MESOSPHERE INVERSION LAYERS AND STRATOSPHERE TEMPERATURE ENHANCEMENTS

Transcription

1 MESOSPHERE INVERSION LAYERS AND STRATOSPHERE TEMPERATURE ENHANCEMENTS John W. Meriwether and Andrew J. Gerrard Department of Physics and Astronomy Clemson University Clemson, South Carolina, USA Received 21 April 2003; revised 5 May 2004; accepted 14 July 2004; published 21 September [1] It has been known for at least 30 years that vertically narrow thermal layers form within the middle atmosphere. Two types of temperature enhancements, the low-latitude to midlatitude mesosphere inversion layer (MIL) and the highlatitude winter stratosphere temperature enhancement (STE), have both received much attention within the atmospheric science community because of their unexplained formation mechanisms and potential impacts on the middle-atmosphere global circulation. Numerous experimental, numerical, and theoretical studies have attempted to explain certain aspects of these respective thermal layers, but no one theory consistently and satisfactorily describes all the features observed. We present a review of the literature and explicitly propose a classification scheme based on the different formation mechanisms suspected to cause these events. For the MIL we demonstrate that there are two subtypes. The first one is tidally driven and tends to occur above 85 km. This MIL originates from large-amplitude tidal waves propagating into the mesosphere and their subsequent nonlinear interactions with gravity waves, which can often create the appearance of a double MIL separated by approximately one vertical tidal wavelength (25 km). The other subtype of MIL is formed by a climatological planetary wave dissipation mechanism that occurs at a zerowind line. The dissipation of the planetary wave tends to generate a mesoscale (1000 km) inversion layer in the range of km. These two formation mechanisms explain a host of observed characteristics, including the Citation: Geophys., 42,, doi: /2003rg reason behind the downward progression of some MILs and not others, the different climatological nature of the two forms of MIL events, and the relative scarcity of MIL observations at high latitudes. The STE is believed to be generated by an altogether different process, namely, the nonlinear interaction between the polar vortex and planetary waves/aleutian High. The induced temperatures typically peak around 40 km and often exceed 300 K, generating what appears to be a low, hot stratopause. When vertical temperature profiles are combined with synoptic analyses, one observes that the STE is the consequence of high-latitude vortex interactions creating a baroclinic atmosphere, i.e., a downward adiabatic compression induced by an ageostropic flow. We summarize the details of the relationship between this feature and sudden stratospheric warmings, as well as the potential for in situ gravity wave generation. We close with a review of currently unexplained MIL/STE features and offer new directions for future middle-atmosphere thermal layer research. INDEX TERMS: 3332 Meteorology and Atmospheric Dynamics: Mesospheric dynamics; 3334 Meteorology and Atmospheric Dynamics: Middle atmosphere dynamics (0341, 0342); 3360 Meteorology and Atmospheric Dynamics: Remote sensing; 0342 Atmospheric Composition and Structure: Middle atmosphere energy deposition; 3384 Meteorology and Atmospheric Dynamics: Waves and tides; KEYWORDS: sodium wind and temperature lidar, mesosphere inversion layer, gravity wave coupling, stratosphere temperature enhancement. Meriwether, J. W., and A. J. Gerrard (2004), Mesosphere inversion layers and stratosphere temperature enhancements, Rev. 1. INTRODUCTION [2] In tropospheric meteorology the diurnally driven turbulent atmospheric boundary layer, also known as the mixed layer and typically existing below 3 km, is capped by a very stable region separating it from the free atmosphere. This capped region is often observed as a vertically narrow (i.e., less than 1 km thick) temperature inversion, or an increase in the otherwise decreasing vertical temperature profile, that is associated with strong static stability and reduced vertical mixing [e.g., Wallace and Hobbs, 1977; Salby, 1996; Stull, 2000]. This inversion layer therefore acts like an upper lid or barrier in regard to nearsurface and atmosphere constituent exchange. There has been substantial research on how such thermal inversions contribute to the very practical problems of urban smog, mesoscale weather prediction, airport turbulence, cloud formation mechanisms, etc. [3] Recent research indicates that analogous inversion layers/temperature enhancements also exist at higher alti- Copyright 2004 by the American Geophysical Union /04/2003RG Reviews of Geophysics, 42, / of31 Paper number 2003RG000133

2 Figure 1. (a) Illustration of the temperature profile for the mesosphere and lower thermosphere regions with (solid line) and without (dotted line) typical upper and lower mesosphere inversion layers. (b) Illustration of the temperature profile for the upper stratosphere with (solid line) and without (dotted line) a stratosphere temperature enhancement. Warm and cool are in comparison to the dotted climatological profiles. tudes, particularly in the middle-atmosphere regions of the stratosphere and mesosphere. In spite of intensive research activity over the past few years the formation mechanisms for both of these remain poorly understood. We now know that there exist several different types of middle-atmosphere thermal inversions/enhancements. Though the underlying causes of several of these inversions/enhancements are well known (e.g., the stratosphere itself can be considered a huge inversion layer capping the troposphere), there are two particular types of temperature inversions/enhancements of much interest, namely, the mesosphere inversion layer (MIL) and the stratosphere temperature enhancement (STE) Brief Overview of the MIL [4] The MIL is found as a layer 10 km vertically thick within the upper mesosphere at equatorial and midlatitude regions with an amplitude of K that is superimposed upon the characteristically decreasing temperatures of the upper mesosphere. This phenomenon occurs quite often, especially in the midlatitude winter hemisphere, may last for many days, and is observed to have a broad horizontal Figure 2. (left) An early comparison of mesospheric temperature profiles obtained at Wallops Island, Virginia, during a campaign in July 1973 using three different measurement techniques: falling sphere (i.e., a sphere is dropped from the rocket payload and tracked very precisely with radar), datasonde (i.e., an in situ measurement of temperature mounted on a Viper Dart rocket that is radioed to the ground), and grenade (i.e., a series of detonations generated by explosive charges on the rocket at different altitudes and acoustically detected with an array of ground-based microphones). Each of these profiles demonstrates a relative temperature enhancement at 73 km. (right) A temporal series of temperature profiles from falling sphere and datasonde measurements illustrating the persistence of the mesosphere inversion layer (MIL) phenomenon over the nighttime hours. From Schmidlin [1976]. 2of31

3 Figure 3. (top) Nightly mean temperature profiles at Biscarrosse from 1 December to 5 December 1986 as measured by Rayleigh lidar. The error bars (±s) are indicated by the shaded area. The COSPAR International Reference Atmosphere (CIRA) 1972 climatological profile is shown for comparison (dotted lines). Perturbations with vertical wavelengths shorter than 1.5 km have been filtered. (bottom) Hourly mean temperature profiles on 1 December 1986 at Haute Provence Observatory. The adiabatic lapse rate G is represented for comparison in the layer of strong negative lapse rates. From Hauchecorne et al. [1987]. obtained by the three different techniques in which the same MIL feature was found in the temperature profile derived for each of the three experiments. [6] A classic example of a MIL event as observed over a period of 5 days with a Rayleigh lidar located at Biscarrosse, France (44 N, 1 W), is depicted in Figure 3 [Hauchecorne et al., 1987]. The top plot displays each of the 5 days ranging from 1 December to 5 December 1986 of nightly mean temperatures. The bottom plot displays hourly mean temperature profiles from 1 December 1986 as observed with a Rayleigh lidar system located at Haute Provence Observatory, France (44 N, 6 E). On inspection of these data a temperature inversion(s) is clearly present at 75 km on most all temperature profiles and exhibits changing structure over multiple days and multiple hours. In addition, the topside of the MIL is close to the atmospheric adiabatic lapse rate, indicating the potential for convective and/or dynamic instability. [7] An example of a double MIL event is displayed in Figure 4, which shows the results of simultaneous temperature and ozone measurements in a rocket experiment from White Sands, New Mexico (32 N, 106 W) [Mlynczak et al., 2001]. The appearance of a correlation between the ozone concentration peak and the upper MIL is believed to be a coincidence rather than any indication of a relationship due distribution thousands of kilometers in scale. Figure 1 depicts a schematic showing the typical structure of a MIL profile as might be observed with a Rayleigh or Na lidar system capable of measuring vertical temperature profiles. Almost always, the topside of the MIL will feature a lapse rate that is nearly as steep as the adiabatic lapse rate. [5] The MIL may appear in the mesosphere-lower thermosphere (MLT) (spanning km in altitude) region of the low-latitude and midlatitude middle atmosphere at any time of the year. Historically, rocket experiments and reentry data from U.S. space shuttles demonstrated that there often appears a region of reduced density within the nighttime mesosphere near 75 km, which would imply a band of enhanced temperatures [Schmidlin, 1976; Fritts et al., 1989, 1993]. The discovery of the thermal feature as depicted in Figure 2, which was called an inversion layer, was first reported without explanation by Schmidlin [1976]. The name inversion layer was given to the observed event because its appearance was so similar to that of the inversion layers seen in tropospheric profiles of temperature near the ground. The MIL feature was also observed in the results from five inflatable falling spheres experiments carried out on the same night at different times. Because the falling sphere experiment is known for possible artifacts in the density profile, Schmidlin [1976] demonstrated the credibility of these measurements through the intercomparison of density measurements 3of31 Figure 4. Altitude profiles of (top) temperature and (bottom) ozone concentration observed in a rocket experiment flown on 8 August 1997 at White Sands, New Mexico. The bottom plot compares ozone measurements from two different data inversion methods (solid and dashed lines) and includes the 1 standard deviation uncertainty bars on one of the measurement profiles. The mesosphere ozone layer is seen to overlap with the upper MIL feature at km. From Mlynczak et al. [2001].

4 exceeding 300 K at 45 km and lasting over the course of a week. These anomalous temperature enhancements have often been observed before the onset of the more familiar sudden stratospheric warming, which is by definition a lower stratosphere event. [10] An example of a reported STE as a function of time is depicted in Figure 5 taken from von Zahn et al. [1998]. Figure 5 presents a time-height plot of temperatures as measured by the Arctic Lidar Observatory for Middle Atmosphere Research (ALOMAR) Rayleigh/Mie/Raman lidar (69 N, 16 E) above 30 km. Below 30 km the plot is based on temperature derived from global maps generated by analyses based upon satellite and radiosonde observations. One sees that the early portion of the observations is typical of the climatology for polar winter. On 3 February, however, temperatures of over 300 K at 43 km are observed, 60 K higher than the climatological zonal mean. Over the course of 4 days the altitude region of the warming decreases, and the STE amplitude subsides in magnitude. Figure 5. Time-height section of temperatures [K] measured by the Arctic Lidar Observatory for Middle Atmosphere Research (ALOMAR) Rayleigh/Mie/Raman lidar (69 N, 16 E) from 31 January to 10 February The span of the bars on the top denotes the measurement times over the entire period. European Centre for Medium-Range Weather Forecasts analyses are used below 30 km. From von Zahn et al. [1998]. to the thermal dependence of the chemical kinetics (i.e., the changing composition of atomic oxygen (increasing) and hydrogen (decreasing) with height would produce a secondary maximum of ozone production) Brief Overview of the STE [8] The other type of temperature enhancement that is discussed in this paper, the STE (not to be confused with Stratospheric-Tropospheric Exchange of Holton et al. [1995]), develops in the high-latitude winter upper stratosphere (spanning km in altitude). Figure 1 also illustrates a typical example of an STE profile. These enhancements are sometimes called stratosphere inversion layers (SILs), but technically, they are not inversions as this latter phrase usually indicates a change of sign in the vertical temperature profile gradient. These enhancements have also been called baroclinic zones on the basis of the work presented by Fairlie et al. [1990]. In this paper, we will use STE to include all of the above nomenclature. [9] The STE was first observed via rocket experiments launched at high latitudes during winter periods. The data indicated that occasionally the stratopause will, over a day, descend while its amplitude dramatically increases in magnitude. More recently, lidar observations of the upper stratosphere have also observed such enhancements, often 1.3. Importance of MILs/STEs and Observational Difficulties [11] Understanding both MIL and STE phenomena is important to the understanding of middle-atmosphere dynamics for two primary reasons: stability and energy transfer. Addressing the first reason, at the bottom side of the thermal layers (where the temperature profile is increasing in altitude) the positive temperature gradient with increasing altitude signifies an increase in atmospheric stability and reduction of vertical mixing. In contrast, at the topside of the thermal layer the negative temperature gradient with decreasing altitude implies a reduction in atmospheric stability to the point that the atmosphere may become convectively unstable, thus possibly supporting the development of turbulence. [12] The MLT is a region of transition between a medium where the hydrodynamical fluid equations apply (below 80 km) to the region of free molecular flow (above 110 km). The physics of such a transition region is complicated enough in its own right, and any mechanism, like a MIL, that makes the medium less stable on a mesoscale level adds a significant degree of complexity to the fluid dynamics of this region. Such instability is known to contribute to the production of atmospheric phenomena including that of mesospheric radar echoes and atmospheric bores. In the upper stratosphere the very nature of STE formation is indicative of mesoscale baroclinic instability. The vertical mixing associated with such disturbed circulation can transport oxygen-containing species to ever higher altitudes, suppress the otherwise downward ageostrophic flow, and potentially weaken the global polar vortex. Understanding the birth, evolution, and dissipation of both disturbances is a major modeling challenge, partly because the problem demands the modeling of the atmosphere s threedimensional structure and also because the instability is very time-dependent. [13] Addressing the second reason, it is known that the breaking of atmospheric waves represents the primary 4of31

5 means by which energy is transferred from the lower atmosphere to the upper atmosphere. Such waves have horizontal wavelengths ranging from high wave numbers (several kilometers) to low wave numbers ( km). The MLT region is the primary region where these waves break, and the amount of energy transferred into the mesosphere is known to be comparable with the amount of energy absorbed from the Sun. It had been believed for many years that the MIL phenomenon represented a telltale marker of these breaking waves, but, in fact, recent research has shown that the MIL mechanism is far more complicated than this description implies. As applied to STE events, such topside instability can be an additional, and currently unaccounted for, in situ source of such gravity waves. [14] Several other reasons provide additional motivation for achieving an understanding of MIL and STE phenomenology. There is the question of the MIL influence upon the vertical profile distribution of airglow layer intensities for those emissions in which the chemical reaction rate constants are temperature-dependent. In addition, the study of MIL phenomenology gives insight into how tidal wave variability depends upon gravity wave and planetary wave interactions, as well as the possible effect of nonmigrating tidal forcing upon the variability of tidal wave amplitudes and phase. Meanwhile, a better understanding of the STE allows us to more precisely test our understanding of stratospheric dynamics and subsequently improve the forecasting of high-latitude springtime ozone depletion. We also note that observations of stratosphere temperature enhancements may assist in predicting the onset of sudden stratospheric warmings, which are closely related to the intensity of downward propagating annular modes described in section 5. [15] Three problems account, in part, for the slow progress in reaching an understanding of the physics of the formation of both MIL and STE phenomena. First is the complexity of realistically modeling the energy transfer processes for waves breaking in the middle atmosphere for the range of spatial scales of 300 m to 5000 km. This modeling demands a theoretical understanding of the wavebreaking process for this range of spatial scales, which, in its own right, is difficult to achieve and remains a major challenge in atmospheric fluid dynamics studies. The second problem is one of access in that the altitude regions of the stratosphere and mesosphere, which lie above the reach of the radiosonde, are difficult to study with instrumentation, whether remote sensing or in situ in design, that is capable of making the measurements of the basic atmospheric parameters of temperature, winds, and density for both regions. Without the constraints upon possible mechanisms that such measurements represent, gaining a detailed understanding of the basic physical processes that enter into the production of these temperature layers becomes rather difficult. The third problem is simply the practical difficulty in piecing together the fragmentary observations derived from rockets, satellites, and ground sensors to form a threedimensional picture of both phenomena. [16] In this paper we summarize the recent progress of this work with suggestions in regard to possible future lines of research. Our aim is to draw attention to these phenomena and to provide a framework for future studies. In section 2 we present a qualitative review of the middle-atmosphere dynamics so that the framework in which these anomalous middle-atmosphere thermal layers exist can be more clearly understood. In sections 4 and 5 we present detailed summaries of the MIL and the STE research, respectively. In section 6 we present goals for future work, and we conclude the paper with a summary of conclusions in section A CONCEPTUAL REVIEW OF THE MIDDLE ATMOSPHERE [17] In this section we present a brief review of the middle atmosphere for two primary reasons. First, we need to describe the general structure of the middle atmosphere itself to better appreciate the dynamics of these middleatmosphere inversion layers/temperature enhancements. Second, and more importantly, we want to differentiate between those inversions/enhancements already well understood theoretically and those that remain obscure, the latter being the focus of this paper. MILs/STEs can be regarded as dynamic tracers that give us an excellent tool for quantitatively testing our physics-based numerical models of the middle atmosphere, but this would depend upon the depth of theoretical understanding that research might have achieved. [18] The middle atmosphere is that portion of the Earth s atmosphere that spans from 15 to 100 km in altitude, comprising the stratosphere and mesosphere. Figure 6 presents a schematic of the globally averaged lower- and middle-atmosphere thermal and dynamical structure for solstice-like conditions based on a compilation of data from, for example, Barnett and Corney [1985] and the Garcia and Solomon [1985] model. The altitudes below 70 km are dominated by various seasonal neutral dynamics, for example, the Arctic polar vortex, planetary waves (also called Rossby waves), and the quasi-biennial oscillation, while the higher altitudes contain a complex mixture of both seasonal neutral dynamics and electromagnetic effects, for example, gravity wave breaking, mesospheric clouds, polar mesospheric summer echoes, and sporadic layers. Of particular interest in Figure 6 is how the observed behavior of the temperature profile in the winter, high-latitude, upper stratosphere and the entire global mesosphere departs rather drastically from what is predicted by radiative equilibrium alone, namely, that the mesosphere is warm during the summer and cold in the winter [e.g., Andrews et al., 1987, Figure 1.2]. Such thermal structures of the warm winter and cold summer mesopause temperatures have been observed for over 50 years [e.g., Kellog and Schilling, 1951]. [19] We note that obtaining the temperature information summarized by Figure 6 has required decades of rocket launches, satellite measurements, radiosonde releases (also called rawinsondes, RAOBS for radiosonde observations, or ballonsondes), high-flying aircraft measurements, and lidar 5of31

6 Figure 6. Schematic of the two-dimensional lower and middle atmosphere. Colors indicate relative temperatures, with red being warmer and dark blue being cooler. Ray paths of gravity waves and planetary waves are also shown. The polar vortex is on the left, extending from the upper troposphere into the upper mesosphere. In the upper right corner we see mesospheric clouds forming in the cold summer mesosphere, and in the lower left we see polar stratospheric clouds forming in the cold polar vortex core. observations, all of which provide only pieces of the whole picture. Rocket probes have been successful in measuring temperatures in the middle atmosphere since the 1950s. Such results are sparse because the means for exploring this region are expensive and the development of the instrumentation required for successful measurements generally had represented major technological challenges. The task of analyzing data from these in situ measuring techniques to accurately determine the neutral temperature is made especially difficult by the transitional nature of this region from fluid hydrodynamics to binary collisional interactions. Remote sensing techniques applied to satellite instruments generally are problematic in that these instruments measure the brightness of the source with a path integration through the limb airglow. This means that the vertical resolution is apt to be insufficient to elucidate the details of the thermal structure in the vertical direction. Furthermore, any horizontal structure with spatial scales less than 500 km would be obscured by this path integration. Consequently, for the mesosphere thermal structure the initial results collected with these instrumental techniques exhibited poor precision with limited sampling of the climatology. In view of these difficulties it is not surprising that the first atmospheric models constructed showed only a monotonically decreasing profile of temperatures from the stratopause at km to the mesopause at 100 km. [20] Until the introduction of the two remote-sensing techniques represented by the Rayleigh lidar [Hauchecorne and Chanin, 1980] and metallic wind-temperature lidar [She et al., 1990; Bills et al., 1991], our knowledge of the middle-atmosphere thermal structure had been scanty and fragmentary. This changed when the challenge of accurately measuring the temperature profile between 30 and 105 km was met with sophisticated lidar technology combined with large-diameter steerable telescopes possessing the sensitivity required to obtain the necessary measurements. Beginning with the French pioneering work represented by Hauchecorne and Chanin [1980] and Hauchecorne et al. [1987], Rayleigh lidar systems were developed to obtain accurate density profiles from 30 km to 85 km in altitude as a function of time. These high spatial-temporal resolution profiles could then be used to calculate temperatures assuming hydrostatic equilibrium. By examining individual profiles relative to the nighttime mean density profile, density fluctuation profiles can be derived to study gravity wave activity. For the mesosphere region, which lies above the altitude range of the Rayleigh lidar, the application of the resonance lidar technique to measure winds and temperatures is made possible by a layer of meteoritic debris that exists near the lower boundary of the lower thermosphere region. [21] In addition to the much improved precision of these temperature and wind measurements, perhaps a more significant contribution achieved by the use of these lidar systems was to overcome the relative inaccessibility of the stratosphere and mesosphere regions simultaneously. Balloons reach only to altitudes of 30 km. Satellites using remote-sensing techniques for temperature measurements generally make measurements only for altitudes above 200 km or below km with vertical resolution of the order of several kilometers at best. Ground-based remote-sensing techniques such as the incoherent scatter radar and optical spectroscopy (spectrograph or Fabry-Perot interferometer) observe the temperature structure only for 6of31

7 heights that are defined by the regions of good signal to noise ratios (e.g., the daytime lower E region where collisions dominate) or by the heights of the airglow layers utilized, for example, 87 km ± 5 km for the OH airglow or 97 km ± 5 km for the nm airglow layer. Thus no other means of remote sensing will sample the entire range of altitudes between the surface and the outer edge of the atmosphere as well as the lidar technique. [22] Examination of the global zonal mean wind fields computed via thermal wind considerations from these observed temperatures (overlaid on Figure 6) shows that in the winter hemisphere, there are at least two large eastward jets in the middle atmosphere. The first one is at 70 latitude with an altitude range extending from 20 to 90 km with a peak in velocity located around 40 km. The other one extends from 60 latitude/20 km altitude equatorward to 40 latitude/90 km altitude, with a peak in velocity located around 45 latitude/65 km altitude. These winds are very similar to those presented by Andrews et al. [1987] or Holton [1992]. The first jet is designated as the polar vortex jet (often referred to as the polar night jet in zonal mean studies). Eastward tropospheric jets (namely, the eddy-driven jet or polar front jet, which should not be confused with the polar night jet) and the subtropical jet are confined below 20 km and are found at low to middle latitudes. In the summer hemisphere the middle atmospheric winds are generally westward throughout the middle atmosphere, opposite in direction to those of the troposphere (with only the eddy-driven jet present). Though these winds are derived via the thermal wind equation, they seem to agree qualitatively with the limited observations of global climatological wind fields, for example, the COSPAR International Reference Atmosphere-86 (CIRA-86) or the Mass Spectrometer Incoherent Scatter-90 (MSIS-90) empirical models of Fleming et al. [1990] and Hedin [1991], respectively. [23] However, it is important to note that there have been and continue to be numerous observations of winds and temperatures that demonstrate large discrepancies from this zonal mean/climatological view over most all geographic locations and times of year. This is particularly evident during campaign-based or synoptic observations, when such measurements present only a limited snapshot of the current state of the middle atmosphere. It is only after taking many observational samples that the climatological characteristics depicted in Figure 6 become apparent. Middle atmospheric temperature inversions/enhancements are excellent examples of such synoptic discrepancies from the zonal mean/climatological middle atmosphere. [24] The study of such short-term, synoptic structures is made even more difficult by the fact that even a basic level of understanding and concurrent ability to model the aspects of the climatological global middle-atmosphere thermal structure has only been achieved within the past 20 years. Explanation of the warm winter and the cool summer mesosphere, for example, required that the modeled mesospheric temperatures be forced away from their expected radiative equilibrium values. It was then theorized that an additional dynamical driver had to be incorporated into global circulation models because the then accepted Rayleigh drag parameterizations, which represented frictional or viscous drag, failed to reproduce observed characteristics [Leovy, 1964; Schoeberl and Strobel, 1978; Holton and Wehrbein, 1980]. It was speculated that gravity waves in the mesosphere might provide the forcing needed to induce the thermally indirect circulation necessary to create the observed thermal structure [Lindzen, 1967; Hodges, 1967, 1969; Hines, 1970; Jones and Houghton, 1971; Lindzen, 1971; Jones and Houghton, 1972; Lindzen, 1973]. Indeed, it was only with the effects of gravity wave breaking in the middle atmosphere, creating a source of momentum deposition that slowed down the overly fast zonal winds, that models could begin to reproduce the observed features [Lindzen, 1981]. [25] Since most general circulation model horizontal and temporal resolutions were (and still are) too coarse to model the small-scale features of gravity waves (generally less than 100-km scales), physically crude parameterizations were created in the early 1980s to account for the effects of such gravity wave breaking [Lindzen, 1981; Weinstock, 1982; Dunkerton, 1982a, 1982b]. Further analyses found that the incorporation of such gravity wave parameterizations did indeed yield the correct qualitative thermal and dynamical structure [Matsuno, 1982; Holton, 1982, 1983; Gaertner et al., 1983; Garcia and Solomon, 1985]. Quantitative results for zonal mean/climatological values also matched well, but when applied to full threedimensional, time-dependent general circulation models, the results have tended to be highly variable, especially during seasonal transitions (e.g., equinoctial periods) or for the synoptic observations previously referred to. Since the early 1980s, there have been a number of new gravity wave parameterizations, recently reviewed by Fritts and Alexander [2003]. To date, continued experimental observations are needed to provide suitable boundary conditions that may be applied to tune the various gravity wave parameterizations. [26] However, since these modeled results are seen to be correct qualitatively, one can discuss the gravity wave-mean flow interaction from a conceptual framework. In the summer hemisphere it is believed that the eastward to westward wind reversal between the troposphere and stratosphere filters out much of the gravity wave activity primarily through critical layer absorption [Bretherton, 1966; Holton, 1982, 1983; Salby, 1996; Brasseur and Solomon, 1997]. Therefore very few topographically forced waves (which are stationary waves in the absence of transience or nonlinearity) can penetrate through the wind reversal into these higher altitudes (in fact, many traveling waves discussed next are also filtered out because of this wind reversal). As such, most of the waves observed in the summer stratosphere and mesosphere are thought to be generated from convective processes, geostrophic adjustment processes, barotropic or baroclinic instabilities, etc. (or some combination thereof ), all of which generate varying spectra of traveling gravity waves in the lower atmosphere 7of31

8 [Holton, 1992; Alexander and Pfister, 1995; Alexander et al., 1995; Alexander, 1996; Holton and Alexander, 2000]. The resultant wave spectrum at around 50 km in the summer hemisphere tends to consist of large vertical wavelengths and eastward horizontal phase speeds. It is suspected that the eastward waves can propagate to relatively high altitudes because the longer vertical wavelengths imply spectral saturation would take place at higher altitudes (85 km), where the waves would break and distribute their energy to smaller scales, eventually ending up as small-scale turbulence [Brasseur and Solomon, 1997]. Such body forcing provided by the wave-induced momentum flux divergence acts against the westward zonal wind, slowing down the winds. Such forcing on the anticyclonic summer winds and the resultant upward and outward circulation is often referred to as the emptying of the global high-pressure cell. This forcing therefore reduces the summer hemisphere mass column and sets up an upward and equatorward meridional circulation (which is due to mass conservation arguments on a sphere). The upward motion adiabatically cools the summer mesosphere (predominantly the high-latitude summer mesosphere), eventually allowing for the saturation vapor pressure to drop in the polar summer mesosphere and ultimately allowing mesospheric clouds to form. The whole concept is partially expressed as the principle of downward control [Haynes et al., 1991; Garcia and Boville, 1994; Egger, 1996; Haynes et al., 1996]. [27] In the winter hemisphere the winds in the troposphere and middle atmosphere are both eastward, allowing a wider spectrum of eastward and westward gravity waves with both traveling and stationary components. The eastward propagating gravity waves with shorter vertical wavelengths tend to break at lower altitudes (65 km) because of convective or shear instabilities induced by critical levels. The westward propagating gravity waves with longer vertical wavelengths tend to break at higher altitudes (75 km). In addition, the potential for nonlinear wave-wave interaction also increases, which can further induce gravity wave breaking at lower altitudes. This combined wave forcing slows down the eastward winds, reducing the forcing on the cyclonic winter winds and the resultant filling of the global low-pressure cell. This forcing therefore fills the winter mass column and sets up a downward and poleward meridional circulation (which is again due to mass conservation arguments). Ultimately, the downward motion adiabatically warms the winter mesosphere and high-latitude winter stratopause [Hitchman et al., 1989; Garcia and Boville, 1994]. [28] It is important to note that planetary waves are also considered to be an important driving mechanism in the middle atmosphere, particularly within the stratosphere. However, such waves are suspected to not significantly affect the mesospheric thermal structure for at least two reasons. First, planetary waves, like stationary gravity waves, were found to be strongly filtered in the summer hemisphere, and they cannot deposit a lot of energy into the higher altitudes. Though observations have noted planetary wave effects at these higher altitudes, the associated forcing is relatively weak. Second, in the winter hemisphere these waves were found to be ducted equatorward in the stratosphere, and therefore they could not reach the winter mesosphere to break there [Holton, 1983]. Since these waves could not propagate upward, there was no reduction in the mesosphere winds nor any change in the thermal structure generated. [29] All of this is not to say that planetary waves have no influence in the middle atmosphere, quite the contrary, as such waves transport heat and momentum in the stratosphere to ever higher latitudes. Planetary waves also play a significant role in the high-latitude winter stratosphere when interaction with the polar vortex can lead to subtle synopticscale activity throughout the entire middle atmosphere and can potentially induce stratospheric warmings. They can also cause the generation of in situ planetary wave modes, like the 2-day wave observed in the mesosphere, whose importance increases at higher altitudes in the upper atmosphere. Furthermore, planetary waves can significantly modulate gravity waves passing through them, acting as a variable filter of gravity wave transmission. [30] Meanwhile, the stratosphere is dominantly driven by the thermally direct absorption of solar UV radiation and subsequent collisional energy exchange with the molecular atmosphere (i.e., thermally direct) and to a lesser extent by a northward planetary wave heat flux and gravity wave momentum deposition leading toward a poleward circulation in the high-latitude winter (i.e., thermally indirect circulations). That is, in the summer stratosphere and low- to middle-latitude winter stratosphere, ozone absorbs an incoming UV photon making O 3 + hv =O 2 + O (and to a lesser extent O 2 + hv =O+O). This reaction then allows O + O 2 + M = O 3 + M to proceed. It is M here that transfers the excess energy to the surrounding atmosphere, creating the warm stratopause. However, one notes from Figure 6 that there is a pronounced warming (relative to that expected from radiative balance) in the high-latitude upper stratosphere and lower mesosphere. Since this region is not exposed to sunlight during the winter periods, the classical reasoning of ozone absorption of UV wavelengths for heating of the stratopause would obviously not apply. [31] This warm high-latitude winter stratopause feature was first observed by Barnett [1974] in Selective Chopper Radiometer data on the Nimbus IV satellite. The characteristics of the feature were first modeled in a zonal mean model (using a Lindzen [1981] gravity wave parameterization) by Holton [1983], who noted that gravity wave forcing was a primary contributor to the stratopause and mesosphere warming. Holton [1983] also noted that planetary wave influence was relatively negligible. Kanzawa [1989] later noted that the Southern Hemisphere winter stratopause was noticeably warmer than the Arctic stratopause and that a satisfactory explanation for the difference was lacking in the literature. Hitchman et al. [1989] authored the seminal paper on the warm winter stratopause feature and coined the phrase separated stratopause (separated from the strato- 8of31

9 pause of midlatitudes). Their model also indicated that gravity waves propagated through the strong polar night jet and then broke at mesospheric altitudes. This wave breaking acted in much the same way that the global-scale gravity wave breaking affected the global circulation noted above, specifically, that the breaking waves induced a drag on the zonal flow, slowing down (and closing) the polar night jet, which also forced air poleward into the vortex core (or the region within the vortex jet). Again, by continuity considerations on a sphere the air was compressed downward and warmed adiabatically, heating the lower mesosphere and upper stratopause very much like the downward control arguments noted above. [32] Evidence of increased gravity wave activity within the vortex jet came with the satellite observations of Wu and Waters [1996a, 1996b] showing enhanced gravity wave activity in the 10- to 100-km vertical wavelength range in the vortex jet. Later lidar observations by Whiteway et al. [1997] also showed greater activity in the 2- to 12-km vertical wavelength range of the gravity wave spectrum in the vortex jet compared to the vortex core/outside the vortex altogether. They also noted significantly more activity in the longer vertical wavelength portion of the spectrum (as one would expect considering the vertical wavelength shift of gravity waves in a high background flow). Both sets of observations explained the enhanced gravity wave activity by noting that in the highly directed eastward flow of the vortex jet, there is less potential for critical layer interaction for stationary gravity waves as well as less chance for wave saturation. This theory was further verified in ray-tracing modeling done by Alexander [1998]. This work simulated a spectrum of gravity waves that were allowed to propagate through the tropospheric/stratospheric winds. The resultant variances, once filtered by an observational window, closely resembled those observed in the Wu and Waters [1996a, 1996b] study. [33] Further lidar observations by Duck et al. [1998, 2000a, 2000b] and Gerrard et al. [2002] highlighted the thermal structure of the polar vortex throughout the entire middle atmosphere. They found that in the vortex core, there was a warm upper stratosphere and a cold lower stratosphere. As the lidar observatory position changed from a point inside the vortex core to a point within the vortex jet (and eventually outside of the vortex altogether), temperatures in the upper stratosphere became cooler and temperatures in the lower stratosphere were warmer. These observations and trends are apparent in Figure 6. When the station was inside high-pressure cells (e.g., the Aleutian High [Harvey and Hitchman, 1996]), they found a cold upper stratosphere and warm lower stratosphere, opposite to that of the polar vortex. Duck et al. [1998] also showed evidence for how the temporal evolution of the vortex core temperatures increased throughout the winter season as gravity wave activity in the vortex jet increased. [34] In early winter the Arctic troposphere and lower stratosphere start to radiatively relax to very cold temperatures. The resulting meridional pressure gradient sets up an eastward jet in the lower stratosphere that increases in speed with altitude. Since both the tropospheric and stratospheric winds are eastward and in the same direction, gravity and planetary waves are allowed to propagate to higher altitudes and grow in amplitude. Planetary waves soon encounter the fast jets in the stratosphere and are ducted equatorward (though strong planetary waves can still interact with the vortex structure, inducing subtle to not-so-subtle changes in the vortex location and potentially dramatic stratospheric warmings). Gravity waves, however, propagate upward and break, inducing a flow of air into the vortex core. The air column in the vortex core is compressed and heated adiabatically, thus closing the vortex jet. [35] This conceptual model is the basis for the polar vortex structure [O Neill and Pope, 1988]. This feature is easily observed in stratospheric synoptic maps and plays a large role in the high-latitude circulation. The polar vortex is often noted to interact with other pressure cells (like its interaction with the Aleutian High) and planetary waves. It is during these wave-vorton interactions that a number of things can happen in the high-latitude middle atmosphere, depending on the strength of the forcing involved with the overall interaction. For example, weak interactions are fairly typical in all high-latitude regimes (though perhaps a bit more typical in the Northern Hemisphere because of greater planetary wave activity) and lead to the subtle meandering characteristics of the polar vortex. These smaller movements, however, can a play a large role in synoptic-scale activity over a specific geographic location [Gerrard et al., 2002]. Medium strength interactions usually involve the breaking of planetary waves on the vortex edge and related peeling of high potential vorticity air off the polar vortex [McIntyre and Palmer, 1983, 1984]. During these interactions the polar vortex is often forced off the pole but not so much that a sudden stratospheric warming (considered a strong interaction below) occurs. Air locked in the polar vortex core, which is usually isolated from lower latitudes by the vortex jet, can then mix with midlatitude air and vice versa. These interactions seem to occur less often than subtle meandering movements, and it is oftentimes difficult to distinguish the two without computation of quasi-geostrophic potential vorticity. [36] Strong interactions can lead to dramatic sudden stratospheric warmings. Reviews of sudden stratospheric warmings are given by Labitzke [1981], Andrews et al. [1987], Schoeberl and Hartmann [1991], Labitzke and van Loon [1999], Liu and Roble [2002], and references therein. First noted in 1952 from radiosonde measurements of the lower stratosphere made by R. Scherhag [Labitzke and van Loon, 1999], these events indicated a sudden (from a few days to a couple of weeks) warming (up to K) below the 10-mbar level when analyzed using zonal averages [Andrews et al., 1987]. Warmings are defined as major if at 10 mbar or below the zonal mean temperature increases poleward from 60 latitude and the zonal mean zonal wind reverses; warmings are defined as minor if at 10 mbar or below only the zonal mean temperature increases poleward from 60 latitude. In 9of31

10 addition, there are Canadian warmings, which are an anomalous strengthening of the Aleutian High that looks like a minor warming [Labitzke, 1981], final warmings, which are a minor or major warming that occurs near late winter/early spring and prevents the usual wintertime conditions from returning, and precursor events, which displace or weaken the polar vortex, making it more susceptible to a sudden stratospheric warming. We note that major and minor warmings are also defined by World Meteorological Organization definitions. Major warming events were first studied in the zonal mean mechanistic model by Matsuno [1971], where the warming in the lower stratosphere was attributed to planetary wave breaking that forced warmer air toward the poles. Later, O Neill and Pope [1988] considered the sudden stratospheric warming event as a vorton-vorton interaction, thus giving a synoptic viewpoint to the event and challenging the zonal mean interpretation of planetary wave breaking. [37] Furthermore, mesospheric coolings of K have been associated with major stratospheric warmings, as first illustrated in the modeling study by Matsuno [1971]. Observational evidence of mesospheric cooling during stratospheric warmings has been provided by Labitzke [1972], Walterscheid et al. [2000], and Hernandez [2003], for example. The mesospheric temperature changes are tied to the mesosphere-stratosphere circulation, and measurements in the mesosphere have observed wind reversals from the normal eastward to westward winds during minor and major warmings [Labitzke and van Loon, 1999; Hoffmann et al., 2002; Liu and Roble, 2002]. [38] For completeness, throughout the summer hemisphere the modeled wind fields are close to those observed, and the change in wind direction at 15 km between the troposphere and lower stratosphere filters out most planetary wave influences and a large portion of the gravity wave spectrum. Therefore this region is generally noted for very little daily and monthly variability because of less atmospheric wave activity, especially below 70 km. Above this altitude the major sources of daily variability are due to upwardly propagating atmospheric tides and gravity waves with relatively large vertical wavelengths (i.e., the small portion of the tropospheric gravity wave spectrum that was not removed at lower altitudes). [39] More detailed/analytical reviews of the middle atmospheric composition, thermal and dynamical structure, and associated wave activity may be found in the texts of Andrews et al. [1987], Holton [1992], Salby [1996], and Brasseur and Solomon [1997], with more recent material presented by Holton et al. [1995], Hamilton [1996], McLandress [1998], Hamilton [1999], Holton and Alexander [2000], and Fritts and Alexander [2003]. 3. CLASSIFICATION OF THERMAL INVERSION LAYERS/TEMPERATURE ENHANCEMENTS [40] In addition to providing background material, section 2 helps us to identify and classify thermal inversion layers/temperature enhancements. We identify middle-atmosphere thermal inversion/enhancement layers of interest to this paper as those that are due to a nonlinear and/or dissipative mechanisms that create a vertically narrow layer of enhanced temperatures that is at least 10 K greater than the background profile. Such inversions/enhancements are located between 15 and 100 km and are individually short-lived, lasting on the order of a week or less. [41] Examples of apparent inversions/enhancements that are excluded by this criteria include (1) the radiatively driven stratosphere, (2) the dynamically driven highlatitude winter stratosphere/stable polar vortex configuration, (3) temporary lower stratosphere or troposphere temperature enhancements that are observed during or after an officially defined sudden stratospheric warming, (4) an apparent inversion layer due to the phase of a propagating atmospheric wave, particularly gravity waves or tides, and (5) an apparent inversion layer due to the natural tilt of the polar vortex/aleutian High system. The fourth exclusion prohibits all but the most intense rocket observations, as temporal coverage of any possible phase progression is generally not available unless consecutive rockets are launched. The fifth exclusion requires substantial evidence in the reporting of high-latitude MILs, as one cannot determine if a thermal enhancement is due to a MIL or due to vortex tilt based only on a temperature profile. 4. MIL OBSERVATIONS AND MODELING [42] Owing to poor instrumental sampling of the mesosphere and lower thermosphere region the MIL phenomenon was originally identified as a single layer near 70 km, typically km thick, that is hotter than the atmosphere above or below by K. In recent years, however, more complete measurements of the temperature structure between 30 and 105 km have indicated that multiple layers are more generally observed rather than a single layer near 70 km. It is believed that there are at least two physically different formation mechanisms responsible for MILs: one that is essentially the result of nonlinear gravity wave-tidal interactions and one that is attributed to dissipating planetary waves. These mechanisms are depicted in Figure 7. Though other formation mechanisms may certainly exist, for example, chemical heating as discussed by Meriwether and Mlynczak [1995] and States and Gardner [2000] or gravity wave breaking as discussed by Hauchecorne et al. [1987] and Whiteway et al. [1995], these processes are still under study. Furthermore, we note that the overall effects of these other potential MIL-generating mechanisms seem to be insignificant in the lower mesosphere, i.e., only a few degrees in amplitude. [43] Generally, the MIL generated by gravity wave-tidal interactions occurs at higher altitudes between 85 and 100 km, where tidal amplitudes are much larger, and has often been called the upper MIL. Meanwhile, the MIL generated by planetary waves is believed to generally occur at lower altitudes between 65 and 80 km. At these altitudes, 10 of 31

11 Figure 7. Schematic showing possible mesosphere inversion layer (MIL) formation mechanisms. Solid thin line represents relative zonal winds for equinox conditions (E), dashed thin line represents zonal winds for Northern Hemisphere winter solstice (W), and dotted thin line represents zonal winds for Northern Hemisphere summer solstice (S), based on values from Roble [2000]. (a) Gravity waves (indicated by thin phase lines with a thick arrow denoting the upward group velocity) reach a critical level via interaction with the background flow and/or tides. This initiates the onset of instability. This scenario seems to be possible year-round, with gravity wave activity being spectrally filtered by the seasonally varying lower atmospheric winds. (b) Planetary waves (indicated by westward tilting gray/white anomalies) can reach a zero-wind line in the middle mesospheric during equinox and winter solstice conditions, which, in turn, causes wave dissipation, thus creating a lower-altitude MIL. This scenario seems to correspond only to equinox and winter solstice conditions. From Brown et al. [2004]. planetary waves will begin to break as the phase speed of the wave matches that of the zonal wind flow, thus forming a critical level. The dissipation of these breaking planetary waves forms what has been called the lower MIL. However, we note that the height dependence of these two mechanisms changes as a function of season and geographic location. At times, if the diurnal tidal wave is sufficiently vigorous and the gravity wave field is particularly active, the upper MIL mechanism could be observed at lower MIL altitudes. In addition, it has also become apparent that there could be several gravity wave, tidally induced MIL events occurring simultaneously as in the work of Dao et al. [1995] or that longer-period gravity waves could act in lieu of tides as in the work of Liu and Meriwether [2004]. This complicated mixture of MIL formation mechanisms, observed structures, and seasonal/ geographical dependencies has made achieving an understanding of MIL formation and development rather difficult. [44] A classic example of a multilayered upper MIL event is depicted in Figure 8, taken from Hawaiian observations obtained in a series of simultaneous dual lidar observations that combined Rayleigh and sodium resonance lidar temperature measurements [Dao et al., 1995]. In Figure 8 we can see two inversion layers, one at 90 km and one at 65 km, that are both slowly progressing downward as a function of time. Dao et al. determined that the MIL amplitudes for the two events observed near 65 km and 90 km were nearly a factor of 5 greater than the tidal wave predictions by the Global Scale Wave Model [Hagan et al., 1995, 1997, 1999]. Such observations make the possible connection between the process that enhances the MIL amplitude and the diurnal tidal wave structure more evident. However, they also make the distinction 11 of 31 Figure 8. Combined Rayleigh and sodium temperature lidar observations averaged over 30 min each obtained at Mount Haleakala during the Hawaiian 1993 lidar campaign. Note the maxima near 90 and 65 km and the subsequent downward progression. The Rayleigh observations from 30 to the tie point indicated by a plus symbol at 85 km were obtained with the Air Force Phillips Laboratory Air Force Maui Optical Site 1.6-m telescope and a 3-W neodymium:yttrium/aluminum/garnet transmitter. The sodium resonance temperature measurements were obtained with the University of Illinois sodium 0.5-W transmitter and the Air Force 0.8-m beam director telescope. From Dao et al. [1995].

12 between tidally driven upper MIL events and planetary wave driven lower MIL events more confusing in the literature. [45] In fact, we argue that the only real way to separate the different MIL formation mechanisms is via timeresolved temperature measurements covering the full range from 30 to 105 km. However, this currently requires the operation of a dual Rayleigh/Na resonance temperature lidar system, which owing to expense and complexity is not easy to achieve. The Rayleigh lidar technique is generally limited by the low atmospheric density to altitudes below 90 km. Thus measurements and past observations with this particular instrument generally observe only the lower MIL event. The sodium resonance temperature lidar system, now in regular use by lidar groups located in Hawaii (Haleakala, N, W), Norway (ALOMAR, 69.3 N, 16.0 E), and Colorado (Fort Collins, 41 N, 105 W), measures the thermal structure for altitudes between 80 and 105 km. Thus this instrument has often observed the higherlying MIL that is found within this altitude range, i.e., the upper MIL event. [46] Below, in sections 4.1 and 4.2, we discuss further the dominant characteristics of each of the two types of MILs with the aim of clarifying the observations and discussions found in the literature. For example, the mesospheric temperature inversions observed by Leblanc and Hauchecorne [1997] and Meriwether et al. [1994] and modeled by Salby et al. [2002] and Sassi et al. [2002] are all considered herein as lower MILs because of their altitudes of formation. However, the observations and modeling of Huang et al. [1998, 2002] and Liu and Meriwether [2004] are considered herein as an upper MIL for similar reasons (though Liu and Meriwether deal with the modeling of short-period gravity wave interactions with long-period gravity waves instead of tides). We note that this qualifying information (i.e., a distinct classification of upper or lower MIL) is lacking in all of these studies, but it is vitally important in establishing a comprehensive theory of MIL formation mechanisms. Meriwether and Gardner [2000] were the first to attempt to distinguish these events, and we note for completeness that many of the MILs discussed in that text discuss the upper MIL Tidally Driven MIL or the Upper MIL [47] For future reference, to illustrate what a sophisticated general circulation model would produce for a typical temperature profile of the upper mesosphere and mesopause region, in Figure 9 we show the results of the National Center for Atmospheric Research (NCAR) thermosphereionosphere-mesosphere electrodynamics general circulation model (TIME-GCM) (discussed by Roble and Ridley [1994] and Roble [2000]) plots of the background temperature and wind profiles together with the tidal temperature amplitudes and tidal wave winds for the location of 40 N [Liu and Meriwether, 2004] for the evening of 17 November These model predictions were made more realistic because these calculations, which are similar to those described by Walterscheid et al. [2000], used the National Centers for Environmental Prediction (NCEP) measurements to characterize the wind and temperature fields below 30 km for the lower boundary of the model calculations. One sees that the thermal enhancement that may be seen for the altitude of 90 km is 5 K and arises largely from the diurnal tidal variation. The weak amplitude of this MIL compared with what is normally observed (30 50 K) illustrates that the MIL production mechanism is not completely represented by the NCAR general circulation modeling predictions that are based upon classical tidal theory combined with a gravity wave parameterization scheme necessary for the modeling of tidal dissipation. [48] It therefore appears that upper MIL amplitudes, for example, those shown in section 1, are considerably greater than tidal model predictions attributed to migrating tidal waves alone. The reason for this increase remains a mystery. Both the NCAR general circulation model and the Global Scale Wave Model predict similar amplitudes. Given that the current state of tidal theory is reasonably representative of atmospheric behavior as observed in the lower thermosphere [Burrage et al., 1995; Hagan et al., 1995; Meyer, 1999], the implication is that there is present within the upper MIL event a dynamical amplification process that produces the large MIL amplitudes that are often observed. Because the peak altitudes of MIL events will often show a downward phase progression typical of the diurnal tide, Meriwether et al. [1998] suggested that the enhancement of the MIL amplitude to temperatures considerably elevated above that represented by the tidal structure alone may be indicative of dynamical forcing. A possible source for this was suggested to be gravity waves interacting with tidal wave activity. [49] The summary and examination of various proposed mechanisms for the production of the upper MIL by Meriwether and Gardner [2000] have concluded that the most promising formation mechanism is a nonlinear interaction of gravity waves with tidal waves (Figure 7), the impacts of which are not taken into account in Figure 9. This theory is supported for two primary reasons. First, it was found that when 24-hour observations of the MIL structure were averaged together, the MIL was no longer apparent. This point was discussed in detail by Meriwether and Gardner [2000], where Na resonance lidar observations and subsequent analyses [States and Gardner, 2000] indicated that the upper MIL did indeed wash out. Such would be expected given the 24-hour period and lower harmonics of the atmospheric tide (or long-period inertiogravity waves). [50] The second supporting reason for a gravity wavetidal wave interaction was presented in the modeling analysis by Liu et al. [2000]. This wave-wave coupling idea was first studied by Liu and Hagan [1998] with encouraging results from their nonlinear two-dimensional Navier-Stokes numerical simulations. An example is plotted in Figure 10, which shows the downward propagation of a MIL event generated by the modeling of the breaking of a gravity wave with a 50-km horizontal wavelength and a horizontal phase speed of 30 m s 1 for two latitudes (equator and 39 N) and 12 of 31

13 Figure 9. National Center for Atmospheric Research (NCAR) general circulation model predictions of mesospheric and lower thermosphere zonal wind speeds and temperatures as well as wind speeds and temperatures for the diurnal and semidiurnal tidal components. These calculations utilized results from lower stratosphere global temperature measurements to characterize the stratosphere boundary conditions. From Liu and Meriwether [2004]. two seasons (equinox and winter). One sees in Figure 10 that the model is successful in reproducing the experimental observation that the MIL amplitude is larger at the equator for the equinox and at midlatitudes during winter. Liu et al. [2000] demonstrated that large mean state changes in the thermal structure may indeed be introduced by the process of gravity wave breaking provided that the mean stability of the background atmosphere has been decreased by the propagation of the diurnal tidal wave through the region. A total heating rate of ±10 K h 1 is produced with the dissipative heating rate itself being insignificant when compared with the rate of heating due to turbulent diffusion in the lower part of the breaking wave. In the upper part of the wave-breaking event extending over a large fraction of the vertical wavelength of the wave, cooling is the dominant effect caused by wave advection and also by the turbulent diffusion process. [51] This separation of the heating and cooling functions induced by the dynamics of the wave-breaking process 13 of 31 decreases the lapse rate further, making the medium more unstable. The consequence is the continued development of a shear region at a lower altitude. A critical layer with a Richardson number less than the 0.25 value that is the threshold for the possible development of turbulence would eventually emerge from this nonlinear interaction. The formation of this critical layer can be expected to shift the wave-breaking region to lower altitudes at a rate that is consistent with the diurnal tidal vertical phase speed producing the downward progression of observed MIL events. When this layer reaches an altitude (near the stratopause) at which the tidal wave no longer has much effect upon the stability of the upper mesosphere, then the phase speed would be diminished or the downward motion terminated. This work is based upon a two-dimensional numerical simulation of the nonlinear interactions of a monochromatic gravity wave with the diurnal tidal wave within the mesosphere and lower thermosphere region. It assumes a wave turbulence model that has been successful in geophysical

14 Figure 10. Model calculations at half-hour intervals from 1230 to 2200 LT of temperature profiles for the upper mesosphere for MIL events at the equator and at 39 N for the equinox and winter solstice. The solid curves are profiles including zonal mean temperature, tidal perturbation, and the local mean changes due to gravity wave breaking. The dotted curves are the zonal mean temperature profiles. From Liu et al. [2000]. fluid problems [Mellor and Yamada, 1982]. The details of these nonlinear wave-wave interactions still remain to be further developed, clarified, and understood. A more complete simulation will require the development of a three-dimensional simulation model, the addition of the semidiurnal tidal mode to the diurnal tidal mode, and the inclusion of multiple gravity waves that are mesoscale in scope to achieve a more physically realistic simulation. While Figure 10 does show some evidence of a double MIL event separated by km, the amplitude of the lower event is much weaker than the amplitude of the upper one, which is contrary to what the observations show, namely, that the lower MIL amplitudes can be as large as K. If this modeling result is correct, then the implication is that there is an additional MIL mechanism that explains the events observed for the lower MIL. [52] Other theoretical studies have also found that the wave-wave interaction between gravity waves and tidal waves can be important. Meyer [1999] examined the question of the seasonal tidal wave variability (minima at solstices and maxima at equinoxes) by utilizing the Global Scale Wave Model combined with a hybrid parameterization scheme. This analysis incorporated a new approach to the description of the small-scale interactions of gravity waves with the global scope of tidal waves. The hybrid parameterization procedure developed in the Meyer [1999] analysis unified the parameterization scheme introduced by 14 of 31 the Lindzen [1981] concept of wave breaking at a saturation height (where the vertical temperature lapse rate of the wave and mean temperature fields becomes adiabatic) with the Matsuno [1982] concept of eddy viscosity that is based upon the attenuation of gravity waves, which, in turn, is considered to determine the vertical distribution of eddy diffusion. This viscous dissipation of gravity waves produces deposition of momentum, and the resulting profile of eddy diffusivity can be treated as a parameter that determines the transmissivity of the atmosphere to the passage of gravity waves. The Lindzen [1981] wave-breaking parameterization produces forcing of the medium that dominates in the middle and upper mesosphere, while the Matsuno [1982] formulation of the eddy diffusivity is important in the lower thermosphere where the phase speed of the gravity wave approaches the mean wind speed. The Meyer [1999] analysis found that the semiannual variations, enhanced two times relative to predicted model calculations, in tidal amplitudes depend upon the degree of atmospheric filtering imposed upon the flux of gravity waves by the seasonal variations of background winds. Annual variations in the eddy diffusion and diurnal momentum flux divergence arise as a result of this filtering action, which, in turn, will modify tidal amplitudes by dampening the rate of growth with altitude. [53] An example of the development of an upper MIL event is the set of the University of Illinois sodium lidar

15 Figure 11. Simultaneous (left) wind and (right) temperature lidar measurements illustrating the development of a MIL event observed at Mount Haleakala for the night of 21 October 1993 during the Hawaiian 1993 lidar campaign and (middle) intermediate plots showing the Richardson number computed from these measurements. Shown are successive 30-min averages between 0830 UT and 1030 UT. From Huang et al. [2002]. measurements obtained during the 1993 Hawaiian campaign illustrated in Figure 11 selected from the work reported by Huang et al. [1998, 2002], which shows the sequential development of a MIL event from an initial amplitude that was weak (<15 K) to the magnitude of 15 of K. As Figure 11 illustrates, the progression of the increase in the MIL amplitude was accompanied by the simultaneous development of a vertical shear in the observed horizontal winds. (The magnitude of the horizontal winds in Figure 11 is plotted for the direction of the

16 Figure 12. Starfire Optical Range measurements of temperatures, winds, and stability profiles for the night of 18 February 1999 in New Mexico. The color coding for the stability plot is as follows: Green squares represent regions that are convectively unstable (N 2 < 0), red squares represent regions of dynamical instability (0 < Ri < 1/4), and the black squares are regions of stability. The temperature and wind profiles were provided by A. Liu (personal communication, 2004). The stability profile is from Zhao et al. [2003], reprinted with permission from Elsevier. See Liu et al. [2004] for additional examples of instability regions. propagation of a gravity wave to the NNW direction.) Examination of the variation of the Richardson number, Ri, during this interval showed this parameter decreased to the critical level at which the development of turbulence would be favored prior to the development of the MIL event. These results support the conclusion that the continued development of this MIL event was accompanied by simultaneous increases in the MIL temperature and wind shear for the upper portion of the MIL layer. Similar events observed in this campaign featuring a large MIL amplitude (40 K) accompanied by high wind shears and the development of a sporadic sodium layer were reported by Gardner et al. [1995] for observations on two other nights. Hence the results reported by Huang et al. [1998, 2002] were not for an event that was particularly rare in occurrence. Huang et al. [2002] concluded that the production of the MIL came about as the result of the mesosphere and lower thermosphere region becoming unstable partly because of the overlying tidal wave activity that drove the development of a critical layer that blocked the propagation of a gravity wave progressing to lower altitudes. Observations were not taken long enough to demonstrate that this inversion could average out in 24-hour data sets. [54] Our understanding of the relationship between the appearance of a region of shear and turbulence in the vertical profile of horizontal winds and the development of the MIL, first noted by Thomas et al. [1996], is enhanced further by recent results [Zhao et al., 2003] highlighting sodium lidar temperature and wind measurements that were obtained at the Starfire Optical Range lidar facility in New Mexico (35.1 N, W). Zhao et al. examined the structure and seasonal variations of the convective and shear instabilities in the mesopause region. The results showed that the mesopause is most stable in the summer when the buoyancy frequency is large and the total vertical shear in the horizontal winds is weak. Furthermore, in the absence of gravity waves and tidal temperature and wind 16 of 31 variations their analysis demonstrated that the mesosphere and lower thermosphere region is stable for all seasons. In the winter the probability for the development of instability increases by nearly a factor of 2 compared with summer. Figure 12 illustrates an example of these results for a winter night and shows the relationship between the meridional winds, temperatures, and atmospheric stability. The difference in the semidiurnal tidal wave temperature between the crest and valley is K, which is a factor of 20 larger than the NCAR general circulation semidiurnal amplitude model predictions plotted in Figure 9. Here, as was also observed in the Hawaiian 1993 campaign event, one sees that the MIL located near 90 km at 1000 UT is accompanied by a strong shear in the meridional winds at altitudes that overlap with the altitudes of the topside region of the MIL event. It is no surprise that the lapse rate is nearly adiabatic. Moreover, the altitudes near 85 km that coincide with a region of increased instability overlap with the bottom side of the MIL region. These results are consistent with the findings reported by Thomas et al. [1996] in which two layers of VHF mesosphere daytime echoes were found to coincide with the observed location of a MIL observed in the nighttime Rayleigh lidar measurements. They had suggested that the topside MIL region is characterized by dynamic instability and the bottom side is characterized by static instability, an inference that is well supported by the Starfire results. As the MIL progresses through the mesopause region, the color coding of the N 2 parameter in Figure 12 illustrates that the development of static instability (N 2 < 0) is preceded by a period of dynamic instability (Ri < 1/4). As these results show, the shear zone extends over an altitude range 3 5 km thick, and consequently, the excellent height resolution of this lidar facility was critical to the acquisition of these results. These results are consistent with the findings published recently by Larsen [2000] in which an explanation for the development of quasiperiodic (QP) radar echoes is presented that is based

17 Figure 13. Simultaneous comparisons of temperature profiles for ground-based Rayleigh lidar (at Haute Provence Observatory) with satellite Improved Stratospheric and Mesospheric Sounder and Halogen Occultation Experiment (HALOE) results for three separate UARS overflights. From Leblanc and Hauchecorne [1997]. upon the development of Kelvin-Helmholtz billows from a shear region that is 3 5 km in depth Planetary Wave Driven MIL: The Lower MIL [55] In addition to the tidally driven MIL discussed in section 4.1, it has become clear that there exists in all seasons a characteristic climatological feature of the mesosphere that we now identify as the lower MIL. The results from the Solar Mesospheric Explorer monthly mean profiles [Clancy and Rusch, 1989] were the first to show the consistent appearance of the lower MIL as a climatological feature of the mesosphere thermal structure. These measurements were limited by the technique to altitudes below 90 km, and so the dual MIL phenomenology was not detected in these satellite measurements. In general, it had been true that satellite temperature measurements above 80 km are difficult to obtain with good accuracy. The results regarding the lower MIL were confirmed with the nighttime Rayleigh lidar measurements at the Haute Provence Observatory [Hauchecorne and Chanin, 1980; Hauchecorne et al., 17 of ; Leblanc and Hauchecorne, 1997] and by similar Rayleigh lidar measurements obtained over Aberystwyth, Wales (52.4 N, 4.1 W) [Mitchell et al., 1990, 1991; Thomas et al., 1996]. Low-latitude observations have been presented from Gadanki, India (13.5 N, 79.2 E) [Kumar et al., 2001]. [56] These studies characterized the midlatitude climatology of the lower MIL, showing the persistence of this phenomenon throughout the night and over many nights in sequence. The lower MIL is observed most often during the winter between 70 and 80 km with large amplitudes (typically K) but with higher altitudes in the summer (typically km) and weaker amplitudes [Hauchecorne et al., 1987; Gille et al., 1991; Hauchecorne et al., 1991]. In addition, Hauchecorne et al. [1987] also showed that the phenomenon is mesoscale in dimension with a cross section many hundreds of kilometers. [57] A study of the global distribution of lower MIL events as a function of latitude and season utilizing remotesensing results from the UARS platform [Leblanc and

18 Figure 14. Averaged latitudinal distributions of the MIL amplitudes observed with the UARS/HALOE instrument for the four seasons in the Northern Hemisphere. From Leblanc and Hauchecorne [1997]. 18 of 31 Hauchecorne, 1997] resulted in two interesting conclusions, namely, that MIL events are rarely found at high latitudes and that the MIL events are most often observed at equatorial latitudes during equinoxes and at midlatitudes in the winter hemisphere. This morphological distribution was confirmed by Kumar et al. [2001] for low-latitude MILs. The global distribution of the MIL amplitudes is typically characterized by planetary-scale spatial variations in longitude. Figure 13 illustrates the comparison of groundbased detection of the MIL with observations from two UARS instruments. It is evident from this comparison that the amplitude of the MIL observed from space tends to be diminished by factors of 2 3, an effect interpreted to be a consequence of the slant path averaging through the limb source region. Figure 14 presents the latitudinal distribution of MIL events illustrating that the events with the largest amplitudes are observed at midlatitudes in the winter hemisphere. [58] Recent work by Sassi et al. [2002] has demonstrated that the simulation of the breaking of planetary waves in the mesosphere region would produce a MIL event with an amplitude of K depending upon the strength of the planetary wave (Figure 7). These events arise as a consequence of the very rapid wave-breaking dissipation of these waves that takes place when a critical line of zero wind in the upper mesosphere is encountered as these waves propagate vertically into the upper mesosphere. The model used in these calculations was the Whole Atmosphere Community Climate Model that represents the upward extension of the NCAR Community Climate Model into the lower thermosphere utilizing the physical parameterizations for this region that have been developed for the NCAR general circulation model [Roble and Ridley, 1994]. Parameterization of both orographic waves and a spectrum of traveling gravity waves was included in this analysis to produce the deceleration of the westward mesospheric jet by the deposition of momentum from westward propagating waves. Alternating regions of westward and eastward zonal mean acceleration are produced giving rise to corresponding layers of alternating zonal winds, with reversals (the zerowind line) occurring at about and 110 km at winter midlatitudes. This region of planetary wave breaking between 70 and 80 km at medium and high latitudes is characterized as a mesospheric surf zone to differentiate it from the stratospheric surf zone that occurs at low latitudes when planetary waves encounter a zero-wind line within the low-latitude stratosphere. [59] An example of the results of this NCAR simulation is shown in Figure 15. Figures 15a, 15b, 15c, and 15d depict profiles of temperature, zonal wind, geopotential anomaly (deviation from the zonal mean), and temperature anomaly (deviation from the zonal mean), respectively, for the longitude circle at the latitude of 43 N. Overlaid on Figures 15a 15d is a color shade plot of the temperature gradient. An example of the large MIL amplitude produced by this modeling effort is seen in the altitude region of 70

19 Figure 15. The NCAR general circulation model results for an example of a planetary wave (a) temperature structure, (b) zonal wind distribution, (c) geopotential anomaly, and (d) temperature anomalies. The color overlay in Figures 15a 15d illustrates the temperature lapse rate computed for the topside section of the MIL induced by the breaking planetary wave. From Sassi et al. [2002]. 80 km (Figure 15a) near 60 E, which lies just above the region of maximum positive geopotential anomaly in Figure 15c. The amplitude of the MIL (peak temperature minus the base temperature) is 45 K. Figure 15b shows this longitudinal sector to be a region of weak westward winds between the altitudes of 70 and 80 km. The phase progression of the contours of the geopotential anomaly between 30 and 65 km illustrates a planetary wave of zonal wave number 1 tilted to the west as is typical for a vertical propagating planetary wave. The discontinuity of this phase progression at 65 km near the region of the zero wind is where the rapid dissipation of the planetary wave takes place producing the large temperature anomaly in this range of altitudes. Figure 16 illustrates the temperature profiles between 30 and 110 km for two longitudes, 90 E and 90 W, in this simulation and for the case in which the gravity wave activity in the model was turned off. Figure 16 shows the difference in the MIL amplitudes at the two different longitudes of the planetary wave activity. Figure 16 also illustrates how diminished the MIL amplitude is for the case of weak or no gravity wave activity. The MIL seen near 90 km in the no gravity wave profile can be attributed to the diurnal tidal structure as seen in Figure 9. [60] An example of a lower MIL event as a function of time is depicted in Figure 17, which demonstrates 24-hour observations of a MIL at relatively low altitudes. These data were obtained with the Rayleigh lidar system located at the Millstone Hill/Massachusetts Institute of Technology Haystack Observatory (42.6 N, 71.5 W) and had the rare ability to operate in high solar background conditions [Duck 19 of 31 Figure 16. Comparison of NCAR general circulation model temperature profiles for two different longitudes: 90 E and 90 W. Also shown is a temperature profile for an example in which the gravity wave (GW) simulation is absent from the NCAR general circulation model calculations. From Sassi et al. [2002].

20 Figure 17. A time series of 1.5-hour average temperature measurements taken by Rayleigh lidar over the Millstone Hill/Massachusetts Institute of Technology Haystack Observatory near Westford, Massachusetts. Successive profiles in time are separated by 30 K, and the time stamps indicate the start of each integration. The adiabatic lapse rate is given by the dotted lines for reference. The nighttime profiles have been truncated at the top for this plot. Sunrise was at 1047 UT, and civil twilight began about half an hour earlier. Note that local time lags UT by 5 hours. From Duck et al. [2001]. et al., 2001]. In addition, Duck et al. [2001] pointed out that the topside lapse rate for the MIL was often observed to be nearly adiabatic, a result that was originally noted in Toronto observations of MILs [Whiteway et al., 1995]. This result is very significant as it implies that the atmospheric stability may already be reduced below the limit of the Richardson number of 1/4, in which case dynamical instability leading to the production of atmospheric turbulence is likely to occur. It is to be noted that the name inversion layer is therefore somewhat misleading in regard to high atmospheric stability, as it is colloquially inferred in tropospheric meteorology. Indeed, the bottom side is more stable because the thermal lapse rate is positive. Conversely, the negative lapse rate on the topside makes the atmospheric stability much less stable, and, in fact, most lower MIL events observed suggest that the thermal lapse rate is close to being adiabatic, which would be characteristic of the development of dynamic instability. Such instabilities may provide an in situ source of gravity waves within the middle atmosphere. 5. STE OBSERVATIONS AND MODELING [61] Observations of STEs in the high-latitude winter stratosphere have been present in the literature for at least 30 years [Labitzke, 1972]. This was clearly pointed out by von Zahn et al. [1998]; Figure 18 demonstrates that there have been numerous temperature profiles obtained by rockets that have repeatedly observed STEs, thus demonstrating their altitude occurrence (40 km ± 1 km) and magnitude Figure 18. Vertical temperature profiles ( C) from several Northern Hemisphere rocket and lidar stations during the peak warming period of different stratospheric warming events. Crosses indicate the CIRA 1986 mean January profile at 70 N; squares denote a profile from the Berlin tropospherestratosphere-mesosphere general circulation model. From von Zahn et al. [1998]. 20 of 31

21 Figure 19. Time-height development of the Northern Annular Mode during the winter of The indices have daily resolution and are nondimensional but physically represent relative pressure at each height. Blue corresponds to positive values (strong polar vortex), and red corresponds to negative values (weak polar vortex) of the Northern Annular Mode index. The contour interval is 0.5, with values between 0.5 and 0.5 unshaded. The thin horizontal line indicates the approximate boundary between the troposphere and the stratosphere. Reprinted with permission from Baldwin and Dunkerton [2001]. Copyright 2001 AAAS. (over 303 K). However, owing to the large expense of rockets and the lack of any time resolution the investigations into these thermal features were sorely limited. This has also been true of many current remote-sensing instruments, including radiosondes, which can only get to 30 km and therefore cannot observe STEs. Similarly, it is true for limb scanning measurements from satellites, which usually lack the vertical detail necessary to observe STEs. Recently, Rayleigh lidar, with its high spatial and temporal resolution, has become the sole instrument that can truly capture the vertical evolution of these features with the spatial and temporal resolutions needed. Model analyses are often used in conjunction with these point observations to further investigate the synoptic atmospheric background in which these events take place. [62] Accordingly, the ground-based accessibility to STEs via lidar techniques and the association of STEs with sudden stratospheric warmings and their relationship to the Northern Annular Mode (formerly known as the Arctic Oscillation or North Atlantic Oscillation [Baldwin and Dunkerton, 1999; Baldwin, 2000; Baldwin and Dunkerton, 2001; Thompson et al., 2001]) have created a resurgence of interest in STEs. As also reported by von Zahn et al. [1998], in all cases except two the observed STEs from the rocket measurements were followed by a major sudden stratospheric warming. This observational correlation was further addressed by Braesicke and Langematz [2000] where they inspected model runs of the Berlin Climate Middle Atmosphere Model to look for a similar correspondence. They found that STEs tended to occur in almost all Northern Hemisphere high-latitude winters but that the eventual creation of a sudden stratospheric warming soon after a STE need not be the case. The modeling results of Braesicke and Langematz [2000] do not necessarily contradict the observational results of von Zahn et al. [1998], as the von Zahn et al. [1998] study only focused on the most severe STEs (i.e., greater than 303 K). Hence one can infer that a stronger STE will more likely proceed to a major sudden stratospheric warming. Unfortunately, the Braesicke and Langematz [2000] study was rather cursory in its analysis and did not discuss the physical connection between STEs and sudden stratospheric warmings. Nonetheless, the observations of intense STEs and subsequent forecasting of sudden stratospheric warmings would be of great benefit to the weather forecasting community, as has been pointed out in recent studies of the Northern Annular Mode. [63] To better understand the Northern Annular Mode, we return to our discussion of the polar vortex and sudden stratospheric warmings initiated at the end of section 3. At high latitudes it is theorized that upwardly propagating planetary waves interact with the highaltitude polar vortex jet. These interactions in the upper stratosphere and mesosphere are usually poorly characterized and not well understood, largely because of the paucity of measurements at these altitudes in the high-latitude regions. This higher-altitude region is more susceptible to wave energy flux and responds more dramatically than at lower altitudes. This is because the background density is lower at higher altitudes, and less wave flux is required to force a significant change in the background winds and thermal structure. Thus the initial effects of a Northern Annular Mode signature are believed to appear at higher altitudes even though wave amplitudes may be less because of attenuation by the strong lower stratospheric eastward winds [Liu and Roble, 2002]. [64] Ultimately, the planetary wave polar vortex interaction forces the high-altitude zonal wind flow toward slower/ westward speeds. This reduces the vertical propagation of later planetary waves, forcing them to break at ever lower altitudes and subsequently slowing down the vortex jet at still lower altitudes. If the continual planetary wave forcing is intense enough, it can alter the entire zonal wind flow by 21 of 31

22 help assess these influences. The current research and development version is the T79L54 version extending from 0 to 85 km that is referred to as NOGAPS-ALPHA (NOGAPS with Advanced Level Physics and High Altitude) [Coy et al., 2002]. Figure 20. (left) Temperature profile with 1 standard deviation uncertainty derived at 1.92-km resolution for 30-min integration of Arctic lidar technology (ARCLITE) measurements acquired at UT on 12 December 2000 (J. Thayer, personal communication 2003). (right) Successive 30-min integrated temperature profiles (blue) calculated from ARCLITE measurements acquired during the period 1715 UT on 11 December to 2358 UT on 12 December Red lines connect corresponding National Centers for Environmental Prediction (NCEP)- derived temperatures at constant pressure surfaces of 10, 5, 2, 1, and 0.4 hpa for 1200 UT on December. The thick black line depicts the corresponding Mass Spectrometer Incoherent Scatter (MSIS) model temperature profile calculated for this time period. Image courtesy of J. Thayer December 2000 STE Over Sondrestrom, Greenland [66] An example of a STE that occurred over Sondrestrom, Greenland (67 N, 309 E), on December 2000 is depicted in Figure 20 (J. Thayer, personal communication, 2003). These temperature profiles, based upon Rayleigh lidar relative density measurements obtained at the Sondrestrom Research Facility, demonstrate that the altitude of the stratopause had decreased from its climatological mean as represented by the MSIS-90 model with an amplitude increase to well over 300 K. At the same time one observes a pronounced cooling in the lower mesosphere relative to the climatological mean that is capped by more climatological temperatures at 75 km. We also note that the event is not resolved by typical NCEP analyses. [67] Figure 21 shows the time series of this event as observed over Sondrestrom (J. Thayer, personal communication, 2003). The progression of the warming event is quite rapid over the site. This evolution, however, appears to be continuously forcing planetary waves to break at successively lower altitude levels. This zonal wind anomaly appears to propagate down into the troposphere. However, when there is very little planetary wave activity and the polar vortex is subsequently extremely stable, the cold core of the vortex can also appear to propagate downward and thus further strengthen the vortex. [65] Evidence that the downward propagation of the zonal wind anomaly is representative of the Northern Annular Mode morphology is indicated in Figure 19, where we see a zonal mean proxy of vortex stability as measured by the Northern Annular Mode. Technically, the Northern Annular Mode is a global-scale mode of climate variability, but the weekly impact of this variability is easily seen in Figure 19. Furthermore, this downward influence can have a significant impact on the regional tropospheric conditions, thus showing that the middle-atmosphere dynamics may influence the lower atmosphere. Such impacts would not be accounted for in many of today s numerical weather prediction models, as they have top lids (i.e., upper model domains) that are generally below 30 km. Though the degree and importance of this downward influence is largely unknown, it is, nonetheless, a new area of research that has important implications. In fact, the Naval Research Laboratory s Navy Operational Global Atmospheric Prediction System (NOGAPS) is undergoing current revision to 22 of 31 Figure 21. Color coded height-time contour analysis of temperature (K) derived from ARCLITE 532-nm backscatter measurements acquired during December 2000 (J. Thayer, personal communication, 2003). Arrows indicate the midpoint times of profile averages that varied from 2 to 6 hours. Included in the analysis but not shown are profiles before day 341 and after day 356. The temporal variation in kinetic temperature derived from OH airglow measurements taken at Sondrestrom and centered near 87 km is shown at the top of the plot. Image courtesy of J. Thayer.

23 Figure 22. Simulated cross section of isotherms (contour interval, 10 K) and arrows depicting the ageostrophic velocity at 57.5 N on 31 December The arrow at the bottom left of Figure 22 represents a horizontal velocity of 50 m s 1 and a vertical velocity of 18.4 cm s 1. From Fairlie et al. [1990]. due to horizontal advection of the feature as the ALOMAR station situated 67 to the east observed similar temperatures 2 days prior to Sondrestrom measurements of the warming. At the beginning of this time period (day 341) the stratopause is seen to be rather high in altitude, near 70 km, and its relatively warm temperature extends into the mesosphere, as indicated by the concurrent OH temperatures. The stratopause then descends in altitude relatively rapidly, and its temperature increases to 320 K (depicted in Figure 20). Concurrently, the mesosphere at 60 km cools dramatically by more than 50 K. Five days later (around day 353), we note that a minor warming develops and a splitting of the vortex occurs [Naujokat and Kunze, 2001] Baroclinic Zones [68] As modeled by Fairlie et al. [1990], interactions between the polar vortex, the Aleutian High, and upwardly propagating planetary waves can cause the temporary formation of strong horizontal and vertical gradients in middle atmospheric winds and temperatures over synoptic scales. The otherwise geostrophic and thermal wind-determined flow of the polar vortex is disturbed by the associated localized momentum forcing induced by the Aleutian High or planetary waves. This creates a synoptic ageostrophic circulation that is accompanied with vigorous vertical motion. This vertical motion provides a positive feedback enforcing the secondary circulation, thus providing a form of frontogenesis within the upper stratosphere. The adiabatic heating associated with downward motion (thermally indirect) results in vertically narrow temperature enhancements, thus creating a STE. Fairlie et al. [1990] called these temperature enhancements baroclinic zones because of the nature of the baroclinic atmosphere that is induced by the secondary circulation. [69] An example of a numerically modeled baroclinic zone is depicted in Figure 22, where we see both the characteristic westward tilt of a STE structure, as well as the superimposed ageostrophic circulation [Fairlie et al., 1990]. The event has a peak temperature contour of 290 K and spans from 30 km well into the mesosphere. The ageostrophic circulation is downward in the baroclinic zone and upward to either side of the event. As noted by Fairlie 23 of 31

24 Figure 23. Simulated cross section of temperature (dashed, units of K) and the divergence of the horizontal wind (solid, units of 10 6 s 1 ) at 57.5 N on 25 February Regions of convergence are shaded. From Fairlie et al. [1990]. et al. [1990], it is the vertical motion that is the true cause of the temperature profile not the horizontal advection of lower-latitude air, which is cooler than the observed 290 K anyway. [70] Manney et al. [1994] modeled the three-dimensional evolution of the major sudden stratospheric warming that occurred in They found that their model reproduced the baroclinic zones of Fairlie et al. [1990], noting that the modeled winds to the west of the event induced poleward and upward motion, while winds to the east of the event induced equatorward and downward motion. Hence, to the west, there is a cooling, while to the east there is a warming. Such a circulation would enhance the westward tilting of the baroclinic zones, thus increasing the baroclinic nature of the event. [71] Fairlie et al. [1990] also noted the generation of in situ inertiogravity waves from STEs in his model. Since the STE is an ageostrophic event, the atmosphere attempts to reestablish geostrophy by the radiation of gravity waves. The radiation of gravity waves as part of a geostrophic process in the troposphere has been known for some time, and it is of little surprise that such is the case here. Specifically, in Figure 23 we see that to the east of the STE/baroclinic zone there are downward pointing phase lines in the temperature profile. Such lines imply the upward propagation of gravity waves to the west and to the east. Fairlie et al. [1990] noted that these waves had horizontal wavelengths of the order of 1000 km and vertical wavelengths of 8 km. From the gravity wave dispersion relationship, with information on the background wind fields and temperatures, this yields intrinsic periods of 10 hours and an observed period of 24 hours. It is important to note that direct experimental observations of such gravity wave generation is currently lacking in the literature. [72] The results presented by Fairlie et al. [1990] and Manney et al. [1994] focused on time periods of major sudden stratospheric warmings. However, it is believed that STEs can develop at any time in the high-latitude winter, depending on the amount of forcing that is present and subsequent response to the zonal flow/polar vortex jet. The exact nature of such forcing and its role in the downward propagating Northern Annular Mode, however, remains unanswered. 6. DIRECTIONS FOR FUTURE RESEARCH [73] It is clear that there have been major advancements in the field in regard to the study of the middle atmosphere itself and the study of related thermal enhancements. However, it is also clear that we do not yet fully understand a range of issues, including the topics discussed in the sections below, MIL Formation at High Latitudes [74] It is interesting, as has been noted by Meriwether and Gardner [2000] and Cutler et al. [2001], that there are fewer observations of MILs at high latitudes compared with 24 of 31