We examine simulations of the surface mass balance of the sector of the East Antarctic Ice Sheet between 2.4oW and 110.5oE as produced by the UK Meteorological Office Unified Climate Model. Estimates of the actual mass balance of this sector can be obtained from glaciological observations of snow accumulation and from studies of the atmospheric water vapour budget using radiosonde observations. The former technique gives an average sector accumulation of 104 mm year-1, the latter yields 157 mm year-1. The modelled accumulation in this sector, 122 mm year-1, lies between these two estimates suggesting that the model can accurately represent the processes controlling surface mass balance. However, examination of the atmospheric water vapour budget in the model shows that only 30% of the water vapour precipitated in this sector is carried by resolved-scale transport. Although the model produces the "correct" accumulation in present-day climate simulations, it is not clear that this would change appropriately if the model was used to simulate future climates.
By producing synthetic estimates of water vapour transport at radiosonde station locations, we have used the model data to investigate the uncertainties in estimating sector accumulation from radiosonde data. The results of this study indicate that the radiosonde technique will tend to overestimate sector accumulation, thus reconciling the two observational estimates.
Changes in the mass balance of the Antarctic ice sheets have important implications for sea level rise. Over periods of less than a century or so, the mass balance of these ice sheets responds primarily to changes in precipitation and evaporation which, in turn, are governed by changes in atmospheric circulation. The dynamic response of the ice sheets to these changes will only become important over longer timescales [Warrick and Oerlemans, 1990]. Changes in precipitation and evaporation (and, hence, the short-term changes in mass balance) can be obtained from GCM simulations of future climates. However, in order to have confidence in the predicted changes in mass balance, it is first necessary to ensure that the model being used is capable of producing a good simulation of present-day precipitation and evaporation over Antarctica.
In this paper, we examine simulations of present-day snow accumulation over a sector of East Antarctica using the UK Meteorological Office Unified Climate Model. Our chosen sector lies between 2.4oW and 110.5oE (see figure 1) and covers 40% of the area of Antarctica. This sector was chosen principally because its northern boundary is well-provided with radiosonde stations from which the net water vapour transport into the sector can be estimated [Bromwich, 1979; Connolley and King, 1993]. From a modelling perspective it has the advantage of not including the mountainous terrain of the Antarctic Peninsula or Transantarctic Mountains. The sector includes a large portion of the high plateau of East Antarctica so accumulation rates are lower than the continental average. Over most of the sector, temperatures are below freezing all year round and the mass loss from meltwater runoff is negligible. The rate of net snow accumulation is thus simply the difference between precipitation and sublimation.
Validation of GCM simulations is not straightforward because of uncertainties in the "true" value of snow accumulation in this sector. We compare the modelled accumulation with direct estimates from snow pits and stake arrays and with indirect estimates made by calculating the atmospheric transport of water vapour into this sector. Both techniques have advantages and disadvantages, but there are likely to be larger uncertainties associated with the indirect calculation since the water vapour transport must be estimated from measurements at a small number of radiosonde stations around the sector boundary. As a secondary objective, we use water vapour transport fields from the model to examine the uncertainties in making estimates of sector accumulation from radiosonde data.
Finally, we examine the full water vapour budget of this sector of the Antarctic atmosphere in the model simulation. Although the modelled snow accumulation is in reasonable agreement with observations, it appears that the moisture transport into the sector is accomplished largely by diffusion of water vapour rather than by resolved-scale transport. We discuss reasons for this behaviour and consider its implications for model simulations of Antarctic mass balance.
The surface mass balance of a sector of the Antarctic can be estimated directly from snow accumulation measurements or indirectly by integrating moisture fluxes across the sector boundary. Each method has advantages and disadvantages which we consider in the subsections below.
Ice cores, snow pits and stake arrays provide approximately 1,500 accumulation data points for Antarctica. The coverage is far from complete and there are problems with interpretation of annual layers where the precipitation rate is low. Different compilations of the available data produce very different maps; we use that produced by Giovinetto and Bentley [1985, henceforth G&B] which seems to be of high quality. The accumulation record goes back to the International Geophysical Year of 1957 and in order to obtain good spatial coverage one has to use all data points and assume that accumulation rates have been essentially unchanged over this period. Despite these problems the calculation of areal average accumulation from mass-balance maps is probably the most accurate method of calculating the surface mass balance.
The mass balance at the surface is given by:
dm/dt = P - E - D - M (1)
where
m is the mass of the surface snow (kg m-2),
P is the precipitation rate (kg m-2 s-1, positive downwards),
E is the evaporation rate (kg m-2 s-1, positive upwards),
D is the divergence of snow drift (kg m-2 s-1) and
M is the divergence of meltwater and rainfall runoff (kg m-2 s-1).
The divergence of snow drift, D, is not thought to be significant on a continental scale [Giovinetto et al., 1992] although it is possibly important locally. Over East Antarctica temperatures are well below freezing for much of the year so M is negligible. The mass balance equation thus reduces to
dm/dt = P - E (2)
The left hand side of equation (2) is the accumulation rate measured by glaciological observations.
G&B present an accumulation map and calculate area averages of accumulation for ice drainage basins. The sector 2.4o W to 110.5o E cuts across basins and so, using a planimeter and integrating contours, we have calculated areas and accumulation rates for those basins cut by the sector lines. There is uncertainty in assigning an accumulation rate to the interior cut basins delineated by J''K and J''D in G&B (which we shall refer to as X and Y respectively) since they lie almost within the 50 mm year-1 contour. We have assumed an average of 30 mm year-1 for X and 40 mm year-1 for Y. The area average for the sector is then 104 mm year-1. To test the sensitivity of the result to these values we changed them both by ñ 10 mm year-1, which changes the overall average to 100 and 107 mm year-1. This change is small compared to other uncertainties such as the delineation of the contours or the original measurements. There are other uncertainties particularly near the coast which have similar sensitivities.
The advantages of deducing surface mass balance from glaciological observations are that spatial information about the distribution of accumulation is available that can be usefully compared to GCM results. Individual observations are accurate to about 10% and accuracy is higher in regions of high accumulation where it is easier to identify annual layers. Nonetheless the overall accuracy of the method is hard to assess as there are large areas of the interior with no observations and different maps delineate very different contours especially in the high accumulation regions near the coast. The standard error of 13 balance composites listed in Giovinetto and Bull [1987] for the whole Antarctic ice sheet excluding the Antarctic Peninsula is 5%. G&B estimate an uncertainty of 10% in their value for the whole Antarctic.
The water vapour balance of the Antarctic atmosphere within a region A with border A may be written as:
integral_border(A) F.n ds = integral_A [ dm/dt + dQ/dt ] dA (3)
where
F is the vertically-integrated horizontal moisture flux (kg m-1 s-1),
n is the unit vector normal to the boundary,
Q is the vertically-integrated moisture content of the atmosphere (kg m-2).
As in section 2.1, snowdrift divergence and meltwater runoff have been neglected. The sector used to evaluate the left hand side of equation (3) is shown in figure 1. The northern boundary is latitude 68.4o S, the average latitude of the coastline. Connolley and King [1993] describes the construction of a database of six years of radiosonde data for available Antarctic stations and the computation of vertically-integrated moisture fluxes from the database. The 8 radiosonde stations are roughly equally spread across the northern border whose length is 4,632 km. We assume that these fluxes are representative of the conditions along the whole of the border and that the average flux across the border is the average of the fluxes at the 8 stations. We further assume that the fluxes along the eastern and western borders cancel out. This is done in the absence of any real information concerning these boundaries. Bromwich and Robasky (1993), by applying coastal fluxes along the western boundary with an allowance made for the part of the profile cut off by elevation, conclude that there is a large easterly flux across the western boundary. However, in the absence of direct observational evidence, and referring to figures 2 and 4 which show that in the GCM the fluxes along this boundary change from westerlies in the interior to easterlies at the coast and back to westerlies offshore, we believe that the best possible assumption to make is that the fluxes along the eastern and western borders cancel. The sparsity of the radiosonde chain around Antarctica dictates the choice of the sector between SANAE and Casey stations (2.4o W to 110.5o E) for the radiosonde analysis. Also this sector was used by Bromwich [1979 and 1988] for a similar calculation with only one year's data. The vertical integration of F in equation (3) is limited by the resolution of the reported radiosonde ascents. If equation (3) is averaged over seasonal or longer timescales, dQ/dt becomes negligible and we obtain:
integral_border(A) F.n ds = integral_A [ dm/dt ] dA (4)
All quantities are now appropriate time-averages. The right hand side of equation (4) is the area-averaged accumulation rate whose value was discussed in section 2.1. Observations of the fluxes, F, are at best available at 12-hourly intervals and the averaging may be biased by missing observations. This is discussed in Connolley and King [1993] and Bromwich [1979]. Using equation (4) the implied average accumulation in our sector is 157 mm year-1. The difference between this figure and that quoted in our earlier paper [Connolley and King, 1993] is due to an incorrect calculation of the sector area in the earlier paper. The standard deviation is 30 mm year-1 but this is based on only 6 years data.
Bromwich [1979] applied a similar method to data from 1972, but assigned all the precipitation to the area between the coast and the crest line of the plateau. His estimate of 170 mm year-1 for this smaller region corresponds to an estimate of 122 mm year-1 when averaged over all of our sector. Bromwich [1988] revised his earlier estimate, partly by including climatological information for the eastern and western boundaries, to 113 mm year-1. Although this estimate used only one years data the main reason for lower accumulation estimate is the different assumption about the western border flux , noted above.
The advantages of the radiosonde method are that the result is applicable to a large area and interannual variations can be monitored. There are several disadvantages. Errors may result because missing soundings may be biased towards days with high winds (which may make balloon launch difficult). The inherent accuracy of the flux calculated from any one sounding is about 20 % [Connolley and King, 1993] but this should not be subject to a systematic bias. Problems with the eastern and western boundaries have already been mentioned. The most serious and least determined error arises from the assumption that the 8 stations are typical of conditions along the coast. Conventional observations do not have sufficient coverage to investigate this but a model in which we had confidence would enable this assumption to be tested. We consider this further in section 6.
We use data from the UK Met. Office GCM. An atmosphere-only version of this model was described by Connolley and Cattle [1994] and was found to produce an acceptable simulation of the Antarctic climate. In particular the precipitation was well simulated despite excessive accumulation at a few coastal points. This paper uses 8 decades of a control run which couples atmosphere, deep ocean and sea-ice models (and includes flux-correction). The model is a finite-difference GCM with horizontal resolution of 96 points (longitudinally) by 73 points (latitudinally) corresponding to a grid spacing of approximately 270 km in latitude.
The model accumulation is the difference between precipitation and evaporation given by equation (2). Most model precipitation comes from large-scale (non-convective) snowfall. Evaporation includes (and is mostly made up of) sublimation, which is small relative to the snowfall. Since large-scale and seasonal measurements of these parameters are rare in Antarctica the model results may be of interest; we give them in table 1.
Period Evap/Subl Ppn
Year 7 129
DJF 6 34
JJA -0.5 28
Table 1. Model values (mm) of evaporation (including sublimation) and precipitation for annual, summer (December-January-February, DJF) and winter (June-July-August, JJA).
The accumulation figure for the model sector is 122 mm year-1. Note that these values are area averages for the region outlined in figure 2, which differs slightly from our sector. It is necessary to restrict the region to model land points or the sector mean evaporation is contaminated by large values of evaporation from the sea. This region includes a large area of the cold high plateau; sublimation values for coastal areas or an average over all Antarctica would be higher (e.g. sublimation at Halley (75.5o S, 26.7o W, 32 m) over the winter (JJA) is -2 mm [King et al, 1995]). An effect that may be important but is not included in the model is sublimation of blowing snow. However present estimates of this are too imprecise to be used.
Figure 3 shows the pattern of accumulation around Antarctica. Note the high accumulation band around the coast and the presence of points (eg at 45 E) of very high accumulation. The two areas of net loss (at 10o E and 20o W) are off the coast and so do not contribute to the budgets considered here. This figure may be compared to figure 4.a of Connolley and Cattle [1994] which shows the accumulation from the atmosphere-only model; the difference is slight. It may also be compared to figure 3 of G&B showing that the spatial pattern as well as total value of accumulation is well modelled.
Equation (3) describes the water vapour balance of the atmosphere. In the case of the model atmosphere the equation becomes:
integral_border(A) F.n ds = integral_A [ dm/dt + dQ/dt -R ] dA (5)where the residual term, R (kg m-2 s-1), is sum of the divergence of diffusive transport of moisture (R1) and errors in the moisture scheme (R2). R was not included in equation (3) because in reality R1 represents motion on scales smaller than are resolved by the radiosonde and is negligible. In the model it results from a diffusive term used to ensure computational stability and to represent transport by subgrid scale eddies and is significant. R2 is of course an artefact of the moisture scheme, which arises near sharp gradients in specific humidity (principally near steep orography), when "undershoots" in the specific humidity are filled in by the model to prevent negative humidities.. If equation (5) is averaged over seasonal or longer timescales, dQ/dt becomes negligible and we obtain (substituting dm/dt from equation (2)):
integral_border(A) F.n ds = integral_A [ P - E - R ] dA (6)
All quantities are now multi-annual time-averages.
We shall look at the model moisture budget for the region closely following our sector that was used in section 3. For the (nearly) exact calculation of the model budget it is necessary for the region boundary to follow the model coastline exactly. Figure 2 shows the integration region we shall use in this section together with the sector used in section 2. Model moisture fluxes across zonal and meridional boundaries are shown. The terms in the model budget, all converted into kg m-2 year-1, are given in table 2.
Boundary integral of F: 35
P, Precipitation: 129
E, Evap/Subl: 7
R1, Moisture diffusion: 70
R2, Moisture error: 16
Table 2. Terms in the model budget (kg m-2 year-1) for the region shown in figure 2. Note that the terms do not balance exactly due to rounding errors.
The term R1 is calculated as a residual of the other four according to equation (6). In the model simulation, moisture diffusion is the most important mechanism for water transport into the region, accounting for more than 50% of the sector accumulation. The effect of diffusion and moisture errors is particularly important over East Antarctica; for the entire continent the boundary integral of F is 106 and the areal average of P-E is 184 kg m-2 year-1.
The large contribution from diffusive transport is disquieting. Diffusion is applied to ensure computational stability arising from small-scale noise and the diffusion coefficient used is larger than that required just to represent sub-grid-scale eddy transport. The diffusion coefficient for moisture is chosen so that the e-folding time for 2-gridlength waves is 24 hours. Horizontal diffusion occurs along model levels, which are sigma (terrain-following) surfaces near the ground and smoothly change to pressure surfaces at the upper levels. Over steeply-sloping terrain, such as the coastal slopes of East Antarctica, there will be large temperature gradients along sigma-surfaces which simply reflect the changes in height of the surfaces. Since saturation humidity is a strong function of temperature, there will be corresponding humidity gradients along sigma-surfaces, with lower humidity values inland, resulting in a diffusive flux of moisture towards the interior. Convergence of this flux will generate precipitation where relative humidities exceed 100%. In reality, the atmosphere moves along isentropic surfaces rather than sigma-levels. If the diffusion was calculated along isentropes it would make a much smaller contribution to the total moisture flux. Another approach is to use a semi-Lagrangian transport scheme for moisture advection. Analysis of results from NCAR CCM1 and CCM2 GCMs indicate that use of the semi-Lagrangian scheme leads to small values for the residual terms (D. Bromwich, pers. comm.).
Figure 4 shows modelled moisture fluxes for the whole Antarctic region. Fluxes are omitted at every other point longitudinally and where they exceed 35 kg m-1 s-1 for clarity. The large fluxes over West Antarctica cannot be verified because of the lack of suitable radiosonde stations. However they are consistent with glaciological studies showing high accumulation in this region.
In the previous section we showed that the greater part of the moisture entering our study region in the model was carried by diffusion. We would thus not expect to see a strong correlation between resolved-scale transport in the model and observations of moisture transport at coastal radiosonde stations. Nevertheless, it is instructive to carry out the comparison in order to see whether resolved-scale fluxes in the model exhibit similar spatial variability to that seen in reality. If this is the case then a comparison of the model moisture fluxes at station locations with the total transport into the sector will give some idea of the representivity of the fluxes calculated from radiosonde observations.
Because of the coarseness of the model grid the model and real coastlines do not coincide, making comparison of model results with station data difficult. Stations which are in reality located near the coast are, in the model, situated somewhat inland at excessive height. Modified locations were therefore chosen so that the model height at the modified station location corresponds to the real station height; this was done by moving the point within a grid box towards the coast. The average station latitude changed from 68.3o S to 67.6o S. The fluxes were then bilinearly interpolated to the modified station positions. Figure 5 shows the observed zonal and meridional fluxes plotted against modelled fluxes at modified station locations. For both components the modelled fluxes are less than those observed, more severely so in the zonal case. The correlation coefficient is 0.69 for the meridional fluxes and 0.10 for the zonal. This problem is most severe around the coastal slopes of Antarctica. Further north, at Bellingshausen station (at the tip of the Antarctic Peninsula in a maritime environment), the modelled fluxes are in good agreement with observed: zonally -57.7 kg m-1 s-1 modelled and -61.9 kg m-1 s-1 observed; meridionally 14.4 kg m-1 s-1 modelled and 12.6 kg m-1 s-1 observed (northerly and easterly fluxes are positive).
In order to investigate the difference between modelled and observed fluxes we have compared modelled and observed profiles of winds and humidity at one station, Molodeznaja. This station has a good observational record with two ascents per day and few missing ascents. Data were taken from the study of Connolley and King [1993] and cover the periods 1988-90 and 1980-82. Figure 6 shows annual average vertical plots of zonal and meridional winds and humidity mixing ratio from these observations and from the model. Observational data are plotted at standard pressure levels supplemented by interpolation of significant level data at 20 hPa intervals between 900 hPa and the surface. There are 14 model levels between 1000 and 100 hPa, with the lowest separated by approximately 3, 21 and 44 hPa.
At Molodeznaja the modelled and observed meridional fluxes are in good agreement but the modelled zonal fluxes are less realistic (see figure 5). Figure 6 shows that the model meridional winds are within the observed interannual variation above 960 hPa but too weak below this. The observed average meridional wind shows a nearly linear profile between 850 hPa and the surface. This is because for the earliest three years of the record almost no winds were reported in this interval. In the latter three years more detailed winds were reported and the observed and modelled profiles near the surface are in closer agreement. The model zonal winds exhibit a consistent westerly bias of several m s-1. This effect in the zonal winds is common to other coastal stations although the degree by which the model winds are too westerly varies. Egger [1985] and James [1989] have shown that two dimensional (axisymmetric) models of the Antarctic circulation fail to maintain a katabatic drainage flow and develop excessive upper-level westerlies because of an inability to export cyclonic vorticity at upper levels. Detten and Egger [1994] suggest that this vorticity transport is accomplished by topographic modification of synoptic-scale weather systems approaching Antarctica. We have not carried out a detailed vorticity budget for the Unified Model but it is possible that the westerly bias in the model winds may result from the failure of the model to capture this effect fully. However, the model does produce a reasonable simulation of the broad scale features of the katabatic flow [Connolley and Cattle, 1994]. At stations where the meridional fluxes are too small there is no common cause. The model is slightly too dry above 900 hPa. Above the tropopause there is a slight but significant increase in observed humidity but this is not seen in the model. The dryness in the model does not significantly affect the moisture flux budget since the absolute errors are small, but might be a problem for radiative calculations. In the lowest model levels there is a noticeable increase in moisture that is not seen in the observations.
An average accumulation rate can be calculated from model fluxes at the modified station locations using the procedure described in section 2. The value obtained is 96 mm year-1 which is 40% lower than the rate (157 kg m-2 year-1) deduced from observed fluxes in section 2. If the "real" station locations are used the value is 62 mm year-1, which is lower because the moisture-carrying capacity of the air decreases as temperature decreases as height increases inland. This is equivalent to saying that higher snowfalls occur on the coastal slopes.
In section 2, two major assumptions were made when calculating an implied sector accumulation rate from radiosonde observations. Firstly, the moisture fluxes at the stations were assumed to be representative of the whole of the northern boundary of the sector. Secondly, in the absence of any data, it was assumed that moisture fluxes across the eastern and western sector boundaries cancelled exactly. Since moisture flux values are available at all model gridpoints we can use the model data to test the validity of these assumptions.
The boundary integral in the model can be split into contributions (positive indicating influx into the region) from the northern border (41 kg m-2 year-1 ) and the sum of the fluxes across all north-south boundaries (-6 kg m-2 year-1). This latter term is indeed small as was assumed. That the model eastern and western boundary fluxes nearly cancel is not proof that the same occurs in reality but does lend credence to the idea. This latter term includes the contributions from the one grid point north-south step boundaries along the coast. If we consider only the eastern and western sector boundaries at 1.875o W and 110.625o E the contributions are 10 and -3 kg m-2 year-1 respectively. Notice that at both zonal boundaries the fluxes are westerly not easterly. However the western border extends to 71.25o S and the longer eastern border to 66.25o S. If both are extended to 68.75o S the contributions become 0 and -4 kg m-2 year-1. This shows that the fluxes calculated across these north-south boundaries are sensitive to the northern extent of the boundary, because the absolute value of the fluxes increases rapidly towards the coast and their direction changes from westerly to easterly near the coast. Table 3 summarises these figures.
All North-South boundaries shown in figure 2 -6
Eastern and Western sector boundaries only 13
Eastern and Western sector boundaries extended to 68.75 S -4
Table 3. Contributions to sector mass balance (kg m-2 year-1) from model zonal moisture
convergence across North-South oriented boundaries under various boundary assumptions.
The contribution from flux across the northern boundary (41 kg m-2 year-1) is less than half the value (96 kg m-2 year-1) deduced in section 5 using model data at station locations only. This is partly attributable to non-representativeness of station locations and partly to problems with coastlines. Had we used 68.75o S as our northern border for this region (a model grid- line very close to 68.4o S which is the average coastline latitude) the northern contribution would have increased to 59 kg m-2 year-1. This is very close to 62 kg m-2 year-1, the value calculated using model data interpolated to real station locations. Table 4 summarises these values, which suggest that the stations are indeed representative of the fluxes across 68.4o S (at least in the model) but that fluxes across 68.4o S are greater than those across the coastline. It also indicates that the result is sensitive to the positioning of the boundary.
Northern boundary shown in figure 2 41
Averaged from modified station locations 96
Using 68.75 S as boundary 59
Averaged from original station locations 62
Table 4. Contributions to sector mass balance (kg m-2 year-1) from model meridional moisture
convergence across the northern boundary under various assumptions.
The accumulation estimates from glaciological and radiosonde observations deduced in sections 2.1 and 2.2 are in disagreement. This disparity can be removed by using a larger value for the western boundary flux [Bromwich and Robasky, 1993] leading to a lower radiosonde based estimate of accumulation and therefore closer agreement between the radiosonde and glaciological estimates. However, as noted in section 2.2, in the absence of data for these boundaries we prefer to assume that the eastern and western boundary fluxes cancel. We now use model data to investigate the possibility that the disparity arises because the stations overestimate the fluxes across the coast. The modelled total moisture convergence is 35 kg m-2 year-1 (resolved transport) plus 86 kg m-2 year-1 (residual from diffusion and moisture fixer). The spatial distribution of the diffusive term is not available as a model diagnostic. However, if we assign all of the model residual transport to the northern boundary (a reasonable assumption, since the orographic slope is greatest there) then the total northern boundary flux deduced from the modelled radiosonde stations is 96 kg m-2 year-1 (resolved transport) plus 86 kg m-2 year-1 (residual). Hence the ratio of modelled sector accumulation to northern border flux (calculated using model data at radiosonde station locations only) is 122:(96+86). This is very close to the ratio 104:157 of the observed sector accumulation (from glaciological measurements) to the accumulation deduced from radiosonde observations. The model results thus suggest that the two observational estimates of sector accumulation can be reconciled by assuming that the radiosonde observations overestimate the total water vapour transport into the sector.
We have compared observational estimates of the surface mass budget for the Antarctic sector 2.4o W to 110.5o E with that obtained from a GCM. Accumulation estimated from radiosonde observations (157 mm year-1) is higher than that obtained from direct glaciological measurements (104 mm year-1); the latter value is probably to be preferred. The model accumulation (122 mm year-1) is in reasonable agreement with the glaciological estimate. Connolley and Cattle [1994] have shown that the model produces excessive precipitation in the Antarctic coastal region and it is probably this coastal zone which is contributing to the small overestimate of sector accumulation in the model. The derivation of the accumulation estimate from radiosondes is dependent on assuming that the radiosonde stations are representative of the coast and that the eastern and western boundary fluxes balance (see section 2.2). Bromwich and Robasky [1993] deduce a large easterly flux at the western boundary and consequently a lower radiosonde-derived estimate of accumulation comparable to the glaciological estimate.
Examination of the model water vapour budget reveals that only 30% of the water precipitated in this sector is carried by resolved-scale motions while 57% is carried by modelled diffusion and 13% by the artificial moisture fixer. The low fraction of resolved-scale transport is disturbing and suggests some uncertainty in the ability of this model to predict Antarctic accumulation in a changed climate. Even if all dynamical changes were simulated perfectly, there is no a priori reason to expect the diffusive transport to change appropriately. In order to obtain confidence in model predictions, it will be necessary to conduct a number of present-day climate simulations with reduced values of the water vapour diffusion coefficient. If reduced diffusive transport is compensated for by increased resolved-scale transport then it would appear that the model can simulate total transport accurately and model predictions could be regarded with greater confidence. Such experiments remain to be done.
The use of model data to simulate the estimation of sector accumulation from radiosonde observations has indicated that this technique may overestimate accumulation by approximately the expected amount. It is thus possible to reconcile the two observational estimates. A simple scaling factor can then be applied to the radiosonde estimate in order to obtain the true sector accumulation. Since radiosonde data are available for a number of years, it should be possible to obtain information on interannual variability in sector accumulation using this technique. This cannot currently be obtained from glaciological measurements.
Acknowledgement: thanks are due to the staff of the Hadley centre of the UKMO Met. Office for their help in obtaining and interpreting model runs.
Bromwich, D. H., Precipitation and accumulation estimates for East Antarctica, derived from rawinsonde information, Ph.D thesis, University of Wisconsin-Madison, 1979.
Bromwich, D. H., Snowfall in high southern latitudes, Rev. Geophys., 26, 149-168, 1988.
Bromwich, D. H., and F. M. Robasky, Recent precipitation trends over the polar ice sheets, Meteorol. Atmos. Phys., 51, 259-274, 1993.
Connolley, W. M., and H. Cattle, The Antarctic climate of the UKMO unified model, Ant. Sci., 6, 115-122, 1994.
Connolley, W. M., and J. C. King, Atmospheric water vapour transport to Antarctica inferred from radiosondes, Q. J. R. Meteorol. Soc., 119, 325-342, 1993.
Detten, Y., and J. Egger, Topographic wave modification and the angular momentum balance of the Antarctic Troposphere. Part II: Baroclinic flows, J. Atmos. Sci., 51, 3351-3359, 1994.
Egger, J., Slope winds and the axisymmetric circulation over Antarctica, J. Atmos. Sci., 42, 1859-1867, 1985.
Giovinetto, M. B., and C. R. Bentley, Surface balance in ice drainage systems of Antarctica, Ant. J. U.S., 20(4), 6-13, 1985.
Giovinetto, M. B., D. H. Bromwich, and G. Wendler, Atmospheric net transport of water vapour and latent heat across 70S, J. Geophys. Res., 97(D1), 917-930, 1992.
Giovinetto, M. B. and C. Bull, (1987): Summary and analysis of surface mass balance compilations for Antarctica, 1960-1985. Byrd Polar Research Center Report No. 1. Ohio State University, Columbus.
James, I. N., The Antarctic drainage flow: implications for hemispheric flow on the southern hemisphere, Ant. Sci., 1, 279-290, 1989.
King, J. C., Anderson, P.S., Smith, M.C. and Mobbs, S.D. Surface energy and water balance over an Antarctic ice shelf in winter. Preprints, 4th AMS Conference on Polar Meteorology and Oceanography, 79-81, 1995.
Warrick, R., and J. Oerlemans, Sea level rise, in Climate change - the IPCC scientific assessment, edited by J. T. Houghton et al., pp. 257-281, CUP, Cambridge, 1990.
Figure 1. Location map showing Antarctica, orography (smoothed to 2.5o latitude by 3.75o longitude) at 500 m intervals, the sector 2.4o W to 110.5o E to 68.4o S, station locations (diamonds) modified locations (crosses) and other stations (stars).
Figure 2. Model moisture fluxes (kg m-1 s-1) and outline of model sector.
Figure 3: Average model accumulation, contour intervals 0, 50, 100, 200, 400, 600, 800 mm year-1. Accumulation less than 0 is shaded, accumulation greater than 800 is stippled.
Figure 4. Model moisture fluxes (kg m-1 s-1). Fluxes are omitted at alternate points latitudinally and where they exceed 35 kg m-1 s-1.
Figure 5. Observed (a) zonal and (b) meridional fluxes (kg m-1 s-1, northerly and easterly fluxes positive) plotted against those modelled for 8 coastal stations.
Figure 6. Vertical plots of (a) zonal, (b) meridional winds (m s-1) and (c) moisture (g kg-1) from observations and the model, for Molodeznaja. Thick solid line: observed average. Thin dotted line: observed, years 1980-1982. Thin dot-dashed line: observed years 1988-1990. Thick dashed line: model.
