations of pink (P) and green (G) dispersalarea and the AWS snow depth gauge (Willie Field microspherehorizons, and the location of the glass shard (S) AWS only). Snow pit samplingat the end of the first study horizon from samplingconducted at Willie Field in November period at Willie Field failed to identify the glass shard 1995. Profile Al (Figure 3) was designatedas the reference horizon, but the first microspherehorizon from February 1, profile with the glass shard horizon identified at 37.8 cm. 1994, indicated an accumulation of 520. + 18. mm, and the Comparingthe depth of the glassshard horizon to that of the snow depth gauge measurementlocated 70 m south of the pink microsphere(38.4 cm) and green microsphere(35.2 cm) dispersalarea gave an accumulationof 558. mm for this same horizonsin profile B1 (Figure 3), it is apparentthat by simply period. Snow stakespositioned 30 m north, west and east of combining the results of the two snow profile produces a the dispersalarea provided an accumulationmeasurement of sequencewhich is out of chronologicalorder and indicatesthat 424. +__42. mm for a period beginning 2 weeks earlier on some adjustmentmust be made. Normalizing profile B 1 to the January 18, 1994. During the secondperiod the glass shard referenceprofile was accomplishedby mappingthe reference horizon was identified and indicated an accumulation of 363. _+ profile snow density pattern versus depth near 34 cm to a 15. mm, with the snow depth gauge located 80 m west, similar pattern near 36 cm in profile B l, resulting in a measuringan accumulationof 367. mm. Snow stakeslocated correction of-2 cm to the depth of the two microsphere 30 m north, south, east and west of the dispersalarea provided horizonsin profile B 1. Upon further examinationof the snow an accumulation measurement of 354. + 17. mm. At Ferrell density profiles in Figure 6, it was recognizedthat ice layers during the first period, the glass shard horizon indicated an and density profile patterns closer to the surface in both accumulation of 655 + 5. mm, while an array of snow stakes profiles matched but were offset by approximately +5 cm. located 30 m north, south, east and west of the dispersalarea Using microsphere horizon information from other snow gave an accumulationof 539. + 29. mm. During the second period, the glass shard horizon indicatedan accumulationof 363 + 16. mm, while a similar array of snow stakesas the first SnowDensity (g cm -3) SnowDensity (g cm -3) period gave an accumulation of 373. _+ 41. mm. Snow 0.3 0.4 0.5 0.6 0.7 0.3 0.4 0.5 0.6 0.7 o i i i i i i i i i i i i i i i i accumulationresults at Willie Field AWS show a consistently iiii i• iii, iiii 11/95 good relationshipbetween the microspheredispersal area and - ? t• -• . the snow depth gauge; however, the snow stake agreementat both sites is mixed. This probably results from the small lO • +5(cm number of snow stakes deployed as compared to other snow stake arrays [Mosley-Thompsonet al., 1995]. For a more detailed look at the snow accumulationprocess within the dispersalarea, the microspherehorizons identified 2o in the snow pits were used. Assigningdates to microsphere 6/95 horizons identified is not a simple process, since not all microsphere horizons are found during snow pit sampling - (even though they might be found during the snow core survey) and the local snow surfaceheight can vary by several cm centimetersdue to the intermittentpresence of surfacefeatures • •• -2c (e.g., ripples and sastrugi). Therefore horizonsof the same microspherecolor found at slightly different depthsin two 4o 2/95 different snow pit profiles are not necessarilyfrom different P dispersalevents, and conversely,the samecolor microspheres found at roughly the same depth might actually be from different dispersal events. To overcome this inherent 50 uncertaintyin dating microspherehorizons, the vertical snow density distribution and the depth of visually observed ice Figure 6. Example of snow pit (left) A1 (reference) and layers were used to normalize the depth of microsphere (right) B1 at Willie Field (1995) showing measured snow horizons to a reference snow pit profile. The primary density with depth and the locationsof the visible ice layers consideration in choosing the reference profile was (dashed lines), yellow glass shard horizon (S), pink identification of the glass shard horizon indicating the start of microspherehorizons (P) and green microspherehorizon (G). BRAATEN: SNOW ACCUMULATION IN KATABATIC WIND AREAS 30,055

SnowDensity (g cm -3) SnowDensity (g cm-3) SnowDensity (g cm -3) 0.3 0.4 0.5 0.6 0.3 0.4 0.5 0.6 0.3 0.4 0.5 0.6 0 I .... I .... I .... I 0 • .... I .... • .... I 0 iiii iiii Iiii . • o

10 _ p 10 10 Figure 7. Exampleof snowpits (left) A1 (reference),(middle) D1, and (right) A2 at Willie Field (1995) showingmeasured snow density with depthand the locationsof the pink microspherehorizons (P) andorange microspherehorizon (O). profiles, the earlier of these ice layers is estimatedto have core slices instead of snow pit cuvette sampling to identify formedaround the beginningof June 1995. This indicatesthat microsphere horizons. the accumulation rate for the reference profile is greater than For a more quantitative comparison of accumulation for profileB1 for the first 4 monthsof the surveyperiod, and between the two sites, the detailed snow density measurements the accumulation rate of profile B1 was greater than the were used to calculate mass accumulation rate integrated over referenceprofile for the remainderof the period. Overall,the periods dated by the microspherehorizons, the initial snow net accumulationobserved in profile B I was 3 cm more than surface indicated by the glass shards, or the snow surface the reference profile. height at the time of snow pit sampling. Figure 10 gives the During periods with little snow accumulation,accurate calculated mass accumulation rate for all integration periods at datingof the microspherehorizons identified in differentsnow each site during both survey periods. The dashed lines give pit profilesagain must rely on snowdensity profile patterns the weighted mean mass accumulation rate for each survey and/or visible stratigraphicfeatures to obtain a normalized period. The period-integrated mass accumulation rate depthfor thesehorizons. Figure7 showsthe snowdensity increasedat Willie Field betweenperiods 1 and 2 (0.59 to 0.64 profiles,and locationsof pink (P) microspherehorizons at kg m-2 d'l), anddecreased between periods 1 and2 at Ferrell 2.2 cm (referenceprofile A l) and 3.0 cm (profile D l), and an (0.78 to 0.58 kg m-2 d'l). The weightedmean for the entire orange(O) microspherehorizon at 3.9 cm. ProfilesA I andD1 studyperiod at WillieField was 0.61 kg m-2 d -1 andat Ferrell (Figure3) are separatedby a distanceof 13 m, andprofiles A1 was 0.69 kg m-2 d-• Although the weighted mean and A2 (Figure3) are only I m apart. Normalizingthe depths accumulationrate at Ferrell is greater than at Willie Field, the of the pink horizonfound in profileD1 andthe orangehorizon difference in the two mean values is not statistically in profile A2 to the referenceprofile was basedon similar significant at the 5% level. The integrationperiods of 14 to snow density patternsnear 2.5 cm in the referenceprofile, 1.0 28 days shown in Figure 10 provide some indication of the cm in profile D1 and 3.4 cm in profile A2. Therefore the variability of the short-term mass accumulationrate ranging normalized pink microspherehorizon depth in profile D I is from0 to morethan 2.0 kg m-2 d'• Consideringboth sites, 4.5 cm, indicating that the two pink microsphere horizons none of the seasonsduring the study period were found to have shown in Figure 7 are from different activations (August 2, consistently low or consistently high mass accumulation 1995, and September 27, 1995). The normalized orange rates. In addition, no consistent relationship was found microspherehorizon depth in profile A2 is 3.0 cm, placing it between mass accumulation rate and mean wind speed for the between the two pink microspherehorizons and dating the integration periods, implying that high wind speeds do not horizon at August 30, 1995. necessarilyreduce the mass accumulationrate in areas with an Once the microsphere horizons identified from snow pit unlimited upwind supply of mobile snow grains. Differences sampling have been normalized, dates are assignedto the in the mass accumulation rates at the two sites are more likely horizons based on the known times of dispersal for each governed by differences in the mesoscale precipitation microspherecolor, counting forward in time from the glass patternsassociated with stormsmoving throughthe area. shard horizon, and counting back in time from the snow The microsphere horizons and correspondingsnow density surfaceat the time of snow pit sampling. The ultrasonicsnow profile provide an opportunity to relate snow surface height gaugedata, when available,are also usedas input in assigning measuredby an ultrasonic snow depth gauge with actual mass dates to the microsphere horizons. The locations and accumulation rates for individual periods. Data have already uncertaintyof microspherehorizons along with the date of been presentedshowing that snow surface height changes at dispersalto the snow surfaceat Willie Field and Ferrell during the snow depth gaugeare in good agreementwith snow surface the two field study surveys are shown in Figures 8 and 9, height changesat the microspheredispersal area. To obtain respectively. Figure 8 also includesthe ultrasonicsnow gauge mass accumulation rate from a snow surface height change, a deriveddepths at Willie Field (when available)around the time snow density profile must be either known, modeled or of a scheduledmicrosphere activation for comparison. Ferrell assumed. In this case the snow density profile is known at the did not have an ultrasonic snow depth gauge installed on site. microspheredispersal site, and this same profile is assumed The uncertaintyin placing the depth of a microspherehorizon for the snow depth gauge site. Two characteristicallydifferent ranged from the dimensionsof the cuvette samplingtube (1 snow accumulationperiods are examinedin Table 4 to quantify cm) to several centimeters caused by encountering a mixed the differences in the actual mass accumulation rate determined horizon, or by observed depth variations of the horizon in in the microsphere dispersal area and the estimated mass different snow pit profiles. The noticeably larger accumulationrate at the snow depth gauge. The first period uncertaintiesin the secondsurvey period at Ferrell (Figure 9) (April 12, 1995 to July 5, 1995) is a period with a large snow comparedto the first surveyperiod are the resultof usingsnow depth change, and the second period (August 2, 1995 to 30,056 BRAATEN: SNOW ACCUMULATION IN KATABATIC WIND AREAS

lO A

AA A

,- 30 AA

60 A ' I ' I ' I ' I ' I ' I 1/1/94 3/1/94 5/1/94 7/1/94 9/1/94 11/1/94 1/1/95

0- A " ,•

.-. 15

O - v ,- 20 A A

Q 25

401/1/95 , 3/1/95i , 5/1/95• , 7/1/95, , 9/1/95• , 11//951•

Figure 8. Microspherehorizons identified at Willie Field for the two field periods(top) endingJanuary 28, 1995, and (bottom) endingNovember 2, 1995, are shownas verticalbars which indicatethe date of dispersal and the depth uncertaintyin the snow profile. The trianglesgive the ultrasonicsnow depth gaugederived depthsaround the time of a scheduledmicrosphere activation (when available).

20 ÷

E • 30

• 40

70 I I ' I ' I i 1/1/94 3/1/94 5/1}94 7/1/94 9/1/94 11/1/94 1/1/95

20:

35 Ii 40 I ' I ' I " I ' I 1/1/95 3/1/95 5/1/95 7/1/95 9/1/95 11/1/95 Figure 9. Microspherehorizons identified at Ferrell for the two field periods(top) endingJanuary 23, 1995, and (bottom)ending November 23, 1995, are shownas verticalbars which indicate the date of dispersaland the depth uncertainty in the snow profile. BRAATEN: SNOW ACCUMULATION IN KATABATIC WIND AREAS 30,057

2.5-- Willie Field

2.0

1.5 H

1.o I I I 0.5 I-I

0 I I

-0.5 1/1/94' 7/1/94' I ' ; 0/1•/9•

2.5-- Ferrell

2.0

1.5 I I

1.o i

I I I 0.5 I I

-0.5 ' I ' ' I ' 7/1/94 10/1/94

Figure 10. The solid horizontal bars give the mass accumulationrate integrated over periods dated by the microspherehorizons, the glass shard horizon, or the snow surfaceheight at the time of snow pit sampling. The dashedlines indicatethe weightedmean massaccumulation rate for each survey period.

November 2, 1995) is a period with a small snow depth mass accumulationby the snow depth gauge is the inability of change. The length of time covered by both periods allows this instrument to identify when decreases in snow surface short-term fluctuations in the snow surface height to be height causedby densificationand metamorphosisof the snow ignored. During the first period the microsphere-derivedsnow profile are being offset by the accumulation of new depth change was observed to be 4.6 cm less than that precipitation and/or wind-blown snow. measuredby the snow depth gauge,and if the samemean snow densityfor these layers is assumed,the massaccumulation rate 4. Conclusion for this period is overestimatedby 34%. During the second period, microspherederived snow depth changewas observed The aim of this research has been to quantity snow to be 3.3 cm greater than that measured by the snow depth accumulation in two windswept ice shelf locations in gauge, indicating the snow depth gauge underpredictedthe Antarctica with good time resolution. To accomplishthis, a massaccumulation rate for this period by 375%. The primary newly developed instrument was used which dispersedcolored cause of both the overestimations and the underestimations of glass microspheres at fixed intervals to act as time markers

Table 4. Comparisonof MeasuredMass Accumulation Rates Using Snow Horizons Dated Using Microspheres and Changesin SnowSurface Height Using an UltrasonicSnow Depth Gauge

Microsphere- Snow Gauge Microsphere- SnowGauge MeasuredMean DerivedMass DerivedMass Derived Snow Derived Depth SnowDensity, AccumulationRate, AccumulationRate, Period DepthChange, cm Change,cm kgm -3 kgm '2 d 'l kgm -2 d -I

April 12, 1995 13.7 18.3 510. 0.832 1.111 to July 5, 1995

Aug. 2, 1995 4.5 1.2 448. 0.217 0.058 to Nov. 2, 1995 30,058 BRAATEN: SNOW ACCUMULATION IN KATABATIC WIND AREAS within the growing ice sheet and as wind-blown tracers. The determinedusing microspheresduring periodswith large two study sites were located adjacent to automatic weather (small)precipitation inputs. This is attributedto the inability stations. Both sites were influenced by katabatic outflow of the instrumentto identify when decreasesin snow surface winds with one site (Willie Field AWS) having lower average heightcaused by densificationand metamorphosis of the snow wind speeds than the other (Ferrell AWS). Microspheres profile are being offset by snowaccumulation. chosen for these experiments had a similar terminal settling speedas that of typical snow grains and thereforeare a good Acknowledgments. The author is indebtedto Edward Zeller and analog in analyzing snow grain wind erosion. Meteorological GiselaDreschhoff for sharingtheir polar experiences and for providing data were provided by the automatic weather stations, and theinspiration and encouragement to pursue the subject of thispaper. The authoralso thanksC. R. Stearns,G. Weidner,and R. Holmesof the precipitation observations were obtained from nearby Space Scienceand EngineeringCenter, University of Wisconsin- McMurdo Station. In addition, an ultrasonicsnow depthgauge Madison for providing the automaticweather station and ultrasonic was operatedat one of the automaticweather stations. snowdepth gauge data, as well as frequentlyproviding helpful advice. Weighted mean mass accumulationrates calculatedfor the The authorgratefully acknowledges the assistanceof C. Yess and S. two sites showedthat Ferrell, with higher overall wind speeds, Delfelderin field samplingand laboratoryanalysis of snowsamples, andthe helpful suggestions and comments of two anonymousreviewers had the largest accumulationrate. However, the difference in thatimproved the text. This investigationwas supported by National the weighted mean mass accumulation rates was not ScienceFoundation grants DPP-9218868 and OPP-9417255. statistically significant at the 5% level of significance. These resultsgo againstthe conventionalwisdom that the magnitude References of the accumulationis generally inversely related to the wind speed magnitude, but additional data to improve statistical Braaten, D.A., Instrumentationto quantify snow accumulationand confidence are needed to settle this issue. One possible transportdynamics at two locationson the RossIce Shelf, Antarct.J. U.S., 29(5), 86-87, 1994. explanation of these results is that the sites have a virtually Braaten, D.A., A new techniqueto provide high time resolution unlimited upwind supply of erodible snow which replaces snowpackdating for stratigraphyand chemistryassessments, Atmos. snow eroded from the site and that observed accumulation Environ., 18, 2535-2539, 1995. differences are due mainly to differences in mesoscale Bromwich,D.H., Snowfallin highsouthern latitudes, Rev. Geophys., 26, 149-168, 1988. precipitation patterns affecting the sites. Another important Budd,W.F., W.R.J.Dingle, and U. Radok,The Byrdsnowdrift project: question which should be examined in future researchis the Outlineand basic results, Studies in AntarcticMeteorology, Antarct. effect of even strongerwinds on local snow accumulationrate. Res. Set., vol. 9, edited by M.J. Rubin, pp. 71-134, AGU, Over shorter 2 and 4 week integration periods the mass Washington,D.C., 1966. accumulationis highly variable ranging from 0 for periodsof Colbeck,S.C., Theory of metamorphismof dry snow, J. Geophys.Res., 88, 5475-5482, 1983. either no precipitation or wind erosion to more than 2.0 kg Fujii, Y., and K. Kusunoki,The role of sublimationand condensationin m-2 d-1 (Ferrell)for a periodassociated with one or more the formationof ice sheetsurface at Mizuho Station,Antarctica, J. significant precipitation events. For these shorterperiods, no Geophys.Res., 87, 4293-4300, 1982. strong correlation between mean wind speed and mass Holmes, R.E., C.R. Stearns, and G.A. Weidner, Antarctic automatic weatherstations Field Report:1996, 19 pp., SpaceSci. andEng. accumulation rate was found, although this analysis with Cent., Univ. of Wis., Madison, 1996. larger data sets is ongoing. In addition, no consistent Hunter, R.E., Basic types of stratification in small eolian dunes, seasonal trend of mass accumulation was identified in this data Sedimentology,24, 361-387, 1977. set, which is surprising given the influence of sea ice Jacobs,S.S., Is the Antarcticice sheetgrowing?, Nature, 360, 29-33, 1992. development and retreat on cyclonic storm trajectories. Kobayashi,S., Snowtransport by katabaticwinds in MizuhoCamp Additional data are neededto confirm this finding as well, but Area,East Antarctica, J. Meteorol.Soc. Jpn., 56, 130-139,1978. these results point to a large role that wind-blown snow plays Kobayashi, S., and T. lshida, Interaction between wind and snow in overall net accumulation on the Ross Ice Shelf. surface,Boundary Layer Meteoroi., 16, 35-47, 1979. The performance of the ultrasonic snow depth gauge has Maeno, N., K. Nishimura,K. Sugiura,and K. Kosugi,Grain size been assessedusing nearby precipitation observations and dependenceof eolian saltationlengths during snow drifting, Geophys.Res. Lett., 22, 2009-2012, 1995. microsphere-derived snow mass accumulation rates. While Mosley-Thompson,E., J.F. Paskievitch,M. Pourchet,A.J. Gow, M.E. most of the snow accumulationevents identified by the snow Davis, and J. Kleinman, Recent increase in South Pole snow depth gauge were associatedwith an observed precipitation accumulation,Ann. Giaciol, 21, 13l- 138, 1995. event, less than half of the observedprecipitation events were Stearns, C.R., and G. Wendlet, Research results from automatic weatherstations, Rev. Geophys.,26, 45-61, 1988. identified by the snow depth gauge. Although localized Wendlet,G., On the blowingsnow in AdelieLand, Eastern Antarctica, precipitation events could explain some of these missed in GlacierFluctuations and ClimateChange, edited by J. Oerlemans, events, high winds accompanying most of the precipitation pp. 261-279, Kluwer Acad., Norwell, Mass., 1989. events likely limited snow accumulation and mask the precipitation event. These results illustrate the limitations of D. A. Braaten,Department of Physicsand Astronomy, University of Kansas,Lawrence, KS 66045. (e-mail:[email protected]. ukans.edu) using snow gauge data to identify precipitationevents. In addition, snow depth gauge data were shown to significantly (ReceivedNovember 20, 1996;revised August 13, 1997; over (under) state snow accumulation rates independently acceptedAugust 15, 1997.)