Introduction and Methodology

Two generations of the NCAR (U.S. National Center for Atmospheric Research) community climate model (CCM versions 1 and 2) are widely used global-climate models. However, comparatively little effort has been devoted to evaluating the models’ simulations of the present climate in polar regions. Not only are the polar regions the major energy sinks in the global atmospheric circulation but they also play a crucial role in climate change. Recently, (Reference Tzeng. and Parish.Tzeng and others 1993, Reference Tzeng. and Bromwich1994), Reference Bromwich, Tzeng and Parish.Bromwich and others (1994) and Reference Tzeng., Bromwich, Parish and Chen.Tzeng and Bromwich (1994) have analyzed the simulated climate of the Arctic and Antarctica by CCM1 and CCM2 in terms of atmospheric circulation, storm tracks, moisture and energy budgets, cloud coverage, radiation fluxes, etc. In this summary paper, we will focus on the hydrologic cycle, i.e. moisture budget of the simulated atmosphere in the polar regions by CCM1 and CCM2, with horizontal resolutions of R15 and T42 (about 7.5° x 4.5° and 2.8° x 2.8° in longitude and latitude), respectively. There are 12 and 18σ it levels in the vertical for CCM1 and CCM2, respectively. One of the important differences between the two versions of the model, which is directly related to moisture budgets, is the method used to predict moisture contents. CCM1 uses the spectral-transform method to predict the moisture with a positive-moisture fixer to correct negative values of moist-ure content in arid areas due to the spectral truncation in the model. CCM2 uses a shape-preserving semi-Lagrangian transport for all positive-definite quantities (including water vapor, cloud water, etc.) to alleviate the negative values created by spectral truncation (Reference Williamson and Rasch.Williamson and Rasch, 1994) .

The moisture budget is calculated from the temporally and spatially averaged moisture-balance equation (Reference Peixoto, Oort., Streei Perron, Benin and Ratelille.Peixoto and Oort, 1983). The equation has the form,

where ϕ is the latitude, the angled brackets denote a zonal average, and the overbar a time average, Q _ϕ is the vertically averaged moisture, transport (flux) in the meridional direction, E and P are evaporation and precipitation, respectively, and R is a residual term which represents the lack of balance in the moisture equation over long time averages. For the polar caps, the fluxconvergence term of the equation can be rewritten using Gauss’s theorem:

where A is the surface area of the polar cap and the square brackets denote an area average of the polar cap.

Arctic Moisture Budget

The mean annual precipitation rate [P] averaged from 70° N to the North Pole is 29.4 cm a¹ from a new analysis (Reference Walsh, Zhou, Portis and Serreze.Walsh and others, 1994; hereafter referred to as W94), which is about twice that of (Reference Peixoto and Oort.Peixoto and Oort’s 1992; hereafter referred to as PO92) observations (Table 1). The simulated precipitation in CCM1 is 51.9 cm a^-1, which is about 1.9 and 3.2 limes more precipitation than observed by W94 and PO92, respectively. The simulated evaporation rate [E] (13.8 cm a^-1) approximates the observed. However, the CCM1’s net annual precipitation [P — E] is 2.3 and 6.6 times larger than that of the observations of W94and PO92, respectively. In addition, CCM1 transports 10.5 x 10¹⁵kg a^-1 of moisture across 70° N (or 67.8 cm a^-1 water equivalent), but the observed (W94) is 2.6 x 10¹⁵kg a^-1 (or 16.3 cm a ^-1 w.c.). At least four times more moisture is transported into the North Polar ice cap by CCM1 than is observed.

Table 1. Annual values of the Arctic areally averaged (70° N to NP) precipitation ([P]), and evaporation ([E]) rates, and of the moisture-flux convergence poleward of 70° N ([FQ]) from the NCAR CCM1 and CCM2 and observations. Unit is cm a^-1

However, the CCMl’s annual cycle of precipitable water (W) is consistent with the observations in terms of both amplitude and phase. The annual average of W is 5.4 kg m^-2 in CCM1 and 6.0 kg m^-2 in PO92 (not shown). The maximum of W is in mid-summer (July—August) and the minimum in mid-winter (January—February) in both CCM1 and PO92. Therefore, the excessive precipitation rate in CCM1 ([P], 51.9 vs 29.4 cm a^-1) is due to the excessive moisture transported into this region ([FQ], 67.8 vs 16.3 cm a^-1). However, CCM1’s moisture flux into the Arctic basin ([FQ], 67.8 cm a^-1) is much larger than its net annual precipitation ([P — E]. 38.1 cm a^-1). Obviously, this excessive moisture (-29.7 cm a^-1) must be transported out of this region, which is done in an artificial manner by the positive-moislure-fixer scheme in order to remedy negative moisture values spuriously produced by the spectral-transform method (Reference Rasch and Williamson.Rasch and Williamson, 1990). On the other hand, the causes of the excessive simulated poleward moisture transport by the advection term ([FQ], 67.8 vs 16.3 cm a^-1) are attributed to the mertdional-wind component (v) at 70° N. The bias in the simulated v is dearly attributable to the mass-field (sea-level-pressure) simulation (Reference Bromwich, Tzeng and Parish.Bromwich and others, 1994), which is caused by the low horizontal R15 spectral-1runcation error that distorts the representation of Greenland topography in the model. We conclude that the errors in the moisture budget of CCM1 are due to the positive-moisture-fixer scheme and the low horizontal resolution.

Using the semi-Lagrangian transport scheme to predict water vapor instead of the spectral-transform method with positive-moisture fixer, CCM2 definitely can more realistically simulate the various moisture components (Table 1). The moisture-budget equation is now almost balanced but with a positive residual, and the moisture transport into the Arctic basin is reasonably well simulated. There is substantial uncertainty in the observed value of [P — E] (Reference Peixoto and Oort.Peixoto and Oort, 1992; Reference Walsh, Zhou, Portis and Serreze.Walsh and others, 1994; Reference Serreze, Barry and Walsh.Serreze and others, in press), with Walsh and others’ result being much closer to those modeled by CCM2. The primary cause of the [P — E] discrepancy is the very different observed values of [P] which need to be rationalized. Even so, the precipitation simulated by CCM2 appears to be high, most likely resulting from the deficiencies in the simulated atmospheric circulation of this region (Reference Bromwich, Tzeng and Parish.Bromwich and others, 1994; Reference Tzeng., Bromwich, Parish and Chen.Tzeng and Bromwich, 1994).

The annual variations of precipitation, evaporation and [P — E] over the Arctic basin are shown in Figure 1 along with the new observations from Reference Walsh, Zhou, Portis and Serreze.Walsh and others (1994). It is evident that CCM2 greatly overestimates the precipitation (circles) from late winter to summer, and the simulated maximum of [P] comes 1 month earlier than the observed maximum in August. Though the annual average evaporation of CCM2 is close to the observations, the annual variation of CCM2 evaporation (triangles) is shifted 2 months earlier than the observed. These result in large biases of simulated [P — E] (squares) during summer, almost 30% larger than the observations. The observations illustrate that the maxima of evaporation, precipitation and [P — E] occur in July, August and September, respectively. The COM2 captures this order quite well but is phase-shifted. For the summer months of July and August, [E] exceeds [P — E] (approximately equivalent to moisture-flux convergence) in the observations, implying that surface evaporation is the source of more moisture than the net import of water vapor from outside of the North Polar ice cap. By comparison to the moisture budget over the Arctic Ocean, which is dominated by convergence of moisture fluxes year-round. Reference Walsh, Zhou, Portis and Serreze.Walsh and others (1994) pointed out that the hydrologic budget of the entire polar cap (70° N to NP) is influenced significantly by the icefree parts of the North Atlantic sub-polar seas and by the snow-free (during summer) land areas. Evaporation in the CCM2 also becomes rite dominant source from April to June over the Arctic basin, which however, is 2 months ahead of the observed. The cause of these phase shifts in the simulated moisture budget will be investigated in the near future.

Fig. 1. The areally averaged (Arctic) annual variations of precipitation, surface evaporation, and (P — E) from CCM2 and the observations (from Reference Walsh, Zhou, Portis and Serreze.Walsh and others, 1994). Unit is cm month^-1. The circle represents precipitation, the triangle evaporation, and the square P - E; observations are thick solid lines and COM2 results are dotted.

Antarctic Moisture Budget

The accumulation of snowfall on the Antarctic continent is one of the major factors determining the response of the ice sheet to climatic change. The amount of annual snow accumulation simulated over Antarctica and along the coastline by CCM1 and CCM2 is shown in Figure 2. The simulation in CCM2 is close to tiie observations. In particular, a very arid climate over the continental interior (less than 5 cm a^-1) is modeled. Also, the accumulation minima (less than 15 cm a^-1) over the Ross and Filchner-Ronne Ice Shelves are to some extent captured by the model (Reference Tzeng. and BromwichTzeng and others, 1994). Although these two accumulation minima are also shown in the CGMI-R15 simulation, the cause of these features is different from that in the CCM2 T42. Reference Tzeng. and Parish.Tzeng and others (1993) indicated that these accumulation minima and the maximum Center around the South Pole in CCM1-R15 are mainly caused by the spectral truncation error in such a low- (R15) resolution model, T42 resolution of CCM1, however, does not seriously suffer from this type of error.

Fig. 2. The annual snowfall accumulation (P–E) in 100 mm a^-1 from (a) CCM1, (b) CCM2, and (c) the observations (after Reference BromwichBromwich, 1988). (a) Contour interval is 1.0; values less than 4.0 are stippled and greater than 6.0 are hatched, (b) C.l. is 0.5; values less than 2.0 are stippled. (c) C.l. is variable; 0.5 and 2.0 contours are bolded.

The moisture-budget analyses of CCM1 and CCM2 For the Antarctic are given in Table 2. The simulated precipitation by CCM1 is 3.4 times larger than the observations given by PO92. The large bias of [P — E] (about 3.5 times larger than the observed from Reference Giovinetto, Bromwich and Wendler.Giovinetto and others (1992)) is attributable to the precipitation. Although the poleward moisture transport across 70° S in CCM1 is much closer to the observations than those across 70° N, there is a huge residual in CCM1 (37.7 cm a^-1) which is apparently artificially "transported" into Antarctica from outside by the moisturefixer scheme due to the extremely dry and cold conditions over the Antarctic continent on a year-round basis. Therefore, it is clear that CCM1 is not appropriate for climate-change studies that rely on snowfall accumulation (Reference Tzeng. and Parish.Tzeng and others, 1993). By contrast, the moisturebudget analysis of CCM2 is significantly improved over Antarctica. Although the simulated precipitation ([P]) and evaporation ([E]) rates are slightly greater than PO92's climatological values, the simulated net precipitation [P — E] (15.3 cm a^-1) is slightly better than PO92’s result (14.7 cm a^-1) in comparison to the latest observational estimate 18.4 ± 3.7 cm a^-1 by Reference Giovinetto, Bromwich and Wendler.Giovinetto and others (1992). The CCM2 moisture-transport convergence poleward ot 70° S is very close to the observed value quoted by PO92 but significantly smaller than Reference YamazakiYamazaki’s (1992) diagnosis which matches Reference Giovinetto, Bromwich and Wendler.Giovinetto and others (1992) surface-based [P — E]. The residual of the CCM2 budget equation is quite small (3.3 cm a^-1) and approximates that of Reference Peixoto and Oort.Peixoto and Oort (1992). The improvement of the moisture budget in CCM2 is attributed to the use of the semi-Lagrangian transport scheme instead of CCM1’s positivc-moisturefixer scheme (Reference Tzeng. and BromwichTzeng and others, 1994). Definitely, the semi-Lagrangian scheme is very good at reproducing the moisture transport into the polar cap.

Table 2. Annual values of the Antarctic areally averaged (70° S to SP) precipitation ([P]), and evaporation ([E]) rates, and of the moisture-flux convergence poleward of 70° S ([FQ]) from the NCAR CCM1 and CCM2 and observations. Unit is cm a ^-1

Though the total snowfall ([P]) and its accumulation ([P - E]) over Antarctica are wrong in the CCM1, the annual cycle of the simulated precipitation rate over the continent is consistent with the observations (Fig. 3). Like observations from surface stations, the maximum of [P] appears during winter while the minimum of [P] occurs during summer in the simulation by CCM1. In addition, CCM1 even captures the semi-annual variation in precipitation rate as some observations show. Figure 4 displays the annual variations of precipitation, evaporation and net precipitation [P — E] for the region poleward, of 70° S in CCM2 along with the observational estimate of [P — E] by Reference YamazakiYamazaki (1992). The simulated stiow-accumulation rate ([P — E]) is a low 0.6 cm month^-1 in early summer and reaches maxima in fall (1.8 cm month^-1) and late winter (1.6 cm month^-1), which is in good agreement with the observations. It seems that this weak semi-annual oscillation is well represented in CCM2 for area-averaged [P — E] over Antarctica in terms of magnitude and phase apart from a variable 1 month phase shift (Reference Tzeng. and BromwichTzeng and others, 1994).

Fig. 3. The areally averaged annual variations of precipitation from CCM1. (a) The annual average is given at the edge oj each panel. Observed precipitation rate at stations (b) along the coast of East Antarctica and (c) in the interior of the continent (from Reference BromwichBromwich, 1988); the number in parentheses after each station is the annual average in cm a .

Fig. 4. The areally averaged (Antarctica) annual variations of precipitation, surface evaporation, and (P — E) from CCM2 and the observations [FQ] (from Reference YamazakiYamazaki. 1992). Unit is cm month^-1. The circle represents precipitation, the square evaporation, and the hatched square P — E; observations of [FQ] are shown by the thick solid line.

Conclusions

We have examined two generations of the National Center for Atmospheric Research’s community climate model (CCM1 and CCM2) for the simulated hydrologtc cycle over the polar regions. The moisture-budget analyses reveal that the different levels of bias in simulations are associated with the specific type of moisture scheme, horizontal resolution and other shortcomings in the general circulations of the models.

rhe moisture budgets over the Arctic and Antarctica are shown again side by side in Table 3 along with the observed analyses to compare clearly the features of the two poles. It is clear that CCM1 and CCM2 estimate evaporation rate quite well in both polar regions. There are large discrepancies in [P — E] in CCM1. however, which arc attributable to the huge errors in simulated precipitation [P]. CCM1 also significantly over-simulated the moisture-flux convergence across 70° N (S) so that the residual of the moisture budget is on the order of the net precipitation ([P — E]) over the Arctic and Antarctica. These errors are primarily due to the simulated meridional-wind component, because the model fails to simulate adequately the large-scale waves and distorts the respresentatidn of topography by the truncation error in such a low horizontal spectral resolution. Nevertheless, CCMl’s precipitable water over the North and South Polar caps approximates the observations. This indicates that the simulated moisture excess (R) in the Arctic ¡moisture deficiency in the Antarctic) has to be "transported" out of (into) the polar region artificially by the positive-moisture fixer of CCM1.

Table 3. Annual values of the areally averaged (70° -P) precipitation ([P]), and evaporation ([E]) rates, and of the moisture-flux convergence poleward of 70° ([FQ]), from the NCAR CCM1 and CCM2 and observations. Unit is cm a^-1. The values for Antarctica are given in parentheses next to the Arctic values

These opposite impacts of the positive-moisture fixer in the Arctic and Antarctica are most likely attributable to lire differences in land sea configuration and topography. Antarctica (poleward of 70° S) is essentially an ice-covered mountain surrounded by ocean with an average elevation of 2500 m and with very low moisture contents (zonal mean precipitable water at 80° S is only 1.5 kg m^-2 in the annual average; from Reference Peixoto, Oort., Streei Perron, Benin and Ratelille.Peixoto and Oort, 1983). Due to the large horizontal gradient of moisture over this area, negative values of moisture generated by the speetral-transform method inside Antarctica have to be continually corrected artificially by the positive-moisture fixer. Meanwhile, the bias in simulated moisture fluxes across 70° S is relatively small. On the other hand, the Arctic basin primarily consists of sea ice and sea surrounded by major continents with relatively high moisture contents (zonal mean precipitable water at 80° N is 4.9 kg m^-2 in the annual average, which is more than three times larger than that over Antarctica). Reference Rasch and Williamson.Rasch and Williamson (1990) showed that the maximum error due to the artificial positiye- moisture-fixer scheme is over high latitudes of the Eurasian continent during winter and that the impact of the moisture fixer in these regions is nearly as large as the largest term in the moisture-balance equation and larger than any other term. Furthermore, Reference Bromwich, Tzeng and Parish.Bromwich and others (1994) found that CCM1 over-simulates the storm activity over the continents and underestimates the storm activity over the oceans in the middle and high latitudes during the non-summer months, which is related to the positive-moisture fixer. The model inputs too much moisture into the continents in the lower troposphere, which in turn increases the latent-heat release over these regions and hence intensifies the cyclone, activity. In order to conserve the model’s total moisture content the specific humidity over the oceans (including the Arctic basin) has to be transported to the continents artificially by the local and global moisture correction.

In contrast to its earlier version, the moisture-budget analyses of" CCM2 are significantly improved over tire polar regions, especially for Antarctica, Although finer horizontal resolution can alleviate the error introduced by spectral truncation, the implementation of the semi-Lagrangian moisture transport is the more fundamental improvement for the much-improved simulation of the hydrologic cycle by CCM2. All components of the moisture-budget equation over Antarctica are very close to the observed from Reference Peixoto and Oort.Peixoto and Oort (1992). However, the precipitation rate simulated by CCM2 over the Arctic appears to be 21% higher than new observations (Reference Walsh, Zhou, Portis and Serreze.Walsh and others, 1994), which partly results from biases in large-scale flow pattern and storm track over the north Pacific Ocean that distorts the moisture transport into the Arctic basin. Furthermore, the annual cycle of the moisture budget over the .North Polar ice cap is generally shifted 1 or 2 months ahead of the observations.