1. Introduction
An increased awareness of the importance of Antarctica in controlling the Southern Hemisphere circulation and an appreciation of the potential role of the southern polar regions in determining the response of the climate system to enhanced “greenhouse” forcing have prompted several investigations of the ability of GCMs to model the present-day Antarctic climate. Studies using low-resolution GCMs (e.g., Tzeng et al. 1993) have met with only limited success, but simulations using modern high-resolution models (e.g., Connolley and Cattle 1994; Tzeng et al. 1994) have been able to reproduce the observed climate of Antarctica with some degree of accuracy.
The validation of these models has largely been restricted to the comparison of simple model variables, such as surface temperature, with observations. However, just because a model is able to reproduce the observed distribution of surface temperature is not, in itself, proof that the model is correctly representing the processes that determine the energy balance at the surface. Proper representation of the present-day surface energy balance is a necessary prerequisite for any model that is to produce useful climate predictions; hence, it is important that all of the surface energy fluxes should be validated in detail.
Few attempts have been made to validate GCM surface energy balance computations, largely because surface energy balance measurements are only available for a limited number of stations worldwide and measurements at these locations may only be representative of a small area around the station rather than a region of comparable size to a GCM grid square. In Antarctica, stations where surface energy balance measurements have been made are typically separated by hundreds of kilometers. However, such stations are often situated on very uniform ice sheets where point measurements of surface fluxes are likely to be representative of a wide area. There is thus some justification for attempting to validate model surface energy balance computations against such observations. In this paper we present the results of a comparison of the surface energy budget in the U.K. Meteorological Office unified climate model (henceforth referred to as the unified model) with observations made at a number of locations on the East Antarctic Ice Sheet.
In section 2, we briefly describe the unified model and the parameterizations used for surface fluxes. We describe the observations used for model validation in section 3 and compare these with model computations in section 4. The results of the comparison are discussed in section 5, and we present our conclusions in section 6.
2. The unified model
Cullen (1993) and Johns et al. (1997) provide a general overview of the unified model and the physical parameterizations used within it. The model uses a conservative split–explicit integration scheme with fourth-order horizontal advection (Cullen and Davies 1991) and, in climate mode, is run on a 2.5° lat × 3.75° long grid. A hybrid 19-level vertical coordinate system is used, which changes from sigma coordinates near the ground to a pressure coordinate in the upper troposphere. The vertical mesh is stretched to give improved resolution in the boundary layer and at the tropopause; the lowest model level is at σ = 0.997, which is approximately 20 m above terrain level over the interior of Antarctica.
The model employs a sophisticated radiation scheme in which cloud optical properties (in both the longwave and shortwave regions of the spectrum) depend on cloud water and ice content. The longwave radiation scheme employs a band model to calculate absorption by water vapor, carbon dioxide and ozone lines, and the water vapor continuum, broadly following the treatment of Slingo and Wilderspin (1986). This model has been shown to perform well when validated against a spectrally detailed model (Slingo and Wilderspin 1986). Upward longwave radiation from the surface is calculated from (3) with the surface emissivity set to 1.0 everywhere, so the reflected longwave radiation term vanishes.
The representations of surface, subsurface, and boundary layer processes in the model are described by Smith (1990). The subsurface model uses 4 levels to give a good simulated response to forcing periods from half a day to a year. The albedo of a snow-covered land surface depends on snow depth and varies from a bare surface value to a limiting deep snow value. Over Antarctica this latter value of 0.8 always applies. The albedo is further modified by a linear reduction as temperature varies between −2°C and 0°C; this has a very small effect during summer at Antarctic coastal stations. We discuss the details of the parameterizations used for surface fluxes in section 5.
The Antarctic climatology of the unified model was examined by Connolley and Cattle (1994). Model surface temperatures are reasonable with good agreement with observations at the South Pole, although farther into the interior temperatures are too cold in winter. Surface wind speeds are higher than in earlier versions of the model (Mitchell and Senior 1989) and are in better agreement with observations. In this paper we shall examine the model surface energy balance averaged over a 5-yr model run in which sea surface temperatures and sea ice extent were prescribed according to observed climatology over the period 1951–80, as in the run considered by Connolley and Cattle (1994). A number of minor changes have been made to the model since this earlier study was carried out.
3. Validation data
a. Radiation balance data
Surface radiation measurements have been published for a number of Antarctic stations (e.g., Rusin 1961; Schwerdtfeger 1984) but are of rather variable quality. For this study, the following criteria were used to select data.
Instruments and measurement procedures should be well documented to ensure consistency of the measurements.
Only measurements made over permanent snow and ice should be used since these conditions are representative of most of the surface of the Antarctic continent and the radiational regime of isolated areas of bare rock may be anomalous. All areas of permanently exposed rock in Antarctica are considerably smaller than the grid box size of the model runs considered in this paper.
Measurements should be available for all months of the year.
Longwave radiation fluxes should be measured directly rather than computed from measurements of the shortwave fluxes and the net radiation balance.
We have chosen to use radiation data from South Pole (SPO), Plateau (PLA), Mizuho (MIZ), and Neumayer (NEU) stations in this study. Data from SPO, MIZ, and NEU fulfill all of the criteria given above, but only the shortwave components and net radiation balance were measured at PLA. Some details of these observations are given in Table 1. The published radiation data for these stations are in the form of monthly means and, where data were available for more than 1 yr, averages were formed to produce a dataset containing 12-monthly averages for each station. The series of measurements from NEU is sufficiently long to provide an indication of the level of interannual variability. At this station, the interannual standard deviations of monthly means of downward shortwave and longwave radiation are less than 10% of the respective long-term means. Measurements from a single year at other stations should thus be reasonably representative of climatological conditions.
The four stations considered span the entire range of climatological conditions encountered on the East Antarctic Ice Sheet. SPO and PLA are typical of the high, cold, interior plateau. MIZ is situated in the steeply sloping peripheral region and is subject to strong katabatic winds, while NEU is typical of the coastal zone. All four stations are situated on uniform ice sheets so measurements of radiation (and other energy fluxes) at the stations should be representative of a wide area.
b. Heat flux through the snowpack
We have used (5) to calculate monthly mean values of G for all stations for which we have radiation measurements. In the absence of surface temperature data, a cyclic fit was applied to the monthly mean air temperatures to estimate the amplitudes and phases of the Fourier components, and it was found that truncation at n = 4 gave an adequate fit to the annual cycle of temperature at all stations. During the Antarctic winter, the surface layer is usually stably stratified and the snow surface may be significantly colder than the air temperature measured at a nominal 2-m height, while in summer the temperature difference is smaller or even reversed. Schwerdtfeger (1984) quotes monthly means of air minus snow temperature at SPO of up to +2.4°C in winter and up to −1.4°C in summer. Since these differences are small compared with the annual temperature range (about 32°C at SPO), the use of air temperature rather than surface temperature in (5) will make little difference to the calculated values of G. However, we have accounted crudely for the seasonal variation in air–snow surface temperature difference by increasing the amplitude of the annual temperature component, A1, by 1°C at all stations before using (5) to calculate G.
Our procedure for calculating G uses a highly simplified model of heat transport within the snowpack in which snow properties are constant with depth and heat transport takes place purely by conduction. In reality, meltwater percolation, wind pumping (Clarke et al. 1987), and shortwave radiative transfer (Schwerdtfeger and Weller 1977) may contribute to the heat transport. These processes can be accounted for in more sophisticated snow models (e.g., Morris et al. 1994). However, our calculations show that G is a relatively minor component of the surface energy balance, and we believe that the errors in our estimates are less than the uncertainties in the observations of the radiation components. We thus believe that our simple model is adequate for the purpose of validating GCM simulations.
c. Sensible and latent heat flux
Very few direct measurements of the atmospheric fluxes of sensible and latent heat have been made in the Antarctic. The spatial and temporal coverage of the few measurements that do exist (e.g., King et al. 1996; Stearns and Weidner 1993; Ohata et al. 1985) is not sufficient for these measurements to be of direct use in validating GCM simulations. We have chosen to calculate HS + HL as the residual of the other terms in (1), all of which can be measured or calculated. We then compare this residual, which we shall henceforth refer to as the atmospheric heat flux, with HS + HL calculated by the GCM.
At the low temperatures experienced in Antarctica the atmosphere holds little water vapor even when saturated. Consequently, the latent heat flux is also small and, in winter, makes an insignificant contribution to the surface energy balance. At Halley, a coastal station 700 km southwest of NEU, the latent heat flux is typically an order of magnitude smaller than the sensible heat flux during the winter months (King and Anderson 1994). Measurements at MIZ (Ohata et al. 1985) show that, at this colder station, latent heat flux is negligible during the winter but is of comparable magnitude to the sensible heat flux during summer.
4. Validation of the unified model
a. Problems with collocation
In general, model grid points will not be collocated with stations from which observations are available. This presents no great difficulties over the interior of Antarctica where the topography is relatively smooth and model variables can be interpolated onto the locations of stations such as SPO, PLA, and MIZ in a relatively simple way. At coastal stations such as NEU, however, matters are not quite so straightforward. Because the coastal orography slopes steeply, the nearest land point to a coastal station is often at an elevation of several hundred meters (595 m in the case of NEU). The nearest grid points with an elevation comparable to that of the station are generally ocean points that, irrespective of the degree of sea ice cover, will have very different surface energy balance characteristics to the snow-covered land surface at the station location. It is clearly not possible to derive representative surface fluxes for the station location by simple interpolation between such a point and an elevated land point, and this problem has restricted the utility of data from coastal regions (where the majority of Antarctic research stations are located) in validating model performance. Recently, the unified model has been run with a higher horizontal resolution than was used in the present study. Such runs may offer a potential solution to the problem of collocation but were not available to us when the work reported here was carried out.
In the following subsections, we compare the components of the surface energy balance in the model with observations at SPO, PLA, MIZ, and NEU. SPO is collocated with a model grid point that has an elevation of 2708 m—approximately the same elevation as the real station (2800 m). Model data from the grid points surrounding PLA were linearly interpolated onto the station location, giving a model point with an elevation of 3380 m compared to the actual station elevation of 3625 m. Around MIZ, there are significant differences between the model and real topography. We have chosen to interpolate model results onto a point that has the same elevation as the real MIZ station (2230 m). This interpolated point lies 1.23° to the south of the real location of MIZ. For comparison with NEU observations, we have taken data from the nearest model grid point, which lies at an elevation of 595 m, compared with a station elevation of 36 m.
b. Shortwave radiation
Figure 1 shows model calculations and observations of the annual course of incoming and reflected shortwave radiation at SPO, MIZ, and NEU. Although there is some tendency for the model to overestimate S↓ at SPO, MIZ, and PLA (not shown), the differences between modeled and observed values are no greater than those that might be expected from interannual variability, assuming that the interannual variability of around 10% of the mean seen in the long series of measurements from NEU is typical of these stations. The model results also indicate this level of interannual variability over the 5-yr model run. At NEU, the overestimate of S↓ is larger, particularly from spring to midsummer. However, this may simply result from the higher elevation, and hence smaller airmass factor, of the point chosen to represent NEU in the model.
The model uses an albedo of 0.80 throughout the year at SPO, PLA, and MIZ and in all months except December and January at NEU, when a fractionally smaller value is used. Observations indicate that albedos at these stations range between 0.81 and 0.85, so the model value may be slightly too low. Changing to a slightly higher value of, say, 0.83 would increase S↑ proportionately but would only have a small impact on the modeled surface energy balance. At SPO, the annual average modeled net shortwave balance (i.e., S↓ − S↑) would be reduced from 28.1 to 23.9 W m−2.
c. Longwave radiation
Observations and model calculations of the longwave radiation components at SPO, MIZ, and NEU are shown in Fig. 2. At SPO and MIZ, modeled L↓ values are significantly smaller than observed values by 10–20 W m−2 throughout most of the year. At NEU the difference is even more marked, but it must be remembered that the model point is at a higher elevation than the real station. The interannual variability in modeled monthly means is around 10% of the 5-yr mean at all stations, which is comparable with the observed variability at NEU. Modeled L↑ values at SPO are in reasonable agreement with observations, indicating that the modeled surface temperature at this location is also realistic. Modeled 1.5-m air temperatures at SPO and PLA (Fig. 3) are also close to observed values. At MIZ, however, modeled L↑ is significantly smaller than that observed, and Fig. 3 shows that modeled air temperatures are up to 8°C colder than observations during the winter months at this station. Modeled L↑ is too low at NEU because the model temperature at this station is also too cold but, again, some of this discrepancy may result from the excess elevation of the model grid point.
The absolute accuracy of monthly means of longwave radiation measurements is probably about ±5 W m−2 (Wild et al. 1995; Garratt and Prata 1996), so the differences between modeled and observed values seen in Fig. 2 are possibly of marginal statistical significance. However, the similar results seen at all stations (where different instruments and procedures were used), and the appearance of a bias in all months of the year strongly suggests that the observed bias is genuine.
d. Snow heat flux
The heat flux, G, at the surface of the snowpack is not available directly as a model diagnostic. This is because the model has an upper snow layer of finite thickness so G is equal to the heat flux into the base of this layer less the rate of heat storage within the layer. It is simpler to calculate G as the residual of the other components of the surface energy balance in the model using (1), and this is what we have chosen to do.
Figure 4 shows model calculations and “observed” values of G (i.e., those calculated using the method described in section 3b) at SPO, PLA, MIZ, and NEU. Monthly mean values of G are seen to be much smaller than the components of the net radiation balance, so G is only of secondary importance in the overall surface energy balance. Modeled and observed values of G are in very good agreement at all four stations. This indicates that, as seen in Fig. 3, the model correctly reproduces the amplitude and form of the annual variation of surface temperature at these locations.
e. Atmospheric heat flux
As described in section 3, we have calculated the observed atmospheric heat flux (i.e., the sum of the sensible and latent heat fluxes) as the residual of the radiation fluxes and the snow heat flux using (1). In Fig. 5, this quantity is compared with the sum of the model sensible and latent heat fluxes at SPO, PLA, MIZ, and NEU. At SPO there are significant differences between the model and observations in all months. The observations show the heat flux going negative (i.e., carrying heat away from the surface) between November and March and averaging around +10 W m−2 during the winter months. In the model, the heat flux only becomes slightly negative during December and January and averages about +26 W m−2 during the winter. The excessive model heat flux at SPO thus compensates almost exactly for the deficiency in L↓ and results in a modeled surface temperature close to that that is observed. A similar behavior is seen at PLA, although here the model heat flux exceeds observations by only some 10 W m−2 for most of the year. The model thus seems to generate an excessive heat flux over much of the Antarctic plateau.
At MIZ, observed and modeled heat fluxes are in good agreement throughout the year. We have already seen that at this station the deficiency in L↓ in the model is compensated for by a lower surface temperature and hence a smaller value of L↑. At NEU, the model heat flux is somewhat greater than that observed, although, once again, there are problems with making a direct comparison between model results and observations at this coastal station.
f. Summary
The results presented above are summarized in Table 2, which shows annual means of the observed and modeled surface energy balance components at SPO, PLA, MIZ, and NEU (G is not shown because its annual average is close to zero by definition). In this table we have attempted to correct the modeled longwave fluxes at NEU for the unrealistically high elevation of this point in the model; L↓ has been corrected using the factor of −2.8 W m−2 per 100 m of elevation suggested by Wild et al. (1995), while L↑ has been recomputed by taking model monthly mean surface temperatures for NEU, adding a correction based on the lapse rate of −5.102°C km−1 calculated by Fortuin and Oerlemans (1990) for the coastal escarpment region of Antarctica, and then recomputing L↑ using (3) with ε = 1.0. These corrections can only be regarded as approximate, but they facilitate the comparison of modeled and observed data at this station.
At SPO, PLA, and NEU the model slightly overestimates net shortwave radiation. More seriously, the downward longwave component is underestimated at all stations. In the interior of Antarctica, characterized by SPO and PLA, the model compensates for the reduction of 10–15 W m−2 in downward longwave radiation by maintaining an atmospheric heat flux that is considerably larger than that observed and thus maintains a surface temperature that is in good agreement with observations. Connolley and Cattle (1994) noted that while near-surface temperatures at SPO were simulated quite well by the unified model, winter temperatures at Vostok, Antarctica (78.5°S, 106.9°E; 3488 m), were too cold by up to 5°C. Radiation data for Vostok (Schwerdtfeger 1984) are not well documented and have been excluded from our detailed study. However, a tentative comparison with unified model results suggests that modeled values of L↓ at Vostok are about 20 W m−2 less than observed values, a significantly greater deficiency than that seen at SPO. It is likely that the unified model overestimates HS + HL at Vostok to a similar degree as at SPO since the climatologies of both near-surface wind speed and temperature gradient are similar at these two stations. However, it would appear that in the case of Vostok, the enhanced surface heat fluxes are not sufficient to balance the deficiency in L↓, resulting in an anomalously cold surface.
At MIZ, on the steep coastal slopes, modeled and observed heat fluxes are in good agreement and the model achieves balance by maintaining a lower surface temperature than that that is observed thus reducing the radiative cooling of the snow surface. It is not possible to compare observations from coastal stations with model results directly but the indications are that the underestimate of downward longwave radiation in the model is also present at the coast and is balanced partly by excessive heat flux in the model and partly by a reduction in surface temperature. In the following section we investigate the reasons for the discrepancies between modeled and observed surface energy balance components.
5. Discussion
Wild et al. (1995) compared values of L↓ calculated by the ECHAM3 climate model with observations from a global network of stations and concluded that this model underestimates L↓ by an average of 14 W m−2. They attributed this underestimation partly to the model’s failure to generate sufficient low-level cloud and partly to underestimation of clear-sky radiation. Calculations using the ECHAM3 radiation scheme and the LOWTRAN7 narrowband model forced by the same atmospheric profiles showed that the value of L↓ predicted by ECHAM3 was consistently lower, by about 10 W m−2, than that produced by the more sophisticated LOWTRAN7 model. Wild et al. (1985) also compared L↓ predicted by other GCMs with observations and concluded that the underestimation of L↓ was a problem common to many models, a result confirmed in a study of four more GCMs by Garratt and Prata (1996). Our study shows that the unified model also suffers from this problem. Wild et al. (1995) showed that, in ECHAM3, the underestimate of L↓ was, in a global average, almost exactly balanced by an overestimate of S↓, leading to a global surface energy balance in good agreement with observations. We have shown that the unified model slightly overestimates S↓ in Antarctica, but this is not sufficient to balance the underestimate in L↓. Furthermore, during the Antarctic winter S↓ is zero and the model has to compensate for the lack of L↓ by either reducing L↑ (i.e., reducing the surface temperature) or generating a larger surface heat flux.
The anomalously low L↓ values calculated by the unified model may arise either as a result of deficiencies in the longwave radiation scheme used in the model or from incorrect modeling of the atmospheric temperature and humidity structure or cloudiness. An analysis of the performance of the model radiation scheme is beyond the scope of the present paper, but we may make a few remarks on errors of the latter kind. We have compared modeled lower-tropospheric temperatures and total column moisture with observations (Connolley and King 1993) at NEU and SPO. At NEU, the modeled mean July temperature of the lowest 200 hPa of the troposphere is 1.0°C colder than in the observations and the total column moisture is only 83% of that observed. The corresponding figures for SPO are 1.7°C and 47%, respectively. Thus the lower troposphere in the model appears to be anomalously cold and dry; this is likely to contribute to an underestimate in L↓. At Vostok, the anomaly in lower-tropospheric temperatures is even greater (5.6°C too cold in July), which may explain the larger L↓ anomaly seen at this station. We have not attempted to validate the representation of cloud cover in the model. This would be a difficult task, partly because of the difficulty of obtaining reliable cloud observations during the polar night but also because of the uncertainty as to how cloud fractions calculated by the model should be related to actual observations. The best approach might be to compare cloud longwave forcing in the model with observations, but this would require daily (or even shorter period) data that were not available for the present study.
Dutton et al. (1989) give monthly means of resultant wind speed at 10 m and of air temperature at heights of 2 and 20 m at the South Pole along with their radiation measurements. Conveniently, the height of the lowest model level at the South Pole is very close to 20 m and the model produces a diagnostic temperature at 1.5 m, so it is possible to compare the measured temperatures directly with model results. The observed 2–20-m and modeled 1.5-m–level-one temperature differences (Fig. 6) are seen to be in good agreement. Furthermore, the difference between modeled 1.5-m and surface temperatures is consistent with measurements from this location quoted by Schwerdtfeger (1984). The anomalously large model heat flux thus does not appear to be driven by excessive temperature gradients.
Comparison of modeled and observed wind speeds is somewhat more problematic since wind speed observations are made at a height of 10 m. The unified model does produce a 10-m wind speed as a diagnostic, but calculation of this quantity involves the use of the model’s surface flux scheme and the difference between model level-one and 10-m wind speeds can be quite large. In what follows, we demonstrate that this scheme may not be appropriate to Antarctic conditions, and we thus have some reservations about the validity of the diagnostic 10-m winds produced by the model. Similar reservations apply to the model diagnostic 1.5-m temperature used in Fig. 6; however, the correction applied to derive this from model surface temperature is relatively small. Using Monin–Obukhov similarity theory together with the observed monthly mean atmospheric heat fluxes (Fig. 5), we tentatively suggest that SPO winds from the first model level can be reduced to 10 m by multiplying by 0.95 from November to February and by 0.80 for the remainder of the year. Figure 7 shows model winds thus reduced together with observations. During the winter months, the ratio of model to observed wind speed is about 1.5. The excessive strength of the model winds may contribute to the large modeled heat flux but is probably not sufficient to explain it all. During the winter months, the modeled heat flux is about 25 W m−2. Other things being equal, we would expect the heat flux to change approximately proportionally to the wind speed; thus, if the model wind speed were reduced to the observed value, we might expect the model flux to drop to about 17 W m−2. This is still significantly larger than the observed mean winter heat flux of about 10 W m−2, suggesting that the excess modeled heat flux at the South Pole results from the use of too large a transfer coefficient in the model.
We shall now compare this functional form with observations. Using hourly mean observations of surface fluxes and surface layer profiles of wind speed and temperature made at Halley, Antarctica (King and Anderson 1994), we can study the variation of observed bulk transfer coefficients with stability. Figure 8 shows observations of the stability function for momentum, fM(RiB) as a function of RiB. Here, fM(RiB) is defined in an analogous manner to fH(RiB) as the ratio of the bulk transfer coefficient for momentum CD to its mean value under neutral conditions. The transfer coefficient and bulk Richardson number have both been calculated from the measured snow surface temperature and wind and temperature measurements at approximately 18 m. This is close to the height of the first level of the unified model over Antarctica, so the measurements should be directly comparable with model results. Also shown in Fig. 8 is the function fH(RiB) as given by (10) and (11). (The unified model actually uses the same stability function for momentum as for heat so this is a valid comparison.) The data points are somewhat scattered but the observed transfer coefficient clearly decreases much more rapidly with increasing stability than is predicted by (11). It thus seems likely that the overprediction of atmospheric heat flux over the interior of Antarctica in the unified model results largely from the use of an inappropriate stability dependence of the model bulk transfer coefficients.
In section 4, we noted that model atmospheric heat fluxes were only excessive in the interior while on the coastal slopes (MIZ) they were in reasonable agreement with observations. On the coastal slopes, strong katabatic winds act to keep the surface layer well mixed and the bulk Richardson number is generally smaller than at interior stations. If RiB is small, the stability correction to the neutral bulk transfer coefficients is small and the transfer coefficients calculated using (11) will differ little from those determined from a more appropriate stability function. We have investigated the effect of changing the model surface flux parameterization by taking time series of model data from SPO and MIZ and recomputing surface fluxes using alternative schemes. The time series contain modeled winds and temperatures at level 1 and at the surface for each model time step (0.5 h) during the month of July. From these data, heat fluxes have been computed using both the unified model scheme (11) and a modified Monin–Obukhov scheme, described in the appendix. This latter scheme has been shown to perform well when validated against measurements made at Halley.
The results of this recomputation are shown in Table 3. At MIZ the mean wind speed is quite high and the mean temperature difference between the surface and model level 1 is relatively small, leading, on average, to a small value of RiB. The bulk transfer coefficients thus remain close to their neutral values, and there is only a small difference between the heat fluxes calculated using the unified model and modified schemes. At SPO, however, the level-1 winds are somewhat weaker and the temperature difference is much larger, making RiB of order 0.1 for much of the time. Under these very stable conditions the unified model scheme generates a heat flux that is substantially larger than that predicted by the modified scheme.
6. Conclusions
Comparison of surface energy fluxes computed by the unified model with observations at Antarctic stations has shown that the modeled shortwave radiative fluxes are in reasonable agreement with observations but that the downward flux of longwave radiation is consistently too small. A study by Wild et al. (1995) suggests that this problem is common to many GCMs and is not restricted to Antarctica. However, the impact of this error is likely to be greater in Antarctica than elsewhere. This is because there is no compensating excess of absorbed shortwave radiation during the Antarctic winter and because the error in downward longwave radiation is a significant fraction of all individual components of the surface energy balance. At low latitudes, all energy fluxes are larger but, according to Wild et al. (1995), the deficit in L↓ varies little with location and will thus be relatively less important in the Tropics.
Over the interior of Antarctica, the unified model compensates for the lack of downward longwave radiation by generating an atmospheric heat flux that is significantly larger than that that is observed. This excessive heat flux is partly sustained by somewhat anomalously strong model surface winds, but we have also demonstrated that the surface heat flux parameterization used in the model is not appropriate to the very stable conditions that prevail in Antarctica. Since the errors in the modeled heat flux and downward longwave radiation approximately cancel each other over much of this region, the modeled surface temperature is in good agreement with observations (Connolley and Cattle 1994). If only one of these errors was rectified, the agreement between modeled and observed climate might not be so good, so it would not be appropriate to change the surface heat flux parameterization without first modifying the radiation scheme. However, both problems need to be addressed, particularly if we wish to remain confident in the model’s ability to produce accurate predictions of future Antarctic regional climates.
Although this study has considered results from only one model, our conclusions are probably broadly applicable to many GCMs in current use. The underestimate of L↓ in a variety of GCM radiation schemes has already been noted by Wild et al. (1995) and by Garratt and Prata (1996). Furthermore, the surface heat flux parameterization used in the unified model is similar to that used in many GCMs. We might thus expect other GCMs to show similar errors in the surface energy balance over Antarctica and in other regions where strongly stable stratification prevails in the boundary layer. Ohmura et al. (1994) found that the surface heat flux scheme employed in the ECHAM3 GCM (which is similar to that used in the unified model) systematically overestimated sensible heat fluxes under stable conditions over the Greenland Ice Sheet for similar reasons to those found in this study, indicating that our findings are not restricted to one particular model or to a single geographic region.
Acknowledgments
We thank P. S. Anderson for suggesting how we might calculate the heat flux through the snowpack, R. L. H. Essery for extracting time series data from the unified model, and S. H. Derbyshire for useful discussions. The model results described in this paper were obtained from model runs carried out by the Hadley Centre for Climate Prediction and Research at the U.K. Meteorological Office, and we thank Dr. H. Cattle and colleagues for making these data available to us.
REFERENCES
Anderson, P. S., 1994: Aspects of the Antarctic boundary layer. Ph.D. dissertation, University of Lancaster, 196 pp. [Available from University of Lancaster, Lancaster LA1 4YQ, United Kingdom.].
Arya, S. P., 1988: Introduction to Micrometeorology. Academic Press, 307 pp.
Businger, J. A., J. C. Wyngaard, Y. Izumi, and E. F. Bradley, 1971: Flux-profile relationships in the atmospheric surface layer. J. Atmos. Sci.,28, 181–189.
Carroll, J. J., 1982: Long-term means and short-term variability of the surface energy balance components at the South Pole. J. Geophys. Res.,87, 4277–4286.
Carson, D. J., and P. J. R. Richards, 1978: Modelling surface turbulent fluxes in stable conditions. Bound.-Layer Meteor.,14, 67–81.
Clarke, G. K. C., D. A. Fisher, and E. D. Waddington, 1987: Wind pumping: A potentially significant heat source in ice sheets. The Physical Basis of Ice Sheet Modelling, E. D. Waddington and J. S. Walder, Eds., International Association of Hydrological Studies Press, 169–180.
Connolley, W. M., and J. C. King, 1993: Atmospheric water-vapour transport to Antarctica inferred from radiosonde data. Quart. J. Roy. Meteor. Soc.,119, 325–342.
——, and H. Cattle, 1994: The Antarctic climate of the UKMO unified model. Antarct. Sci.,6, 115–122.
Cullen, M. J. P., 1993: The unified forecast/climate model. Meteor. Mag.,122, 81–94.
——, and T. Davies, 1991: A conservative split–explicit integration scheme with fourth-order horizontal advection. Quart. J. Roy. Meteor. Soc.,117, 993–1002.
Dutton, E. G., R. S. Stone, and J. J. DeLuisi, 1989: South Pole surface radiation balance measurements April 1986 to February 1988. NOAA Data Rep. ERL ARL-17, 149 pp. [Available from National Technical Information Services, 5285 Port Royal Rd., Springfield, VA 22161.].
Fortuin, J. P. F., and J. Oerlemans, 1990: Parameterisation of the annual surface temperature and mass balance of Antarctica. Ann. Glaciol.,14, 78–84.
Garratt, J. R., 1992: The Atmospheric Boundary Layer. Cambridge University Press, 316 pp.
——, and A. J. Prata, 1996: Downwelling longwave fluxes at continental surfaces—A comparison of observations with GCM simulations and implications for the global land-surface radiation budget. J. Climate,9, 646–655.
Högström, U., 1988: Non-dimensional wind and temperature profiles in the atmospheric surface layer: A re-evaluation. Bound.-Layer Meteor.,42, 55–78.
Ishikawa, N., S. Kobayashi, T. Ohata, and S. Kawaguchi, 1982: Radiation data at Mizuho station in Antarctica in 1980. JARE Data Rep. 73, 145 pp. [Available from National Institute of Polar Research, 9-10, Kaga 1-chome, Itabashi-ku, Tokyo 173, Japan.].
Johns, T. J., R. E. Carnell, J. F. Crossley, J. M. Gregory, J. F. B. Mitchell, C. A. Senior, S. F. B. Tett, and R. A. Wood, 1997: The second Hadley Centre coupled ocean–atmosphere GCM: Model description, spinup and validation. Climate Dyn., in press.
King, J. C., 1990: Some measurements of turbulence over an Antarctic ice shelf. Quart. J. Roy. Meteor. Soc.,116, 379–400.
——, and P. S. Anderson, 1994: Heat and water vapour fluxes and scalar roughness lengths over an Antarctic ice shelf. Bound.-Layer Meteor.,69, 101–121.
——, S. D. Mobbs, and N. R. Edwards, 1994: Surface boundary conditions in stably stratified environmental flows. Stably Stratified Flows: Flow and Dispersion over Topography, I. P. Castro and N. J. Rockliff, Eds., Oxford University Press, 93–103.
——, P. S. Anderson, M. C. Smith, and S. D. Mobbs, 1996: The surface energy and mass balance at Halley, Antarctica during winter. J. Geophys. Res.,101, 19 119–19 128.
Kuhn, M., L. S. Kundla, and L. A. Stroschein, 1977: The radiation budget at Plateau Station, Antarctica. Meteorological Studies at Plateau Station, J. A. Businger, Ed., Antarctic Research Series, Vol. 25, Amer. Geophys. Union, 41–73.
Louis, J. F., 1979: A parametric model of vertical eddy fluxes in the atmosphere. Bound.-Layer Meteor.,17, 187–202.
Mahrt, L., 1987: Grid-averaged surface fluxes. Mon. Wea. Rev.,115, 1550–1560.
Mitchell, J. F. B., and C. A. Senior, 1989: The Antarctic winter: Simulations with climatological and reduced sea-ice extents. Quart. J. Roy. Meteor. Soc.,115, 225–246.
Morris, E. M., 1989: Turbulent transfer over snow and ice. J. Hydrol.,105, 205–223.
——, P. S. Anderson, H.-P. Bader, P. Weilenman, and C. Blight, 1994: Modelling mass and energy exchange over polar snow using the DIASY model. Snow and Ice Covers: Interactions with the Atmosphere and Ecosystems, H. G. Jones, T. D. Davies, A. Ohmura, and E. M. Morris, Eds., International Association of Hydrological Studies Press, 53–60.
Ohata, T., N. Ishikawa, S. Kobayashi, and S. Kawaguchi, 1985: Heat balance at the snow surface in a katabatic wind zone, east Antarctica. Ann. Glaciol.,6, 174–177.
Ohmura, A., T. Konzelmann, M. Rotach, J. Forrer, M. Wild, A. Abe-Ouchi, and H. Toritani, 1994: Energy balance for the Greenland ice sheet by observation and model computation. Snow and Ice Covers: Interactions with the Atmosphere and Ecosystems, H. G. Jones, T. D. Davies, A. Ohmura, and E. M. Morris, Eds., International Association of Hydrological Studies Press, 85–94.
Rusin, N. P., 1961: Meteorological and Radiational Regime of Antarctica. Israel Program for Scientific Translations, 355 pp.
Schmidt, T., and G. König-Langlo, 1994: Radiation Measurements at the German Antarctic Station Neumayer 1982–1992. Berichte zur Polarforschung 146. Alfred-Wegener-Institute fur Polar- und Meeresforschung, 66 pp.
Schwerdtfeger, P., and G. Weller, 1977: Radiative heat transfer processes in snow and ice. Meteorological Studies at Plateau Station, J. A. Businger, Ed., Antarctic Research Series, Vol. 25, Amer. Geophys. Union, 35–39.
Schwerdtfeger, W., 1984: Weather and Climate of the Antarctic, Elsevier, 261 pp.
Slingo, A., and R. C. Wilderspin, 1986: Development of a revised longwave radiation scheme for an atmospheric general circulation model. Quart. J. Roy. Meteor. Soc.,112, 371–386.
Smith, R. N. B., 1990: Subsurface, surface and boundary layer processes. Unified Model Paper 245 75 pp. [Available from the National Meteorological Library, Bracknell, RG12 25Z, United Kingdom.].
Stearns, C. R., and G. A. Weidner, 1993: Sensible and latent heat flux estimates in Antarctica. Antarctic Meteorology and Climatology: Studies Based on Automatic Weather Stations, D. H. Bromwich and C. R. Stearns, Eds., Antarctic Research Series, Vol. 61, Amer. Geophys. Union, 109–138.
Tzeng, R.-Y., D. H. Bromwich, and T. R. Parish, 1993: Present-day Antarctic climatology of the NCAR Community Climate Model version 1. J. Climate,6, 205–226.
——, ——, ——, and B. Chen, 1994: NCAR CCM2 simulation of the modern Antarctic climate. J. Geophys. Res.,99, 23 131–23 148.
Webb, E. K., 1970: Profile relationships: The log-linear range, and extension to strong stability. Quart. J. Roy. Meteor. Soc.,96, 67–90.
Wild, M., A. Ohmura, and H. Gilgen, 1995: Validation of general circulation model radiative fluxes using surface observations. J. Climate,8, 1309–1324.
Yamanouchi, T., M. Wada, S. Mae, and S. Kawaguchi, 1981: POLEX-South data. Part I: Radiation data at Mizuho station in 1979. JARE Data Rep. 61, 350 pp. [Available from National Institute of Polar Research, 9-10, Kaga 1-chome, Itabashi-ku, Tokyo 173, Japan.].
APPENDIX
The Modified Monin–Obukhov Surface Flux Scheme
The annual cycle of incoming and reflected shortwave radiation at SPO, MIZ, and NEU as measured (broken line) and as modeled in the unified model (solid line).
Citation: Journal of Climate 10, 6; 10.1175/1520-0442(1997)010<1273:VOTSEB>2.0.CO;2
The annual cycle of downward and upward longwave radiation at SPO, MIZ, and NEU as measured (broken line) and as modeled in the unified model (solid line).
Citation: Journal of Climate 10, 6; 10.1175/1520-0442(1997)010<1273:VOTSEB>2.0.CO;2
The annual cycle of 1.5-m air temperature at SPO, PLA, MIZ, and NEU as measured (broken line) and as modeled in the unified model (solid line).
Citation: Journal of Climate 10, 6; 10.1175/1520-0442(1997)010<1273:VOTSEB>2.0.CO;2
The annual cycle of heat flux at the surface of the snowpack at SPO, PLA, MIZ, and NEU as calculated using the method described in section 3b (broken line) and as modeled in the unified model (solid line).
Citation: Journal of Climate 10, 6; 10.1175/1520-0442(1997)010<1273:VOTSEB>2.0.CO;2
The annual cycle of total atmospheric heat flux (HS + HL) at SPO, PLA, MIZ, and NEU inferred from the difference between the observed net radiation and calculated snow heat flux (broken line) and as modeled in the unified model (solid line).
Citation: Journal of Climate 10, 6; 10.1175/1520-0442(1997)010<1273:VOTSEB>2.0.CO;2
A comparison of modeled and observed temperature gradients at SPO. The solid line shows monthly mean values of the temperature difference between the lowest model level (approximately 20 m above the surface) and 1.5 m. The broken line shows monthly mean values of the observed temperature difference between 20 and 2 m.
Citation: Journal of Climate 10, 6; 10.1175/1520-0442(1997)010<1273:VOTSEB>2.0.CO;2
The annual cycle of 10-m vector mean wind speed at SPO as observed and as calculated from unified model level-1 winds.
Citation: Journal of Climate 10, 6; 10.1175/1520-0442(1997)010<1273:VOTSEB>2.0.CO;2
Observations made at Halley, Antarctica, of the variation of the stability function for momentum, fM(RiB), with bulk Richardson number. Each cross represents a 1-h mean. Also shown are the forms of the function fH(RiB) given by (10) (broken line) and (11) (solid line).
Citation: Journal of Climate 10, 6; 10.1175/1520-0442(1997)010<1273:VOTSEB>2.0.CO;2
Details of stations from which radiation measurements were used for this study.
Annual mean surface energy fluxes at SPO, PLA, MIZ, and NEU.
Mean sensible heat fluxes for July recalculated from model time series data using the unified model scheme and a modified scheme.
