• Andrews, D. G., J. R. Holton, and C. B. Leovy, 1987: Middle Atmosphere Dynamics. Academic Press, 489 pp.

  • Austin, J., and N. Butchart, 1994: The influence of climate change and the timing of stratospheric warmings on Arctic ozone depletion. J. Geophys. Res.,99, 1127–1145.

  • ——, ——, and R. Swinbank, 1997: Sensitivity of ozone and temperature to vertical resolution in a GCM with coupled stratospheric chemistry. Quart. J. Roy. Meteor. Soc.,123, 1405–1431.

  • Bailey, M. J., A. O’Neill, and V. D. Pope, 1993: Stratospheric analyses produced by the United Kingdom Meteorological Office. J. Appl. Meteor.,32, 1472–1483.

  • Beagley, S. R., J. de Grandpré, J. N. Koshyk, N. A. McFarlane, and T. G. Shepherd, 1997: Radiative-dynamical climatology of the first-generation Canadian Middle Atmosphere Model. Atmos.-Ocean,35, 293–331.

  • Boville, B. A., 1995: Middle atmosphere version of CCM2 (MACCM2): Annual cycle and interannual variability. J. Geophys. Res.,100, 9017–9039.

  • Butchart, N., J. Austin, C. S. Douglas, and R. Swinbank, 1997: Comparison of climatologies from the middle atmosphere versions of the UKMO’s GCM and data assimilation. Proc. First SPARC General Assembly, Melbourne, Australia, World Meteorological Organization, 24–27.

  • Clough, S. A., N. S. Grahame, and A. O’Neill, 1985: Potential vorticity in the stratosphere derived using data from satellites. Quart. J. Roy. Meteor. Soc.,111, 335–358.

  • Crutzen, P. J., and F. Arnold, 1986: Nitric acid cloud formation in the cold Antarctic stratosphere: A major cause of the springtime“ozone hole.” Nature,324, 651–655.

  • 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.

  • ——, ——, M. H. Mawson, J. A. James, S. C. Coulter, and A. Malcolm, 1998: Some issues in numerical methods for the next generation of NWP and climate models. Atmos.-Ocean,38, 425–444.

  • Edwards, J. M., and A. Slingo, 1996: Studies with a flexible new radiation code. Part I: Choosing a configuration for a large-scale model. Quart. J. Roy. Meteor. Soc.,122, 689–719.

  • Fritts, D. C., 1984: Gravity wave saturation in the middle atmosphere:A review of theory and observations. Rev. Geophys. Space Phys.,22, 275–308.

  • Hamilton, K., 1995: Interannual variability in the Northern Hemisphere winter middle atmosphere in control and perturbed experiments with the GFDL SKYHI general circulation model. J. Atmos. Sci.,52, 44–66.

  • ——, 1996: Comprehensive meteorological modelling of the middle atmosphere: A tutorial review. J. Atmos. Terr. Phys.,58, 1591–1627.

  • ——, R. J. Wilson, J. D. Mahlman, and L. J. Umscheid, 1995: Climatology of the SKYHI troposphere–stratosphere–mesosphere general circulation model. J. Atmos. Sci.,52, 5–43.

  • Hanson, D., and K. Mauersberger, 1988: Laboratory studies of the nitric acid trihydrate: Implications for the south polar stratosphere. Geophys. Res. Lett.,15, 855–858.

  • Haynes, P. H., C. J. Marks, M. E. McIntyre, T. G. Shepherd, and K. P. Shine, 1991: On the “downward control” of extratropical diabatic circulations by eddy-induced mean zonal forces. J. Atmos. Sci.,48, 651–678.

  • Hofmann, D. J., and S. Solomon, 1989: Ozone destruction through heterogeneous chemistry following the eruption of El Chichón. J. Geophys. Res.,94, 5029–5041.

  • Jenkins, D. B., D. P. Wareing, L. Thomas, and G. Vaughan, 1987: Upper stratospheric and mesospheric temperatures derived from lidar observations at Aberystwyth. J. Atmos. Terr. Phys.,49, 287–298.

  • Kanzawa, H., and S. Kawaguchi, 1990: Large stratospheric sudden warming in Antarctic late winter and shallow ozone hole in 1988. Geophys. Res. Lett.,17, 77–80.

  • Manney, G. L., R. Swinbank, S. T. Massie, M. E. Gelman, A. J. Miller, R. Nagatani, A. O’Neill, and R. W. Zurek, 1996: Comparison of U. K. Meteorological Office and U. S. National Meteorological Center stratospheric analyses during northern and southern winter. J. Geophys. Res.,101, 10 311–10 334.

  • Manzini, E., and L. Bengtsson, 1996: Stratospheric climate and variability from a general circulation model and observations. Climate Dyn.,12, 615–639.

  • McIntyre, M. E., and T. N. Palmer, 1983: Breaking planetary waves in the stratosphere. Nature,305, 593–600.

  • Milton, S. F., and C. A. Wilson, 1996: The impact of parameterized subgrid-scale orographic forcing on systematic errors in a global NWP model. Mon. Wea. Rev.,124, 2023–2045.

  • Mitchell, J. F. B., T. C. Johns, J. M. Gregory, and S. F. B. Tett, 1995:Climate response to increasing levels of greenhouse gases and sulphate aerosols. Nature,376, 501–504.

  • Morcrette, J.-J., L. Smith, and Y. Fouquart, 1986: Pressure and temperature dependence of the absorption in longwave radiation parameterizations. Contrib. Atmos. Phys.,59, 455–469.

  • Nash, J., and J. L. Brownscombe, 1983: Validation of the stratospheric sounding unit. Adv. Space Res.,2, 59–62.

  • Palmer, T. N., G. J. Shutts, and R. Swinbank, 1986: Alleviation of a systematic westerly bias in general circulation and numerical weather prediction models through an orographic gravity wave drag parametrization. Quart. J. Roy. Meteor. Soc.,112, 1001–1039.

  • Randel, W. J., 1987: The evaluation of winds from geopotential height data in the stratosphere. J. Atmos. Sci.,44, 3097–3120.

  • Reber, C. A., 1993: The Upper Atmosphere Research Satellite (UARS). Geophys. Res. Lett.,20, 1215–1218.

  • Rind, D., R. Suozzo, and N. K. Balachandran, 1988a: The GISS global climate-middle atmosphere model. Part II: Model variability due to interactions between planetary waves, the mean circulation and gravity wave drag. J. Atmos. Sci.,45, 371–386.

  • ——, ——, ——, A. Lacis, and G. Russell, 1988b: The GISS global climate-middle atmosphere model. Part I: Model structure and climatology. J. Atmos. Sci.,45, 329–370.

  • Rood, R. B., and M. A. Geller, 1994: Editorial. J. Atmos. Sci.,51, 2781–2782.

  • Rosenlof, K. H., 1995: Seasonal cycle of the residual mean meridional circulation in the stratosphere. J. Geophys. Res.,100, 5173–5191.

  • Sassi, F., R. R. Garcia, and B. A. Boville, 1993: The stratopause semiannual oscillation in the NCAR Community Climate Model. J. Atmos. Sci.,50, 3608–3624.

  • Shepherd, T. G., K. Semeniuk, and J. N. Koshyk, 1996: Sponge layer feedbacks in middle-atmosphere models. J. Geophys. Res.,101, 23 447–23 464.

  • Shibata, K., and M. Chiba, 1990: A simulation of seasonal variation of the stratospheric circulation with a general circulation model. J. Meteor. Soc. Japan,68, 687–703.

  • Swinbank, R., and A. O’Neill, 1994a: A stratosphere-troposphere data assimilation system. Mon. Wea. Rev.,122, 686–702.

  • ——, and ——, 1994b: Quasi-biennial and semi-annual oscillations in equatorial wind fields constructed by data assimilation. Geophys. Res. Lett.,21, 2099–2102.

  • ——, W. A. Lahoz, A. O’Neill, C. S. Douglas, A. Heaps, and D. Podd, 1998: Middle atmosphere variability in the Met. Office unified model. Quart. J. Roy. Meteor. Soc., in press.

  • Turco, R. P., O. B. Toon, and P. Hamil, 1989: Hetrogeneous physicochemistry of the polar ozone hole. J. Geophys. Res.,94, 16 493–16 510.

  • Warner, C. D., and M. E. McIntyre, 1996: On the propagation and dissipation of gravity wave spectra through a realistic middle atmosphere. J. Atmos. Sci.,53, 3213–3235.

  • White, A. A., and R. A. Bromley, 1995: Dynamically consistent, quasi-hydrostatic equations for global models with a complete representation of the Coriolis force. Quart. J. Roy. Meteor. Soc.,121, 399–418.

  • World Meteorological Organization, 1995: Scientific assessment of ozone depletion: 1994. Rep. 37, WMO Global Ozone Research and Monitoring Project, Geneva.

  • View in gallery

    Vertical profile of Rayleigh friction coefficient. Solid and dashed curves are for the two UM simulations. The UM upper boundary is at 0.1 mb but the solid curve also indicates the drag used in the SMM and, therefore, to show the drag used in that model, is extended higher. The dotted curve is the friction coefficient used by Boville (1995) in the MACCM2.

  • View in gallery

    January monthly and zonally averaged westerly winds. (a) and (b) The 10-yr means from the UM simulations with Rayleigh friction coefficients given by the solid and dashed curves, respectively, in Fig. 1. (c) The 5-yr mean from the UKMO’s stratospheric data assimilation for 1992–96. (d) The 18-yr mean geostrophic winds calculated from the UKMO’s stratospheric analyses for 1979–96. The contour interval is 10 m s−1 and shading denotes easterlies.

  • View in gallery

    January monthly and zonally averaged temperatures and temperature differences. (a) The 10-yr mean from the UM simulation with Rayleigh friction coefficient given by the solid curve in Fig. 1. (b) The 5-yr mean from the UKMO’s stratospheric data assimilation for 1992–96. (c) Simulated temperatures minus assimilated temperatures. (d) Simulated temperatures minus the simulated temperatures from the integration with increased Rayleigh friction (see dashed curve in Fig. 1). (a) and (b) The contour interval is 5 K. Dotted shading indicates temperatures above 270 K, and the hatched shading indicates temperatures below 190 K. (c) and (d) The contour interval is 2 K. Dotted shading indicates a cold bias of more than 16 K, and the hatched shading indicates warm biases.

  • View in gallery

    As in Fig. 2 but the standard deviation of the individual January means from the corresponding multiannual mean [10, 10, 5, and 18 yr, respectively for (a), (b), (c), and (d)]. The contour interval is 2 m s−1 and light shading indicates a standard deviation less than 4, and dark shading a standard deviation less than 2.

  • View in gallery

    Standard deviations of the individual January, monthly, and zonally averaged, temperatures from the multiyear average as (a) a function of height at the North Pole and (b), (c), and (d) a function of latitude at 100, 10, and 1 mb, respectively. Solid and dashed curves:10-yr model simulations using the friction coefficients given by the solid and dashed curves in Fig. 1, respectively. Dot–dashed curve: UKMO SSU analyses for the years 1979–96. Dotted curve: UKMO stratospheric data assimilation for the years 1992–96.

  • View in gallery

    Average amplitudes of the January stationary waves. (a) and (b) Wavenumbers 1 and 2, respectively, from the 10-yr model simulation using the Rayleigh friction coefficient given by the solid curve in Fig. 1. (c) and (d) Wavenumbers 1 and 2 calculated from assimilated observations for the years 1992–96. The contour intervals above and below 100 mb are 100 and 20 m, respectively.

  • View in gallery

    January mean residual streamfunction Ψ calculated from υ* for (a) the UKMO stratospheric data assimilation for 1992–96, and (b) the 10-yr model simulation with Rayleigh friction coefficient given by the solid curve in Fig. 1. Positive contours represent a circulation in the counterclockwise sense. Negative contours are indicated by the dashed lines. Contour levels are ±2, ±5, ±10, ±20, ±50, ±100, ±200, ±500, ±1000, and ±2000 kg m−1 s−1. The results with the Rayleigh friction given by the dashed curve in Fig. 1 were found to be similar and therefore not shown.

  • View in gallery

    As for Fig. 2 except that July zonal winds are shown.

  • View in gallery

    As for Fig. 3 except that July temperatures and temperature differences are shown.

  • View in gallery

    As for Fig. 4 except that standard deviations of the July zonal winds are shown.

  • View in gallery

    As for Fig. 5 except that standard deviations of the July, monthly and zonally averaged, temperatures are shown, but note the different scale for some of the axes.

  • View in gallery

    As for Fig. 6 except that the average amplitudes of the July stationary waves are shown.

  • View in gallery

    Amplitudes of stationary wavenumber 1 for July at (a) 100 mb and (b) 1.468 mb. The solid curve is the average amplitude from 10 years of model results (using the friction profile given by the solid curve in Fig. 1); shading denotes the range of amplitudes. The corresponding information for the 18 years of SSU observations is given by the three dashed curves, with the thickest line denoting the average.

  • View in gallery

    As for Fig. 7 except that the July mean residual streamfunctions Ψ are shown.

  • View in gallery

    >Fig. 15. Evolution of the zonal-mean zonal wind at 10 mb for the model simulation with the smaller Rayleigh friction coefficient (see solid curve in Fig. 1). Results shown are from 5-day running means. The contour interval is 10 m s−1, and regions of easterlies are shaded.

  • View in gallery

    As for Fig. 15 except that the winds from the UKMO stratospheric data assimilation for the period 17 October 1991 to 31 December 1996 are shown.

  • View in gallery

    Zonal-mean zonal wind (m s−1) at the equator and at 1 mb. The solid and dotted curves are for the last 5 years of the two UM simulations using the friction profiles given by the solid and dashed curves in Fig. 1, respectively. The dashed curve is the assimilated wind for the period 1 January 1992 to 31 December 1996. A 5-day running mean has been applied to all the data.

  • View in gallery

    Time series of daily North Pole temperatures (K) at 10 mb for November–April for (a) 10 consecutive winters of the UM simulation with the smaller Rayleigh friction coefficient (see solid curve in Fig. 1) and (b) 10 consecutive years of SSU observations from 1985 to 1995. The dashed curve is the 10-yr average on that calendar day and shading denotes the anomalously warm periods.

  • View in gallery

    As for Fig. 18 but daily South Pole temperatures (K) at 10 mb for April–September.

  • View in gallery

    Minimum temperatures at 46.42 mb poleward of 43.75° for (a) the northern and (b) southern winters. The solid curves indicate the minimum temperatures in each of the 10 years of the simulation using the smaller Rayleigh friction coefficient (see solid curve in Fig. 1). The observed range from November 1991 to October 1996 is given by the yellow shading. The red and green dashed lines are thermodynamic equilibrium temperatures for the formation of NAT and ice PSCs, respectively.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 22 22 4
PDF Downloads 9 9 0

Middle Atmosphere Climatologies from the Troposphere–Stratosphere Configuration of the UKMO’s Unified Model

View More View Less
  • 1 Stratospheric Chemistry and Dynamics Group, U.K. Meteorological Office, Bracknell, Berkshire, United Kingdom
© Get Permissions
Full access

Abstract

A climatology of the middle atmosphere is determined from 11-yr integrations of the U.K. Meteorological Office Unified Model and compared with 18 years of satellite observations and 5 years of data assimilation fields. The model has an upper boundary at 0.1 mb, and above 20 mb uses Rayleigh friction as a substitute for gravity wave drag. Many of the results are, however, found to be relatively insensitive to enhancing the damping above 0.3 mb. As with most general circulation models, the polar night jet in both hemispheres is too strong and does not have the observed equatorward slope with height. The model suffers from the common “cold pole” problem and, apart from a local warm pool centered just below 100 mb in northern high latitudes in January, and another at about 30 mb at 70°S in July, has a cold bias throughout the stratosphere. At the level where polar stratospheric clouds occur, the temperature bias is about −4 K in the Northern Hemisphere and up to +6 K in the Southern Hemisphere. For the majority of the southern winters, local minimum temperatures in the lower stratosphere agree well with observations but in some years the behavior is more like the Northern Hemisphere with values rising rapidly in late winter. This feature of the simulation is also seen in the South Pole temperatures at 10 mb with midwinter warmings occurring in two of the years. At 10 mb, midwinter warming behavior at the North Pole is quite well reproduced, as is the annual cycle in extratropical circulation. In the Tropics, there is no quasi-biennial oscillation, and the semiannual oscillation in the upper stratosphere has a poorly simulated westerly phase, while the easterly phase lacks the observed seasonal asymmetry. Simulated stationary wave amplitudes in the upper stratosphere lack a strong hemispheric asymmetry and are overpredicted in both hemispheres despite having roughly the correct amplitudes at 100 mb. Interannual variability in the winter stratosphere is underestimated, and again there is evidence that the model does not produce the proper hemispheric asymmetries.

Corresponding author address: Neal Butchart, Stratospheric Chemistry and Dynamics Group, Meteorological Office, London Rd., Bracknell, Berkshire, RG12 2SZ, U.K.

Email: nbutchart@meto.gov.uk

Abstract

A climatology of the middle atmosphere is determined from 11-yr integrations of the U.K. Meteorological Office Unified Model and compared with 18 years of satellite observations and 5 years of data assimilation fields. The model has an upper boundary at 0.1 mb, and above 20 mb uses Rayleigh friction as a substitute for gravity wave drag. Many of the results are, however, found to be relatively insensitive to enhancing the damping above 0.3 mb. As with most general circulation models, the polar night jet in both hemispheres is too strong and does not have the observed equatorward slope with height. The model suffers from the common “cold pole” problem and, apart from a local warm pool centered just below 100 mb in northern high latitudes in January, and another at about 30 mb at 70°S in July, has a cold bias throughout the stratosphere. At the level where polar stratospheric clouds occur, the temperature bias is about −4 K in the Northern Hemisphere and up to +6 K in the Southern Hemisphere. For the majority of the southern winters, local minimum temperatures in the lower stratosphere agree well with observations but in some years the behavior is more like the Northern Hemisphere with values rising rapidly in late winter. This feature of the simulation is also seen in the South Pole temperatures at 10 mb with midwinter warmings occurring in two of the years. At 10 mb, midwinter warming behavior at the North Pole is quite well reproduced, as is the annual cycle in extratropical circulation. In the Tropics, there is no quasi-biennial oscillation, and the semiannual oscillation in the upper stratosphere has a poorly simulated westerly phase, while the easterly phase lacks the observed seasonal asymmetry. Simulated stationary wave amplitudes in the upper stratosphere lack a strong hemispheric asymmetry and are overpredicted in both hemispheres despite having roughly the correct amplitudes at 100 mb. Interannual variability in the winter stratosphere is underestimated, and again there is evidence that the model does not produce the proper hemispheric asymmetries.

Corresponding author address: Neal Butchart, Stratospheric Chemistry and Dynamics Group, Meteorological Office, London Rd., Bracknell, Berkshire, RG12 2SZ, U.K.

Email: nbutchart@meto.gov.uk

1. Introduction

The U.K. Meteorological Office (UKMO) “Unified Model” (UM) (Cullen 1993) is used in a wide range of operational and research applications extending from mesoscale forecasting to climate prediction, and also data assimilation. These many requirements are satisfied by using well tested standard versions, or “configurations,” of a common basic numerical model. Improvements incorporated in new “builds” of this common model can therefore benefit all the applications.

Here we report the first results from multiyear integrations of a troposphere–stratosphere configuration of the UM and document, in particular, its simulation of the middle atmosphere. The model configuration is similar to the global atmosphere version used in climate studies by the Hadley Centre (e.g., Mitchell et al. 1995) but with an extended vertical domain and increased vertical resolution near and above the tropopause. It has been developed for studying stratospheric processes and their role in climate, and joins an increasing number of general circulation models (GCMs), which include proper representations of the middle atmosphere [e.g., see Table 1 of Hamilton (1996)]. As yet little has been published on the performance of such models during long integrations. The most notable exceptions are the Geophysical Fluid Dynamics Laboratory SKYHI model (Hamilton et al. 1995; Hamilton 1995), the Goddard Institute for Space Science (GISS) model (Rind et al. 1988a,b), and the middle atmosphere version of the National Center for Atmospheric Research (NCAR) Community Climate Model, version 2 (MACCM2) (Boville 1995).

A particular objective of the current work is to establish the strengths and weaknesses of the troposphere–stratosphere configuration of the UM for future chemistry and climate studies. Using a slightly earlier build of the model, Austin et al. (1997) have already shown that the extended vertical domain and increased vertical resolution we use here are important for the inclusion of stratospheric chemistry in the UM, at least on a seasonal timescale. In this paper we concentrate on the model’s climatology of the wintertime extratropical stratospheric circulation and its interannual variability. We also consider the accuracy of the simulated annual cycle and, from a synoptic point of view, examine the model’s ability to reproduce the correct local conditions suitable for the formation of polar stratospheric clouds (PSCs).

Since September 1991 a troposphere–stratosphere configuration of the UM (with slightly fewer vertical levels) has also been used for assimilating observations to provide correlative data for the Upper Atmosphere Research Satellite (UARS) project (Swinbank and O’Neill 1994a). However, despite widespread use, only a few aspects of the quality of this new dataset have been investigated (Manney et al. 1996). Nevertheless, we compare a preliminary 5-yr climatology with the GCM results but calculate additional model validation diagnostics from analyses produced since December 1978 in the UKMO, using data from the stratospheric sounding units (SSUs) (Bailey et al. 1993). A shortcoming of data assimilation is that in regions of limited observations, or observation type (such as in the stratosphere where there is essentially only temperature and very few wind measurements), certain features can be dominated by model errors. Consequently, the quality of the UKMO’s UARS correlative dataset is likely to depend, in turn, on the performance of the troposphere–stratosphere configuration of the UM.

2. Model formulation and integrations

The UM is a gridpoint model with a hybrid sigma-pressure vertical coordinate based on an extended version of the traditional hydrostatic primitive equations (White and Bromley 1995). In “climate configuration” the horizontal resolution is 3.75° long × 2.5° lat and there are 19 vertical levels extending from the surface to 4.6 mb. For the troposphere–stratosphere configuration used here, the number of vertical levels was increased to 49 by improving the vertical resolution to approximately 1.3 km throughout the stratosphere and raising the upper boundary to 0.1 mb. More details on the location of the vertical levels in these two configurations are given in Austin et al. (1997; see, in particular, their Fig. 1).

The model is formulated on a staggered Arakawa B-grid with split-explicit time integration using a forward-backward approach for the adjustment step and a Heun scheme for the advection step (Cullen and Davies 1991). Fourier filtering in polar latitudes eliminates undesirable restrictions on the choice of time step. Thirty minutes is used for the climate configuration but, because of the much stronger winds at higher levels, this was reduced to 15 min for integrations of the 49-level model reported here. On three occasions (August of years 5 and 7, and January of year 10), however, the time step was further halved for a period of between 20 and 30 days in order to prevent the integration failing due to numerical instability at upper levels within the polar night jet. A second integration with increased friction (see below) did not suffer from this problem. Horizontal diffusion on model levels of the form [(K)]n represents dissipation by unresolved eddies and removes unwanted grid-scale noise. For the integrations reported here, n = 3 and K = 0.547 × 109 s−1/3 (1 and 0.4 × 107 s−1, respectively, at the top level) for the horizontal wind and temperature fields.

Most of the physical parameterizations employed in the UM have been extensively described in separate papers [see Cullen (1993) or Mitchell et al. (1995) for appropriate references]. Apart from the differences noted below, the physical parameterizations included in the model described here are from a build of the climate configuration referred to as the “third climate version” or “3CV.” Some changes had to be made, however, because of the extended vertical domain of the troposphere–stratosphere configuration. Notably, following Morcrette et al. (1986), the treatment of gaseous effects in the longwave radiation scheme was modified to give smoother, more accurate heating rates in the stratosphere. Doppler line broadening and the effects of the atmosphere above the top of the model were also taken into account.

Other changes from 3CV involved parameterizing the momentum deposition from breaking internal gravity waves. This is known to play a crucial role in the middle atmosphere circulation but is subject to great uncertainties, especially the contribution from nonorographic waves (e.g., Fritts 1984). With the extended vertical domain used here, the parameterization of the effects of subgrid-scale orographic gravity waves used in 3CV [see section 5 of Milton and Wilson (1996) for a description of that scheme], and also an earlier scheme based on Palmer et al. (1986), introduced too much noise into the stratosphere (Swinbank 1996, personal communication). Therefore, we decided to use the less sophisticated, though better tested, Palmer et al. based scheme up to the 20-mb level with a simple Rayleigh friction above that. The friction coefficient used was the same as that used successfully in the UKMO’s Stratosphere–Mesosphere Model (SMM) when a proper treatment of radiative processes is included (e.g., Austin and Butchart 1994). Between 20 and 0.79 mb the coefficient was 10−7 s−1 and then increased with height as 10−7 + 15 × 10−6 [1 − 3.49(p/1000)7/40] s−1, where p is the pressure in mb (see Fig. 1, solid curve).

To test the sensitivity to the choice of upper-level friction, and the effects this might have on the interannual variability, the integration was repeated with higher values of the coefficient above about 0.3 mb. The coefficient was again 10−7 s−1 between 20 and 5.84 mb but now increased with height as (2/3){[36 + 7 ln(p/1000)]/28}4 day−1 (see Fig. 1, dashed curve). The assumption was that increasing the friction would crudely compensate for the relatively low upper boundary of the UM, in comparison with, say, the SMM or NCAR’s MACCM2 (Boville 1995), by representing some of the effects of the gravity wave momentum deposition that would normally occur above the top of the model (0.1 mb) and contribute, by the downward control principle, to determining the seasonal mean meridional circulation lower down (Haynes et al. 1991). The purpose then was to examine what effect, if any, this had on the accuracy of the simulation of the lower and middle stratosphere. For comparison the friction coefficient used by Boville (1995) in the MACCM2 is also shown in Fig. 1 (dotted curve), but note that Boville also included orographic gravity wave drag in the upper stratosphere and mesosphere. The effects of subgrid-scale gravity waves with nonzero phase speeds where not included in the MACCM2 nor in the integrations reported here.

Both the UM integrations were initialized on 15 January 1992 with observations extracted from the UKMO’s stratospheric data assimilation archive (Swinbank and O’Neill 1994a). Climatological sea surface temperatures were specified at the model lower boundary and the solar heating was calculated using a zonally symmetric ozone climatology. Both the sea surface temperatures and ozone values were updated every 5 days by interpolating from monthly means. The model was integrated for almost 11 yr and the results shown are from the 10-yr periods starting from the model date 1 January 1993.

3. Observational datasets

a. UKMO stratospheric data assimilation

Since September 1991 the UKMO has produced daily analyses from a stratospheric data assimilation system as a contribution to the UARS project (Reber 1993). Observations from a variety of sources are assimilated into a global version of the UM using the same techniques as used in operational numerical weather forecasting (Swinbank and O’Neill 1994a). The UM configuration used has 42 levels between the ground and 0.3 mb and horizontal grid intervals of 2.5° lat and 3.75° long. Apart from experimental periods, no actual measurements from the instruments onboard UARS are used. The system therefore provides “correlative” data for validating the UARS measurements and the analyses have been extensively used in many UARS-related studies (Rood and Geller 1994). At a reduced vertical resolution, the data have been made available at both the British Atmosphere Data Centre (BADC) at the Rutherford Appleton Laboratory (RAL), United Kingdom, and the Distributed Active Archive Center at Goddard Space Flight Center. Also since October 1995 they have been produced operationally in the UKMO.

In the stratosphere the assimilation relies almost entirely on temperature soundings from operational satellites augmented by a few radiosonde ascents, and hence the quality of the other meteorological fields is likely to be dependent upon the accuracy of the assimilation model. This dependence has not yet been fully investigated but the data have been validated against National Centers for Environmental Prediction (NCEP, formerly the National Meteorological Center) analyses (Swinbank and O’Neill 1994a; Manney et al. 1996). In particular, the latter study noted that the temperature gradients at the edge of the polar vortex are stronger in the UKMO analyses than in the NCEP analyses, but that the UKMO analyses are probably more realistic in this respect. Also during rapid changes, such as stratospheric warmings, the UKMO analyses are probably more realistic in representing the temporal behavior because each datum is taken at the correct observation time, rather than binned in the 12-h periods that NCEP use. On the other hand, the UKMO assimilations are often inferior in times of cross-polar flow, and at high levels unrealistic small-scale features sometimes occur. Of particular relevance to heterogeneous chemistry in the lower stratosphere, Manney et al. (1996) noted that neither analysis system captures the very lowest temperatures observed, except during early winter for the UKMO analyses in the Southern Hemisphere. The UKMO data assimilation also provides error statistics [see Swinbank and O’Neill (1994a) for details] from which biases in the global average temperature were found to be less than 0.2 K with standard deviations varying between 1 K in the lower stratosphere to 2 K in the lower mesosphere.

In this work we have used the 5 years of analyses now available from the UKMO data assimilation to construct a climatology of the more basic quantities for comparison with the UM simulations, but supplement these with fields derived from 18 years of UKMO’s SSU analyses (see below). Note however that comparison between the SSU data for the 5- and 18-yr periods indicated that the data assimilation period sometimes gave a false picture of the full 18-yr climatology and interannual variability.

b. UKMO SSU analyses

Between 1978 and 1997, radiance measurements were obtained from SSUs, microwave sounding units, and high resolution infrared sounders, onboard the National Oceanic and Atmospheric Administration series of operational satellites. From these measurements the UKMO derived atmospheric thicknesses using multilinear regression analysis with rocket profiles (Bailey et al. 1993). The observed 100-mb height field was added to the basic thicknesses to give a set of height fields at 20, 10, 5, 2, and 1 mb on a 5° lat × 5° long grid. In the data used here, these levels were supplemented in the troposphere and lower stratosphere by analyses produced for operational weather forecasting [see Bailey et al. (1993) for details]. The data have been deposited at the BADC at RAL. Temperatures and winds were derived from the geopotential heights by assuming hydrostatic and geostrophic balance.

The satellite instruments scan to the nadir, and also scan at certain other fixed angles to improve horizontal resolution, but the vertical resolution of some 8 km is rather poorer than may be obtained with limb scanning instruments. All the observations obtained in a single 24-h period are incorporated into a single daily analysis at 1200 UT. This poor temporal and vertical resolution can make analyses during baroclinic and/or rapidly changing events such as stratospheric warmings, subject to large error. The data have been validated by Nash and Brownscombe (1983), who concluded that the SSU radiance measurements had a mean error of less than 0.2 K in brightness temperature, while Jenkins et al. (1987) demonstrated agreement to within 1 K between SSU data and the radiance synthesized from a lidar instrument.

Despite the inherent weaknesses in the SSU analyses, it is now generally accepted that it is possible to derive various meteorological quantities from the basic data. For example, Clough et al. (1985) convincingly demonstrated that the potential vorticity fields that had been calculated from SSU data were qualitatively accurate. This had previously been tacitly assumed by McIntyre and Palmer (1983) when they used the global coverage afforded by the then newly available SSU data to present the first observational evidence of planetary wave breaking in the stratosphere. Use of these data has also featured prominently in other key investigations of the behavior of the stratosphere. SSU instruments have now provided almost continuous global observations of the stratosphere since 1978 and we use the corresponding UKMO analyses to establish a global climatology of the stratosphere to compare with results obtained from the 5 years of assimilated data. We also use the long-term record of the SSU analyses for validating the interannual variability of the UM simulations.

4. January climatology

a. Zonal mean winds and temperatures

Ten-year means of the zonally averaged zonal wind for January for the two simulations with different friction are shown in Fig. 2. Also shown are 5- and 18-yr means obtained from the UKMO’s data assimilation and SSU analyses, respectively. In general, there was little difference between the simulations, and the model reproduced all the main climatological features quite well. The strengths and positions of the subtropical jets were well simulated, as was the observed separation of the Northern Hemisphere stratospheric and tropospheric westerly jets. However, at higher levels the westerlies in the model were somewhat stronger than those assimilated (∼49 m s−1 compared to ∼35 m s−1 for the jet maximum at about 2 mb) and agreed better with the SSU analyses (∼48 m s−1, again for the jet maximum at roughly 2 mb). On the other hand, the geostrophic winds in Fig. 2d probably substantially overestimate the strength of the polar night jet (e.g., Randel 1987) and SSU analyses (not shown) confirmed that for the years 1992–96 (Fig. 2c) the jet was, on average, weaker than normal, thereby, implying that the simulated polar night jet was, indeed, too strong in the upper stratosphere. Moreover, there was little sensitivity to the amount of friction used. The maximum westerlies in the upper stratosphere and mesosphere also occurred too close to the pole, as is typical of many other GCM simulations (e.g., Boville 1995; Shibata and Chiba 1990). Consequently, the UM failed to reproduce the observed equatorward tilt with height of the stratospheric westerly jet core. In contrast, an equatorward tilt can be found in the MACCM2 (Boville 1995), although this is not at all clear in his figures.

In the summer hemisphere the upper-level easterlies in both simulations were generally weaker than observed apart from the localized maximum near 1 mb. This feature was accompanied by a protrusion of the easterlies into the Northern Hemisphere but was less prominent with the increased friction, possibly because the relative increase in friction was quite significant at 1 mb (see Fig. 1). Similar features are also seen in the December–February GCM results of Rind et al. (1988b) and the January MACCM2 results presented in Fig. 2 of Boville (1995), who noted an association with the semi-annual oscillation (SAO) (see also Sassi et al. 1993). A link with the SAO in the UM is less certain for, despite the success of the assimilation version of the UM in capturing the SAO (Swinbank and O’Neill 1994b), a significant protrusion of easterlies into the winter hemisphere is not apparent in Fig. 2c, possibly because the assimilation extends only up to 0.3 mb compared to 0.1 mb for the GCM.

Consistent with the model’s overprediction of the strength of the polar night jet, stratospheric temperatures were too low at the North Pole (see Fig. 3). Here, there was little sensitivity to the choice of friction coefficient (∼1 K, see Fig. 3d) or the different observational analyses (again ∼1 K, results not shown) but the model’s cold bias peaked at more than 26 K in the upper stratosphere (see Fig. 3c). Lower down the cold bias diminished so that by 100 mb the model temperatures were 2–3 K too high at the pole. In contrast the model’s tropical tropopause was too cold with temperatures below 190 K. Throughout the Southern Hemisphere stratosphere the model was again too cold with a bias generally in excess of 12 K in high latitudes. Increasing the friction had very little effect on the model’s temperature biases and, if anything, tended to exacerbate the deficiencies in the extratropical stratosphere in both hemispheres.

Sensitivity to the upper-level friction was, however, apparent in the interannual variability of the January mean polar night jet (Fig. 4). Greater friction reduced variability above 1 mb but compensated with an increase throughout the stratosphere (cf. Figs. 4a and 4b). Nonetheless, in the lower and middle stratosphere it remained below that derived from the 18 years of SSU observations (Fig. 4d). Also the peak in the variability in Figs. 4a and 4b is too close to the pole, though there is a secondary peak near 30°N, which was strengthened by increasing the friction. The success of the data assimilation configuration of the UM in capturing the quasi-biennial oscillation (QBO) in the equatorial lower stratosphere (Swinbank and O’Neill 1994b) is evident in Fig. 4c but in common with most other GCM simulations there is no sign of this component of the low-frequency variability in either Figs. 4a or 4b.

Interannual variability of the January zonal mean stratospheric temperatures was mainly observed in the high-latitude Northern Hemisphere with the maximum at the pole [see dot–dashed and also dotted curves in Fig. 5—note, the significantly smaller variability in the assimilated temperatures (dotted curve) at 1 mb is the result of the availability of only 5 years of data]. With the smaller friction coefficient the model results agreed quite well with the observations (cf. solid and dot-dashed curves in Fig. 5), though, as with the positioning and variability of the polar night jet, the simulated temperature variability in the upper stratosphere was concentrated too close to the pole (see Figs. 5a and 5d). Increasing the friction coefficient significantly reduced the variability above 10 mb and, again, there was a compensating increase lower down.

b. Stationary waves

The left-hand panels of Fig. 6 show the average amplitudes, for the 10 years, of the first two zonal harmonics of the January mean geopotential height fields from the simulation with the smaller Rayleigh friction coefficient (see solid curve in Fig. 1). Results (not shown) with increased friction (see dashed curve in Fig. 1) were very similar. The main difference was a reduction in the average amplitude of wavenumber 1 in the upper stratosphere and mesosphere from a maximum of 1304 m in Fig. 6a to 1063 m, though this occurred at roughly the same altitude. In the troposphere and lower stratosphere there was good agreement between the model and the corresponding results from 5 years of assimilated observations (right-hand panels), but noticeable differences occurred above about 10 mb. In particular, the observed wavenumber 1 amplitude reached a maximum at about 40 km,1 whereas the amplitude of the simulated wave continued to increase with height into the mesosphere (cf. Figs. 6a and 6c). Possibly, this was because of insufficient wave breaking in the stratosphere and incorrect equatorward refraction, which may have resulted from deficiencies in the simulated polar night jet (see section 4a). Inadequate planetary wave dissipation in the model’s upper levels is also likely to have contributed to the model error. Thus, increasing the friction slightly alleviated the problem. A more effective way to alleviate the problem could be to use a lower order diffusion in more than just the top layer but this can lead to problems lower down where the diffusion changes from ∇6 to ∇2 (Davies 1997, personal communication). The failure to reproduce the observed equatorward tilt with height for the maximum amplitude of wavenumber 2 (cf. Figs. 6b and 6d) was, again, possibly caused by deficiencies in the simulated polar night jet.

Results for the individual model years (not shown here) indicate that at 100 mb the UM was unable to reproduce the range of January stationary wave amplitudes found in 18 years of SSU analyses for either wavenumbers 1 or 2. For example, in the 10 simulated years the maximum amplitudes at 100 mb were 248 and 205 m, for wavenumbers 1 and 2, respectively, compared to 319 and 325 m for the SSU analyses. Interestingly, increasing the upper-level friction reduced the simulated January stationary waves at 100 mb so that the maximum amplitudes over the 10 years were changed to 221 and 159 m for wavenumbers 1 and 2, respectively, with a roughly comparable reduction in the average amplitudes.

c. Meridional circulation

The climatology of the stratosphere is linked to the diabatic heating through the mean meridional circulation, which, when expressed in terms of the transformed Eulerian-mean (TEM) residual velocities (υ*, w*), also provides a useful proxy for the Lagrangian-mean transport (e.g., Andrews et al. 1987). In the notation of Andrews et al. [their Eq. (3.5.1)],
i1520-0469-55-17-2782-eq1
and, by the continuity equation, the mass streamfunction Ψ is defined as:
i1520-0469-55-17-2782-eq2
Reliable direct measurements of υ* and w* are not generally available but TEM residual circulations can be estimated (e.g., Rosenlof 1995) from, either a diabatic heating calculation, or using the downward-control control principle (Haynes et al. 1991). Using output from the Canadian Middle Atmosphere Model (CMAM), Beagley et al. (1997) found the first method largely unsuccessful, whereas the downward control estimate worked quite well outside the Tropics. Here we use the direct method of Beagley et al. to obtain an observational estimate of the January TEM residual circulation by calculating υ* directly from the output of the UKMO data assimilation model, then obtaining Ψ by integrating downward from the model lid, assuming Ψ vanishes there. The same method was used to obtain Ψ from the UM simulation with the smaller Rayleigh friction coefficient (see solid curve in Fig. 1) and the results are compared in Fig. 7.

In general the model results agreed well with the observed estimate, with both displaying the distinct Brewer–Dobson circulation in the stratosphere (cf. Figs. 7a and 7b). A single-cell summer-to-winter-pole circulation existed in the mesosphere in the model and observations, and although the observed circulation appears slightly weaker, this was probably just a consequence of the different levels at which Ψ was assumed to vanish (cf. 0.3 and 0.1 mb). The main difference is in the high-latitude Northern Hemisphere where throughout the stratosphere the estimated observed streamfunction shows first sinking, then rising motion, ongoing from 60° to 70°N. It is not clear what causes this, or even whether it is a real feature. At the higher levels it is possibly just contamination from the somewhat noisy assimilated fields at 0.3 mb but this is unlikely to affect the streamfunction below 10 mb. Poleward of 60°N the model results give only a slight hint of upward motion and the streamfunction remains more or less horizontal. Nonetheless this contrasts with the CMAM (Beagley et al. 1997), which shows descent throughout that region. The January 1993 TEM residual circulations estimated from diabatic heating rates by Rosenlof (1995) confirm this but the streamfunction she presents, by horizontally integrating the January 1993 w* from the UKMO assimilation, turns horizontal toward the North Pole. Because the UKMO assimilated fields in the stratosphere rely almost entirely on satellite temperature soundings, derived quantities, such as υ* and w*, are likely to be sensitive to the accuracy of the model. The results presented here therefore suggest that there are possible deficiencies in the TEM meridional circulation of the UM over the winter pole, which may also be corrupting the UKMO stratospheric data assimilation.

5. July climatology

a. Zonal mean winds and temperatures

As for January, the strengths and positions of the subtropical jets in July (Fig. 8) were accurately modeled. However, in the Southern Hemisphere, the tropospheric jet was not as clearly separated from the polar night jet as observed (cf. Figs. 8a and 8c). Nonetheless, because of the relatively strong Rayleigh friction the strength of the stratospheric westerly jet was much more realistic than in some other GCMs [e.g., Boville (1995), or the“N45” results of Hamilton et al. (1995)], and improved with increased friction (Fig. 8b). On the other hand, the jet was confined too close to the pole and the maximum occurred too high up. Hence, as in the Northern Hemisphere, the observed equatorward tilt with height of the polar night jet was not reproduced by the UM. The simulated localized easterly maximum and protrusion of easterlies into the winter hemisphere near 1 mb (Figs. 8a and 8b) were similar to January (Figs. 2a and 2b) and agree reasonably well with the MACCM2 results and observations presented by Boville (1995). Nonetheless it is absent from the observations assimilated with the UM (Figs. 2c and 8c) despite the evidence that it is a climatological feature of the model, common to other middle atmosphere GCMs, such as the GISS model (Rind et al. 1988b).

Minimum simulated temperatures at the South Pole in July were ∼180 K at about 20 mb, irrespective of the amount of friction used (see Figs. 9a and 9d), and compared favorably with the observed minimum of about 181 K, again at roughly 20 mb (Figs. 9b and 9c). Above the minimum, the polar temperatures increased less rapidly with height than observed so that the simulated stratopause was about 24 K too cold (232 K and 254 K in Figs. 9a and 9b, respectively), and occurred 1–2 km higher than observed. Increasing the friction reduced the cold bias at the stratopause to ∼13 K. In general, the effects of the increased friction on the July zonal mean temperatures were confined to the upper stratosphere and mesosphere poleward of 60°S, apart, that is, from a slight cooling over the North Pole at the top of the model (see Fig. 9d). The UM’s inability to reproduce the equatorward tilt with height of the jet core resulted in the largest horizontal temperature gradients in the upper stratosphere occurring poleward of 60°S (Fig. 9a) and not, as observed, between 30° and 60°S (Fig. 9b). In the stratosphere below 30 km the model’s horizontal temperature gradients were again too strong near the South Pole and correspondingly too weak in midlatitudes. The slack midlatitude gradients are a manifestation of the poor separation of the tropospheric and stratospheric westerly jets, whereas the poleward shift in the band of strong temperature gradients resulted in a local warm bias between 50° and 80°S, and between 100 and 10 mb (see Fig. 9c). This was the only warm bias in the simulated stratospheric temperatures for July. In the Northern Hemisphere there was a cold bias that exceeded 14 K in the high-latitude upper stratosphere, whereas the tropical tropopause was about 6 K too cold.

The observed interannual variability of the Southern Hemisphere polar night jet differs significantly from its Northern Hemisphere counterpart by the presence of a double maximum in the upper stratosphere (cf. Figs. 10c or 10d, with Figs. 4c or 4d). The UM failed to reproduce this interhemispheric difference (cf. Figs. 10a or 10b, with Figs. 4a or 4b) and the pattern of variability for July was very similar to that for January, albeit with reduced magnitude. In general, the model’s variability for July was less than observed and was further reduced with increased friction (cf. Figs. 10a and 10b). The absence of the interhemispheric differences are probably associated with the lack of a latitudinal dependence to the empirical representation of gravity wave momentum deposition by Rayleigh friction.2 Moreover this appears to be much more crucial for the simulation of the Southern Hemisphere winter than the Northern Hemisphere, particularly, when considering interannual variability. Outside the winter extratropical region there is little interannual variability in the simulated zonal winds for July but, again, the success of the data assimilation in capturing the tropical QBO (Swinbank and O’Neill 1994b) is evident by the peak at about 30 mb at the equator in Fig. 10c.

Differences in interannual variability between the northern and southern winters are also seen in the observed temperatures (cf. Figs. 5 and 11). Whereas in January the greatest variability occurs at the North Pole throughout the stratosphere (e.g., dot–dashed curves in Figs. 5b–d), at 10 mb in July the greatest variability is near 50°S (dotted and dot–dashed curves in Fig. 11c). At 1 mb most of the variability is found in the high latitudes but unlike the Northern Hemisphere there is no observed sharp increase toward the pole (cf. dot–dashed curves in Figs. 5d and 11d). In contrast to the observations there was little difference in the variability of the simulated temperatures between the two hemispheres (cf. solid curves in Figs. 5 and 11). In the Southern Hemisphere there was a slight reduction in the amount of variability and, at 10 mb, the maximum variability was displaced from the pole by about 20°. Also, increasing the friction had more of an effect on the variability in midlatitudes than at the pole and thereby had the undesirable effect of almost eliminating the off pole maximum (cf. solid and dashed curves in Fig. 11c). Overall, however, the variability of the simulated zonal mean temperatures was less sensitive to the choice of friction coefficient in the Southern Hemisphere winter than in the Northern Hemisphere.

b. Stationary waves

A significant weakness of the UM is its poor simulation of the July stationary wave amplitudes in the Southern Hemisphere stratosphere. Figure 12 shows that, whereas the average simulated amplitudes of the first two zonal harmonics agreed rather well with those observed below 100 mb, in the upper stratosphere the amplitude of wavenumber 1 was overpredicted by roughly a factor of 3, and was similar even to the Northern Hemisphere in January (cf. Figs. 6a and 12a). Even the increased friction reduced the maximum amplitude only from 1519 m in Fig. 12a to 1343 m. Furthermore, the SSU analyses indicate that, for the 18-yr period, the mean amplitude of wavenumber 1 had a maximum of ∼360 m, compared with 492 m in Fig. 12c. For wavenumber 2 there was better agreement between model and observations up to about 1 mb but at the higher levels there is a suggestion that the wave was confined too close to the pole (cf. Figs. 12b and 12d).

At 100 mb the UM did not quite reproduce in July the range of stationary wavenumber 1 amplitudes found in the 18 years of SSU analyses (see Fig. 13a). The greatest discrepancies in the range were between 30° and 90°N and 10° and 40°S, where the model did least well in predicting the average amplitudes. However, poleward of 40°S the UM performed well in predicting both the average and range of amplitudes for wavenumber 1 at 100 mb. Therefore, the absence, noted above, of interhemispheric differences in stationary wave amplitudes in the upper stratosphere cannot simply be attributed to problems at the tropopause but must be related directly to errors in the simulated stratospheric circulation and, in particular, the deficiencies in the polar night jet. At 1.468 mb (Fig. 13b) the range of amplitudes was much greater than observed.

c. Meridional circulation

Figure 14 shows the modeled TEM residual circulation for July and that computed from observations. The model results again showed little sensitivity to the friction and only results from the simulation using the smaller coefficient (see solid curve in Fig. 1) are shown. In general, the residual circulation in the model agreed well with that estimated from the observations and was very similar to that for January (cf. Fig. 7), with a Brewer-Dobson circulation in the stratosphere, and single-cell summer-to-winter-pole circulation in the mesosphere. As for January, the different upper-boundary at which Ψ was assumed to vanish, has resulted in the simulated circulation in the mesosphere appearing stronger than that derived from the observations. In the summer stratosphere, consistent with the findings of Rosenlof (1995), the circulation computed from observations is shallower and weaker in July than in January, but this hemispheric asymmetry is less well captured by the UM. However, in both Figs. 14a and 14b the descent in the stratosphere in the winter polar region is rather more than that seen in either Figs. 7a or 7b, which is somewhat contrary to the findings of Rosenlof (1995), who found stronger descent in the northern winter. This discrepancy probably results from deficiencies in the UM’s ability to reproduce the correct residual circulation, which, because of the model’s role in the data assimilation system, could also affect the observational estimates of Ψ.

6. Seasonal and annual cycles

The seasonal variations of the simulated and observed zonal winds at 10 mb are shown in Figs. 15 and 16, respectively. A 5-day running mean has been applied to both datasets. The model results are from the simulation with the smaller Rayleigh friction coefficient (see solid curve in Fig. 1). The observations are from the UKMO’s stratospheric data assimilation for the period 1 October 1991–31 December 1996.

In the extratropics the model performed well in reproducing the basic components of the seasonal cycle. In both hemispheres the timings of the transitions from the summer easterlies to winter westerlies were essentially correct. In the Southern Hemisphere all the breakdowns of the simulated polar vortex occurred, as observed, between late November and early December. In the Northern Hemisphere the large interannual variations in the breakdown of the polar vortex were realistically modeled. Despite this, there were also obvious deficiencies. The summer easterlies were too weak in mid to high latitudes. The Southern Hemisphere polar night jet generally remained too close to the pole as it intensified and did not quite reach the same strengths as observed. The variability in strength of the polar night jet during each northern winter was, on average, much less than observed. Nonetheless this did not prevent the model from spontaneously producing major midwinter warmings in some years, as, for example, occurred in the fifth February in Fig. 15 (see also remarks on sudden warmings later in this section). Changing the upper-level friction did not significantly alter the strengths or weaknesses in the simulated seasonal cycle in the extratropics at 10 mb.

The most notable feature of the simulated zonal wind at the equator compared to the observations was the absence of episodes of westerlies. At 10 mb this was simply a manifestation of the absence of a QBO in the model, but persistent equatorial easterlies were also present throughout the model’s stratosphere. Hence, although the model reproduced an SAO in the tropical upper stratosphere and mesosphere (see Fig. 17) with a reasonably successful easterly phase, it had a poor westerly phase. Shibata and Chiba (1990) found a similar SAO in their spectral GCM and attributed the weakness of the westerly phase to the Rayleigh friction, which can only damp the equatorial easterlies and not reverse the direction of the flow. On the other hand, the simulated SAO in the UM was relatively insensitive to the actual strength of the Rayleigh friction (cf. solid and dotted curves in Fig. 17). In addition, the distinct seasonal asymmetry seen in the observations, with stronger easterlies during the Northern Hemisphere winter, was not reproduced by the model. The MACCM2 SAO simulation also suffered from this defect (Boville 1995).

In the UM, the lack of a seasonal asymmetry in the SAO can probably be attributed to the poor contrast in behavior between the two hemispheres already noted in sections 4 and 5. This is also evident in the seasonal cycle of the amplitude of the monthly mean planetary wavenumber 1 in the upper stratosphere [results not shown, but see Fig. 3 of Butchart et al. (1997)]. In particular, during the southern winter, the stationary wave amplitudes were too large and the UM failed to reproduce the significant growth in amplitude that is observed to occur during spring, as the vortex weakens (Butchart et al. 1997). Excessive easterly momentum deposited in the equatorial upper stratosphere by these waves could then account for the stronger than observed easterly phase of the SAO during this season. In the Northern Hemisphere the largest amplitude stationary wave occurred in February a month later than observed probably due to a relative lack of early and midwinter stratospheric warmings in the model (Butchart et al. 1997).

Confirmation of this is given in Fig. 18a, which shows the daily time series of the North Pole temperatures at 10 mb for 10 consecutive winters of the simulation with the smaller Rayleigh friction. The dotted curves are the same in each panel and represent the 10-yr mean on each calendar date. Anomalously warm periods are denoted by shading. Figure 18b shows the same information for 10 consecutive years of SSU observations. Midwinter warmings with realistic amplitude and suddenness occurred over the North Pole in the simulation in roughly the same number of years as observed. The durations of the anomalously warm periods were, broadly, compatible with the observations though there was less daily variability in the model results.3 The UM also managed to produce “cold-winters” (e.g., 1992/93 or 1997/98) in which the flow remained largely undisturbed for long periods as was observed, for example, in 1987/88. The main deficiency of the simulation was, however, the tendency for the warming events to occur, on average, later than observed. Indeed, the last five simulated winters remained fairly undisturbed until the end of February, and unlike the atmosphere, the model did not produce a significant warming in any of the Decembers. This bias toward the late winter probably accounts, at least in part, for the reduced interannual variability in the simulated January mean circulation noted in section 4. In this respect the simulation differs from that of both the SKYHI model (Hamilton 1995) and the CMAM (Beagley et al. 1997), which had biases toward early winter warmings. Increasing the friction in the UM had little impact on the character of these results.

During the Southern Hemisphere winter there is very little variability in the polar temperatures at 10 mb (see Fig. 19b) and in most years the UM reproduced this behavior quite well (Fig. 19a). The coldest time of the year tended to be around the end of July or the beginning of August slightly later than observed and similar to the results of Beagley et al. (1997). In two consecutive simulated southern winters (1995 and 1996), however, the UM produced anomalously warm periods comparable in amplitude to those observed, and simulated, for the Northern Hemisphere. The other simulation with increased friction (results not shown) also produced consecutive winters (1994 and 1995) with anomalously warm periods, with smaller warm anomalies occurring in 1997 and 1998. Polar temperatures in the other winters were relatively undisturbed, consistent with observed Southern Hemisphere behavior. The largest observed polar warming over the 10 years was in 1988 (Kanzawa and Kawaguchi 1990) but this was less extensive and occurred much later in winter than those found in the simulations. In general, the UM’s behavior in southern winter falls into two regimes, irrespective of the choice of friction. For the majority of the winters the circulation in the polar stratosphere was relatively undisturbed. In the others, temperature variability at the South Pole was poorly simulated and was similar in character to that at the North Pole during winter.

7. Polar stratospheric clouds

Heterogeneous chemistry and denitrification associated with the presence of volcanic aerosol and the formation of PSCs is now considered important for stratospheric ozone depletion (e.g., Crutzen and Arnold 1986; Turco et al. 1989; Hofmann and Solomon 1989). In this section we consider only the model’s ability to predict the correct frequency of occurrences of PSCs as indicated by the daily minimum temperatures at 46 mb, poleward of 43.75°, and shown in Fig. 20 (solid curves) for the 10 years of the simulation with the smaller Rayleigh friction coefficient (see the solid curve in Fig. 1). The corresponding observed range from the 5 years of UKMO stratospheric data assimilation is indicated by the yellow shading. The red dashed line at 195.3 K denotes the thermodynamic equilibrium temperature for the formation of PSCs of nitric acid trihydrate at this pressure with 10 ppbv of nitric acid and 5 ppmv of water vapor present (Hanson and Mauersberger 1988). Similarly the green dashed line at 188.5 K is the thermodynamic equilibrium temperature for ice clouds.

During early and midwinter the UM produced a similar range of minimum temperatures for the polar lower stratosphere as observed, with the differences in variability between the two hemispheres well reproduced, at least until early spring. On the other hand, the Northern Hemisphere results suggest that conditions for the occurrences of PSCs in December and January may be more prevalent in the model than in the atmosphere. From March onward the envelope of simulated minimum temperatures is lower than observed due to the model’s overall stratospheric cold bias (see subsections 4a and 5a) and delayed final warmings (see section 6). Nonetheless, with the exception of one year, conditions suitable for PSC formation did not persist any later into spring than observed during the years 1991–96 (see Fig. 20a). When the upper-level friction was increased, the simulated minimum temperatures were generally similar in character to those shown in Fig. 20a, but temperatures below 195.3 K persisted throughout March in three of the years. Also, simulated temperatures below 188.5 K tended to occur more often and in more years in January and February.

In the Southern Hemisphere spring, the observed range of minimum temperatures in Fig. 20b agrees with other results using more years of data (e.g., Fig. 3-3 of World Meteorological Organization 1995), and the greater simulated range in late winter indicated that the model’s evolution was determined more by dynamical processes than is the case for the atmosphere. This suggests that the UM reproduced a Southern Hemisphere circulation that was perhaps more representative of the observed behavior in the north, but not for every year. Indeed, if the four years with the highest minimum temperatures in September and October are discarded, the range over the remaining years is then comparable to that observed. The more gradual rise in minimum temperatures in these years is then consistent with observed Southern Hemisphere behavior but with conditions suitable for PSC formation persisting later into spring. However, as in the Northern Hemisphere there was some sensitivity to the choice of upper-level friction. In early and midwinter the increased friction tended to produce higher minimum temperatures (sometimes 2–3 K above the highest observed) and also increased the daily range. In spring the distinction between years with Northern and Southern hemisphere-like behavior was now less clearly apparent in this diagnostic.

8. Summary and discussion

In this paper we have presented the first results from multiyear integrations of the troposphere–stratosphere configuration of the UKMO UM. Unlike many other GCMs with full representations of the stratosphere, the model has a relatively low upper boundary at 0.1 mb. There are a number of reasons for this. First, as it was based on a version of the UM used in producing stratospheric analyses by assimilating observations (Swinbank and O’Neill 1994a), data coverage was a key factor in determining the height of the upper boundary. Second, it was developed primarily for studying climate and not stratospheric processes per se. Therefore, the possible benefits of adding more levels above 0.1 mb had to be considered in the context of the overall computing requirements and the need to improve other aspects of the model. On the other hand, it should be noted that the configuration of the UM we used had many levels giving a relatively high vertical resolution of roughly 1.3 km throughout the stratosphere. In principle these could be redistributed to extend the vertical domain. However, it was decided not to do this but, instead, to examine the quality of the simulation that could be achieved with the existing upper boundary at 0.1 mb.

Below the 20-mb level a detailed parameterization of the momentum deposition by orographic gravity waves was used but this proved too noisy in the stratosphere (Swinbank 1996, personal communication). It was therefore decided to use a simple Rayleigh friction as a substitute, with the coefficient chosen on the basis of experience with a “mechanistic” model of the stratosphere and mesosphere. The absence of model levels above 0.1 mb, and its effects on the stratosphere, were crudely compensated for in a second multiyear integration with considerably enhanced Rayleigh damping above 0.3 mb (see Fig. 1). Ten-year climatologies were obtained using both friction coefficients, but our results were found to be largely independent of the choice of Rayleigh friction. Given the absence of a physical basis for determining the coefficient, this then provides confidence in the other conclusions of this study.

The UM did well in capturing the basic features of the monthly and zonally averaged circulation in both January and July though, throughout most of the stratosphere there was a cold bias that was most serious at the winter pole. This “cold pole” problem afflicts nearly all middle atmosphere GCMs without parameterized nonorographic gravity wave drag (e.g., Beagley et al. 1997; Boville 1995; Hamilton et al. 1995). Also, in common with many other models (e.g., Boville 1995; Shibata and Chiba 1990), the polar night jets were tilted poleward with height and not equatorward, as observed. On the other hand, in contrast to other models (e.g., Beagley et al. 1997; Boville 1995; Hamilton et al. 1995) the Southern Hemisphere polar night jet had roughly the correct strength, most likely as a consequence of the rather strong Rayleigh damping in the upper stratosphere.

Interannual variability of the January and July monthly mean circulation was generally less than observed in the lower and middle stratosphere. In the northern winter this was partially due to the relative lack of early and midwinter warmings. Also, as with the positioning of the polar night jet, the variability tended to occur too close to the pole in the upper stratosphere. In the southern winter the problems were more serious with the patterns of variability looking much more like those of the Northern Hemisphere winter, albeit with reduced magnitude, than those observed in the Southern Hemisphere in July.

January and July stationary wave amplitudes and, in particular those for wavenumber 1, were overpredicted in the upper stratosphere. This was mainly the result of the poor stratospheric simulation as amplitudes at 100 mb were roughly correct (though the range over the 10 years was underestimated). Also the seasonal cycle of the upper-stratospheric stationary wavenumber 1 amplitude during the southern winter was poorly simulated and agreed better with the observed behavior for the northern winter.

A brief examination of the TEM residual circulation showed that, whereas the model did reproduce the familiar Brewer–Dobson circulation in the stratosphere, descent rates over the winter polar regions were likely to be inaccurate. In particular, for January, the streamfunction was almost horizontal poleward of 60°N, in the region where strong descent would normally be expected. Some preliminary long integrations of the UM including chemistry suggest that this has a detrimental effect on the simulated total ozone. Our results also suggested a possible detrimental effect on the observational estimates of the residual circulation obtained from the UKMO’s stratospheric data assimilation, which depends on the model.

In the extratropical middle stratosphere the UM reproduced the annual cycle very well with the timing and interannual variability of the transitions between the seasons in good agreement with observations. In the Tropics, in common with other GCMs, there was no QBO in the UM simulation but an SAO was found in the upper stratosphere and lower mesosphere. This, however, did not have a proper westerly phase as zonal mean winds at the equator remained easterly throughout the year at 1 mb. This was attributed to the poor representation of the effects of subgrid-scale momentum deposition by the simple Rayleigh damping. The easterly phase lacked the seasonal asymmetry found in the atmosphere and it was argued that this resulted from the UM’s inability to reproduce the correct contrast between the extratropical stratospheric circulation in the northern and southern winters.

Daily variability was generally less than observed in the winter stratosphere. On the other hand, in the Northern Hemisphere there was reasonable midwinter stratospheric warming behavior, apart from a slight bias toward the late winter. This contrasted with other models (e.g., Beagley et al. 1997; Hamilton 1995) that had biases toward early winter warmings. Performance in the southern winter was once again not as good, with the polar temperatures at 10 mb in some years displaying a more Northern Hemisphere–like behavior. Nonetheless, in the majority of the winters the simulated temperature behavior at the South Pole was quite realistic. It is worth emphasizing that this characteristic of the UM, in producing two distinct flow regimes for the Southern Hemisphere, was rather independent of the choice of friction and therefore it would be important to discover whether other stratospheric GCMs exhibit the same behavior. Certainly, for the UM, it indicates the dangers of obtaining conclusions from integrations of less than, say, 5 years (e.g., Swinbank et al. 1998), at least regarding the performance of the model in the Southern Hemisphere.

The final aspects of the model performance we considered were minimum temperatures in the polar lower stratosphere and the implications for PSC formation. Unlike many other diagnostics these were found to be quite sensitive to the choice of Rayleigh friction. Thus, in the Northern Hemisphere conditions suitable for PSC formation in December and January were most prevalent in the simulation with the strongest friction, and in both simulations more prevalent than in the atmosphere. Increasing the friction caused the model to produce conditions suitable for PSC formation throughout March in some years, contrary to observations. On the other hand, in the Southern Hemisphere, the increased friction produced, on average, higher minimum temperatures and thereby reduced the prospect of PSC formation at the beginning and end of winter. The southern winters could also be separated into two groups. In the majority of the years the model behavior was consistent with the observations but with conditions suitable for PSC formation persisting later into spring. In the remaining years the minimum temperatures rose rapidly above the observed range in late winter.

In common with most other comprehensive GCMs the individual components of the UM are continually being improved. Three aspects are worth noting. First, other applications of the UM are already using a new parameterization of the effects of subgrid-scale orographic gravity waves (e.g., Milton and Wilson 1996), but, with the extended vertical domain, this again introduced too much noise into the stratosphere. Nonetheless, preliminary results suggest that aspects of the model’s performance, such as the poor separation of the subtropical and polar night jets in July, could be improved by using the new scheme below the 20-mb level (Swinbank 1997, personal communication). Second, a new state-of-the-art radiation code, which can take account of a whole range of radiative gases is now available (Edwards and Slingo 1996) and being tested in the climate configuration. Third, a new integration scheme with semi-Lagrangian advection is under development (Cullen et al. 1998). Introducing these last two schemes is expected to help in eliminating the overall stratospheric cold bias and the wintertime local warm biases in the high-latitude lower stratosphere that we found in this study. This should benefit chemistry-climate simulations by improving both the simulation of PSCs and the meridional circulation for tracer transport.

Other deficiencies that we found, such as the incorrect tilt with height of the polar-night jets, the “cold pole,” and the less than observed distinction between the northern and southern winters, are unlikely to be eliminated by improvement in the physical parameterizations and integration scheme. Indeed, all these deficiencies are almost certainly related to the absence above 20 mb of a proper parameterization of the momentum deposition from subgrid-scale waves. However, this is known to occur mostly above 0.1 mb (the UM’s upper boundary) and, according to Shepherd et al. (1996), its impact on the circulation lower down could only be properly represented if it were occurring well below the model’s upper boundary. Therefore, although the climatologies of the lower and middle stratosphere obtained from these first multiyear integrations of the troposphere–stratosphere configuration UM are highly respectable, major improvements are only likely to be obtained by raising the upper boundary in order to allow for a better representation of gravity wave momentum deposition in the mesosphere (e.g., Warner and McIntyre 1996).

Acknowledgments

Discussions with Terry Davies on the numerical formulation of the UM are gratefully acknowledged. Darren Podd and Adam Scaife provided assistance with the observational datasets.

REFERENCES

  • Andrews, D. G., J. R. Holton, and C. B. Leovy, 1987: Middle Atmosphere Dynamics. Academic Press, 489 pp.

  • Austin, J., and N. Butchart, 1994: The influence of climate change and the timing of stratospheric warmings on Arctic ozone depletion. J. Geophys. Res.,99, 1127–1145.

  • ——, ——, and R. Swinbank, 1997: Sensitivity of ozone and temperature to vertical resolution in a GCM with coupled stratospheric chemistry. Quart. J. Roy. Meteor. Soc.,123, 1405–1431.

  • Bailey, M. J., A. O’Neill, and V. D. Pope, 1993: Stratospheric analyses produced by the United Kingdom Meteorological Office. J. Appl. Meteor.,32, 1472–1483.

  • Beagley, S. R., J. de Grandpré, J. N. Koshyk, N. A. McFarlane, and T. G. Shepherd, 1997: Radiative-dynamical climatology of the first-generation Canadian Middle Atmosphere Model. Atmos.-Ocean,35, 293–331.

  • Boville, B. A., 1995: Middle atmosphere version of CCM2 (MACCM2): Annual cycle and interannual variability. J. Geophys. Res.,100, 9017–9039.

  • Butchart, N., J. Austin, C. S. Douglas, and R. Swinbank, 1997: Comparison of climatologies from the middle atmosphere versions of the UKMO’s GCM and data assimilation. Proc. First SPARC General Assembly, Melbourne, Australia, World Meteorological Organization, 24–27.

  • Clough, S. A., N. S. Grahame, and A. O’Neill, 1985: Potential vorticity in the stratosphere derived using data from satellites. Quart. J. Roy. Meteor. Soc.,111, 335–358.

  • Crutzen, P. J., and F. Arnold, 1986: Nitric acid cloud formation in the cold Antarctic stratosphere: A major cause of the springtime“ozone hole.” Nature,324, 651–655.

  • 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.

  • ——, ——, M. H. Mawson, J. A. James, S. C. Coulter, and A. Malcolm, 1998: Some issues in numerical methods for the next generation of NWP and climate models. Atmos.-Ocean,38, 425–444.

  • Edwards, J. M., and A. Slingo, 1996: Studies with a flexible new radiation code. Part I: Choosing a configuration for a large-scale model. Quart. J. Roy. Meteor. Soc.,122, 689–719.

  • Fritts, D. C., 1984: Gravity wave saturation in the middle atmosphere:A review of theory and observations. Rev. Geophys. Space Phys.,22, 275–308.

  • Hamilton, K., 1995: Interannual variability in the Northern Hemisphere winter middle atmosphere in control and perturbed experiments with the GFDL SKYHI general circulation model. J. Atmos. Sci.,52, 44–66.

  • ——, 1996: Comprehensive meteorological modelling of the middle atmosphere: A tutorial review. J. Atmos. Terr. Phys.,58, 1591–1627.

  • ——, R. J. Wilson, J. D. Mahlman, and L. J. Umscheid, 1995: Climatology of the SKYHI troposphere–stratosphere–mesosphere general circulation model. J. Atmos. Sci.,52, 5–43.

  • Hanson, D., and K. Mauersberger, 1988: Laboratory studies of the nitric acid trihydrate: Implications for the south polar stratosphere. Geophys. Res. Lett.,15, 855–858.

  • Haynes, P. H., C. J. Marks, M. E. McIntyre, T. G. Shepherd, and K. P. Shine, 1991: On the “downward control” of extratropical diabatic circulations by eddy-induced mean zonal forces. J. Atmos. Sci.,48, 651–678.

  • Hofmann, D. J., and S. Solomon, 1989: Ozone destruction through heterogeneous chemistry following the eruption of El Chichón. J. Geophys. Res.,94, 5029–5041.

  • Jenkins, D. B., D. P. Wareing, L. Thomas, and G. Vaughan, 1987: Upper stratospheric and mesospheric temperatures derived from lidar observations at Aberystwyth. J. Atmos. Terr. Phys.,49, 287–298.

  • Kanzawa, H., and S. Kawaguchi, 1990: Large stratospheric sudden warming in Antarctic late winter and shallow ozone hole in 1988. Geophys. Res. Lett.,17, 77–80.

  • Manney, G. L., R. Swinbank, S. T. Massie, M. E. Gelman, A. J. Miller, R. Nagatani, A. O’Neill, and R. W. Zurek, 1996: Comparison of U. K. Meteorological Office and U. S. National Meteorological Center stratospheric analyses during northern and southern winter. J. Geophys. Res.,101, 10 311–10 334.

  • Manzini, E., and L. Bengtsson, 1996: Stratospheric climate and variability from a general circulation model and observations. Climate Dyn.,12, 615–639.

  • McIntyre, M. E., and T. N. Palmer, 1983: Breaking planetary waves in the stratosphere. Nature,305, 593–600.

  • Milton, S. F., and C. A. Wilson, 1996: The impact of parameterized subgrid-scale orographic forcing on systematic errors in a global NWP model. Mon. Wea. Rev.,124, 2023–2045.

  • Mitchell, J. F. B., T. C. Johns, J. M. Gregory, and S. F. B. Tett, 1995:Climate response to increasing levels of greenhouse gases and sulphate aerosols. Nature,376, 501–504.

  • Morcrette, J.-J., L. Smith, and Y. Fouquart, 1986: Pressure and temperature dependence of the absorption in longwave radiation parameterizations. Contrib. Atmos. Phys.,59, 455–469.

  • Nash, J., and J. L. Brownscombe, 1983: Validation of the stratospheric sounding unit. Adv. Space Res.,2, 59–62.

  • Palmer, T. N., G. J. Shutts, and R. Swinbank, 1986: Alleviation of a systematic westerly bias in general circulation and numerical weather prediction models through an orographic gravity wave drag parametrization. Quart. J. Roy. Meteor. Soc.,112, 1001–1039.

  • Randel, W. J., 1987: The evaluation of winds from geopotential height data in the stratosphere. J. Atmos. Sci.,44, 3097–3120.

  • Reber, C. A., 1993: The Upper Atmosphere Research Satellite (UARS). Geophys. Res. Lett.,20, 1215–1218.

  • Rind, D., R. Suozzo, and N. K. Balachandran, 1988a: The GISS global climate-middle atmosphere model. Part II: Model variability due to interactions between planetary waves, the mean circulation and gravity wave drag. J. Atmos. Sci.,45, 371–386.

  • ——, ——, ——, A. Lacis, and G. Russell, 1988b: The GISS global climate-middle atmosphere model. Part I: Model structure and climatology. J. Atmos. Sci.,45, 329–370.

  • Rood, R. B., and M. A. Geller, 1994: Editorial. J. Atmos. Sci.,51, 2781–2782.

  • Rosenlof, K. H., 1995: Seasonal cycle of the residual mean meridional circulation in the stratosphere. J. Geophys. Res.,100, 5173–5191.

  • Sassi, F., R. R. Garcia, and B. A. Boville, 1993: The stratopause semiannual oscillation in the NCAR Community Climate Model. J. Atmos. Sci.,50, 3608–3624.

  • Shepherd, T. G., K. Semeniuk, and J. N. Koshyk, 1996: Sponge layer feedbacks in middle-atmosphere models. J. Geophys. Res.,101, 23 447–23 464.

  • Shibata, K., and M. Chiba, 1990: A simulation of seasonal variation of the stratospheric circulation with a general circulation model. J. Meteor. Soc. Japan,68, 687–703.

  • Swinbank, R., and A. O’Neill, 1994a: A stratosphere-troposphere data assimilation system. Mon. Wea. Rev.,122, 686–702.

  • ——, and ——, 1994b: Quasi-biennial and semi-annual oscillations in equatorial wind fields constructed by data assimilation. Geophys. Res. Lett.,21, 2099–2102.

  • ——, W. A. Lahoz, A. O’Neill, C. S. Douglas, A. Heaps, and D. Podd, 1998: Middle atmosphere variability in the Met. Office unified model. Quart. J. Roy. Meteor. Soc., in press.

  • Turco, R. P., O. B. Toon, and P. Hamil, 1989: Hetrogeneous physicochemistry of the polar ozone hole. J. Geophys. Res.,94, 16 493–16 510.

  • Warner, C. D., and M. E. McIntyre, 1996: On the propagation and dissipation of gravity wave spectra through a realistic middle atmosphere. J. Atmos. Sci.,53, 3213–3235.

  • White, A. A., and R. A. Bromley, 1995: Dynamically consistent, quasi-hydrostatic equations for global models with a complete representation of the Coriolis force. Quart. J. Roy. Meteor. Soc.,121, 399–418.

  • World Meteorological Organization, 1995: Scientific assessment of ozone depletion: 1994. Rep. 37, WMO Global Ozone Research and Monitoring Project, Geneva.

Fig. 1.
Fig. 1.

Vertical profile of Rayleigh friction coefficient. Solid and dashed curves are for the two UM simulations. The UM upper boundary is at 0.1 mb but the solid curve also indicates the drag used in the SMM and, therefore, to show the drag used in that model, is extended higher. The dotted curve is the friction coefficient used by Boville (1995) in the MACCM2.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 2.
Fig. 2.

January monthly and zonally averaged westerly winds. (a) and (b) The 10-yr means from the UM simulations with Rayleigh friction coefficients given by the solid and dashed curves, respectively, in Fig. 1. (c) The 5-yr mean from the UKMO’s stratospheric data assimilation for 1992–96. (d) The 18-yr mean geostrophic winds calculated from the UKMO’s stratospheric analyses for 1979–96. The contour interval is 10 m s−1 and shading denotes easterlies.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 3.
Fig. 3.

January monthly and zonally averaged temperatures and temperature differences. (a) The 10-yr mean from the UM simulation with Rayleigh friction coefficient given by the solid curve in Fig. 1. (b) The 5-yr mean from the UKMO’s stratospheric data assimilation for 1992–96. (c) Simulated temperatures minus assimilated temperatures. (d) Simulated temperatures minus the simulated temperatures from the integration with increased Rayleigh friction (see dashed curve in Fig. 1). (a) and (b) The contour interval is 5 K. Dotted shading indicates temperatures above 270 K, and the hatched shading indicates temperatures below 190 K. (c) and (d) The contour interval is 2 K. Dotted shading indicates a cold bias of more than 16 K, and the hatched shading indicates warm biases.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 4.
Fig. 4.

As in Fig. 2 but the standard deviation of the individual January means from the corresponding multiannual mean [10, 10, 5, and 18 yr, respectively for (a), (b), (c), and (d)]. The contour interval is 2 m s−1 and light shading indicates a standard deviation less than 4, and dark shading a standard deviation less than 2.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 5.
Fig. 5.

Standard deviations of the individual January, monthly, and zonally averaged, temperatures from the multiyear average as (a) a function of height at the North Pole and (b), (c), and (d) a function of latitude at 100, 10, and 1 mb, respectively. Solid and dashed curves:10-yr model simulations using the friction coefficients given by the solid and dashed curves in Fig. 1, respectively. Dot–dashed curve: UKMO SSU analyses for the years 1979–96. Dotted curve: UKMO stratospheric data assimilation for the years 1992–96.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 6.
Fig. 6.

Average amplitudes of the January stationary waves. (a) and (b) Wavenumbers 1 and 2, respectively, from the 10-yr model simulation using the Rayleigh friction coefficient given by the solid curve in Fig. 1. (c) and (d) Wavenumbers 1 and 2 calculated from assimilated observations for the years 1992–96. The contour intervals above and below 100 mb are 100 and 20 m, respectively.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 7.
Fig. 7.

January mean residual streamfunction Ψ calculated from υ* for (a) the UKMO stratospheric data assimilation for 1992–96, and (b) the 10-yr model simulation with Rayleigh friction coefficient given by the solid curve in Fig. 1. Positive contours represent a circulation in the counterclockwise sense. Negative contours are indicated by the dashed lines. Contour levels are ±2, ±5, ±10, ±20, ±50, ±100, ±200, ±500, ±1000, and ±2000 kg m−1 s−1. The results with the Rayleigh friction given by the dashed curve in Fig. 1 were found to be similar and therefore not shown.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 8.
Fig. 8.

As for Fig. 2 except that July zonal winds are shown.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 9.
Fig. 9.

As for Fig. 3 except that July temperatures and temperature differences are shown.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 10.
Fig. 10.

As for Fig. 4 except that standard deviations of the July zonal winds are shown.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 11.
Fig. 11.

As for Fig. 5 except that standard deviations of the July, monthly and zonally averaged, temperatures are shown, but note the different scale for some of the axes.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 12.
Fig. 12.

As for Fig. 6 except that the average amplitudes of the July stationary waves are shown.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 13.
Fig. 13.

Amplitudes of stationary wavenumber 1 for July at (a) 100 mb and (b) 1.468 mb. The solid curve is the average amplitude from 10 years of model results (using the friction profile given by the solid curve in Fig. 1); shading denotes the range of amplitudes. The corresponding information for the 18 years of SSU observations is given by the three dashed curves, with the thickest line denoting the average.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 14.
Fig. 14.

As for Fig. 7 except that the July mean residual streamfunctions Ψ are shown.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

i1520-0469-55-17-2782-f15

>Fig. 15. Evolution of the zonal-mean zonal wind at 10 mb for the model simulation with the smaller Rayleigh friction coefficient (see solid curve in Fig. 1). Results shown are from 5-day running means. The contour interval is 10 m s−1, and regions of easterlies are shaded.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 16.
Fig. 16.

As for Fig. 15 except that the winds from the UKMO stratospheric data assimilation for the period 17 October 1991 to 31 December 1996 are shown.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 17.
Fig. 17.

Zonal-mean zonal wind (m s−1) at the equator and at 1 mb. The solid and dotted curves are for the last 5 years of the two UM simulations using the friction profiles given by the solid and dashed curves in Fig. 1, respectively. The dashed curve is the assimilated wind for the period 1 January 1992 to 31 December 1996. A 5-day running mean has been applied to all the data.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 18.
Fig. 18.

Time series of daily North Pole temperatures (K) at 10 mb for November–April for (a) 10 consecutive winters of the UM simulation with the smaller Rayleigh friction coefficient (see solid curve in Fig. 1) and (b) 10 consecutive years of SSU observations from 1985 to 1995. The dashed curve is the 10-yr average on that calendar day and shading denotes the anomalously warm periods.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 19.
Fig. 19.

As for Fig. 18 but daily South Pole temperatures (K) at 10 mb for April–September.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

Fig. 20.
Fig. 20.

Minimum temperatures at 46.42 mb poleward of 43.75° for (a) the northern and (b) southern winters. The solid curves indicate the minimum temperatures in each of the 10 years of the simulation using the smaller Rayleigh friction coefficient (see solid curve in Fig. 1). The observed range from November 1991 to October 1996 is given by the yellow shading. The red and green dashed lines are thermodynamic equilibrium temperatures for the formation of NAT and ice PSCs, respectively.

Citation: Journal of the Atmospheric Sciences 55, 17; 10.1175/1520-0469(1998)055<2782:MACFTT>2.0.CO;2

1

The 18 years of SSU analyses suggest that during the 5-yr period in Fig. 6c, the maximum was about 20% higher than the long-term mean, although the position of the maximum was correctly located.

2

In contrast to the UM, newly published results (Manzini and Bengtsson 1996) from a T21 spectral GCM with upper boundary at 0.1 mb, and also using Rayleigh friction as a substitute for gravity wave drag, indicate the correct pattern of variability for July.

3

Note that some of the more prominent “spikes” in the observed temperatures in Fig. 18b and, in particular, those at the beginning of the 1985/86, 1988/89, and 1991/92 winters are probably not real but, arise from problems with data on those days. The same caveat also applies to Fig. 19b.

Save