W. M. Connolley1, J. C. King1 and H. Cattle2
1 British Antarctic Survey, Cambridge CB3 OET, UK.
2 Hadley Centre for Climate Prediction and Research, Bracknell
RG12 2SZ, UK.
The mass balance of Antarctica has two major components: atmospheric (precipitation and evaporation) and glaciological (discharge of glaciers and ice sheets). The atmospheric component is likely to be the most important on short (<100 year) timescales.
We use observations from radiosonde stations1 and GCM simulations of the current climate2 to compute horizontal moisture fluxes, from which we can deduce estimates of the accumulation rate. These can also be compared to glaciological estimates3,4.
The use of radiosonde or glaciological data for monitoring accumulation has drawbacks. Individual ice cores or pits may provide accurate measurements but these are typically of a single place and time-period. Maps of accumulation use non-cotemporal data because of the paucity of measurements; different maps show widely varying features. By integrating moisure fluxes around coastal boundaries5, the average accumulation over the Antarctic interior may be inferred. Potentialy, yearly or decadal variations over large areas may be measured. However the sparsity of the observing network may introduce large and poorly quantified errors. Model data are complete, but the accuracy of the model needs to be verified. To assess the errors the flux- integral technique introduces we can compare this method using model data interpolated to station locations with modelled accumulation.
The amount of moisture available for precipitation over Antarctica is indicated by the total column moisture (TCM). This is shown in figure 1 at the 16 stations, as observed and as modelled. The stations fall into three groups. On the Antarctic Peninsula the TCM is relatively high reflecting the relatively warm and moist conditions prevailing over the Southern Ocean. The 13 stations around the coast of East Antarctica all have comparable values of TCM which are significantly lower. Finally, the two stations high on the Antarctic plateau have very low values of TCM, reflecting the height of the stations.
The TCM values simulated by the GCM are compared to the radiosonde data in table 1. There is good agreement except at the two interior stations, where values are too low. This probably reflects the cold bias in the model atmosphere. 3. Moisture Fluxes
Figure 2 shows the column moisture flux vectors at the 16 stations, with ellipses defining interannual variability. Table 2 compares our results with those of Bromwich6, which were for the single year 1972, and the model. Our total fluxes are in close agreement with Bromwich, but our data indicate a larger value for the eddy components and a corresponding decrease in the mean flux.
The model data show much weaker transports, especially the easterly component. Some of this may be due to a shallower easterly layer above the coast in the model. In reality the easterly layer is up to 500 hPa thick, whereas in the model the layer is thinner, leading to cancellation between the easterly fluxes near the ground and westerly fluxes aloft. However, this cannot explain the entire discrepancy. A further possibility is that the large easterly values derived from the radiosonde stations are caused by the large topographic gradients at the coast, which the model cannot adequately resolve.
Further north, at the tip of the Peninsula in the belt of westerlies, the magnitude of the modelled fluxes (70 kg m-1s-1) and radiosonde-derived fluxes (77 kg m-1s-1) are much closer, although the model's northerly component is greater.
We compare our radiosonde derived accumulation estimate, model data, and glaciological estimates. The glaciological data are probably the benchmark against which the others are judged. The model moisture fluxes are interpolated to station locations to mimic the flux-integral method for radiosonde data, and model accumulation is used to compare with glaciological data.
Comparing the glaciological and model-glaciological values in table 3, we see that the model over-estimates accumulation by approximately 1/3. Sensitivity experiments have shown that the model is affected by many factors, e.g. changes to the gravity wave drag scheme that alter the structure of the circumpolar trough (which is too deep in this simulation). Decreasing climatological sea-ice extents or thicknesses increases precipitation. Halving the parameter for horizontal diffusion decreased accumulation by more that 25%.
The modelled and observed flux-integral budgets match approximately. Because of the shape of East Antarctica, the method mostly uses the northerly flux components which table 2 shows are in close agreement, even though the modelled easterly components are too small. Comparing this method to the glaciological one, it would seem that the flux method is not providing an accurate value for the accumulation. However, it may still be possible to use this method to diagnose relative year- to-year variations. 5. Conclusion
We have compared moisture fluxes from observations and from the UKMO GCM and deduced corresponding estimates of accumulation. These two sources produce comparable accumulation values, because averaged modelled and observed northerly fluxes are in agreement, even though the easterly fluxes are underestimated by the model. This might be considered as a warning on the use of climate models - even when they produce acceptable results this may hide errors in the simulation. On the other hand, we might conclude that even imperfect models can produce useful information.
We hope to extend this work to examine the model-observation differences by looking at variation seasonally and on the synoptic timescale. The weakness of the easterly fluxes, and the low TCM of the interior, may be explained by the vertical structure. When data from climate change runs becomes available it will be very interesting to look at the model response over Antarctica. Since such a run cannot be "verified", it is very important to ensure that the simulation of the current climate is right and for the right reasons.
1. Connolley, W. M. and King, J. C., 1993: Atmospheric water vapour transport to Antarctica inferred from radiosondes. Q. J. R. Meteorol. Soc., 119, 325-342
2. Connolley, W. M. and Cattle, H., 1993: The Antarctic climate of the UKMO Unified Model, submitted to Antarctic Science.
3. Giovinetto, M. B. and Bentley, C. R., 1985: Surface balance in ice drainage systems of Antarctica. Ant. J. U.S., 20, 6-13
4. Frolich, R. M., 1992: Surface mass balance of the Antarctic peninsula ice sheet. In: The contribution of Antarctic peninsula ice to sea level rise, ed. Morris, E. M., EC report EPOC-CT90- 0015
5. Bromwich, D. H., 1988: Snowfall in high southern latitudes. Rev. Geophys., 26, 149-168.
6. Bromwich, D. H., 1979: Precipitation and accumulation estimates for East Antarctica, derived from rawinsonde information. Ph.D thesis, University of Wisconsin-Madison.
Figure 1. Mean (dark bar) and interannual variability (light bar on top) of TCM from observations, and from the model (light bars). Scale - 5 kg m-2
Figure 2. Moisture fluxes from observations with interannual variability defined by ellipses (thick arrows), and from the model (thin arrows). 20 kg m-1 s-1
Figure 3. Expanded view of two areas, showing modelled (thin arrows) and observed (thick arrows) moisture fluxes. The model coastline is indicated.
Figure 4. Location map showing sectors A and B used in table 3, and areas (a) and (b) of figure 3.
Location Obs Model Coast 3.7 ñ0.45 3.25 ñ0.43 Amundsen-Scott 0.5 0.3 Vostok 0.6 0.2 Bellingshausen 8.2 8.2
Connolley & King Bromwich Model
Total N 6.1 ñ1.1 5.9 5.6
E 20.9 ñ3.5 22.3 7.5
Mean N 1.8 ñ0.8 2.6 1.1
E 14.0 ñ2.7 18.6 7.7
Eddy N 4.3 ñ0.7 3.3 4.4
E 6.7 ñ1.7 3.6 0.2
Observations Model
Area Radiosonde Glaciological Model-RS Model-Glac
A 168 ñ 32 101 150 142
B 71 ñ 29 106 97 153
C - 154 - 218
Area A = Sector between 0 and 110o E as shown on figure 4
Area B = Sector between 26o W and 167o E as shown on figure 4
Area C = All Antarctica, including ice shelves
Sources
Radiosonde data from ref. 1.
Glaciological data from ref. 3, estimated in the case of areas
A and B, modified according to ref. 4 for area C.
Model data from the UKMO GCM3 atmosphere-only.
Note
The areas used for the different methods vary slightly.