JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 97, NO. D1, PAGES 917-930, JANUARY 20, 1992

Atmospheric Net Transport of Water Vapor and Latent Heat Across 70øS

M. B. GIOVINETrO

Department of Geography, University of Calgary, Calgary, Alberta, Canada

D. H. BROMWICH

Byrd Polar Research Center, Ohio State University, Columbus

G. WENDLER

Geophysical Institute, University of Alaska, Fairbanks

The annual net atmospheric transports of water vapor and latent heat poleward across 70øS are estimated using the latest compilation of surface mass balance for the Antarctic ice sheet and new estimates of precipitation and evaporation in sectors of the southern oceans and of seaward drifting snow transport in particular sectors of the ice sheet. The mass and energy exchange rates at the ice sheet-atmosphere and ocean-atmosphere interfaces are integrated strictly for areas within that latitude. The estimates of net southward water vapor transport (6.6 + 1.3 kg m -1 s -1) and latent heat transport (18.9 + 3.6 MJ m -1 s -1) are larger than reported in all preceding studies, based on atmospheric advection and moisture data collected at stations located between 66øS and 80øS, and are generally in agreement with those based on surface mass balance data and seaward drifting snow transport across the ice terminus which extends between 65øS and 79øS.

INTRODUCTION sectors 61øW- 76øW and 1øW (eastward) - 160øE, and

The water and energy budgets of the area poleward of 70øS including ocean areas lying south of the reference latitude in the are important elements of models used in atmospheric, sectors 1øW - 61øW and 76øW (westward) - 160øE. It is based on glaciological, and oceanographic studies. Two basic terms in estimates of the mean annual net difference between surface estimates of those budgets are the atmospheric net transports of accumulation and ablation, or surface mass balance (herein water vapor and latent heat, the latter treated as a function of the referred as balance), in the ice sheet, and on the difference former. In the last three decades, estimates of net vapor transport between precipitation and evaporation, or outflow, in the ocean [e.g., Bryan and Oort, 1984] have been made using data for areas. Therefore drifting snow transport needs to be considered Antarctica, excluding Graham Land (Figure 1). Most estimates in two contexts: as an output term in the case of transport have been based on upper air data, hereafter referred to as northward across 70øS on the ice sheet and as an input term in atmospheric data, from data sets of approximately 10 stations [e.g., the case of transport across the periphery of the ice sheet, or ice Peixoto and Oort, 1983] located at relatively short and regular terminus, to the ocean areas lying within the reference latitude distances from each other and presenting an acceptable fit relative (Figure 2). to the reference latitude from 45øW (eastward) to 160øW [Rubin and Weyant, 1963]. Such is not the case in the remaining sector, BACKGROUND where only data from Byrd Station at a latitude of 80øS have been The available estimates of poleward moisture transport from used. A few studies have included comparisons of estimates atmospheric measurements result generally in underestimates. based on atmospheric data with those based on various Bromwich [1990] compared the moisture transport convergence combinations of surface accumulation and ablation terms and other values for Antarctica from two recent climatological mass budget terms such as calving, hereafter referred to as surface atmospheric analyses with those inferable from the balance data, integrated for the ice sheet [e.g., Rubin, 1962; Bromwich, compilation of Giovinetto and Bentley [1985]. For the continent 1990]. Thus the atmospheric data are directly representative of as a whole the atmospheric-derived estimates were only one third vapor transport across the boundary of an area of approximately to one half the surface-based values; for 80øS - 90øS the two 12 x 106 km 2, which boundary extends between 66øS and 80øS, approaches yielded comparable results. Bromwich [1990] and surface data are integrated for an area of approximately 13.7 x concluded that the primary causes of the large discrepancy for 106 km 2, which boundary extends between 65øS and almost 79øS. the continent were deficiencies in atmospheric diagnoses of the These estimates are then extrapolated to apply to the area within transport contributions of cyclones and surface winds near the 70øS or 15.51 x 106 km 2.

The assessment of net vapor transport presented in this work is made specifically for the area within 70øS (Table 1) excluding the areas of the ice sheet lying north of the reference latitude in

Copyright 1992 by the American Geophysical Union.

Paper number 911D02485. 0148-0227/92/91 JD-02485505.00

coast. A secondary cause contributing to the discrepancy was the small systematic underestimate by radiosondes of the atmospheric moisture content during the cold polar night.

A chronological overview of estimates of water vapor transport across 70øS based on either atmospheric or surface data is presented in Table 2. The values are arranged by publication date, and only analyses drawing upon data collected during or after the International Geophysical Year 1957-1958

917

918 GIOVINETTO ET AL.: ATMOSPHERIC WATER VAPOR TRANSPORT ACROSS 70oS

90øW -]-

'"...GRAHAM .. LAND

oC• '• • .....& Norway

Halley

Bay.:/F-"'" ½• .. .•_ -',- 80'S Ellsworth

ß '" • FILCHNER I. S. ....... .," ............... ; "" RONNE I, S.{•

> -',-

! Novolazarevskay•

Roi

EA S T .... '¾2

South Pole Sta. , F 2(c) ..... t'"' --90"E -,; • ,- -:; -',- J- T ' s;;;s. -'- 7o'0s ,o,s• .os / • '!,,,,,r•, / • d / l'• ...... : \ .'.,..e ,,,, • • T :/ / •t' .'"

..• \ ,, :.\ '%/ J GF o8 '),.• \ ':::',• .•Y'• • • -,- / ' •f:

O\ c.. ' _.•:•A N T A R C T • C A o 'r • "- • /- _ • % '•.•,• •o / / • %. A N T A R C T I C A/

" \ • '_-'., :,,, • \ I \. / • ?.Casey

• • u• 4,,' ...o.• •, r'x, I . soøs •' / ,.,2' 0 •, • ...... .•. • -- e _•..' ,, .. . , .•• • • .. O • ...'•.. ,. •.?••(• •.•j,,•....d L.•i. ' .............. •. McMurdo /

• ......... Ameri• V .......... • % • / • J.;

- ,ga,,e, / ,,o, ,,-

Fig. 1. Antarctica, showing the location of places mentioned in the text, ice sheet and ocean areas within the 70th parallel (full circle), stations for which upper air data have been used in studies of vapor transport (Port Martin, D 80, and GF 08, are stations discussed in the context of surface wind speed), and transects (a)-(c) used as examples in Figure 2. The Southern Ocean surrounds the continent. Areas are listed in Table 1.

(IGY) are considered. Estimates preceding the IGY were small equatorward to a poleward transport of 5 kg m -1 s -1' The primarily conjectural because of very limitedobservations. There trend in atmospheric estimates parallels the degree of is a marked contrast between the atmospheric and surface-based monitoring of atmospheric processes over the ocean areas estimates. The latter reveal little trend with publication year, surrounding the continent, ranging from a few oceanic fluctuating around a poleward transport of 6 kg m -1 s -1; whereas, observations during the IGY winter to the comprehensive the atmospheric estimates show a strong trend, changing from drifting buoy program in the Southern Ocean during 1979

TABLE 1. Terms Used in the Estimate of Net Water Vapor Transport Southward Across 70øS (V)

AgT•rms

•m '2 a '1 1015 t•g a '1 V Terms,

•otS •g ,4

Ice sheet (area: 11.95 x 106 km 2) Balance (B)* Drdting snow (Di)

Ocean sectors (area: 3.56 x 106 km 2) Precipitation (P s) Evaporation (Es) Outflow (F) Drifting snow (Di)

Total of terms used in the estimate of V

* From Giovinetto and Bentley [1985].

124 1.48 1.48 ... 0.12 0.12

531 1.89 ... 170 0.61

... <0.01 ...

GIOVINETTO ET AL.: ATMOSPHERIC WATER VAPOR TRANSPORT ACROSS 70oS 919

(S I E I Df I M I ) 5 I"" ' ( PI + Svsl+ C I + DPl)' + + + -• Ps-Es >11 •

' 1 I Grounded ice •' i,•,/, ',terminus ,I 0 '-'

V I ! Ice shelf barrier -,',-,, •,V • , I % I I

......... , , % [ I [ I I [ [ I

70 ø 80 ø 90 ø S 80 ø 70 ø •

Fig. 2. Schematic cross sections of Antarctica (along lines (a)-(c) shown in Figure 1) depicting the disposition of terms discussed in the text.

[Hollingsworth, 1989]. It appears that a substantial fraction of Antarctica as a whole the balance data offer better areal the poleward moisture transport shortfall in early atmospheric diagnoses was due to limited data in the eastern Pacific sector, where approximately 40% of the moisture transport into Antarctica occurs [Lettau, 1969] and which was well monitored only during 1979. Interannual variability complicates this

coverage and at present are more reliable than the atmospheric data. In the last two decades the widespread use of stable- and radioactive-isotope methods, the development of sampling techniques that incorporate deeper strata representing longer periods and the proliferation of stake networks have conlxibuted

comparison because atmospheric estimates tend to be based on to substantial reductions in the error of point balance rate data for a single year and surface estimates are based on data determinations [e.g., Giovinetto and Bull, 1987]. Moreover, that are generally the mean of overlapping multiyear periods. It increases in the number of point data, which in the last two can be concluded that, although the atmospheric analyses show steady improvement, they are not yet suitable for continent-wide studies of the atmospheric water balance. However, for 80øS - 90øS and perhaps a large fraction of East Antarctica they may be used for this purtx)se [Bromwich, 1990].

The underestimate may be explained in part by the paucity of atmospheric data in the coastal sector 70øW - 160øW, for which Lettau [1969] indicated the probability of large tropospheric

decades increased from approximately 350 to over 1200, and areal coverage, together with the production of detailed topographic maps, have contributed to substantial reductions in the error of interpolation and extrapolation of those rates and thus of areally integrated means, which may now be estimated with an error of approximately +_.10% [Giovinetto and Bentley, 1985].

Estimates of net vapor lxansport based exclusively on surface eddy and time-averaged circulation vapor lxansport. It has been data for the ice sheet are a direct assessment of particular known for some time that the vapor transport into that sector is accumulation and ablation terms, typically combined under the large based on both meteorological studies [Alt et al., 1959; headings of precipitation and evaporation in the production of Wexler, 1959; Rubin and Giovinetto, 1962] and balance water budget atlasses of the world showing continental and compilations. On the basis of the latter, Giovinetto [1961] oceanic summaries [e.g., Baumgartner and Reichel, 1975; suggested that the largest balance rates in Antarctica would be Korzoun et al., 1977]. Estimates of net vapor lxansport based in found in the coastal zone of the Bellingshausen Sea sector part on assessments of drifting snow seaward across the ice (Figure 3). This inference was later supported by field terminus must be made with caution because the range of observations of Shimizu [1964] and Behrendt [1965], as studied estimates of drifting snow reaching the sea as reported in the by Lettau [1969], and substantiated by Bishop and Walton [1981] literature is very large [e.g., Rubin, 1962; Budd et al., 1966]. and Potter et al. [1985]. For this sector in particular as well as for Methods to improve snow drift measurements at particular

TABLE 2. Estimates of Mean Annual Net Water Vapor Transport Across 70øS (V) Using Data Collected During and After the International Geophysical Year 1957-1958

Data Base, period or V *, number of years kg m '1 s '1

Based on atmospheric data Starr et al. [1969] 1958 (+)0.73 Peixoto and Oort [1983] 1963-1973 (-)3.0 Howarth [ 1983 ]; Howarth and Raynet [ 1986] 1973-78 and 1980-84 (-)3.7 Masuda [1990] 1979 (-)5.3

Based on surface data

Rubin [1962] multiyear (-)5.6 to (-)7.2•' Baumgartner and Reichel [ 1975] multiyear (-)5.4 Bromwich [1990] multiyear (-)5.8 to (-)6.0t This study multiyear (-)6.6 + 1.3

* Negative southward (following standard meteorological convention). t Transport across the coast of the continent converted for this study to transport across 70øS using the ratio of the

two values given by Baumgartner and Reichel [1975].

920 GIOVINETTO ET AL.' ATMOSPHERIC WATER VAPOR TRANSPORT ACROSS ?0oS

o..

0.5

-I-

3

3

-0.5

0

iii I I I I

.,! !

I',. / ' i

t

Fig. 3. Distribution of surface balance on the ice sheet (modified after Giovinetto and Bentley [1985]) and its extrapolation to show a probable distribution of precipitation in the ocean areas poleward of 70øS. Units are in 100 kg m '2 a -1'

locations and for different snow surface and boundary layer conditions have been developed [e.g., Budd et al., 1966; Kobayashi, 1978; Wendler, 1989a], but uncertainties in the interpolation and extrapolation of the mean rate, as well as of the actual length of the periphery to which the mean rate may be assumed to apply, remain the principal cause of the wide range of estimates. Typically, measurements made at locations in the coastal area are qualified for the effects of topography on surface winds, as these may be deduced for areas of the order of 103 km 2 but not for the effects of broad ridge and trough topography extending upslope, respectively, of capes and bays at or near the grounded ice terminus. For a detailed topography, see Drewry [1983]. The broad ridge and trough topography induce fanning and channeling of the airflow in areas of the order of 105 km 2 and consequently large uncertainty in interpolation and extrapolation. Furthermore, the varying availability of fine-grained dry snow, both as precipitation and surface material dislodged by saltation and corrasion and the frequency of strong surface winds sustained by air drainage over long slopes of diverse steepness are difficult phenomena and elements to synthesize. Thus, a reduction in the composite error would be achieved if all or substantial parts of zones contiguous

to the grounded ice terminus could be excluded, particularly as they contain the steepest parts of terminal slopes where katabatic flow is sustained by surface inversion conditions in the interior [Lettau, 1966; Lettau and Schwerdtfeger, 1967]; these conditions lead to the highest frequencies of strong winds [e.g., Schwerdtfeger, 1970; Parish and Waight, 1987] and largest rams of drifting snow transport. We obtain that objective by limiting our study to the area strictly within the 70th parallel because it excludes a sector approximately from 30oE to 160oE, where the grounded ice sheet extends northward of that latitude (an area of approximately 1.76 x 106 km 2) and where more than 3/4 of the periphery of the ice sheet characterized by steep terminal slopes and strong katabatic winds is found.

APPROACH

Our approach to the estimate of net vapor transport across 70øS (V), based on mass exchange data for the surface of both the ice sheet and ocean areas south of that latitude (Figure 2), includes:

1. Estimates of balance in the ice sheet (B) and of drifting snow losses from the ice sheet northward across 70øS (Do).

2. Estimates of the difference between precipitation (Ps) and

GIOVINETTO ET AL.: ATMOSPHERIC WATER VAPOR TRANSPORT ACROSS 70ø5 921

evaporation (Es) in the ocean areas, or outflow (F), and of In climatological studies of water balance at the air-ocean drifting snow transport from the ice sheet across the ice terminus interface for whole oceans or their hemispheric parts and at to the ocean areas lying within 70øS (Di). Then smaller scales for most seas it is customary to show the

distribution of the precipitation rate by extending isohyets from

V = (B + Do) + (F + Di) (1) land to sea areas. In the case of the ocean areas lying poleward of 70øS the distribution of precipitation is inferred by seaward

In the ice sheet, B is obtained by integrating contoured extrapolations of balance isopleths as determined for the coastal distributions of the balance rate as determined for a particular zone of the ice sheet (Figure 3); Ps is obtained by integrating location (b). The specific rates of mass exchange processes at the those contoured distributions [e.g., Baurngartner and Reichel, surface are not integrated separately; rather, b is determined as a 1975; Korzoun et al., 1977]. The extrapolation of balance net value on the basis of one or more annual accumulation isopleths as isohyets deserve some qualifications, which follow. layers: It is well known that in the coastal areas of the grounded ice

sheet the balance rate is smaller than the pr•ipitation rate [e.g., b=(Pl+Svsl+Cl+Dpl+I1)-(Sl+El+Dfl+M1) (2) Bromwich, 1990]. Over distances of the order of l00 km from

the grounded ice terminus the annual balance is positive because where P1 is precipitation, Svs 1 is vapor to solid sublimation, C1 is condensation, Dpl is drifting snow deposition, I 1 is superimposed ice, S 1 is sublimation, E 1 is evaporation, Dfl is snow deflation, and M1 is melt runoff.

An update of a previous review of snow stratigraphy in the ice sheet [Giovinetto, 1964a] shows that southward of 70øS there is

the summation of the rates of precipitation and drifting snow deposition is greater than the rate of snow deflation. This situation changes in a relatively short distance across the grounded ice terminus, since over the relatively flat sea ice, where wind speed typically d•reases beyond the zone contiguous to the terminus with a width of approximately 15 km [e.g.,

no melting at the surface in approximately 90% of the area. In a Schwerdtfeger, 1970], it may be assumed that the rams of drifting narrow zone along the coast of West Antarctica, as well as in the Ross Ice Shelf and the Siple Coast [Alley and Bentley, 1988], surface melting occurs sporadically in summer; this water percolates into the underlying winter accumulation layer and refreezes. In relatively few locations and only for a few years in long series does the stratigraphy show that percolation has taken place into other annual accumulation layers. Thus the potential for percolation to introduce errors in the determination of the balance rate is restricted to a relatively small area, and it may be totally ignored because most determinations of the rate are made on the basis of several annual accumulation layers. In the remaining area enclosed by the dry snow line, some snow and ice surfaces may be exposed for one or more years [e.g., Fujiwara and Endo, 1971; Picciotto et al., 1972; Nishio et al., 1985]. These types of surfaces have been observed in parts of some of the regions depicted in Figure 3 by the zero and 0.2 x 102 kg m -2 a -1 balance isoplcths; the regions lie well above the dry snow line. There are some conflicting reports as to what the actual surface type and balance may be in all or part of some of those regions [e.g., Mcintyre, 1985; Allison et al., 1985; Seko et al., in press]. However, the regions are relatively small and differences in the estimate of b for their areas are not significant in the estimate of B. From this brief discussion of snow stratigraphy and surface types found in the ice sheet southward of 70øS, it may be stated that in the area of interest the rates of vapor to solid sublimation, condensation, superimposed ice, evaporation, and melt runoff are negligible and that the rates of sublimation and snow deflation are particularly large in some regions. In the context of the latter term it may be noted that an increase in drifting snow leads to an increase in the density of air both by

snow deposition and snow deflation are approximately equal, and hence balance and precipitation rates should also be nearly equal. (It should be noted that complex distributions of rates of drifting snow deposition and deflation, and balance, over distances of the order of 10 km on either side of grounded and floating ice termini [e.g., Giovinetto, 1964b] involve relatively small areas and do not affect our estimate of terms such as B or

P s.) It remains to determine a correction to bring the (smaller) balance rate, as estimated or measured over the terminal slopes to a value that approximates the actual (larger) rate of precipitation.

A correction for the difference between the rates of

precipitation and balance in the terminal slopes is incidentally applied when the balance isopleths are extrapolated as precipitation isohyets. The compilations of balance are produced on small-scale maps on which the complexity of the balance rate distribution in the terminal slopes cannot be shown [Giovinetto and Bull, 1987], with the exception of a few relatively small s•tors which lie outside the area of interest, such as those where

stations Syowa, Mawson, Davis, and Casey are located (Figures 1 and 3). In the areal integration of balance from small-scale maps a deflation correction is applied to most sectors characterized by steep terminal slopes where deflation is much greater than deposition [Giovinetto, 1964b; Giovinetto and Bentley, 1985], but the balance isopleths as mapped do not include the deflation correction, so that their balance rates approximate the precipitation rates. Moreover, precipitation is large in the zone of terminal slopes due to orographically induced uplift of poleward moving air masses [e.g., Rubin and Giovinetto, 1962; Oerlemans, 1982; Brornwich, 1988], a precipitation enhancing

cooling from sublimation of drifting snow and by the phenomenon that is absent beyond approximately 100 km incorporation of particles, thus increasing the speed of katabatic offshore, where for much of the year sensible heat flux from air flow [Wentiler, 1989a]. This suggests a potential for positive feedback episodes that, however short, may produce larger rates of sublimation and snow deflation than have been assessed in

previous studies. Nevertheless, these losses are implicitly accounted for in the estimate of balance.

In the ocean areas, as stated above in 2, F is estimated as the difference between precipitation and evaporation:

F=Ps- Es (3)

to either water or sea ice promotes stability in the lower tropospheric layers inhibiting precipitation [e.g., Zillman, 1972; Andreas and Makshtas, 1985]. We assume that these overestimate and underestimate biasses introduced by the extrapolation procedure cancel each other out.

E s is estimated for open water and sea ice areas as a function of Le (latent heat eddy flux) which is in turn estimated in several studies using modifications of the basic equation, for example as given by Zillman [1972]:

922 GIOVINETTO ET AL.: ATMOSPHERIC WATER VAPOR TRANSPORT ACROSS 70oS

Le = k (dL/p) (vpsf - vph) (u) (4)

where k is a factor determined by the units used and levels at which the observations are made, d is density of air, L is latent heat of vaporization, p is atmospheric pressure, vps f is vapor pressure at the surface, vph is vapor pressure at level h, and u is wind speed at a particular level. Thus Le may be integrated to estimate the net heat energy flux (positive upwards), which at a point and time includes all processes involving latent heat transfer, such as condensation and vapor to solid sublimation on sea ice in winter, evaporation from ocean and (wet) sea ice surfaces in summer [e.g., Gow and Tucker, 1990]. We estimate E s using existing estimates of Le, as described in the following section.

ESTIMATES OF N•:'r WATER VAPOR AND

LATENT HEAT TRANSPORTS

Estimates of B and F

The compilation of balance for the ice sheet by G iovinetto and Bentley [1985] is used to estimate B (Figure 3); that compilation indicates a balance for the area south of 70øS of approximately 1.48 x1015 kg a '1 or 124 kg m -2 a -1 (Table 1).

The projections of balance isopleths from that compilation over the ocean areas, where technically they become isohyets, are shown in Figure 3. They are drawn showing orientations parallel to trends in the frequency and areal distributions of cyclones [Taljaard, 1972; Kep, 1984] and cloud vortices [Streten and Troup, 1973; Carleton, 1981], and larger rates downwind

from areas where large polynyas and leads form frequently in autumn through spring [Streten, 1973; Carsey, 1980; Zwally et al., 1983; Comiso and Gordon, 1987] as well as ice free areas in summer [Naval Oceanography Command Detachment (NOCD), 1985]. The isohyets pattern indicates a probable distribution of precipitation in the ocean areas south of 70øS, the main features of which are: (1) The low precipitation pockets in areas adjacent to the coast in the southern Weddell Sea and southwestern Ross Sea inferred from well-substantiated low balance values on the ice sheet; (2) The relatively low precipitation in the western areas of both seas inferred from high elevation in Palmer Land and Victoria Land oriented perpendicular to the cyclonic tracks, in part deflecting cyclones and in part inducing "rain shadow" effects on the ocean areas

downtrack; (3) The high precipitation in the northeastern Weddell Sea inferred from the relatively high frequency of cyclolysis and the probability of polynya formation north and east of the 600 kg m -2 a -1 isohyet; and (4) The high precipitation in the eastern Amundsen Sea - Bellingshausen Sea area, inferred from the well-substantiated high balance values on the ice sheet, as well as meteorological analyses applicable to this sector as mentioned in a preceding section.

The distribution of precipitation indicates mean rates for the ocean areas of approximately 544 kg m '2 a '1 in the Pacific sector and 504 kg m '2 a -1 in the Atlantic sector, for a combined-sector mean rate of approximately 531 kg m '2 a -1 and a total precipitation of approximately 1.89 x 1015 kg a '1 (Table 1).

The estimate of E s is derived in tabular form using existing estimates of Le for ice-free ocean, open water when ice is present, and sea ice (Tables 3a and 3b). In our estimate the particular surface types just described are each assumed to be relatively homogeneous regarding surface parameters such as albedo, roughness length, and substrate thermal diffusivity because of their roughly circumpolar distribution.

GIOVINETTO ET AL.' ATMOSPHERIC WATER VAPOR TRANSPORT ACROSS 70oS 923

TABLE 3b. Estimate of Latent Heat of Evaporation (Le, W m -2) in the Ocean Sectors South of 70øS

Summer Autumn Winter Spring XII-H III-V VI-VIH IX-XI

Mean for water* 20•' 40 100:[: 40

Mean for ice* 17õ 9õ (nil)• 9

37

Mean for area 18 16 10 13 14

* Latent heat flux values are factored by the weighted indexes listed in Table 3a. t Extrapolated from Zillman [1972]. $ Interpolated from Andreas et al. [ 1979] and Brown [ 1990]. õ Adopted with modifications from Allison [1972] and Maykut [1978]. • Adopted with modifications from Maykut [1978] and Weller [1980].

The first step in the estimate of Le is an examination of the The second step is to estimate Le for particular surface types areal and temporal distribution of ice-free ocean, open water and seasons to ensure reliable interpolations and extrapolations when ice is present, and sea ice. Using semimonthly charts applicable to the mean latitude for the ocean area south of the prepared by the NOCD [1985] for the period 1973-1982 based on 70th parallel, i.e., approximately 73øS. Weller [ 1980] estimated passive-microwave data from Nimbus 5 and Nimbus 7, as well as latent heat flux for sea ice in midwinter in the sector 20øE - visible and infrared imagery particularly during summer months, 90øE, between 58øS and 70øS with concentrations of 15-85% it is possible to summarize the occurrence of each of the three and > 85% (37 and 28 W m -2, respectively). Allison [1972] types of surfaces (Figure 4). It is necessary, however, to estimated latent heat flux at 67.5øS during freeze-up time in differentiate between particular sea ice concentration categories early autumn, and early stages of ice formation to a thickness of (.g,. 20%, 20-50%, 50-80%, and> 80%)to estimate the frequency 0.5 m (values ranging from approximately 80 to 20 W m -2, and area of open water. In this context it should be noted that respectively). Zillman [1972] estimated latent heat flux for ice- both the range of sea ice concentration and open water are free ocean in summer in the sector 110øE - 160øE, between 58øS simplified for this study by assuming a middle range value, and 68øS, listing 6-hour mean fluxes for bands of 4 ø of latitude although the data compiled by the NOCD only represent the overlapped every 2 ø (values between 15 and 17.5 W m -2, range of sea ice concentration in each category. The sea ice increasing southward with length of day). concentration data summarized in Figure 4 are used to obtain Estimates of latent heat flux for sea ice at high latitude in the seasonal area and time-weighted indices measuring the area of central Arctic in winter show small negative values [Maykut, each particular sea ice concentration in each particular season 1978]. Estimates of latent heat flux for open water leads in relative to the total graph area (Table 3a). For brevity, the winter at approximately 71øN near Point Barrow, Alaska, have results of this work are shown for the Pacific and Atlantic sectors combined. Moreover, for easy comparison with existing estimates been estimated at 50 W m -2 [Andreas et al., 1979], and elsewhere of Le (particularly those of Zillman [1972]) the results are in the northem polar region at 150 W m -2 [Brown, 1990]. summarized on the basis of four seasons, each consisting of three Maykut [1978] has shown that latent heat flux, among other calendar months.

SUMMER I- AUTUMN I WINTER • SPRING • SU.-" 100

90

• 8o

• 70

• 50 _> 80

40

'• 3o

Z• 20 50-80 o

lO •

o

j F J A S O N D

Fig. 4. Composite annual distribution of mean sea ice concentration in the ocean areas poleward of 70øS based on an analysis of semimonthly charts for the period 1973-1982. The seasonal labels correspond to the distribution of latent heat flux discussed in the text.

terms of the energy budget of sea ice, is highly dependent on ice thickness in the range of from a few millimeters to approximately 0.5 m. This and other models show little change in latent heat flux after thickness increases beyond 1 m [Weller, 1980]. In the ocean area poleward of 70øS, ice thickness exceeds 1 m in at least 9 months of the year [Washington et al., 1976; Hiikkinen, 1990]. Therefore the values obtained by Weller [ 1980] for midwinter at a latitude lower than 70øS and relatively thin ice should be drastically reduced to reflect the lower values obtained by Maykut [1978] for higher latitudes and thicker ice, particularly as earlier studies suggested that net condensation and vapor to solid sublimation may occur in winter on sea ice in the Southern Ocean [Fletcher, 1969]. Also, the values obtained by Allison [ 1972] for early autumn at a latitude lower than 70øS and relatively thin ice, should be reduced, although not as drastically as Weller's, in light of Maykut's results.

The values we adopted for our estimate of latent heat flux are shown in Table 3b and are implicitly the values assigned to a latitude of 73øS. For ice-free ocean and open water when ice is present they range between 20 W m -2 in summer (extrapolated from Zillman [1972]) and 100 W m -2 in winter (adopted from Arctic data studies by Andreas et al. [1979] and Brown [ 1990]). For sea ice they range between negligible in winter (adopted from Maykut [1978]) and 17 W m -2 in summer (modified from Allison [1972] allowing for differences in time of the year,

924 GIOVINETTO ET AL.: ATMOSPHERIC WATER VAPOR TRANSPORT ACROSS 70oS

latitude, and ice thickness). In each of the surface types we Even though a fixed relationship between wind speed and arbitrarily assigned intermediate values for autumn and spring, transport cannot be expected, we assumed such a relationship to although there are studies that suggest different values for the obtain order-of-magnitude values for the amount of drifting transitional seasons [e.g., Fletcher, 1965, 1969; Washington et al., snow from estimates of wind speed. For Antarctica, where most 1976; Maykut, 1986; Brown, 1990; Gow and Tucker, 1990]; of the zone along 70øS lies within the area covered by dry snow however, their discussion is beyond the scope of our present work. [Giovinetto, 1964a], the errors introduced by this assumption are The weighted means for each season, as well as the weighted minimized, as hardness of the snow cover and size distribution of annual means obtained for water and sea ice surfaces, 37 and 8

W m -2, respectively, indicate a mean latent heat flux for the area of 14 W m '2. Using values of latent heat of vaporization for phase changes at 0 C, i.e., 2495 J g-1 for water, and 2830 J g-1 for ice (increased to account for sublimation or melting), we estimate E s to be approximately 170 kg m -2 a -1 or 0.61 x 1015 kg a '1. Thus F is estimated to be approximately 361 kg m -2 a -1 or 1.28 x 1015 kg a -1 (Table 1).

Estimates of Do and Di

To complete the estimate of V, it is necessary to estimate Do northward across 70øS in the ice sheet sectors 61øW - 76øW and

1 øW (eastward) - 160øE, and Di from the ice sheet to ocean areas in the sectors 1øW - 61øW and 76øW (westward) - 160øE. Ideally, these estimates of drifting snow transport should be based on a large number of actual transport measurements. However,

snow grains vary less than in coastal areas. Wind velocity observations are easier to perform than those

of drifting snow, and furthermore, they can be carried out remotely [Stearns and Wendlet, 1988]. Nevertheless, at or close to 70øS there is an insufficient number of stations to infer a

reasonable distribution of the drifting snow transport. In part, this is due to the pronounced spatial variability of the average surface-wind speeds [Parish and Bromwich, 1991].

We resorted to a numerical model of katabatic airflow over

Antarctica [Parish and Bromwich, 1991] to provide the mean wind speeds. This approach is justified because the surface winds in Antarctica are mostly of a katabatic nature. Schwerdtfeger [ 1970] pointed out that in no other continent is surface climate so strongly dominated by a single meteorological parameter as that of Antarctica is by the katabatic wind. Mather and Miller [1967] used the observed winds and sastrugi orientations to map the time-averaged surface winds over Antarctica, and later

few such determinations exist and as measurements of drifting Parish [1984] and Parish and Bromwich [1987, 1991] used snow are difficult to perform, they cannot be carried out numerical models of katabafic airflow to provide a more detailed remotely. depiction of these winds. The streamline maps derived from the

Observations of drifting snow have been carried out over the observations and from the models showed good overall years at several locations in Antarctica using both mechanical agreement. Furthermore, the constancy of wind direction is very traps [e.g., Mellor and Radok, 1960; Budd, 1966; Loewe, 1970; high; mean monthly values are around 0.9, which confirm the Naruse, 1970; Kobayashi, 1972, 1978] and electronic devices overwhelming predominance of katabatic surface winds. The [e.g., Landon-Smith and Woodberry, 1965' Tag, 1988; Wendler, observed winds exhibit an annual directional range of 1989a, b]; the former catch grains and crystals, while the latter approximately 10 ø with more downslope flow in winter either detect individual snow particles by their shadows on [Wendler, 1991]' neglect of this effect will introduce only a minor photosensitive semiconductors [Schmidt, 1977] or depend on the change in the calculated transport of drifting snow. pulse counting of an oscillating crystal. Such measurements The primitive equation model [Parish and Bromwich, 1991] relate wind speed to the flux of drifting snow at one or more was integrated from a state of rest to near steady conditions after heights; integration with respect to height yields the transport of 48 model hours. The resulting three-dimensional mass drifting snow. Normally, the logarithm of the transport is circulation describes the gravity-driven drainage of cold, near- expressed as a function of wind speed or equivalently the cube of surface air at a spatial resolution of 100 km (Figure 5), as well as the wind speed is related to the drifting snow transport. This the tropospheric response to this boundary layer divergence. The means that relatively minor errors in wind speed can lead to simulation represents well-developed drainage during the polar substantial errors in the calculated transport of drifting snow. night (March-September) without the complications introduced Furthermore, the height variation of flux above an altitude of a by cyclones and clouds. Comparisons between grid point surface few meters is not well known, as such measurements are difficult winds and observed annual resultant winds at adjacent stations to obtain. This makes the relationship between transport and wind speed uncertain for very strong wind speeds, when the flux at greater altitudes becomes important.

Also, a single relationship between drifting snow transport and wind speed cannot be fixed because, in addition to wind

showed that directions were well simulated, but modeled speeds (as expected) tended to be stronger than the observed annual values. Model wind speeds were multiplied by 0.62 (average of seven ratios of observed speed to model speed, with a mean deviation of 0.16) to more closely approximate climatological

speed, the size distribution of snow grains (a grain may consist of conditions. one or more crystals), the surface roughness, and the availability of loose snow are important. For example, Loewe [1970, p. 5], who made observations in Ad61ie Land where extremely strong winds are experienced, states:

Occasionally, the density of the drift can become very small, even with very strong winds. After a prolonged period of gale from the same direction, the snow surface will become firmly compressed and so perfectly streamlined that no more snow can be removed. Such a

case was experienced at Port-Martin on 8 March, 1951, when the drift almost ceased although the wind remained of the order of 40 m s -1 with unchanging direction.

The mean annual wind speed is not well suited to estimate the amount of drifting snow. The wind-speed distribution becomes very important, as the transport is highly sensitive to wind-speed changes. Close to 70øS, there are two automatic weather stations, which report via satellite [Stearns, 1984]: GF 08 in western Wilkes Land at 68.5øS, 102.2øE at an elevation of 2118 m [Allison and Morrissey, 1983]; and D 80 in Ad61ie Land at 70.0øS, 134.7øE at an elevation of 2500 m [Wendler et al., 1983; Andrd et al., 1986]. Hourly values from GF 08 for 1987 and from D 80 for 1987 and 1988 were used. The model gave only mean annual wind velocities and no information on the wind speed distribution. Hence we used these actual measurements to

GIOVINETTO ET AL.' ATMOSPHERIC WATER VAPOR TRANSPORT ACROSS 70oS 925

Fig. 5. Combined streamline and isotach depictions of surface winds (March-September) on the ice sheet simulated by the model of Parish and Bromwich [ 1991 ].

describe the wind speed distribution as a function of the mean the slopes and downslope forcing on the cold air goes to zero annual wind speed. A frequency distribution is not well described [Parish and Bromwich, 1991]. In reality, the adjustment of by a single parameter, but this approach was deemed acceptable katabatic winds beyond the grounding lines appears to be a in view of the approximate transport estimates being sought. function of the upstream boundary layer mass flux [e.g.,

To summarize, the following simple method was used to Bromwich, 1989], and can have a length scale of up to several provide estimates of drifting snow transport. First, the corrected hundred kilometers. These types of models appear to realistically wind-speed data from the model of Parish and Bromwich [1991] represent many aspects of Antarctic katabatic wind behavior were applied to obtain an approximation of the mean annual [e.g., Bromwich et al., 1990] and qualitatively exhibit the above wind speeds. Second, hourly data from the automatic weather behavior beyond the slope break; therefore it is assumed that the stations near 70øS were utilized to obtain the wind speed model results provide a first-order depiction of the windfield near distribution in terms of the mean annual wind speed. Third, the barrier of those ice shelves. drifting snow models from three sources were used: namely Budd The corrected model winds were also used to estimate Di by et al. [1966] as reported by Radok [1970], Lister [1960] as prescribing the variation of annual speeds along the coast lying reported by Loewe [1970], and Kobayashi [1978]; these models southward of 70øS, with the exception of sectors along the predict the drifting snow transport given the wind speed. From barriers of the Ross, Ron_ne, and Filchner ice shelves. These large these data the net transport of drifting snow was calculated. ic, e shelves have relatively flat areas with length scales of several

In the calculation of Do the corrected model winds were used hundred kilometers, and a model that simulates downslope as estimates of the annual winds along 70øS from 1øW airflow cannot supply even an approximate description of typical (eastward) to 160øE and from 61øW to 76øW. The use of model wind conditions. The observed winds along the barriers of those winds between 1øW and 30øE, approximately along the barrier ice shelves are generally directed offshore with annual resultant of ice shelves some 100 km to the north of the grounding lines speeds being much less than 5 m s -1 [Mather and Miller, 1967]. where surface slopes characteristic of grounded ice topography There are notable exceptions to this in the westem Ross Ice Shelf end, is believed to be reliable because in the model the airstream [Savage and Stearns, 1985], and in the Filchnet Ice Shelf, where undergoes marked deceleration once drainage winds blow beyond marked offshore airflow with vector resultant speeds of

926 GIOVINETTO ET AL.: ATMOSPHERIC WATER VAPOR TRANSPORT ACROSS 70øS

approximately 6 m s '1 axe observed, paxtly fed by inland katabatic airflow [Bromwich, 1989]. More exceptions may become known as the surface airflow along the barrier of the Ronne Ice Shelf is studied in more detail. In this context our

estimate of Di is a minimum that may be compensated to an uncertain extent by the overestimates discussed below.

There are some approximations which must be made with the method described above. In constructing the wind speed distribution, all measurements with wind speeds smaller than the annual mean or laxget than three times the annual mean were ignored, i.e., the distribution is zero outside this range. The low wind speeds are unimportant because they have little influence on drifting snow transport. The high wind speeds, however, cause problems for two reasons: first, they are rare, thus it is difficult to fit a distribution to them; second, the drifting snow models are not valid at these high speeds. To fit the data, we produced a histogram according to the logarithm of the number of observations (Figure 6). A linear least squares fit was then applied and converted to give a probability distribution of the form

p(u) = (wW•-(b/W) u (5)

20 3

18

16

¾ 14

E 12

• 10 o

• 8

•- 6 o

-J 4

for the wind speed, where W is the mean annual wind speed, u is the wind speed, and a and b are constants. When integrated from W to 3W, the value obtained represents the fraction of wind speeds which fall in this interval.

The drifting snow models we used are all of the form

In(D) = cu + d (6)

0 10 20 30 40 50

Wind speed ( m s -1) Fig. 7. Comparison of drifting snow transport as a function of wind speed predicted by the models discussed in the text (a: Budd et al. model from Radok [1970]; b: Lister model from Loewe [1970]; c and d: Kobayashi models 1 and 2 from Kobayashi [1978]).

where D is the transport in grams per meter per second, u is the wind speed in meters per second, and c and d are constants. These models are all derived from actual measurements of drifting snow. Because there are considerable difficulties in measuring drifting snow and most of the measurements were

-2

m -3 ß

ß .Xx300 -7 , , , - • 100 150 200 250

% of mean annual windspeed Fig. 6. Truncated wind speed distribution used in the estimate of drifting snow transport. Data are from stations D 80 and GF 08.

made in a fairly small range of wind speeds, these models do not show close agreement, particularly at wind speeds above 25 m s -1 (Figure 7). For consistency, the original formulae were all converted to use the natural logarithm and units of grams per meter per second for transport and meters per second for wind speed. Wind speeds were adjusted to a height of 3 m, assuming a roughness length of z 0 = 1 mm.

Kobayashi [ 1978] gives a range rather than a single formula; we processed the extremes of the range and refer to them as Kobayashi 1 and 2. Lister's formula is given by Loewe [1970], and the formula from Budd et al. is given by Radok [1970]. The adjusted formulae are as follows: Budd et al.

In(D) = 0.235u + 2.72 (7) Lister

In(D) = 0.407u - 0.345 (8) Kobayashi 1

In(D) = 0.397u - 0.276 (9) Kobayashi 2

In(D) = 0.397u - 2.12 (10)

The estimates of Do across 70øS are presented in Table 4. Values for each of the drifting snow transport models are given for segments of 10 ø of longitude. The transports are all northward, except in the northern Palmer Land. The values for Kobayashi 1 and 2 indicate that this model would result in a middle value of 10 TM kg a 'l, and therefore the estimates range between the models of Budd et al. and Lister, respectively 43 and 206 x 1012 kg a '1. This laxge range emphasizes the order-of- magnitude nature of the estimates. It was indicated above that the models give unrealistically large transports at very high speeds, and thus the method probably overestimates the actual transports. Another aspect that leads to overestimation is the

GIOVINETTO ET AL.' ATMOSPHERIC WATER VAPOR TRANSPORT ACROSS ?0oS 927

TABLE 4. Meridional Drifting Snow Transport Across 70øS (Do) for the Sectors 0 ø (Eastward) - 160øE and 60øW (Westward) - 80øW (Mean Values for Subsectors of 10 ø of Longitude)

Longitude Budd et al. p Transport in 1012 k• a '1 (Ne•,ative Southward)*

Lister.• Kobayashi 1õ Kobayashi 2õ

0ø-10øE 0.69 0.39 0.36 0.057

10ø-20øE 0.59 0.24 0.22 0.035

20ø-30øE 0.13 0.11 0.10 0.016

30ø-40øE 0.61 1.10 0.89 0.140

40ø-50øE 0.98 1.50 1.30 0.210

50ø-60øE 2.00 Z20 1.90 0.300

60ø-70øE 2.80 4.10 3.50 0.560

70ø-80øE 0.10 0.02 0.02 0.003

80ø-90øE 3.90 14.00 11.00 1.800

90 ø- 100øE 8.70 65.00 51.00 8.000

100 ø- 110øE 4.90 28.00 22.00 3.500

110 ø- 120øE 3.00 8.90 7.30 1.200

120 ø- 130øE 3.80 17.00 14.00 2.200

130 ø- 1 40øE 1.90 4.00 3.30 0.530

1 40o. 150øE 8.80 59.00 46.00 7.300

150 ø- 160øE 0.87 0.93 0.80 0.130

60ø-70øW (-)0.59 (-)0.29 (-)0.26 (-)0.042

70ø-80øW (-)0.35 (-)0.13 (-)0.12 (-)0.019

Totals 42.83 206.07 16331 25.92

Adopted 43 206 95

* Using wind data from the model of Parish and Bromwich [ 1991 ]. '• From model described by Radok [1970]. $ From model described by Loewe [ 1970]. õ Maximum and minimum transports from a range described by Kobayashi [1978].

observed limited supply of snow at the surface, already described. distributions with height and through the seasons indicates that The preceding caveats withstanding, we estimate Do at a the mean temperature at condensation levels and up to the 300- midrange value of the models, i.e., approximately 0.12 x 1015kg mbar level is approximately -30 C for at least 9 months of the a ~1 (Table 1). The same procedures were used to estimate Di. year [e.g., Weyant, 1966; Taljaard et al., 1969]. Condensation The values obtained using the four models were < 2 x 1012 kg a '1 occurs at a relatively large temperature range and the latent heat and are therefore ignored in this study. release (Lc) increases as the temperature at which condensation

It should be noted that Loewe [1970], using wind data from 17 occurs decreases following a simple relationship [e.g., Hare and stations, estimated snow drift loss across the ice terminus for the Thomas, 1979]: whole ice sheet at 0.11 x 1015 kg a -1. However, our estimate excludes the ice sheet sector north of 70øS where it should be

expected that drifting snow loss is much greater.

Atmospheric Net Water Vapor and Latent Heat Transport Across 70øS

Lc = 3075- 2.13 T (11 )

where Lc is in Joules per gram and T is the air temperature in kelvin at the moment of condensation. Assuming that condensation occurs at 0 C would result in a minimum estimate

of Lc; this conservative estimate of Lc needs to be increased,

however, because south of 70øS most precipitation occurs in solid The summation of surface balance in the ice sheet, outflow in form (i.e., there is a need to add the latent heat released at 'the

the ocean, and drifting snow as discussed in preceding sections moment of freezing, or 335 J g-l). Using a latent heat release results in a net southward vapor transport of 2.88 x 1015 kg a -1 or figure of approximately 2830 J g-1 results in a net transport of 6.6 kg m -1 s -1 (Table 1). An examination of air temperature approximately 8.2 x1021 J a -1 or 18.9x 106j m -1 s -1 .

928 GIOVINETTO ET AL.' ATMOSPHERIC WATER VAPOR TRANSPORT ACROSS 70oS

A first approximation of the composite error in the estimates may be assessed considering the relative standard errors in each of the principal terms:

1. An error of 10% in the estimate of B (.'k 0.15 x 1015 kg a -1) as was assessed for the whole ice sheet [Giovinetto and Bentley, 1985].

2. An error of 70% in the estimate of Do, as indicated by the range of drifting snow transport using the models discussed above (• 0.08 x 1015kg

3. An error in the estimate of F that is larger than that for B

This research was supported in part by the Natural Sciences and Engineering Research Council of Canada under grant OGP0036595 to the University of Calgary and by the National Science Foundation under grants DPP-8916134 to the Ohio State University and DPP-8714828 to the University of Alaska. Byrd Polar Research Center, Ohio State University, contribution 770.

REFERENCES

Alley, R. B., and C. R. Bentley, Siple Coast fire and ice studies'

and smaller than that for Do, say 40% (ñ 0.51 x 1015 kg a-l). Conclusion and prospects, Antarc. J. U.S.,XXIII, 5,58-59,1988. This admittedly cursory approach indicates a relative error Allison, I. F., A sample study of the energy fluxes preceding and nearly equal to the error that would be derived from accompanying the formation of Antarctic sea ice, Energy Fluxes over uncertainties discussed in the estimates of precipitation over large ocean areas [e.g., Baurngartner and Reichel, 1975] and latent heat flux over ice-free ocean and sea ice areas [e.g., Zillrnan, 1972; Weller, 1980].

The summation of errors in the estimates of B, Do, and F,

treated as standard errors, indicates a composite error of + 0.54 x

Polar Surfaces, Tech. Note 129, 115-132, World Meteorol. Org. , Geneva, 1972.

Allison, I. F., and J. V. Morrissey, Automatic weather stations in the Antarctic, Aust. Meteorol. Mag., 31,71-76, 1983.

Allison, I. F., N. W. Young, and T. Medhurst, On reassessment of the mass balance of the Lambert Glacier drainage basin, Antarctica, J. Glaciol., 31,378-381, 1985. 1015 kg a '1 or :t: 19% for the bulk rate estimate of net water vapor

transport. This composite error is smaller than that which could Alt, J., P. D. Astapenko, and N.J. Ropar, Some Aspects of the Antarctic be assessed, at present, for an estimate of V using atmospheric Circulation in 1958, Int. Geophys. Year Gen. Rep. Set., vol. 4, data. Applied to the other estimates, the relative error indicates errors of + 1.6 x 1021 J a 'l for the bulk rate estimate of net latent heat transport and of + 1.3 kg m 'l s -1 and + 3.6 x 106j m -1 s 'l for the corresponding rate-over-length estimates.

In the comparison with previous estimates of V listed in Table 2 the composite error indicates a range between - 5.3 and - 7.9 kg m -1 s -1' The lower value of this range is equal to the latest estimate using atmospheric data [Masuda, 1990], and the upper value is approximately 10% larger than the upper value of the first estimate using surface data [Rubin, 1962]. Overall, the results of our study fit the chronological (as opposed to real or physical) trend of increasing values in the estimates of net water vapor transport across 70øS reported in the last two decades, which were based on atmospheric data. In the context of physical trends, however, it appears that in the last decades surface balance has increased significanfiy at a number of locations in the ice sheet [Pourchet et al., 1983' Peel and Davison, 1988; D. M. Ethridge, personal communication, 1991]. Since the balance data used to draw isopleths as shown in Figure 3 consist of means of from a few to several years distributed between 1940 and 1980 [Giovinetto and Bull, 19871, the possibility of an actual increase in the annual water vapor transport across 70øS should be examined in the future. At present we note that the results of our study are in general agreement with estimates of net water vapor transport across 70øS reported in the last three decades which were based on surface data. This agreement is explained, in part,

National Academy of Sciences, Washington, D.C., 1959. Andrt, J., G. Wendler, and M. Zephoris, The IAGO Katabatic Programme,

in Antarctic Climate Research, vol. l, 17-18, SCAR Group of Specialists on Antarctic Climate Research, Antarctic Division, Kingston, Australia, 1986.

Andreas, E. L., and A. P. Makshtas, Energy exchange over Antarctic sea ice in the spring, J. Geophys. Res., 90, 7199-7212, 1985.

Andreas, E. L., C. A. Paulson, R. M. Williams, R. W. Lindsay, and J. A. Businger, The turbulent heat flux from Arctic leads, Boundary Layer Meteorol., 17, 57-91, 1979.

Baumgarmer, A., and E. Reichel, The World Water Balance, Elsevier, New York, 1975.

Behrendt, J. C., Antarctic Peninsula traverse, geophysical results of glaciological and geological studies, Res. Rep. 64-1, Geophys. and Polar Res. Cent. Univ. of Wis., Madison, 1965.

Bishop, J. F., and J. L. W. Walton, Bottom melting under George VI Ice Shelf, Antarctica, J. Glaciol.,27, 429-447, 1981.

Bromwich, D. H., Snowfall in high southern latitudes, Rev. Geophys., 26, 149-168, 1988.

Bromwich, D. H., Satellite analyses of Antarctic katabatic wind behavior, Bull. Am. Meteorol. Soc., 70,738-749, 1989.

Bromwich, D. H., Estimates of Antarctic precipitation, Nature, 343,627- 629, 1990.

Bromwich, D. H., T. R. Parish, and C. A. Zorman, The confluence zone of

the intense katabatic winds at Terra Nova Bay, Antarctica, as derived from airborne sastrugi surveys and mesoscale numerical modeling, J. Geophys. Res., 95(D5), 5495-5509, 1990.

by the use in all estimates of largely the same surface balance Brown. R. A., Meteorology, in Polar Oceanography, Part A, Physical data base, as stated in a preceding section, and in part by our extrapolation of the values of V estimated for Antarctica by Rubin [1962] and Brornwich [1990] to the area poleward of 70 ø - 90øS, using ratios obtained from the analysis of Baumgartner and Reichel [1975]. Moreover, the upper value of V given by Rubin [1962] is based on the largest estimate of net northward transport of drifting snow available at that time; this upper value would be greatly reduced if the drifting snow transport models used in our study were applied.

Acknowledgments. The authors wish to thank C. R. Bentley, U. Radok, and C. R. Steams for information leading to sources, T. R. Parish for assistance with extraction of wind output from the numerical model, and three anonymous reviewers for their suggestions.

Science, edited by W. O. Smith Jr., 1-46, Academic, San Diego, Calif., 1990.

Bryan, F., and A. Oort, Seasonal variation of the global water balance based on aerological data, J.Geophys,Res., 89, 11,717-11,730, 1984.

Budd, W. F., The drifting of nonuniform snow particles, in Studies in Antarctic Meteorology, Antarc. Res. Ser., vol. 9, edited by M. J. Rubin, 59-70, AGU, Washington, D.C., 1966.

Budd, W. F., W. R. J. Dingle, and U. Radok, The Byrd Snow Drift Project: outline and basic results, in Studies in Antarctic Meteorology, Antarc. Res. Serv., vol. 9, edited by M. J. Rubin, 71-134, AGU, Washington, D.C., 1966.

Carleton, A.M., Ice-ocean-atmosphere interactions at high southern latitudes in winter from satellite observation, Aust. Meteorol. Mag.,

29, 183-195, 1981.

GIOVINETTO ET AL.: ATMOSPHERIC WATER VAPOR TRANSPORT ACROSS 70oS 929

Carsey, F., Microwave observations of the Weddell Sea polynya, Mon. Weather Rev., 108, 2032-2044, 1980.

Comiso, J. C., and A. L. Gordon, Recurring polynyas over the Cosmonaut Sea and the Maud Rise, J. Geophys. Res., 92(C3), 2819-2833, 1987.

Drewry, D. J., The surface of the Antarctic ice sheet, in Antarctica: Glaciological and Geophysical Folio, edited by D. J. Drewry, sheet 2, Scott Polar Research Institute, University of Cambridge, Cambridge, England, 1983.

Fletcher, J. O., The heat budget of the Arctic Basin and its relation to climate, R-444-PR, The Rand Corporation, Santa Monica, Calif., 1965.

Fletcher, J. O., The heat budget of the Southem Ocean and its relation to climate, RM-5793-NSF, The Rand Corporation, Santa Monica, Calif., 1969.

Fujiwara, K., and Y. Endo, Preliminary report of glaciological studies, in Report of the Japanese Traverse Syowa-South Pole 1968-69, edited by M. Murayama, 68-109, Polar Research Center, National Science Museum, Tokyo, 1971.

Giovinetto, M. B., Mass accumulation in West Antarctica, Eos Trans.,

AGU, 42,386-389, 1961.

Giovinetto, M. B., Distribution of diagenetic snow facies in Antarctica and in Greenland, Arctic, 17, 3240, 1964a.

Giovinetto, M. B., The drainage systems of Antarctica: accumulation, in Antarctic Snow and Ice Studies, Antarc. Res. Serv., vol. 2, edited by

M. Melior, 127-155, AGU, Washington, D.C., 1964b. Giovinetto, M. B., and C. R. Bentley, Surface balance in ice drainage

systems of Antarctica, Antarc. J. U.S., 20 (4), 6-13, 1985. Giovinetto, M. B., and C. Bull, Summary and analysis of surface mass

balance compilations for Antarctica, 1960-1985, Rep. 1, Byrd Polar Res. Cent., Ohio State Univ., Columbus, 1987.

Gow, A. J., and W. B. Tucker llI, Sea ice in the polar regions, in Polar Oceanography, Part A, Physical Science, edited by W. O. Smith Jr., 47-122, Academic, San Diego, Calif., 1990.

H•ikkinen, S., Models and their applications to polar oceanography, in Polar Oceanography, Part A, Physical Science, edited by W. O. Smith Jr., 335-384, Academic, San Diego, Calif., 1990.

Hare, F. K., and M. K. Thomas, Climate Canada, John Wiley, 2nd. ed., New York, 1979.

Hollingsworth, A., The global weather experiment - 10 years on, Weather, 44,278-285, 1989.

Howarth, D. A., Seasonal variations in the vertically integrated water

vapor transport fields over the southern hemisphere, Mon. Weather Rev., 111, 1259-1272, 1983.

Howarth, D. A., and J. N. Rayner, Estimates of sources and sinks of atmospheric moisture in the southern hemisphere, in Second International Conference on Southern Hemisphere Meteorology, 163-166, American Meteorological Society, Boston, Mass., 1986.

Kep, S. L., A climatology of cyclogenesis, cyclone tracks and cyclolysis in the Southern Hemisphere for the period 1972-1981, Publ. 25, Meteorol. Dep., Univ. of Melboume, Parkville, Victoria, 1984.

Kobayashi, D., Studies of snow transport in low-level drifting snow, Contrib. Inst. Low Temp. Sci., Hokkaido Univ., Set. A (24), 1-58, 197Z

Kobayashi, S., Snow transport by katabafic winds in Mizuho Camp area, East Antarctica, J. Meteorol. Soc. Jpn., 56, 130-139, 1978.

Korzoun, V. J., A. A. Sokolov, M. I. Budyko, K. P. Voskresensky, G. P. Kalinin, A. A. Konoplyantsev, E. S. Korotkevich, and M. I. Lvovich (Eds.), Atlas of the Worm Water Balance, USSR Committee for the International Hydrological Decade, Hydro- meteorological Publishing House, UNESCO, Paris, 1977.

Landon-Smith, I. H., and B. Woodberry, The photo-electric metering of windblown snow, Australian National Antarctic Research

Expeditions Inter. Rep. Ser. A(IV), Glaciol. Publ., 79, 1-21, 1965.

Lettau, B., The transport of moisture into the Antarctic interior, Tellus, 21,331-340, 1969.

Lettau, H. H., A case study of katabatic flow on the south polar plateau, in Studies in Antarctic Meteorology, Antarc. Res. Serv., vol. 9, edited by M. J. Rubin, 1-11, AGU, Washington, D.C., 1966.

Lettau, H. H., and W. Schwerdtfeger, Dynamics of the surface-wind regime over the interior of Antarctica, Antarc. J. U.S., 2, 155-158, 1967.

Lister, H., Glaciology, Part I: Solid precipitation and drift snow, Trans- Antarctic Expedition, 1955-1958, Scient. Rep. No. 5, Trans-Antarctic Expedition Committee, London, 1960.

Loewe, F., The transport of snow on ice sheets by the wind, in Studies on Drifting Snow, publ. 13, 1-69, Department of Meteorology, Univ. of Melboume, Australia, 1970.

Masuda, K., Atmospheric heat and water budgets of polar regions: Analysis of FGGE data, in Proceedings of the National Institute of Polar Resarch Symposium on Polar Meteorology and Glaciology, vol. 3, 79-88, National Institute of Polar Research, Tokyo, 1990.

Mather, K. B., and G. S. Miller, Notes on topographic factors affecting the surface wind in Antarctica, with special reference to katabatic winds, and bibliography, Rep. UAG-189, Geophys. Inst., Univ. of Alaska, Fairbanks, 1967.

Maykut, G. A., Energy exchange over young se ice in the central Arctic, J. Geophys. Res., 83, 3646-3658, 1978.

Maykut, G. A., The surface heat and mass balance, in The Geophysics of Sea Ice, edited by N. Untersteiner, 385-463, Plenum, New York, 1986.

Mellor, M., and U. Radok, Some properties of drifting snow, in Antarctic Meteorology, 333-346, Pergamon, New York, 1960.

Mcintyre, N., A re-assessment of the mass balance of the Lambert Glacier drainage basin, Antarctica, J. Glaciol., 31,381-382, 1985.

Naruse, R., Measurement of drifting snow on the coastal region of Antarctica, near Syowa Station (in Japanese with English abstract), Contrib. Inst. Low Temp. Sci., Hokkaido Univ. Set. A (28), 147-154, 1970.

Naval Oceanography Command Detachment, Sea ice climatic arias, vol. I, Antarctic, Rep. NAVAIR 50-1C-540, 1985.

Nishio, F., T. Katusushima, and H. Ohmae, Volcanic ash layers in bare ice areas near the Yamato Mountains, Dronning Maud Land and AUan Hills, Victoria Land, Antarctica, Ann. Glaciol., 7, 34-41, 1985.

Oefiemans, J., Response of the Antarctic ice sheet to a climatic warming: a model study, J. Clirnatol., 2, 1-11, 1982.

Parish, T. R., A numerical study of strong katabatic winds over Antarctica, Mon. Weather Rev., 112,545-554, 1984.

Parish, T. R., and D. H. Bromwich, The surface windfield over the

Antarctic ice sheets, Nature, 328, 51-54, 1987.

Parish, T. R., and D. H. Bromwich, Continental-scale simulation of the

Antarctic katabatic wind regime, J. Climate, 4, 135-146, 1991. Parish, T. R., and K. T. Waight, The forcing of Antarctic katabatic winds,

Mon. Weather Rev., 115, 2214-2226, 1987.

Peel, D. A., and B. M. Davison, Ice cores and climate, Ice, 88, 23-25, 1988. Peixoto, J.P. and A. H. Oort, The atmospheric branch of the hydrological

cycle and climate, in Variations in the Global Water Budget, edited by A. Street-Perrott, 5-65, D. Reidel, Hingham, Mass., 1983.

Picciotto, E., G. Crozaz, and W. DeBreuck, Accumulation on the South Pole-Queen Maud Land traverse, 1964-1968, in Antarctic Snow and Ice Studies H, Antarc. Res. Set., vol. 16, edited by A. P. Crary, 257- 315, AGU, Washington, D.C., 1972.

Potter, J. R., J. G. Paren, and J. Loynes, Glaciological and oceanographic calculations of the mass balance and oxygen-isotope ratio of a melting ice shelf, J. Glaciol., 30, 161-170, 1985.

Pourchet, M., F. Pinglot, and C. Lorius, Some meteorological applications

930 GIOVINETTO ET AL.: ATMOSPHERIC WATER VAPOR TRANSPORT ACROSS 70oS

of radioactive fallout in Antarctic snows, J. Geophys. Res., 88 (C10), 6013-6020, 1983.

Radok, U., Boundary processes of drifting snow, in Studies on Drifting Snow, Department of Meteorology, University of Melbourne, Australia, publ. 13, 1-20, 1970.

Rubin, M. J., Atmospheric advection and the Antarctic mass and heat budget, in Antarctic Rest arch: The Matthew Fontaine Maury Memorial Symposium, Geophys. Monogr. Ser., vol. 7, edited by H. Wexler, et al., 149-159, AGU, Washington, D.C., 1962.

Rubin, M. J., and M. B. Giovinetto, Snow accumulation in central West

Antarctica as related to atmospheric and topographic factors, J. Geophys. Res., 67, 5163-5170, 1962.

Rubin, M. J., and W. S. Weyant, The mass and heat budget of the Antarctic atmosphere, Mon. Weather Rev., 91,487-493, 1963.

Savage, M. L., and C. R. Steams, Climate in the vicinity of Ross Island, Antarctica, Antarc. J. U.S., 20 (1), 1-9, 1985.

Seko, K., T. Furukawa, and O. Watanabe, The variation of surface

conditions on the Antarctic ice sheet, in Proceedings of the International Conference on the Role of the Polar Regions in Global Change, June 11-15, 1990, Univ. of Alaska, Fairbanks, Alaska, in press.

Shimizu, H., Glaciological studies in West Antarctica, 1960-1962, in Antarctic Snow and Ice Studies, Antarc. Res. Serv., vol. 2, edited by

M. Mellor, 37-64, AGU, Washington, D.C., 1964. Schmidt, R. A., A system that measures blowing snow, Res. Paper, RM-

194, U.S. Dep. of Agric. Forest Serv., Washington, D.C., 1977.

Taljaard, J. J., H. van Loon, H. L. Cmtcher, and R. L. Jenne, Climate of the

upper air: southern hemisphere, vol. I, Temperatures, dew points, and heights at selected pressure levels, Rep. NAVAIR 50-1C-55, Nay. Weather Serv. Command, Washington, D.C., 1969.

Trig, H., A pulse-counting technique for the measurement of drifting snow, Ann. Glaciol., 11, 184-186, 1988.

Washington, W. M., A. J. Semtner Jr., C. Parkinson, and L. Morrison, On the development of a seasonal change sea-ice model, J. Phys. Oceanog., 6, 679-685, 1976.

Weller, G., Spatial and temporal variations in the south polar surface energy balance, Mon. Weather Rev., 108, 2006-2014, 1980.

Wendler, G., On the blowing snow in Adtlie Land, Eastem Antarctica, in Proceedings of Symposium on Glacier Fluctuations and Climatic Change, Amsterdam, June 1987, edited by J. Oedemans, 261-279, Kluwer Academic, Boston, Mass., 1989a.

Wendler, G., Measuring blowing snow with a photoelectric particle counter at Pole Station, Antarctica, Polarforschung, 59, 9-16, 1989b.

Wendler, G., Strong gravity flow observed along the slope of eastern Antarctica, Meteorol. Atmos. Phys., 43, 127-135, 1991.

Wendlet, G., Y. Kodama, and A. Poggi, Katabatic wind in Ad61ie Land, Antarc. J. U.S., 18(5), 236-238, 1983.

Wexler, H., Seasonal and other temperature changes in the Antarctic atmosphere, Q. J. R. Meteorol. Soc., 85, 196-208, 1959.

Weyant, W. S., The Antarctic atmosphere: Climatology of the surface environment, in Antarctic Map Folio Series, Folio 8, edited by

V. Bushnell, American Geographical Society, Washington, D.C., 1966.

Schwerdtfeger, W., The climate of the Antarctic, in Climates of the Polar Zillman, J. W., Solar radiation and sea-air interaction south of Australia, Regions, World Surv. Climatol., vol. 14, edited by S. Orvig, 253-355, Elsevier, New York, 1970.

Starr, V. P., J.P. Peixoto, and R. G. McKean, Pole-to-pole moisture conditions for the IGY, Pure Appl. Geophysi., 75,300-331, 1969.

Steams, C. R., Antarctic automatic weather stations, austral summer

1983-1984, Antarc. J. U.S., •9(5), 189-191, 1984. Steams, C. R., and G. Wendler, Research results from Antarctic automatic

weather stations, Rev. Geophys., 26, 45-61, 1988. Streten, N. A., Satellite observations of the summer decay of the

Antarctic sea-ice, Arch. Meteorol. Geophys. Bioklimatol., Ser. A, 22, 119-134, 1973.

Streten, N. A., and A. J. Troup, A synoptic climatology of satellite cloud vortices over the southem hemisphere, Q. J. R. Meteorol. Soc., 99, 56- 72, 1973.

Taljaard, J. J., Synoptic meteorology of the southern hemisphere, in Meteorology of the Southern Hemisphere, Meteorol. Monogr., vol. 13, no. 35, edited by C. W. Newton, 139-213, American Meteorological Society, Boston, Mass., 1972.

in Antarctic Oceanology H, The Australian-New Zealand Sector, Antarc. Res. Ser., vol. 19, edited by D. E. Hayes, 11-40, AGU, Washington, D.C., 1972.

Zwally, H. J., J. C. Comiso, C. L. Parkinson, W. J. Campbell, F. D. Carsey, and P. Gloersen, Antarctic Sea Ice, 1973-1976: Satellite passive- microwave observations, NASA Spec. Publ., NASA SP-459, 1983.

D. H. Bromwich, Byrd Polar Research Center, Ohio State University, Columbus, OH 43210.

M. B. Giovinetto, Earth Sciences 356, University of Calgary, Calgary, Alberta, Canada T2N 1N4.

G. Wendlet, Geophysical Institute, University of Alaska, Fairbanks, Alaska 99701.

(Received April 25, 1991; revised September 13, 1991;

accepted September 17, 1991.)