 |
Stratospheric Processes And their Role in Climate
|
||||||||
| Home | Initiatives | Organisation | Publications | Meetings | Acronyms and Abbreviations | Useful Links |
|
Stratospheric Processes And their Role in Climate
|
||||||||
| Home | Initiatives | Organisation | Publications | Meetings | Acronyms and Abbreviations | Useful Links |
Gravity waves can act to transfer mean horizontal momentum from the ground to levels aloft in the atmosphere or from one layer of the atmosphere to another. Flow over topography can generate stationary gravity waves that break nonlinearly in the troposphere and lower stratosphere. Such waves transfer momentum from the breaking region to the earth's surface, and this process is thought to act as a significant drag on the eastward mean winds in the midlatitude troposphere. It is known that numerical simulation models of the global atmosphere run without any attempt to impose such a drag tend to produce results characterized by unrealistically intense midlatitude eastward surface winds. Other processes (such as convection, jet stream instabilities etc.) can produce gravity waves with nonzero horizontal phase speeds and which act to transfer mean momentum between the troposphere and the stratosphere/mesosphere. Comprehensive numerical models of the atmosphere are known to produce unrealistic simulations of the extratropical stratospheric/mesospheric circulation unless some account is taken of the effects of these gravity waves. If credible simulations of climate (and predictions of climate response to anthropogenic forcing) are to be obtained, some physically justifiable parameterization of the momentum transport due to unresolvable gravity waves needs to be formulated. This issue is now recognized as one of the most important challenges in dynamical meteorology.
In order to begin addressing this problem a NATO Advanced Research Workshop was held in Santa Fe from April 1-5. The meeting was cosponsored by SPARC and the Solar Terrestrial Energy Program (STEP) and was attended by a total of 36 scientists from 10 countries. The meeting was notable in bringing together experts in observations and theory of gravity waves along with representatives of several climate modelling groups. The proceedings of this workshop will be published later this year by Springer-Verlag.
This meeting included the first widespread discussion of the performance of currently proposed parameterization schemes. There are three main parameterization schemes now being tested. Each is essentially one-dimensional and explicitly considers only statistically-steady conditions. Each makes some assumption concerning the spectrum of upward- propagating gravity waves at the lowest level considered (typically taken near the tropopause) and then uses a simplified treatment of the dynamics to compute the propagation and dissipation of the waves throughout the middle atmosphere (and hence the mean flow modifications induced by the waves).
The "Lindzen" scheme is based on a quasi-linear treatment of a discrete set of waves (each with prescribed frequency and horizontal phase velocity relative to the ground). The waves are treated independently and are assumed to grow linearly until an instability criterion (typically characterized by some Richardson number criterion) is exceeded at some height. Then the growth of the wave is assumed to be limited to just maintain the saturation of the instability condition. The "Hines" scheme treats a continuous spectrum of waves. Again each component of the spectrum is assumed to grow linearly until some nonlinearity condition is met, but Hines explicitly involves the entire spectrum in the determination of this condition. In particular, he imagines that the high vertical wavenumber (denoted here as m) end of the spectrum is controlled by nonlinearity and that the break between the linear and nonlinear m regimes occurs at lower wavenumber as the total horizontal wind variance in the spectrum increases. The third approach advocated by Fritts and Lu also treats the entire spectrum, but uses empirical evidence to guide the imposition of the saturation of the wave amplitudes.
Both the Lindzen and Hines schemes predict little wave forcing of the mean flow below heights at which their proposed saturation mechanisms become important. The stronger the imposed waves at the lower boundary, the lower down the significant nonlinearities (and consequent wave driving of the mean flow) will occur. This observation has led many modellers to a particular modification of these schemes, which can be described as follows. Imagine first a standard application of, say, the Lindzen scheme with a single prescribed wave with momentum flux Fo imposed at the lower boundary. Then one could imagine another application based on the assumption that the wave is present for a fraction alpha of the time and completely absent for the remaining time. If the wave flux at the lower boundary is multiplied by 1/alpha, then the time-average wave flux imposed in the two cases is the same, but the region of significant forcing of the mean flow will extend lower down in the second case. The factor alpha has been referred to by many investigators as the "efficiency factor", but might be better termed an "intermittancy factor". Typically in the Lindzen scheme the same efficiency factor is assumed to apply to each of the waves considered. The same kind of modification can be introduced into the Hines scheme.
I. Hirota (Kyoto University) opened the meeting with a historical review of our observational knowledge of the gravity wave field in the middle atmosphere. He emphasized the value of single station data in characterizing the details of the gravity wave field. Hirota also reported on recent work concerning the relation between observed stratospheric waves and localized tropospheric sources such as jet streaks.
R. Vincent (University of Adelaide) discussed the use of high-resolution radiosondes to investigate the gravity wave field in the lower stratosphere. He noted that as data from more stations has become available, the geographical and seasonal variations in wave activity have been found to be more complicated. For example at high NH latitudes the temperature variance spectrum seems to have a m-3 behavior, but at the South Pole the observed spectral slope is less steep. Recent analysis of Japanese ship data also suggests some significant interhemispheric asymmetry in the gravity wave properties. Vincent described an analysis of the temperature and horizontal wind observations at some Australian stations designed to determine the intrinsic frequencies and dominant directions of propagation of lower stratospheric gravity waves. He ended with a description of some calculations he has conducted using idealized random wave spectra to produce wind and temperature fields that were then sampled in the same manner as a radiosonde. This allowed him to test the limitations of using single station for determination of gravity wave properties.
R. Sica (University of Western Ontario) discussed observations of the vertical temperature structure in the upper stratosphere and mesosphere obtained with the Purple Crow Rayleigh lidar located near London, Canada. One of his striking results was a high vertical and temporal resolution determination of the regions of negative static stability. An example for one night is shown here as Fig. 1. It appears that very thin regions (~100 m) of unstable lapse rate appear rather randomly, and typically persist for an hour or less. In general about 5% of the measured column (30-70 km) is found to be statically-unstable at any time, with the incidence of unstable situations observed to increase with height.
Figure 1
N. Gavrilov (St. Petersburg University) reviewed observations of gravity wave climatology focussing particularly on radar observations of the horizontal wind in the upper mesosphere and lower thermosphere. He pointed out a number of interesting features of the analysis that has been conducted on the MF radar data from Saskatoon, Canada. In particular, both the seasonal amplitude behavior and dominant wind orientations of the highest frequency variations (periods less than 30 minutes) were found to differ from those of lower frequency waves.
K. Sato (Kyoto University) discussed the analysis of radiosonde data to examine the equatorial waves in the lower stratosphere. She found prominent peaks in zonal wind and temperature variance at periods near 10 days during the westerly shear phase of the QBO and in the 1- 3 day period range throughout the entire QBO cycle. She discussed different approaches to estimating the vertical flux of zonal momentum associated with these prominent waves. The result for the 10 day wave suggested that this spectral peak is largely associated with eastward-propagating Kelvin waves. For the 1-3 day waves there was a clear indication of a rough balance between eastward and westward propagating components. Sato also discussed observations of higher frequency gravity waves made with the Kyoto MU radar during the passage of a typhoon within 100 km of the facility.
C. Gardner (University of Illinois) discussed observations of wind and temperature in the ~80-105 km altitude range obtained during several nights with a Na lidar. The combination of the Na laser and the telescope at the Starfire Observatory near Albuquerque, New Mexico) allowed temperature, vertical and horizontal wind measurements to be obtained with 24 m vertical resolution and 3, 6 and 12 minute time resolution, respectively. This high quality data allowed a very detailed picture of the vertical wavenumber and temporal spectra to be obtained. Gardner found that the temperature variance spectrum varies like m-2 and w-2. The horizontal and vertical power spectra vary roughly as m-3 and m-1, respectively. Gardner pointed out that the shallowness of the spectrum of vertical velocity relative to that of horizontal velocity invalidates the usual assumption of a spectrum of wave separable in vertical wavenumber and frequency. He also reported on calculations of the vertical eddy heat fluxes obtained from the observed wind and temperature fluctuations. This flux showed a great deal of day-to-day variability.
F.-J. Lubken described results from a large number of rocket-borne falling sphere profiles taken in the mesosphere near 70oN over the period 1983-95. The spheres were instrumented with ion probes that allowed a very high-resolution determination of the density and temperature profiles. Lubken noted that the summer mesopause is generally near 88 km and is characterized by temperatures near 130K. The profiles display considerable small-scale temperature variability and thin regions with statically-unstable lapse rate are found to occupy about 4% of the region. In summer turbulent layers were found to be largely restricted to 80-95 km, while in winter significant turbulence was observed throughout the mesosphere.
J. Bacmeister (Naval Research Laboratory) discussed the horizontal spectra of wind and temperature in the lower stratosphere determined from ER-2 flights in the AASE and SPADE as well as the ASHOE/MAESA campaigns. The total data set consists of about 70 flights of ~6-8 hours each and analysis focussed on the 1-100 km wavelength range. The wind and temperature spectra differ significantly from the k- 5/3 power law often taken as a reference for the mesoscale regime. At wavelengths shorter than about 3 km the observed spectra are steeper than k-5/3, while at wavelengths longer than about 6 km the spectra are shallower than k-5/3. By contrast, the ozone and nitrous oxide spectra appear to follow a simple k-5/3 power law over the entire range considered.
L. Pfister (NASA Ames Research Center) described both in situ and microwave profiling observations from the ER2 aircraft in the ASHOE/MAESA campaign. He discussed a particular case of a flight south from New Zealand which passed directly over a rather localized tropospheric storm. Over the storm region there was a clear maximum in the variance in both temperature and vertical wind. The temperature and wind perturbations were consistent with the lower stratospheric response being mainly in linear, vertically-trapped waves. The largest displacements of isentropic surfaces seen above the extratropical storms in the ASHOE/MAESA campaign were ~250 m, which Pfister points out is smaller by a factor of ~2 than the largest displacements seen above tropical convective events.
D. Wu (Jet Propulsion Laboratory) discussed the horizontal variations seen in limb-scanning microwave measurements of atmospheric temperature from the UARS satellite. The observations are particularly sensitive to those waves with vertical wavelengths greater than ~10 km and horizontal wavelengths less than ~100 km. The variance in this wavelength range is found to increase with height throughout the middle atmosphere and also shows strong geographic modulation. Particularly striking are maxima in variance seen over the continental areas in low latitudes and over the midlatitude jet streams. Wu pointed out that this is consistent with the possibility of a strong geographical modulation of tropospheric wave sources.
F. Sassi (University of L'Aquilia and NCAR) discussed a simple mechanistic model of the flow in the tropical middle atmosphere. In particular he considered an equatorial beta-plane model stretching from the ground to the lower thermosphere and truncated to zonal wavenumber 15. This was forced by a random tropospheric heating meant to account for the effects of tropical convection. The spectral form for this heating was chosen with guidance from time series of satellite observations of tropical cloudiness. The model also included a zonal-mean forcing in the middle atmosphere designed to account for the easterly drag connected with cross-equatorial flow and planetary waves in the real atmosphere. Sassi was able to produce a quite realistic semiannual oscillation (SAO) near the stratopause in this model. He showed that the intermediate scale waves (zonal waves 4-15) account for a very significant fraction of the mean flow forcing in the westerly acceleration phase of the stratopause SAO. The model also produces a second peak in SAO amplitude near 90 km. This is even more strongly dependent on the eddy fluxes from intermediate scale waves.
J. Holton (University of Washington) discussed calculations of convectively-forced gravity waves in a 2D, nonlinear, moist, limited-area model. The model is initialized with tropical ambient conditions and a localized disturbance. This led to a rather realistic simulation of a squall line which involves the modulation of a series of propagating precipitation cells of brief duration. This squall line also produced gravity waves radiating into the model stratosphere. Holton repeated his calculations with stratospheric mean winds characteristic of the easterly and westerly phases of the quasi-biennial oscillation (QBO). The tropospheric simulation was largely unaffected by the stratospheric winds, but the gravity wave field within the stratosphere itself was found to be strongly dependent on the mean wind distribution employed. When averaged over the limited area of the model (800 km in the zonal direction) Holton finds that the mean flow forcing from his simulated gravity waves in the lower stratosphere can be ~5 m-s-day-1. He argues that the small horizontal scale, high-frequency gravity waves produced by such tropical weather systems could make a significant contribution to the global forcing of the QBO.
M. Reeder (Monash University) described some calculations of the stratospheric wave response to tropospheric mesoscale systems in the context of a dry 2D anelastic model. He initialized his model with a large-scale deformation wind field taken from the classic semigeostrophic frontogenesis solutions, but with an added perturbation. The integration produced a very well-developed tropopause fold and cross-frontal circulation in the troposphere, and a train of inertia- gravity waves radiating into the stratosphere (see Fig. 2 which shows the horizontal and vertical wind fields after 36 hours of integration). In a separate presentation Reeder discussed his analysis of radiosonde data taken at Macquarie Island (54oS) to determine aspects of the gravity wave field in the lower stratosphere. He then described a ray tracing calculation (using Australian Bureau of Meteorology analyses to define the mean state) designed to determine the origins for the low frequency inertia-gravity waves seen at Macquarie.
Figure 2
J. Alexander (University of Washington) began by reviewing attempts that have been made to use large- scale data and radiative transfer calculations to make a residual calculation of the required gravity wave drag. She then went on to discuss calculations of the gravity wave field and mean flow forcing from a somewhat modified version of the Lindzen parameterization with an input spectrum consisting of 60 discrete waves widely distributed in horizontal wavelength (6 to 800 km) and period (7 minutes to 4 hours). The waves were assumed to have a globally-uniform source, but then propagated through a mean state based on the full 3D monthly-mean UK Meteorological Office observational analyses. Alexander was able to show that in the stratosphere and mesosphere the filtering effects of the mean flow were able to produce very significant geographical variations in the computed wave activity. With her choice of parameters and an efficiency factor ~0.001 Alexander was able to produce a distribution of zonal-mean drag that is comparable to that estimated in the residual calculations.
C. Hines (Toronto) described his work on formulating a parameterization scheme for the gravity wave field that considers the Doppler-shifting of the high vertical wavenumber waves by the wind fluctuations from the entire spectrum of waves. He began with a very basic justification of his approach in the simple case of no mean flow. He imagined a spectrum initially nonzero only in a restricted region of m space. As the waves propagate upward the horizontal winds associated with the waves increase and it is reasonable to expect that nonlinearities will begin to fill in the tail of the spectrum. Hines notes that he has performed detailed calculations that indicate that the asymptotic form of the tail should be roughly m-3, but that this expectation is also consistent with the simplest similitude arguments and thus seems independent of the details of his theory. Hines suggests that the break between the linear low m part of the spectrum and the saturated tail should occur at a wavenumber proportional to the Brunt-Vaisala frequency divided by the rms horizontal wind in the entire spectrum. Hines then went on to consider the implications of this picture of the spectral propagation for practical parameterization schemes and to outline his specific proposal in this regard.
H. Mayr (NASA Goddard Space Flight Center) described the incorporation of the Hines gravity wave parameterization scheme into a 2D nonlinear model of the middle atmosphere (forced also by a diabatic heating computed from a realistic radiative transfer model). The gravity wave source at tropopause levels was specified simply as globally-uniform and isotropic (actually discretized in 8 uniformly-distributed directions). The incorporation of the gravity wave drag led to a number of realistic features in the simulation. The winter and summer mesospheric jets closed off in the upper mesosphere and reverse in the thermosphere, in basic agreement with observations. The imposed gravity wave drag also generates a quite reasonable semiannual oscillation in the equatorial mesosphere. There is even a quasi-biennial oscillation (though with unrealistically weak amplitude) appearing in the equatorial stratosphere of the model.
M. McIntyre (Cambridge University) reviewed some recent work on the maintenance of the large-scale mean circulation in the middle atmosphere, which forms the basic background to the gravity wave parameterization problem. He then went on to consider some simple ray theory calculations for nonhydrostatic, inertia-gravity waves, noting that the usual ray theory fails near critical lines and near caustics (loci of ray crossings). McIntyre noted that ray theory provides an interesting view of some wave-wave interactions. For example, he showed that the propagation of a large m wave through a larger-scale wave packet can actually lead to its transformation into a wave with smaller vertical wavenumber. This represents a counterexample to the simplest notions that suggest that sufficiently large m waves will inevitably be shifted to larger wavenumber and ultimately obliterated.
A. Medvedev (St. Petersburg University and York University) described his modification of the Hines scheme to account for the effects of the high m end of the spectrum on the lower m waves. The damping of low frequency waves in his scheme is accomplished by an effective diffusion due to the motions associated with the high frequency waves. He presented some preliminary results from the inclusion of his scheme into a version of the NCAR Community Climate Model.
C. Warner (Cambridge University) reported on calculations of the propagation of a spectrum of waves with a very detailed quasi-linear model. The model considered a continuous spectrum (treated numerically by a high-resolution discretization) and allowed the waves to propagate conservatively except for absorption near critical surfaces, back-reflection when appropriate, and a nonlinear chopping of the spectrum at high m and high frequency to impose a broadband saturation criterion. Warner investigated the decay of total wave energy with height and its sensitivity to the shape of the spectrum imposed at the lower boundary.
M. Geller (State University of New York at Stony Brook) discussed the use of a tidal model in conjunction with recent UARS satellite observations of tidal winds to estimate dissipation rates in the upper mesosphere and lower thermosphere. The procedure started with a linear calculation of the sun-synchronous tidal fields. The results then were adjusted to fit the UARS observations. Consideration of the energy equation then determined the implied dissipation. When regarded as a second order vertical diffusion, the implied dissipation would require a peak diffusion coefficient of ~200 m2-s-1.
O. Andreassen (Norwegian Defence Research Establishment) described some very detailed calculations of the flow associated with gravity wave breaking. He considered a 2D upward-propagating gravity wave that is allowed to grow until the lapse rate becomes unstable. In a pure 2D calculation the wave will achieve considerable supersaturation before breaking. Andreassen showed, however, that if a small amount of 3D noise is added to the calculation the breaking occurs much more rapidly and involves development of very strong 3D circulations. Andreassen presented a large number of impressive flow visualizations for his simulations.
F. Lott (Laboratoire de Meteorologie Dynamique du CNRS) reviewed the treatment of subgrid-scale topography in GCMs. He then described work with a new scheme for the effects of topography now being incorporated into the ECMWF model. This scheme considers both the blocking effects of the topography on the low-level flow as well as the generation of subgrid- scale gravity waves by the flow over the topography. Tests of this scheme in the ECMWF model were promising.
J. Vanneste (Laboratoire de Meteorologie Dynamique du CNRS and University of Toronto) described his work on the nonlinear interaction among gravity waves in the presence of strong mean flow shear. He was able to show that the usual criterion for instability (any wave is unstable if it can be the highest frequency wave of a resonant triad) can be generalized to the strong shear case. Vanneste also discussed the case of very strong wave growth ("explosive interactions") that is possible in the resonant triads in mean shear.
C. McLandress (Institute for Space and Terrestrial Science, York University) began with a review of the Hines scheme and a discussion of some practical issues involved in its implementation. He noted that the predicted drag from the Hines scheme at high levels depends strongly on the low m behavior of the assumed input spectrum. McLandress then compared the performance of a simple 2D model of the zonally- averaged flow when the Lindzen, Hines, and Fritts and Lu schemes are used to computed the wave drag. He found that the Lindzen and Hines schemes produced qualitatively quite similar results and could easily account for the mean wind reversals in the extratropical lower thermosphere. The results obtained using the Fritts and Lu scheme were substantially different. McLandress then considered the same comparison but now for a model of the diurnal tide. He found that inclusion of either the Hines or Lindzen scheme results in an increase in the tidal amplitude, while the Fritts and Lu scheme acts to damp the diurnal tide.
B. Lawrence (Oxford University) discussed some simulations that were conducted with a mechanistic model based on the UK Meteorological Office stratosphere-mesosphere model. This was forced both by a realistic radiation and by an imposed planetary wave geopotential perturbation at the lower boundary (~16 km) taken from observations. The boundary forcing for the simulations discussed by Lawrence were for the winter of 1991/92 and the results were compared with observations of the middle atmospheric temperature field during that winter from the ISAMS instrument on the UARS satellite. The simulations were conducted with different parameterizations of gravity wave drag. A simple Rayleigh friction, the Fritts and Lu schemes and the Hines scheme were all tested. The Hines scheme was found to produce reasonable results when altered to allow the waves to break at lower levels.
D. Fritts (University of Colorado) began with a brief review of earlier work on using aircraft observations to characterize the relative importance of topographic, convective, and jet stream sources for the near- tropopause gravity wave field. He then went on to describe the Fritts and Lu gravity wave drag parametrization. Fritts discussed the application of this parametrization to the NCAR TIME model of the middle atmosphere and thermosphere. While many features of the mean flow simulation are quite realistic, there was a strong sensitivity to the low m part of the assumed spectrum. Fritts also noted that the inclusion of the gravity wave parameterization strongly affected the tidal amplitudes in the model.
S. Pawson (Free University of Berlin) described some simulations conducted with the Berlin troposphere-stratosphere-mesosphere GCM with different imposed Rayleigh drag distributions. Major changes were found between simulations with drag confined near the mesopause and those including drag distributed throughout the mesosphere. The differences extend into the troposphere and involve the planetary waves as well as the zonal-mean state.
B. Boville (National Center for Atmospheric Research) began with a review of the practical parametrization problem for GCMs and emphasized the need for physically-based schemes. He discussed his own experience with incorporating a version of the Lindzen scheme into the NCAR Community Climate Model middle atmosphere GCM. The scheme he adopted had 9 separate waves each with a prescribed phase speed varying from 0 to 40 m-s-1. The waves are launched upwards from the upper troposphere and the stress vector associated with each of the waves is aligned with the horizontal wind at the source level. Boville found that with judicious tuning of his input parameters he could quite closely reproduce the observed climatological seasonal cycle throughout the middle atmosphere. He needed a fairly small (less than 0.1) efficiency factor in order to produce these realistic simulations, however. Boville noted that it was interesting that the efficiency factor needed in the full GCM was essentially the same as that required in an earlier calculation he had performed with a zonally- averaged model. In the zonally-averaged model the efficiency factor was adopted to spread the region of significant wave drag. One might anticipate that - given the geographical and temporal variability of the resolved flow in the GCM - the zonal-mean gravity wave drag would naturally be smoothed out in the vertical. In practice, however, this seems not to significantly affect the efficiency factor required.
R. Garcia (National Center for Atmospheric Research) discussed high-resolution 2D calculations of the propagation and breaking of gravity waves. The waves were forced by flow over a transient deflection of the lower boundary imposed over a limited horizontal extent. This should excite a localized wavepacket, but Garcia showed that by the time the waves had reached breaking heights the dispersion had transformed the disturbance into a rather monochromatic wavetrain. The details of the wave breaking itself were found to depend strongly on the horizontal wavelength involved. For sufficiently short wavelengths the breaking involves roughly one-half of the wavelength.
E. Manzini (Max Planck Institute Hamburg) described simulations obtained with the 39-level ECHAM GCM when a version of the Hines scheme was included. Two different experiments were discussed. In both runs the parametrized wave spectrum was introduced near the 15 km level. In one case the spectrum was assumed to have an rms wind velocity of 1.75 m-s-1 everywhere at this level, while in the second case included an enhancement to this rms with a term dependent on the precipitation rate. The results obtained for the NH winter season were fairly realistic, but the SH winter polar region was still found to be too cold in both simulations.
N. McFarlane (Canadian Climate Center) discussed results obtained with the Canadian Middle Atmosphere Model when the Hines scheme is included. In this approach the waves are introduced at the ground with a globally-uniform rms wind speed of 1 m-s-1. He also examined the sensitivity of the results to the form of the input spectrum assumed and to the characteristic horizontal wavelength for the input spectrum. McFarlane found that he could obtain realistic results in the upper mesosphere and lower thermosphere as long as the imposed spectrum at low m was not too weak (for example if the spectrum varied as m2 for small m, then the mean wind reversals in the extratropical lower thermosphere are not captured in the simulation). He found that the SH winter simulation was characterized by an unrealistically cold polar region even with the inclusion of the Hines scheme. The SH winter polar simulation improved somewhat when a longer characteristic horizontal wavelength was used for the input spectrum.
W. Norton (Oxford University) discussed simulations obtained using the UGAMP middle atmosphere general circulation model with a version of the Lindzen parameterization included. The parameterization was tuned to allow the model to produce a rather realistic zonal-mean state. This required a rather small (~0.01 or less) efficiency factor for the parameterization. Norton found that when the model was run at fairly high resolution (triangular 42 truncation) and the parameterized drag was included, a strong two-day wave is produced in the summer mesosphere. This simulated wave has many of the features of the observed two-day wave, notably a dominant zonal-wave three structure and a tendency to be particularly strong for a limited period around the solstices. The gravity wave drag seems to be crucial in producing the two day wave in the simulation, and Norton noted that when the drag is included a region of negative meridional gradient of zonal mean potential vorticity is created in the summer mesosphere. This then presumably leads to the excitation of the two-day wave by baroclinic instability.
S. Webster (UK Meteorological Office) described the topographic gravity wave drag scheme now being developed for the UK Meteorological Office operational model. This parameterization considers three distinct regimes. At sufficiently large Froude number the surface drag is proportional to the surface wind and the usual saturation condition for the eddy momentum fluxes is used. For lower Froude number the flow is assumed to be in a hydraulic jump regime and the surface stress varies as the square of the surface wind, At still smaller Froude numbers the effects of low level flow blocking are also considered, and the surface stress is found to vary as the cube of the surface wind.
K. Hamilton (Geophysical Fluid Dynamics Laboratory/NOAA) described a preliminary experiment using the GFDL "SKYHI" GCM with a simple Lindzen parameterization of nonstationary gravity wave drag included. The inclusion of the drag resulted in an improved simulation, particularly in the SH winter upper mesosphere. Hamilton noted that the tuning of the parameterization to further through trial-and-error to further improve the simulation appeared to be a daunting task. He then described a numerical simulation with the SKYHI GCM with zonal-mean zonal flow constrained to agree with climatological observations. This was used to diagnose the drag needed to obtain realistic mean states. While this approach has some important limitations (notably in suppressing the day-to-day and interannual variability in the model), the results may can still be a useful guide to formulating a detailed drag parameterization. The results from this experiment suggested that in order to obtain good agreement with observations the drag must extend to down to near stratopause levels in the SH late winter and spring.
There was lively discussion of several aspects of the currently proposed parameterizations. One problem apparent in the Lindzen and Hines schemes is a tendency for the waves to break significantly only at high altitudes (near and above the mesopause). This leads to a predicted gravity wave drag that has a distribution that is centered at higher altitudes than that required by numerical simulation models in order to obtain realistic results for the global circulation. There are two obvious approaches to dealing with this. One is to impose an efficiency factor as described earlier. The other is to adjust the horizontal wavenumber spectrum assumed for the input waves. In general, a longer horizontal wavelength wave will break lower down than a shorter wavelength wave (given the same horizontal phase velocity and momentum flux). The adjustment of the input wavelengths is very straightforward for the Lindzen scheme, of course. The Hines scheme requires some specification of the horizontal spectrum of the input waves and this involves a characteristic horizontal wavelength that can be adjusted (thus scaling the whole spectrum). In practice, in order to get realistic results, modellers have been forced both to impose a fairly small efficiency factor and to consider quite long horizontal scale waves (typically at least several hundred km). The use of small efficiency factors raises the question of whether such a degree of intermittancy is plausible. The use of long horizontal scales for the (allegedly) subgrid- scale waves begs the question of why such waves are not explicitly resolved in current GCMs. Further discussion focussed on some basic observational requirements for the formulation of realistic middle atmosphere gravity wave drag parameterizations. Notable was the issue of the horizontal spectrum of waves near the tropopause (needed as input to any parameterization). There is currently a lack of detailed observations to characterize the spectrum and its relation to possible wave sources. Another important question arose concerning the very longest vertical wavelength components of the spectrum emerging into the middle atmosphere. This end of the spectrum is extremely hard to characterize from observations, but (as was seen in numerous presentations) it is very important high in the atmosphere, since the longest vertical wavelengths correspond to the highest phase speeds and vertical group velocities.
There was discussion of some practical initiatives for advancing research on the gravity wave parameterization problem. R. Vincent described plans for a project to compare the gravity wave statistics determined from radiosonde observations at as many stations as possible worldwide. At the meeting it was possible to identify at least 7 nations that have taken some regular high-resolution soundings for a long period. The project will involve the joint efforts of groups led by a "principal investigator" for each of the participating nations. More complete plans for this initiative will be described in the next issue of the Newsletter.
N. McFarlane proposed an informal project to compare the predictions of various simplified gravity wave models for a set of standard 1D cases. There was general support for this initiative and a number of individual researchers were identified as potential participants.
The possibility of a large field program to examine middle atmospheric gravity waves in relation to their sources was also discussed. C. Gardner briefly described his experience with the ALOHA-93 experiment in Hawaii and his views on how an even larger experiment might be organized. There seemed to be considerable support for the notion of an experiment to examine the middle atmospheric response to tropical convection. Two possible sites suggested were Hawaii and Darwin, Australia. There was general agreement on a number of points that would need to be considered in a detailed plan for such an experiment. The necessity of numerical 3D simulations for both planning and interpretation of the field experiment was emphasized. It was also felt that the presence of a state-of-the-art Doppler meteorological radar would be crucial for adequate observational coverage of the tropospheric convection. There was an overall concern that the scientific issues to be addressed and hypotheses to be tested be described very clearly as part of the planning procedure. The possible role of SPARC in such a field program will be discussed again in December at the First SPARC General Assembly and the SPARC SSG meeting.