1. Introduction
To study processes responsible for Arctic stratospheric ozone loss, two major planned field campaigns were combined during the Arctic winter of 1999/2000. The U.S. component was the National Aeronautics and Space Administration (NASA) sponsored SAGE III Ozone Loss and Validation Experiment (SOLVE), the acronym referring to the additional goal of acquiring suborbital validation data for the third Stratospheric Aerosol and Gas Experiment (SAGE III) on the Meteor-3M spacecraft. Unfortunately, the launch of Meteor-3M was delayed until December 2001, leading SOLVE to focus more on the stratospheric science goals. The European component was the European Commission sponsored Third European Stratospheric Experiment on Ozone (THESEO 2000). More than 500 scientists participated in SOLVE/THESEO 2000. Activities were coordinated from the Arena Arctica hangar at Kiruna airport, in northern Sweden (68°N, 20°E), where six instrumented research aircraft were based and operated at various times during the campaign. Newman et al. (2002) provide a comprehensive overview of SOLVE/THESEO 2000, including goals, logistics, implementation, and some initial research highlights.
Science flights studying ozone loss processes require careful planning to acquire the requisite scientific data using available onboard instruments and to maximize their overall scientific impact. For example, following the discovery of an “ozone hole” over Antarctica, NASA DC-8 and ER-2 research aircraft acquired stratospheric data during the Airborne Antarctic Ozone Experiment (AAOE) of August–October 1987 that helped confirm one of the several competing theories for its formation: namely, enhanced photochemical ozone destruction by free chlorine radicals supplied by anthropogenic chlorofluorocarbon emissions (Solomon 1999). The success of AAOE owed much to careful flight planning, based in large measure on stratospheric forecast fields issued by operational global numerical weather prediction (NWP) models. NWP fields, once available, were postprocessed into specialized AAOE mission products, such as isentropic potential vorticity maps and parcel trajectories, which were used to devise flight tracks along isentropes or to resample air measured during earlier flights (Tuck et al. 1989). During the 1990s, progressively higher resolution global NWP model forecasts played important and more elaborate roles in flight planning for subsequent airborne measurement campaigns focusing on ozone loss in both the Arctic and Antarctic stratospheres (Wofsy et al. 1994; Pyle and Harris 1995; Stefanutti et al. 1999a; Newman et al. 1999). As understanding of the chemistry associated with polar ozone loss improved, forecast winds from global NWP models were interfaced to global chemical transport models (CTMs) to provide stratospheric chemical forecasts for some later missions prior to SOLVE/THESEO 2000 (e.g., Lee et al. 1997). In preparation for SOLVE/THESEO 2000, it was clear that accurate daily forecasts of the meteorology of the polar stratosphere would once again be essential for scientific flight planning (Newman et al. 2002).
The importance of stratospheric meteorology for these ozone-focused missions is due not just to transport, but also to the strong control that dynamics exert on stratospheric temperatures. The progressive chlorine activation within and denitrification of stratospheric air that lead to rapid ozone loss in sunlight occur within polar stratospheric clouds (PSCs). PSCs only form where temperatures are very cold (≲190–195 K), and the geographical location and depth of these coldest temperatures are controlled mostly by synoptic-scale stratospheric meteorology. Thus, airborne measurements of PSCs and their attendant chemistry and microphysics require accurate forecasts of vortex meteorology to vector flights into geographical zones where temperatures are cold enough for PSCs to exist.
The meteorology of the polar winter stratosphere is characterized by a strong symmetric vortex circulation, which isolates cold polar air from midlatitudes. Planetary waves episodically propagate upward from tropospheric sources, displacing and distorting the vortex and enhancing descent, which warms the stratosphere adiabatically (Newman and Nash 2000). Global models capture these dynamics quite well: indeed, global NWP models have better forecast skill in the stratosphere than the troposphere during polar winter, permitting relatively accurate predictions several days or more into the future (Waugh et al. 1998; Lahoz 1999). While much finer-scale structure is often observed in ozone and related chemicals, CTM studies have shown that this is often the result of isentropic advection by this large-scale meteorology, which can be simulated quite accurately with high-resolution offline trajectories driven by coarser-resolution winds from the global models (see, e.g., Newman et al. 1996).
For these reasons, finer-scale stratospheric dynamics (e.g., gravity waves) were of limited interest when planning flights during early airborne stratospheric ozone campaigns and, so, were not specifically forecast. This situation differs markedly from the troposphere, where mesoscale and planetary-scale circulations both significantly influence chemistry and microphysics (e.g., Webster and Houze 1991; Jacob et al. 2003).
Things began to change during the mid- to late 1990s when polar ozone research focused on the Arctic. Greater topography in the Northern Hemisphere generates stronger planetary wave activity, leading to a more disturbed Arctic vortex and, hence, generally warmer wintertime stratospheric temperatures that exhibit much greater intraseasonal and interannual variability than those in the Antarctic (Pawson and Naujokat 1999; Newman et al. 2001). The warmer Arctic stratosphere produces fewer PSCs and less ozone-loss chemistry (Chipperfield and Jones 1999; Randel and Wu 1999), explaining to first order why the huge losses that yield an ozone hole only occur in the Antarctic.
Yet ozone loss can still be significant in the Arctic. Prior to SOLVE/THESEO 2000, some state-of-the-art global CTM runs were systematically underpredicting the wintertime ozone loss that was observed during previous Arctic winters, due to insufficient PSC formation within the models (e.g., Lutman et al. 1997; Becker et al. 1998). Intercomparisons of minimum stratospheric temperatures in Arctic winter from various global analysis systems, while revealing substantial differences on occasion, did not show any clear systematic warm bias in analysis compared to observations that might explain this PSC underprediction (Knudsen et al. 2002; Manney et al. 2003). This suggested instead that important additional PSC formation mechanisms might be missing from the models.
At about the same time, the role of finer-scale stratospheric dynamics on PSCs and ozone chemistry began receiving serious attention. European aircraft campaigns out of Kiruna during the mid- to late 1990s repeatedly measured PSCs that formed in the cooling phases of stratospheric mountain waves generated by flow across the Scandinavian Mountains (Carslaw et al. 1998b; Wirth et al. 1999). While mountain waves had been observed forming PSCs in earlier airborne Antarctic missions (Cariolle et al. 1989), their effect proved secondary because of the much more extensive synoptic-scale decks of PSC produced by very cold intravortex temperatures. Conversely, minimum synoptic temperatures in the Arctic often hover at or just above the threshold for PSC formation (Schulz et al. 2001). Thus, temperature decreases due to subsynoptic mountain waves could be a much more significant source of PSCs in the Arctic. Since global models generally cannot resolve mountain wave dynamics (see below), Carslaw et al. (1998a) argued that unresolved mountain waves might be the missing ingredient needed in global CTMs to form more PSCs, leading to greater Arctic ozone loss in better agreement with observations. Using a high-resolution global model, Edouard et al. (1996) showed that unresolved small-scale variability in reactive chemical species could have very large influences on modeled Arctic ozone loss. On analyzing aircraft data from earlier missions, Sparling et al. (1998) estimated that unresolved variability in ClO could lead to underestimates in ozone loss near the edge of the Arctic vortex of ∼5%–35%.
Thus, leading up to SOLVE/THESEO 2000 there was new interest in mountain wave–induced PSCs as a priority scientific measurement goal requiring forecasting. Unfortunately, during 1999–2000 the U.S. global NWP models were running at ∼1° horizontal resolution (∼100 km). Since the very smallest spatial scales in global spectral models are unreliable (Lander and Hoskins 1997), model orography is smoothed and spectral divergence damping is imposed to suppress wave forcing at these scales (Derber et al. 1998) so that only at horizontal wavelengths ∼6–10 times this minimum gridpoint resolution of ∼100 km would one expect predictive skill (Davies and Brown 2001; Skamarock 2004). Since mountain waves have horizontal wavelengths ∼5–500 km, forecasts from global NWP systems were not expected to resolve any significant mountain wave activity for SOLVE/THESEO 2000.
Thus, other forecasting approaches were necessary. It soon became clear that no single model could provide the necessary mountain wave predictive capabilities for this campaign. Section 2 describes the three different models that were ultimately used to predict mountain wave–induced PSCs for aircraft flight planning during SOLVE/THESEO 2000. The forecasting effort was carefully planned and successfully executed (see section 2.4 of Newman et al. 2002). In the ∼6 yr that have transpired since, the PSC community has completed and published initial scientific analyses of the PSC data acquired by the aircraft, identifying measured clouds that appear to be mountain wave related (Hu et al. 2002; Dörnbrack et al. 2002; Luo et al. 2003; Fueglistaler et al. 2003; Svendsen et al. 2005). Here we compare the wave PSCs identified in lidar data from the DC-8 and Falcon research aircraft with the models' original forecast guidance. Section 3 describes instruments on the aircraft that acquired PSC and meteorological data that we use to validate our wave PSC forecasts. Detailed comparisons between the models' mountain wave PSC predictions and available PSC and wave measurements are given in section 4. Results are summarized in section 5, along with some recommendations for future mission forecasting.
2. Mountain wave models and forecasting
a. Requirements for PSC forecasting
The mandate of this forecasting effort was to predict the geographical locations, horizontal extents, and altitudes of mountain wave–induced PSCs, to aid flight planning for the scientific research aircraft. Stratospheric mountain waves that did not produce PSCs (e.g., Hertzog et al. 2002) were of limited interest to flight planners.
There are three different types of PSCs: type Ia, consisting of nitric acid trihydrate (NAT) particles; type Ib, consisting of supercooled ternary (HNO3/H2SO4/H2O) solution (STS) droplets; and type II, consisting of water ice particles. At a given altitude each PSC type has a formation threshold temperature, denoted TNAT, TSTS, or TICE. If stratospheric temperatures drop below these thresholds, PSCs of this type can exist. Since TNAT > TSTS > TICE, type I PSCs are most likely to form, whereas type II PSCs require the very coldest stratospheric temperatures and are least common. For typical stratospheric mixing ratios of H2O and HNO3 of 5 ppmv and 10 ppbv, respectively (Toon et al. 1989), the 30-hPa NAT point temperature TNAT ∼193 K (Hanson and Mauersberger 1988), while the frost point temperature TICE ∼186 K (Marti and Mauersberger 1993). Thus, the models had to predict accurately geographical locations and altitudes where mountain waves dropped stratospheric temperatures below these representative TNAT and TICE thresholds.
After preparatory premission meetings, three models were chosen because of their complementary capabilities in forecasting mountain wave–induced stratospheric temperature variability. The models and their forecasting configurations are summarized in Table 1, and each is described in depth in sections 2c–2e. The Mountain Wave Forecast Model (MWFM) provided rapid hemispheric forecasts of stratospheric mountain wave temperature amplitudes and wave-induced turbulence potential. The fifth-generation Pennsylvania State University–National Center for Atmospheric Research Mesoscale Model (MM5) provided detailed forecasts of long-wavelength mountain wave–induced stratospheric temperature variability over Scandinavia, as well as streamlines for planning quasi-Lagrangian flights along forecast particle trajectories (e.g., Wirth et al. 1999). The Vosper Orographic Model (3DVOM) provided high-resolution forecasts of shorter-wavelength mountain wave–induced temperature variability in the stratosphere near Kiruna (and later over Spitzbergen). In a loose sense, these three models provided a kind of nested forecasting capability, with forecasts over the Arctic by MWFM supplemented with additional resolution and detail from MM5 and 3DVOM over Scandinavia and Kiruna, the base for airborne operations.
b. Global NWP models
The starting point for all stratospheric forecasting during SOLVE/THESEO 2000 was global ground-to-stratosphere forecasts and analyses issued operationally by meteorological centers in both the United States and Europe. In addition to being inspected regularly to assess the forecast evolution of synoptic-scale stratospheric temperatures and potential vorticity, these fields provided initial conditions for the mountain wave forecasting models. The specific global NWP model forecasts and analyses that were used to initialize the mountain wave models are summarized in Table 2. Many other forecast/analysis fields were also utilized in other ways during SOLVE/THESEO 2000: see Table 7 of Newman et al. (2002).
c. MWFM
The MWFM is an diagnostic postprocessor of NWP model winds and temperatures that provides regional or global forecasts of mountain wave properties at specifiable atmospheric pressure levels. MWFM mountain wave–induced turbulence forecasts have been used primarily as a flight planning safety tool for high-altitude long-endurance aircraft during previous NASA science campaigns and by the U.S. Department of Defense (Bacmeister et al. 1994; Newman et al. 1999; Eckermann 2002; Eckermann et al. 2004). For SOLVE/THESEO 2000, the MWFM was extended to predict mountain wave temperature amplitudes T̂PEAK for PSC forecasting.
1) MWFM surface forcing
A list of quasi-two-dimensional (2D) ridges characterizes the dominant features in the earth's topography relevant for the MWFM mountain wave calculations. Each ridge feature has, among other properties, a characteristic cross-ridge width L, peak height h, base height zb, and horizontal orientation ϕLONG of its long axis. These parameters define the properties of mountain waves generated by surface flow over these orographic features. Bacmeister et al. (1994) and Eckermann et al. (2004) provide more information on the ridge databases.
Figure 1 shows two ridge databases available for use in the MWFM, one characterizing medium width ridges of L ∼35–150 km (top plot), the other capturing narrower ridges with L ∼20–70 km (lower plot). Note that many ridges lie on top of one another, since this ridge-finding procedure progressively decomposes complex topography into a series of overlapping and overlaying ridge functions, with narrow-width ridges typically isolating peaks that sit atop broader-width ridges defining larger-scale topography. These properties are reflected, for example, in each ridge's base altitude zb, which can be several kilometers above sea level if the identified ridge feature characterizes a peak sitting atop a broader ridge feature with a lower zb.
Winds, temperatures, and geopotential heights from operational NWP models are ingested on pressure levels and used to specify forecast flow conditions at the “surface” of the ridge, zs. From wind speeds |U (zs)| and Brunt–Väisälä frequencies N(zs), the inverse Froude number Fr−1 = N (zs)h/|U (zs)| is computed. If Fr−1 > 1, we assume that the flow here is either blocked upstream of this ridge (if it is a 2D feature) or diverted around it (if it is a more 3D feature); either way, the flow at zs does not pass over the ridge and generate gravity waves. We then compute the reduced component of the ridge height h̃ < h that forces a gravity wave using an Fr-dependent procedure described in the appendix. This final ridge elevation ĥ and other ridge properties are used to specify amplitudes and wavelengths of the mountain wave generated at zs.
2) MWFM-1
Two versions of the MWFM were run during SOLVE/THESEO 2000. Both use the same ridge databases, but different types of wave equations to model the forcing, propagation, and breakdown of the mountain waves by the forecast flow profiles over these ridges.
Version 1 of the MWFM (MWFM-1) is described by Bacmeister et al. (1994) and Eckermann et al. (2004). It uses hydrostatic irrotational gravity wave equations under the assumption that a single linear plane (2D) mountain wave is forced by the surface flow component parallel to the short axis of each ridge feature. A number of minor changes to this original MWFM-1 formulation accumulated prior to SOLVE/THESEO 2000. One important upgrade was to replace earlier pressure-based approximations of atmospheric density scaling terms in the momentum flux calculations with atmospheric densities computed directly from input pressure and temperature profiles.
3) MWFM-2
For SOLVE/THESEO 2000, we issued forecasts for the first time using a next-generation, version 2, MWFM (MWFM-2). Eckermann and Preusse (1999) presented first results from an initial research version developed prior to SOLVE/THESEO 2000, preliminarily validated against satellite data. MWFM-2 has since been developed and used in many subsequent studies, where its evolving formulations are described in more depth (Hertzog et al. 2002; Jiang et al. 2004; Pagan et al. 2004; Eckermann et al. 2004). Here we review its status and salient features for SOLVE/THESEO 2000 forecasting.
Unlike MWFM-1, which launches a single ray from each ridge, MWFM-2 launches a total of I individual rays from each ridge with different horizontal wavenumbers (Kh)i = (ki, li) and horizontal propagation azimuths ϕi = arctan (li/ki), where the subscript i = 1, . . . , I tags each ray. The general approach to ray initialization at the source height zs is explained in detail in section 4.1 of Jiang et al. (2004), and the specifics for SOLVE/THESEO 2000 are summarized in the appendix.
The new nonhydrostatic MWFM-2 ray formulation offered some significant potential improvements for wave PSC forecasting. Previous MWFM-1 simulations of stratospheric mountain waves tended to be “overactive”; that is, wave amplitudes were generally too large compared to other estimates (e.g., Carslaw et al. 1998b, 1999), which could lead to false-alarm wave PSC forecasts. One likely source of this overactivity was the MWFM-1 hydrostatic irrotational wave equations, which do not simulate vertical reflections at turning points zt where ω approaches the high-frequency cutoff ωc ∼ N (see Marks and Eckermann 1995), allowing waves that should be vertically trapped within the troposphere to enter the stratosphere. The nonhydrostatic MWFM-2 ray equations identify rays that reflect vertically (Marks and Eckermann 1995). While ray methods can be used to model these trapped waves (Schoeberl 1985; Broutman et al. 2003), for SOLVE/THESEO 2000 we simply removed rays that reached turning points before entering the stratosphere, since we were only interested in stratospheric mountains waves and this quick removal helped speed up the forecasts. Test MWFM forecasts prior to SOLVE/THESEO 2000 confirmed that peak mountain wave temperature amplitudes predicted by MWFM-2 were lower and within a more realistic range than corresponding MWFM-1 predictions. Thus, MWFM-2 forecasts were our primary MWFM-based guidance for wave PSC forecasts, with MWFM-1 providing backup and cross-validation support.
4) Forecasting configuration
To provide a crude ensemble forecasting capability, MWFM issued separate forecasts initialized with fields from each of the three U.S. global models issuing ground-to-stratosphere forecasts at the time. One set of MWFM forecasts was initialized using global NWP fields issued by the National Centers for Environmental Prediction (NCEP) Aviation (AVN) and Medium-Range Forecast (MRF) models (Kanamitsu 1988). Another was initialized using fields from the Navy Operational Global Atmospheric Prediction System (NOGAPS) run at the Fleet Numerical Meteorology and Oceanography Center (FNMOC) (Hogan and Rosmond 1991). A third MWFM forecast was initialized using output from the GEOS-3 gridpoint global model and Data Assimilation System (DAS) (Swinbank et al. 1999; Dee and Todling 2000), which ran in a reduced-resolution operational forecasting mode during SOLVE/THESEO 2000 at what was then known as NASA's Data Assimilation Office and is now the Global Modeling and Assimilation Office (GMAO). MWFM-1 and MWFM-2 forecasts were both issued in each case.
Since MWFM was the only one of the three mountain wave forecasting models capable of issuing forecasts over the entire Arctic, its standard forecast maps covered the entire SOLVE area of airborne operations, depicted in Fig. 1, at standard reference pressure levels in the 10–100-hPa range. These MWFM forecasts ran automatically at set times each day (staggered based on various global NWP model output times) at the Naval Research Laboratory (NRL) in Washington, D.C. Tailored forecasts for specific scientific or flight planning needs were devised and run quickly in the field when required.
The main forecast maps used by flight planners were MWFM-2 forecasts of stratospheric mountain wave peak temperature amplitudes, T̂PEAK. These ray amplitudes were superimposed onto contours of forecast background temperatures T from the global NWP model, to help flight planners visually locate regions where T − T̂PEAK dropped below TNAT or TICE and thus PSCs of types I or II, respectively, might form. MWFM also forecast the locations and intensities of turbulence due to mountain wave breaking, as a flight-planning safety tool for NASA's DC-8 and high-altitude ER-2 research aircraft (Eckermann et al. 2000, 2004), since previous in situ sampling of mountain wave PSCs near Kiruna with a high-altitude European aircraft encountered substantial turbulence within these waves (Stefanutti et al. 1999b).
d. MM5
The MM5 (Dudhia 1993; Grell et al. 1994) was set up at the Deutsches Zentrum für Luft- und Raumfahrt e.V. (DLR) in Oberpfaffenhofen, Germany, to generate operational forecasts of stratospheric mountain waves over Scandinavia. MM5 uses a finite-difference approximation to solve the fully compressible, nonhydrostatic set of fluid equations in a rotating frame of reference. Due to computational limitations, of all the possible parameterizations available in the model to represent physical processes (e.g., radiation, convection), only a turbulence scheme was used. Thus, the model ran “dry” in stratospheric forecast mode at DLR. The model used 52 levels (σ coordinates) from the ground to 10 hPa, where a radiation boundary condition minimized any spurious downward reflection of vertically propagating gravity waves. The aforementioned model setup was first used at DLR to forecast stratospheric mountain waves over Scandinavia during the winter of 1996/97 (Dörnbrack et al. 1998), verifying its utility for mountain wave PSC forecasting near Kiruna.
For SOLVE/THESEO 2000, the model was configured into a set horizontal domain of 92 × 92 grid cells centered at 65°N and 15°E with horizontal mesh sizes Δx = Δy = 24 km, which spanned all of Scandinavia, as shown later (Fig. 6c). Higher-resolution runs could not be performed, as only 16 CPUs on a CRAY J90 supercomputer were available for the forecasts at DLR.
Forecasts were initialized using +24-h forecast fields from the European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecast System (IFS) 1200 UTC forecast runs (see Table 2) at 15 selected pressure levels. Lateral boundary conditions within the MM5 model were updated every 6 h using subsequent 6-hourly ECMWF forecast fields, so that MM5 boundary conditions at model forecast times of +6 and +12 h were given by the +30-h and +36-h ECMWF forecast fields, respectively. Each MM5 forecast was run out to +60 h. The total CPU time for a complete forecast cycle, including both pre- and postprocessing, was ∼60 h.
These forecasts ran automatically without significant interruptions at DLR from 1 December 1999 until mid-March of 2000. Final forecast products were regularly available at 0600 UTC on the mission Web server. The main maps of interest to flight planners were stratospheric temperatures at given pressure levels, cross sections of temperatures along possible flight segments, and winds and streamlines for planning quasi-Lagrangian flights measuring PSCs along forecast particle trajectories.
e. 3DVOM
1) Model formulation
The 3DVOM is a finite-difference three-dimensional numerical model designed for high-resolution simulations of mountain waves generated by flow over complex terrain. The model is described in detail by Vosper (2003) and Vosper and Worthington (2002), who used it to simulate tropospheric mountain waves over the United Kingdom.
The 3DVOM is based on the time-dependent linearized equations of motion for a dry atmosphere. The equations of motion are linearized about a background wind and potential temperature field (assumed to be dependent on height only) and are integrated forward in time until a steady wave field is obtained. The version of 3DVOM that was used for operational forecasts during SOLVE/THESEO 2000 is based on the shallow Boussinesq version of the equations (Dutton and Fichtl 1969). Although generally valid for tropospheric waves, when a greater depth of the atmosphere is considered, the Boussinesq approximation can give rise to wave amplitude errors (Nance and Durran 1994) in the stratosphere. Thus, the model was extended after SOLVE/THESEO 2000 to an anelastic equation set derived by Lipps and Hemler (1982), which yields more accurate wave solutions over depths of several density scale heights (Nance and Durran 1994).
In the above, ∇h = (∂/∂x, ∂/∂y), k = (0, 0, 1), u′ = (u′, υ′, w′) is the wave perturbation velocity vector, θ′ is the potential temperature perturbation, and 
The model equations are discretized on a staggered mesh using a second-order-accurate centered finite-difference scheme. Steady wave solutions are obtained by integrating (2) and (3) forward in time (from an initial state in which no waves are present) using a modified leapfrog scheme. At each time step a Poisson equation is solved for π̃, which ensures that mass continuity (4) is satisfied. A free-slip lower-boundary condition is applied and zero perturbation conditions are imposed at the upper and side boundaries. Rayleigh damping is applied below the upper boundary to absorb upward-propagating waves. Similarly, damping is applied near all side boundaries to minimize wave reflection. In both cases the damping is increased linearly from zero to a maximum value at the boundaries.
2) Forecasting configuration
High-resolution 3DVOM forecasts were performed using a 256 km × 256 km horizontal domain centered near Kiruna (see Fig. 2). Forecasts here were issued daily between 4 December 1999 and 15 March 2000. These forecasts were designed to provide very high-resolution information about the temperature amplitudes of shorter-wavelength (nonhydrostatic) mountain waves that are not resolved by the coarser-resolution MM5 forecasts over Scandinavia. Previous lidar observations of PSCs had shown wavy signals within the clouds themselves, with wavelengths often as short as 10–20 km (Carslaw et al. 1998b; Wirth et al. 1999), which can make a significant contribution to local cooling and PSC evolution.
The horizontal and vertical grid spacings used for the forecasts were Δx = Δy = 2 km and Δz = 650 m, respectively. The model upper boundary was placed at 40 km and Rayleigh damping was imposed above 28 km. Daily mountain wave forecasts were generated using a vertical profile of winds, densities, and potential temperatures at 68°N, 18°E (the center of the model domain) from ECMWF IFS forecasts. The model equations were integrated forward in time for approximately 2.2 h, by which time a quasi-steady mountain wave field was generally established. Each day the model was run four times to predict wave fields at 1200 UTC (a 24-h forecast), 0000 UTC the following day (a 36-h forecast), and 0000 UTC 2–3 days ahead (60- and 84-h forecasts, respectively). Forecasts were run on a Silicon Graphics Origin 2000 system at the University of Leeds and graphical output was generated and transferred to the mission Web server. The first of the daily forecasts was usually completed by 0800 UTC.
Extra daily forecasts were set up over the island of Spitzbergen (79°N, 18°E) for the period 15 December 1999–15 March 2000, where the coldest climatological stratospheric temperatures are centered (Newman et al. 2002) and mountain waves can often form (Skeie and Grønås 2000). The domain size and grid spacings were the same as for the Scandinavia forecasts. Due to constraints on computing time, only two forecasts were generated for Spitzbergen each day. These were valid at 1200 UTC on successive days (24- and 48-h forecasts).
3. Aircraft validation data
To validate forecasts of mountain wave–induced PSCs, we use data acquired by instruments on three of the six aircraft operated during SOLVE/THESEO 2000.
a. Aerosol lidar data from the NASA DC-8 and DLR Falcon
We use high-resolution PSC data acquired remotely by two similar lidar profilers operated on two aircraft flying near the tropopause. The Langley Research Center (LaRC) aerosol lidar operated on the NASA DC-8 during SOLVE/THESEO 2000 in comanifested form with the Goddard Space Flight Center Airborne Raman Ozone and Temperature lidar. The LaRC lidar emits laser pulses at 1064, 532, and 355 nm, the fundamental, doubled, and tripled frequencies, respectively, from a neodymium:yttrium/aluminum/garnet laser. The Ozone Lidar Experiment (OLEX) (Flentje et al. 2000), operated from the DLR Falcon F-20 aircraft, uses these same laser wavelengths. Both systems transmit vertically and collect backscattered radiation with zenith-viewing telescopes for processing.
Stratospheric ABRs provide first-order discrimination among different PSC types. PSC-free regions yield S1064 ∼1, type I PSCs (NAT/STS) yield S1064 ∼3–30, and type II PSCs (ice) yield S1064 ∼50–500 (e.g., Fueglistaler et al. 2003).
b. Meteorological measurements from the NASA ER-2
We utilize meteorological stratospheric data from two instruments that operated autonomously on NASA's ER-2 research aircraft during SOLVE/THESEO 2000. The ER-2 cruises typically at 50–70 hPa.
The Meteorological Measurement System (MMS) acquires in situ measurements of the three-component wind velocity, temperature, and pressure (Scott et al. 1990). Raw 10-Hz data were averaged to 1 Hz to improve signal to noise, which, for a nominal ER-2 true airspeed of 200 m s−1, yields ∼200 m resolution along track. Bacmeister et al. (1996) have shown that the small-scale variability in 1-Hz MMS stratospheric velocities and temperatures is dominated by gravity waves.
We also utilize data from the Microwave Temperature Profiler (MTP) (Denning et al. 1989). The ER-2 MTP retrieves a temperature profile several kilometers above and below the aircraft from line-of-sight radiances from two pairs of O2 rotational lines observed at a series of different elevation angles. Final profiles are issued every 10 s. We convert temperature profiles into potential temperatures by using MTP-issued background atmospheric number densities to derive pressure profiles from the ideal gas equation.
4. Forecasts of mountain wave–induced PSCs
To provide an observational focus for objectively assessing our mountain wave PSC forecasts, we concentrate on PSC data acquired from the NASA DC-8. Using LaRC aerosol lidar data as their guide, Hu et al. (2002) reviewed all the DC-8 flight data during SOLVE/THESEO 2000 and identified four flights in which LaRC S1064 data showed values and structure that led them to a mountain wave PSC interpretation.
Here we assess the forecasts for all four of these flight dates, since they provide a good range of situations for assessment. Two of these flights, on 23 and 25 January 2000, were meticulously planned based on the models' predictions of wave PSCs over eastern Greenland and Scandinavia, respectively. Conversely, the other two flights of 10 December 1999 and 14 January 2000 were not planned with wave PSCs in mind, since models predicted no mountain wave PSCs, yet wavelike enhancements in S1064 were observed nonetheless. These cases are particularly important to study and reconcile.
A wave PSC was also observed indirectly from the DC-8 on 30 November 1999. We begin by studying this case before progressing to the other four flights.
a. 30 November 1999: PSCs over southern Scandinavia
1) Forecasts
Stratospheric mountain wave forecasting was just starting up on 30 November 1999 as the NASA DC-8 completed its ferry flight from Dryden Flight Research Center (DFRC) in California to Kiruna to commence the first phase of SOLVE/THESEO 2000. On this day, only MWFM was issuing operational forecasts: MM5 forecasts commenced the following day.
Figure 3a shows the plot of the MWFM-2 +18-h forecast of mountain wave–induced peak temperature amplitudes T̂PEAK at 30 hPa, valid on 30 November 1999 at 1800 UTC, based on initialization by operational NOGAPS +18-h forecasts from FNMOC initialized at 0000 UTC on this day. NOGAPS predicted a broad region of very cold 30-hPa temperatures T below ∼190 K over Scotland and southern Scandinavia: see blue contours in Fig. 3a. Similar features emerged in NCEP and GMAO forecasts, although temperatures were a little warmer (see also Hood et al. 2001). Figure 3a shows that the MWFM-2 predicted peak mountain wave temperature amplitudes of up to ∼5–10 K over southern Scandinavia within the core of these low synoptic temperatures. These waves can drop local temperatures (T − T̂PEAK) below TICE ∼186 K, and thus mountain wave–induced type II PSCs were forecast to form in this region.
2) Observational validation
As the DC-8 circled to land at Kiruna airport (green flight track in Fig. 3a), mission scientists on board visually observed thick iridescent PSCs well to the south: a photograph of these clouds from the DC-8 is shown in Fig. 3b. A type I PSC detection was reported by the orbiting Polar Ozone and Aerosol Measurement (POAM) III instrument at the location shown in Fig. 3a, as well as at other locations in the days before and after (Hood et al. 2001). The following day, spectacular banded PSCs were visible from the ground near Oslo, Norway, and were again photographed (Fig. 3c). The strong banding evident in the PSCs in Fig. 3c is a fingerprint of clouds formed in the cooling phases of a gravity wave, and strongly suggests that local stratospheric mountain wave activity was controlling their formation and microphysical evolution. Their excellent visibility to remote photography suggests high opacity, strongly suggesting type II ice PSCs. Thus, these visual observations are consistent with the MWFM-NOGAPS forecast in Fig. 3a of mountain wave–induced type II PSCs forming near Oslo.
As shown in Fig. 3a, Stavanger, Norway, is located near the region of largest predicted stratospheric wave amplitudes. The routine 1200 UTC radiosonde sounding from Stavanger on 30 November 1999 acquired winds and temperatures up to ∼15 hPa. Assuming a typical ascent velocity of 5 m s−1 (e.g., Lane et al. 2000), trajectory calculations based on the radiosonde wind data place the balloon ∼3° to the east and slightly south of Stavanger when it reached 30 hPa, which is within the region of large predicted MWFM-2 ray amplitudes in Fig. 3a
Figure 4 shows a plot of stratospheric temperatures acquired from this radiosonde ascent. At ∼15–50 hPa a coherent wavelike oscillation with a peak amplitude ∼5 K is evident, superimposed on background temperatures ∼190 K. The radiosonde horizontal winds (not shown) also display a coherent stratospheric oscillation that is 90° out of phase with this temperature oscillation, consistent with polarization relations for a gravity wave. Thus, these data support the MWFM-2 forecast of mountain wave amplitudes of up to 5–10 K at 30 hPa, superimposed on background temperatures of ∼189–190 K: see Fig. 3a. The wave oscillation drops temperatures in Fig. 4 below TICE within a narrow layer centered at ∼35 hPa. This confirms the original forecast of synoptic temperatures too warm to form ice, but of ice forming in the cooling phases of mountain waves, further reinforcing the interpretation of the PSCs in Figs. 3b and 3c as mountain wave–induced type II. Given the extensive PSC displays, it suggests larger wave amplitudes at different locations and times than those in Fig. 4 that dropped temperatures within larger volumes further below TICE.
b. 25–27 January 2000: PSCs over northern Scandinavia
1) MWFM
Figure 5 shows MWFM-2 forecasts, using NOGAPS forecasts as initialization, of peak mountain wave temperature amplitudes T̂PEAK at 30 hPa over Scandinavia, all valid for 26 January 2000 at 1200 UTC, ranging in forecast time from +96 h down to +12 h. These plots depict the kind of evolving guidance these forecasts provided for in-field flight planning purposes.
Synoptic 30-hPa NOGAPS temperatures, overplotted as blue contours in Fig. 5, predicted very cold regions over northern Scandinavia, produced by a weak wave-1 disturbance that pushed the core of the coldest intravortex temperatures toward Scandinavia. We noted in the field, however, that NOGAPS stratospheric temperatures were sometimes systematically too cold compared to other NWP models by anywhere up to ∼2–5 K. This is one such case. The origin of this cold bias is not clear, but may be related to poor vertical resolution in the stratosphere, a low 1-hPa model top, cold bias in the operational assimilation, and no parameterized nonorographic gravity wave drag (e.g., Shepherd 2000; Austin et al. 2003). Nonetheless, during SOLVE/THESEO 2000 we found that, apart from this episodic cold bias, NOGAPS prognostic tropospheric and lower-stratospheric meteorology compared quite well with that from other global NWP models. Thus, NOGAPS fields provided reliable initial meteorological conditions for MWFM mountain wave forecasting, so long as occasional systematic stratospheric temperature underpredictions were borne in mind when interpreting forecast maps like those in Fig. 5.
At +96 h (4 days out), MWFM-2 predicted strong mountain wave amplitudes over southern Scandinavia, with T̂PEAK ≈ 10 K, as well as a zone of weaker wave amplitudes to the north near Kiruna: see Fig. 5h. These and other forecasts provided advanced warning to flight planners of the potential for wave PSCs to form over Scandinavia in the coming days. Over the next 24 h, the MWFM-NOGAPS forecasts changed somewhat, with predicted wave amplitudes to the south weakening and those to the north intensifying, such that the +72-h forecast showed T̂PEAK ≈ 8–10 K near Kiruna and T̂PEAK ≈ 6–8 K over southern Scandinavia. Thereafter, the MWFM-2 forecasts did not change substantially, continuing to show strong ∼10 K mountain wave amplitudes over northern Scandinavia on 25–27 January. MWFM-2 forecasts initialized with NCEP AVN/MRF and GMAO GEOS-3 forecasts showed the same basic features, increasing confidence in the forecasts. Combined with the low predicted stratospheric temperatures near Kiruna, these forecasts predicted mountain wave type I–II PSCs forming over this region on 25–27 January.
2) MM5
Once early MWFM forecasts were predicting mountain wave PSCs over Scandinavia, MM5 forecasts received close scrutiny. DLR's MM5 forecasting was pioneered during earlier European campaigns in Scandinavia (Dörnbrack et al. 1998). Those forecasts had shown a stratospheric mountain wave response over the Scandinavian mountains that was often dominated by long horizontal wavelength waves that were well resolved by MM5 and thus could be forecast quite accurately.
MM5 forecasts valid for 1200 UTC 26 January 2000 predicted a large-amplitude long-wavelength (λh ∼400 km) mountain wave propagating into the stratosphere over northern Scandinavia. The +24-h MM5 30-hPa temperature forecast is shown in Fig. 6b. Indeed, the vertical and horizontal wavelengths, λz and λh, were sufficiently long that this wave also appeared in global ECMWF IFS forecast temperatures, as shown in Fig. 6a. Just south of Kiruna, the forecast MM5 wave temperature oscillations were ∼±9–10 K, yielding 30-hPa temperatures as low as 181 K (Fig. 6), some 5 K below the frost point. Thus, type II PSCs were forecast here. Wave temperature amplitudes in the ECMWF forecasts are ∼50% smaller than MM5 near Kiruna, with a coldest forecast temperature ∼184 K. The amplitude of the wave train downstream of Kiruna in ECMWF IFS is ∼75% smaller than in MM5. Serious wave amplitude underpredictions also occurred in the final ECMWF analysis temperatures (Fueglistaler et al. 2003).
Dörnbrack et al. (2002) showed that the downstream penetration of this wave was due to inertial modifications caused by an intrinsic wave frequency |ω| near f, which in turn is a result of the long horizontal wavelength. Since shorter-λh mountain waves are simulated by MWFM-2 using its shorter-width ridge database over Scandinavia (Fig. 1), the waves it predicted had higher frequencies (ω2 ≫ f 2) and thus did not leak downstream due to inertial effects: their limited downstream penetration in Fig. 5 is due mostly to “ship wake” effects due to wave radiation from three-dimensional ridge features.
The MM5 forecast in Fig. 6b also shows a weaker-amplitude mountain wave over southern Scandinavia with a much shorter horizontal wavelength. MM5 was expected to reliably forecast stratospheric gravity waves with horizontal wavelengths ≳150 km (≳6Δx): waves shorter than this were not reliably simulated (Leutbecher and Volkert 2000; Davies and Brown 2001; Skamarock 2004). Thus, the amplitude of this shorter-λh wave over southern Scandinavia was probably underestimated; indeed, the coarser-resolution ECMWF IFS forecast in Fig. 6a did not resolve it at all. Its amplitude in Fig. 6b is ∼3 K, whereas the forecast MWFM-2 wave amplitudes over southern Scandinavia in Fig. 5a are ∼5 K. In either case, the waves here were predicted to form type I but not type II PSCs.
Various Web site options allowed flight planners to profile these forecast temperature fields in various ways. One such view, depicted in Fig. 6c, was a two-dimensional linear transect that approximately followed a forecast 30-hPa streamline, as might be flown by an aircraft acquiring quasi-Lagrangian data (Wirth et al. 1999). Figure 6e shows the two-dimensional forecast temperature cross section as a function of altitude and flight distance along this transect. Upstream of the Scandinavian mountains between ∼19 and 23 km, synoptic temperatures are below TNAT and so can support type I PSCs. Above and downstream of the mountains, cold regions form along the sloping phase of the forecast mountain wave, which descends from ∼27 to ∼18 km on progressing eastward, with the core of coldest temperatures of ∼182 K at 23 km slightly downstream of the peak topography. Since these temperatures lie well below TICE, they can produce type II PSCs.
These and other MM5 forecast maps provided flight planners with extremely detailed information with which to plan science flights. On 26 January 2000, a planned DLR Falcon flight profiled various cold wave phases in these MM5 stratospheric temperature forecasts using OLEX, in hopes of measuring the predicted wave PSCs. Two west-to-east transects were flown, as well as two other transects along the mountains designed to track the coldest forecast phases of the wave.
One of the eastward flight segments flown by the Falcon was almost identical to the one shown in Fig. 6c. The 1064-nm OLEX ABRs, S1064, from this flight segment are plotted in Fig. 6d (after Dörnbrack et al. 2002). As the Falcon flew from left to right in Fig. 6d, OLEX first measured S1064 ∼10–30 at ∼22 km, indicating type I PSCs consistent with TICE < T < TNAT at this altitude in the MM5 forecast. Farther along, two tilted layers containing S1064 ≫ 50–100 reveal type II PSCs that correlate with temperatures below TICE in the cold phases of MM5's predicted mountain wave. Dörnbrack et al. (2002) used a more sophisticated lidar data classification scheme to argue that the thick core of enhanced backscatter is produced by type II ice particles, with type Ia and Ib particles surrounding the upper and lower edges of these ice layers (see also Fueglistaler et al. 2003).
3) 3DVOM
The MM5 and MWFM forecasts of mountain wave PSCs forming quite near Kiruna led us to focus on the high-resolution 3DVOM forecasts centered within the small domain near Kiruna (Fig. 2).
Figure 7a shows the 3DVOM +12-h forecast of mountain wave temperature perturbations T ′(x, y, z) at z = 24 km over northern Scandinavia, valid for 1200 UTC on 25 January 2000. The model predicted a wave field with phase lines oriented approximately perpendicular to the flow, and a dominant horizontal wavelength λh ≈ 15 km. The maximum peak-to-peak amplitude is 3.85 K (T̂PEAK ≈ 2 K).
Figure 7b plots a cross section of stratospheric potential temperatures along the transect shown in pink in Fig. 7a, revealing isentropic wave-induced vertical displacements of ∼300 m peak to peak at 22 km. On this day, a DLR Falcon flight was planned and executed to profile forecast mountain wave–induced PSCs near Kiruna. OLEX data from a flight segment near Kiruna are plotted in Fig. 7c, and show small-scale wavelike vertical displacement oscillations in the measured aerosol layers, much as forecast by 3DVOM. High-resolution MM5 hindcasts for this day by Dörnbrack et al. (2002) simulated wave breaking at ∼26–28 km, some 2 km above the top edge of the PSC layer in Fig. 7c, which led them to associate the “ripples” near the top of the PSC with Kelvin–Helmholtz instabilities generated within the long-λh wave that produces the broad PSC layer. This explanation cannot, however, explain the longer-wavelength larger-amplitude vertical displacement oscillations at z ∼20–23 km in Fig. 7c, which must be short-wavelength gravity waves similar to those forecast here by 3DVOM and MWFM.
However, peak-to-peak vertical displacements evident in the lidar imagery are somewhat larger than the Boussinesq 3DVOM model's forecast isentropic vertical displacements in Fig. 7b. To investigate this, postmission hindcasts were performed using the anelastic 3DVOM model. Figure 8 plots 24-h forecasts for the following day (26 January at 1200 UTC), showing Boussinesq and anelastic 3DVOM results in the left and right columns, respectively. Stratospheric wave amplitudes are noticeably larger in the anelastic simulation: at z = 24 km, maximum and minimum temperature perturbations are 4.4 and −6.1 K, respectively, yielding T̂PEAK ≈ 5 K compared to 2–3 K for the corresponding Boussinesq run. This increased amplitude is reflected in larger vertical displacements of isentropic surfaces in Fig. 8d. Similar increases were noted for an anelastic simulation of 25 January (not shown). These vertical displacements agree better with those seen in the lidar data on these days (Figs. 7c and 6d; see also Dörnbrack et al. 2002). Horizontal wavelengths in the anelastic model runs are also noticeably longer than in the Boussinesq runs.
Forecast wind profiles and associated ℓ2(z) profiles for 1200 UTC 25 January are plotted in Figs. 9a and 9b, respectively. There is a marked reduction in ℓ2(z) in the mid- to upper troposphere, indicating vertical evanescence [ℓ2(z) − K2h < 0] for short-λh waves in this region that will trap these waves in the lower troposphere. Figure 9c plots temperature perturbations from the Boussinesq model cross section in Fig. 7b, from the ground to 30 km. We note strong trapping of wave energy in the troposphere, consistent with the ℓ2(z) profile. Absolute temperature perturbation amplitudes |T ′| are as high as 9 K below 5 km, but decay rapidly above 5 km. The trapping is only partial, however, and a small amount of wave energy is able to “tunnel” through the evanescent layer into the lower stratosphere where ℓ2(z) − K2h > 0 again.
Thus, tropospheric trapping of wave energy had a major effect on the stratospheric wave fields forecast by 3DVOM over Kiruna. The differences in horizontal wavelengths of stratospheric waves forecast in the anelastic solutions are also presumably caused by sensitivity of the wave solution to changes in the expression for the Scorer parameter compared to the simplified Boussinesq expression (7) (Nance 1997). The importance of vertical reflections to these 3DVOM stratospheric forecasts confirmed the perceived importance of including vertical reflections in the MWFM-2 forecast algorithms used during SOLVE/THESEO 2000 (see section 2c).
c. 23 January 2000: PSCs over eastern Greenland
The previous two forecasting examples concentrated on mountain wave PSCs over Scandinavia that were relatively short flight distances from Kiruna. However, the NASA DC-8 can fly continuously for up to 10 h, permitting a large area of potential airborne operations shown in Fig. 1. This area encompassed possible overflights of mountainous Arctic terrain away from Scandinavia, such as Iceland, Greenland, and Svalbard. In addition, permission was granted in some cases for flights into Russian airspace, bringing mountains over Novaya Zemlya, Severnaya Zemlya, Franz Josef Land, and the Urals into play as well (see section 4e). Wider-area MWFM-2 forecasts provided the only forecast guidance on mountain wave PSCs away from Scandinavia and Spitzbergen.
Figure 10 shows a +12-h MWFM-2 forecast of peak mountain wave temperature amplitudes at 30 hPa over an extended geographical region of the Arctic to the west of Kiruna, valid for 23 January 2000 at 1200 UTC. Unlike the previous examples, on this day there was no significant mountain wave activity predicted over Scandinavia. However, large-amplitude stratospheric mountain waves were predicted over the southern and central coasts of Greenland. Indeed, MWFM-2 had been recurrently forecasting large-amplitude waves along the east coast of Greenland for a number of days before this.
Coupled with the cold 30-hPa synoptic temperatures T also predicted here on this day, this forecast predicted mountain wave–induced type II PSCs forming along the eastern coast of central Greenland (i.e., T − T̂PEAK < TICE), which was of considerable scientific interest. Studies of PSCs in previous Arctic winters had suggested, with the aid of MWFM hindcasts, that type II PSCs formed recurrently within mountain waves generated by flow across the east coast of Greenland (e.g., Carslaw et al. 1998a). Furthermore, microphysical trajectory calculations suggested that these eastern Greenland wave-induced ice PSCs produced a surviving wake of solid type I PSC particles nucleated on ice that could be transported long distances downstream, possibly to regions in and around Kiruna (Carslaw et al. 1999; Tsias et al. 1999). Yet no direct experimental verification existed for either stratospheric mountain waves over eastern Greenland, or for PSC particles forming within and surviving downstream of these predicted waves. Thus, a segment of the DC-8 flight for this day was planned to fly beneath and then downstream of these forecast mountain wave ice PSCs. The complete flight track flown on 23 January 2000 is plotted in green in Fig. 10.
Figure 11a focuses on the DC-8 flight segment across the east coast of Greenland and then downstream, from 1500 to 1630 UTC. Figure 11b plots S1064 from the LaRC lidar during this flight period. On either side of the western apex point of the flight at ∼1535 UTC, the lidar measured thick layers of enhanced ABR (S1064 ∼100–500), indicating the presence of type II PSCs (see also Hu et al. 2002; Svendsen et al. 2005). This thick type II PSC layer slopes upward with increasing time along the westward-heading flight segment (1510–1530 UTC), then slopes back downward with time after the plane has reversed course to eastward (1540–1600 UTC). This sloping PSC structure is consistent with ice forming in the coldest temperature phases of a long-wavelength mountain wave propagating westward with respect to the eastward background winds: see, for example, the earlier mountain wave PSC example in Fig. 6d. Smaller-scale wave structure is also visible in Fig. 11b.
Figure 11c shows the MWFM-2 +12-h forecasts of mountain wave peak temperature amplitudes at 30 hPa from Fig. 10, computed along the DC-8 flight track. The solid curves show mean values, computed by averaging all rays located within 0.25° latitude and 0.5° longitude of the current flight position. The dotted curves show the corresponding maximum T̂PEAK value of all the rays within this interval. The gray curves show results using both of the ridge databases in Fig. 1, while black curves use only the narrow-ridge database. All forecast curves show enhanced mountain wave amplitudes from 1520 to 1550 UTC, times which correspond to the measurement of sloping type II PSC layers in Fig. 11b. Figure 11d repeats the average peak amplitude calculations for narrow-ridge runs for forecast times ranging from 12 to 60 h, showing that enhanced wave amplitudes along this flight segment were a robust MWFM prediction out to 2.5 days before the flight.
Thus, our forecasts are consistent with the type II mountain wave PSC layer that was observed from the DC-8 in Fig. 11b. These forecast-guided DC-8 observations provided the first direct observational validation that stratospheric mountain waves form over eastern Greenland and produce type II PSCs. Moreover, we see a similarly thick layer of enhanced but lower ABRs in Fig. 11b downstream of the Greenland coast (1600–1630 UTC). In fact, this layer was observed in the lidar data almost all the way back to Kiruna (∼1730 UTC). Hu et al. (2002) studied this downstream S1064 layer and identified it as a surviving wake of solid NAT particles emanating from the upstream mountain wave PSCs.
d. 14 January 2000: PSCs over northern Greenland
In the three previous cases, mountain wave PSCs were predicted by the models and observed from the DC-8 and Falcon. We now consider two other flights in which mountain wave PSCs were not predicted, but were reported in the DC-8 lidar data by Hu et al. (2002).
On 14 January 2000, both the DC-8 and ER-2 conducted ferry flights to Kiruna from DFRC in California and Westover Air Force Base (AFB) in Massachussetts, respectively, to commence phase 2 of SOLVE/THESEO 2000. To support these flights, MWFM T̂PEAK forecasts were issued over a very broad area of the Arctic encompassing North America. In Fig. 12 we focus on a smaller region encompassing Greenland and Scandinavia, showing MWFM-2 forecasts of peak mountain wave vertical velocities ŴPEAK at 50 hPa, rather than T̂PEAK values at 30 hPa, for reasons that will become clear shortly.
The DC-8 flight track across northern Greenland in Fig. 12 intercepted very little forecast stratospheric mountain wave activity. This led MWFM-2 to predict no mountain wave PSCs along this flight segment. However, as the DC-8 crossed the east coast of Greenland, LaRC lidar data (not shown) recorded S1064 ≳ 100 and strong depolarization at ∼22–24 km (Hu et al. 2002; Luo et al. 2003), which implies the presence of an ice PSC. Since synoptic temperatures were too warm for ice PSCs to exist, Hu et al. (2002) and Luo et al. (2003) identified this cloud as a mountain wave–induced ice PSC. If their intepretation is correct, then the MWFM-2 null PSC prediction in Fig. 12 would imply failure of the forecast guidance in this case.
However, recent work has revealed a nonorographic origin for this PSC that is consistent with the original forecast guidance. A high-resolution mesoscale model simulation by Buss et al. (2004) found that this ice PSC was produced by a stratospheric gravity wave produced by either spontaneous adjustment emission from or unstable vertical shear in the tropospheric jet stream. Their model simulations also revealed little if any collocated mountain wave activity along this DC-8 flight segment across Greenland, since simulations using reduced Greenland topography produced very similar-looking stratospheric gravity wave fields. These findings, then, are in fact consistent with the original MWFM-2 forecast in Fig. 12.
The mesoscale model runs of Buss et al. (2004) did, however, find evidence of stratospheric mountain waves intercepting the ER-2 flight track on the east coast of central Greenland, and little evidence of jet-generated gravity waves there. This too appears to be consistent with the MWFM-2 forecast in Fig. 12, which shows significant mountain wave amplitudes at this portion of the ER-2 flight, with ŴPEAK ≈ 1.5–2 m s−1 at 50 hPa, the approximate cruise altitude of the ER-2 at this time.
To study this more quantiatively, Fig. 13a focuses on the ER-2 flight segment from 1630 to 1815 UTC. Figure 13b plots potential temperatures derived from the ER-2 MTP instrument several kilometers above and below the flight altitude (shown in gray), while Fig. 13c plots MMS vertical velocities W(t) measured at flight altitude. Figure 13d computes the MMS vertical velocity variance using a 50-km along-track sliding window. MTP and MMS data both show enhanced small-scale fluctuations at ∼1710–1735 UTC, and some smaller-amplitude activity immediately before and after this period. There is little or no fluctuating activity from 1630 to 1655 UTC.
Figure 13e shows the variance Ŵ2PEAK/2 of the largest-amplitude MWFM-2 ray in the +12 h forecast located within 0.5° latitude and 1° longitude of the ER-2 flight track, which corresponds roughly to the 50-km averaging window used in Fig. 13d. We see enhanced MWFM-2 ray variance along the flight track from ∼1710 to 1730 UTC with maximum values ∼1.5 m2 s−2, very similar to what is seen in the MMS variances in Fig. 13d. Furthermore, MWFM-2 predicted little or no wave activity from 1600 to 1650 UTC, as observed, despite the fact that this flight segment occurs over elevated surface terrain. There is also evidence of weaker wave activity downstream of Greenland from 1730 to 1800 UTC.
Thus, the ER-2 data validate the salient features of the MWFM-2 forecast along the ER-2 flight track on this day. We note in conclusion that, unlike the “straight shot” flight path of the DC-8 to Kiruna, the ER-2 flight track in Fig. 12 was diverted to the north. This flight path was chosen based on the two regions over southern and southeastern Greenland where MWFM-2 predicted enhanced wave amplitudes in Fig. 12. These wave fields were forecast by MWFM to produce turbulence at 50–70 hPa and lay on the straight-shot ER-2 flight track from Westover AFB to Kiruna. The diverted flight path was allowed to pass through the more northern mountain wave fields predicted by MWFM-2 in Figs. 12 and 13 because these waves were predicted to be nonturbulent. The ER-2 pilot reported light-to-absent turbulence throughout this flight.
e. 10 December 1999: PSCs near Franz Josef Land
Franz Josef Land is an archipelago of over 100 small ice-covered islands in the high Arctic (∼81°N, 45–65°E). A DC-8 flight segment in and around Franz Josef Land on 10 December 1999 recorded moderately enhanced ABRs (S1064 ∼10) and high depolarization, indicative of solid type Ia NAT PSC particles (Hu et al. 2002; Pagan et al. 2004). These flights were designed to sample regions where forecast synoptic stratospheric temperatures were cold enough to support NAT. However, since the ABRs were confined in and around Franz Josef Land and showed structure, Hu et al. (2002) tentatively associated these solid NAT PSCs with mountain waves radiated from Franz Josef Land, with NAT possibly forming on ice produced by mountain wave cooling nearby.
MWFM-2 forecasts predicted essentially no stratospheric mountain wave activity over Franz Josef Land at this time because stratospheric winds were extremely weak (see Fig. 14), presenting critical levels to most mountain waves propagating upward from below. Thus, the MWFM-2 forecasts and postanalyses argue that local mountain wave activity played no role in producing these solid NAT particles, supporting similar conclusions of Hitchman et al. (2003) and Pagan et al. (2004). While Hitchman et al. (2003) found evidence for jet-generated gravity waves organizing type I PSC structures observed in DC-8 lidar data near Novaya Zemlya and Severnaya Zemlya on 7 and 12 December, Pagan et al. (2004) argued that the amplitudes of these waves were too weak to form ice as nucleation sites for NAT. So, while local synoptic temperatures are cold enough to maintain these NAT particles (see Fig. 14c), their nucleation origins remain unclear (Pagan et al. 2004).
5. Summary and conclusions
We have assessed an initial coordinated effort to provide operational multimodel forecasts of mountain wave PSCs for planning scientific measurement flights of instrumented aircraft during the SOLVE/THESEO 2000 campaign in the Arctic winter of 1999/2000. Our observational record of choice for validating these forecasts was lidar aerosol backscatter ratios at 1064 nm, S1064, acquired from the NASA DC-8. On five of these flights, PSCs were observed that appeared to be mountain wave related (Hu et al. 2002). We have compared the forecast guidance with available measurements from the DC-8 and other sources.
The extended wave-induced type I–II PSC event of 25–27 January 2000 was an ideal case for these forecasting models, since it occurred over northern Scandinavia where all three models were issuing forecasts. The models accurately predicted the timing and locations of these mountain wave PSCs up to 3 days in advance. We have shown how the acquired PSC data agreed very closely with the models' predicted PSC locations. These forecasts allowed aircraft and balloons to measure mountain wave–induced PSCs in unprecedented detail (Voigt et al. 2000; Dörnbrack et al. 2002; Hu et al. 2002). These data have since motivated detailed microphysical trajectory modeling studies that have improved our understanding of microphysics and chemical processing of air passing through mountain wave PSCs (Fueglistaler et al. 2003; Svendsen et al. 2005).
Another mountain wave–induced type II PSC was predicted over eastern Greenland on 23 January 2000 in a region where only the MWFM was issuing forecasts. On the DC-8 flight that profiled this predicted mountain wave PSC, onboard lidars measured tilted layers of high S1064 in the stratosphere over eastern Greenland, indicating type II PSCs forming in the cool phases of a mountain wave. Their locations along the flight track correlated closely with regions of enhanced mountain wave temperature amplitudes in the MWFM-2 forecasts. This forecast guidance provided PSC data that verified earlier hypotheses of Carslaw et al. (1999) and Tsias et al. (1999) that Greenland mountain waves can form both type II PSCs within the waves and surviving wakes of solid type Ia PSC particles long distances downstream.
On 30 November 1999, the DC-8 was descending into Kiruna when type II PSCs were observed visually to the south. MWFM-2 forecasts predicted mountain wave–induced ice PSCs forming over southern Scandinavia on this day. Limited available data, based on visual photography and a local radiosonde sounding, all confirm the basic details of the original forecast.
On 14 January 2000, a wavelike PSC was reported in DC-8 lidar data over northeastern Greenland, despite the fact that the MWFM predicted no mountain wave PSCs forming here. Our analysis of this event accords with recent work by Buss et al. (2004), who argued that this PSC was formed by a nonorographic gravity wave radiated from the jet stream, not a mountain wave. We reinforced this conclusion by analyzing MWFM forecasts for the ER-2 flight across the east coast of central Greenland on this same day. Buss et al. (2004) argued that stratospheric mountain waves arose here and intercepted the ER-2 flight track. The MWFM-2 forecasts support this conclusion, and show good agreement with fluctuations in stratospheric MMS and MTP data acquired from the ER-2 along this flight segment.
Finally, on 10 December 1999 type I PSCs were observed in DC-8 lidar data near Franz Josef Land, despite MWFM-2 predicting insignificant mountain wave amplitudes here on this day. Weak stratospheric winds severely suppressed amplitudes of mountain waves that radiated from these islands on this day. Our forecasts support recent conclusions that stratospheric mountain waves from Franz Josef Land played no role in the formation of these clouds (Hitchman et al. 2003; Pagan et al. 2004).
In summary, the PSC data from all five DC-8 flights are consistent with the original forecast guidance. Despite this success, our review has highlighted some areas where future forecasting efforts of this type could be improved.
First, 3DVOM used a linearized set of equations for flow over orography under the Boussinesq approximation, in order to provide computationally fast high-resolution forecasts. Postanalysis of these 3DVOM forecasts revealed systematic underpredictions of wave amplitudes, and better results when anelastic equation sets were used instead. Thus, high-resolution regional orographic flow models of this type should ideally incorporate anelastic rather than Boussinesq formulations for future forecasting efforts.
Second, at least two examples of PSCs measured from the DC-8 during SOLVE/THESEO 2000 were either produced or organized by gravity waves radiated from the jet stream, rather than from mountains (Hitchman et al. 2003; Buss et al. 2004). Nonorographic stratospheric gravity waves were not specifically forecast during SOLVE/THESEO 2000. It is still unclear how common nonorographic gravity wave PSC events are and, thus, how important they are microphysically: some initial diagnostic estimates of jet stream imbalance during the SOLVE/THESEO 2000 winter by Buss et al. (2004) suggest they may be infrequent. Nonetheless, it would be desirable to supplement mountain wave forecasts with forecasts of other nonorographic sources of potentially large-amplitude gravity waves.
These recommendations were integrated into a high-resolution stratospheric forecasting effort for DC-8 flights from Kiruna during the SOLVE-II mission of 2002–03, which was designed to provide the validation data that SOLVE had hoped to provide before the launch of SAGE III was delayed (see, e.g., McCormack et al. 2004). Similar forecasting was also performed in support of NASA's Polar Aura Validation Experiment during January–February 2005. Our assessments here should provide useful guidance for high-resolution forecasting support for future airborne stratospheric science missions, such as the remaining Aura Validation Experiments planned for 2005–07.
Acknowledgments
The second author acknowledges the reliable access to the ECMWF forecast data in the framework of his ECMWF special project. All of us thank the in-field meteorology and flight-planning teams during SOLVE/THESEO 2000 for the open way in which they integrated our new forecasts into overall mission planning. SDE thanks Paul Newman and Leslie Lait of NASA Goddard Space Flight Center for access to GMAO and NCEP fields, and to Tim Hogan and colleagues at NRL's Marine Meteorology Division in Monterey, California, for timely access to operational NOGAPS forecast fields. We also thank Chris Hostetler for his LaRC lidar data that were used in this study. SDE's work was funded by the Office of Naval Research and NASA's Upper Atmosphere Research Program and Geospace Sciences Program. Work by MJM at the Jet Propulsion Laboratory, California Institute of Technology, was performed under contract with the National Aeronautics and Space Administration.
REFERENCES
Austin, J., and Coauthors, 2003: Uncertainties and assessments of chemistry–climate models of the stratosphere. Atmos. Chem. Phys., 3 , 1–27.
Bacmeister, J. T., Newman P. A. , Gary B. L. , and Chan K. R. , 1994: An algorithm for forecasting mountain wave–related turbulence in the stratosphere. Wea. Forecasting, 9 , 241–253.
Bacmeister, J. T., Eckermann S. D. , Newman P. A. , Lait L. , Chan K. R. , Loewenstein M. , Proffitt M. H. , and Gary B. L. , 1996: Stratospheric horizontal wavenumber spectra of winds, potential temperature and atmospheric tracers observed by high-altitude aircraft. J. Geophys. Res., 101 , 9441–9470.
Barker, E., 1992: Design of the navy's multivariate optimum interpolation analysis system. Wea. Forecasting, 7 , 220–231.
Becker, G., Müller R. , McKenna D. S. , Rex M. , and Carslaw K. S. , 1998: Ozone loss rates in the Arctic stratosphere in the winter 1991/1992: Model calculations compared with match results. Geophys. Res. Lett., 25 , 4325–4328. Correction. 26 , 327.
Broutman, D., Rottman J. W. , and Eckermann S. D. , 2001: A hybrid method for analyzing wave propagation from a localized source, with application to mountain waves. Quart. J. Roy. Meteor. Soc., 127 , 129–146.
Broutman, D., Rottman J. W. , and Eckermann S. D. , 2003: A simplified Fourier method for nonhydrostatic mountain waves. J. Atmos. Sci., 60 , 2686–2696.
Buss, S., Hertzog A. , Hostetler C. , Bui T. P. , Lüthi D. , and Wernli H. , 2004: Analysis of a jet stream induced gravity wave associated with an observed stratospheric ice cloud over Greenland. Atmos. Chem. Phys., 4 , 1183–1200.
Caplan, P., and Pan H-L. , 2000: Changes to the 1999 NCEP operational MRF/AVN global analysis/forecast system. National Weather Service Technical Procedures Bull. 452, 15 pp. [Available online at http://www.nws.noaa.gov/om/tpb/indexb.htm.].
Cariolle, D., Müller S. , Cayla F. , and McCormick M. P. , 1989: Mountain waves, polar stratospheric clouds, and the ozone depletion over Antarctica. J. Geophys. Res., 94 , 11233–11240.
Carslaw, K. S., and Coauthors, 1998a: Increased stratospheric ozone depletion due to mountain-induced atmospheric waves. Nature, 391 , 675–678.
Carslaw, K. S., and Coauthors, 1998b: Particle microphysics and chemistry in remotely observed mountain polar stratospheric clouds. J. Geophys. Res., 103 , 5785–5796.
Carslaw, K. S., Peter T. , Bacmeister J. T. , and Eckermann S. D. , 1999: Widespread solid particle formation by mountain waves in the Arctic stratosphere. J. Geophys. Res., 104 , 1827–1836.
Chipperfield, M. P., and Jones R. L. , 1999: Relative influences of atmospheric chemistry and transport on Arctic ozone levels. Nature, 400 , 551–554.
Davies, L. A., and Brown A. R. , 2001: Assessment of which scales of orography can be credibly resolved in a numerical model. Quart. J. Roy. Meteor. Soc., 127 , 1225–1237.
Dee, D. P., and Todling R. , 2000: Data assimilation in the presence of forecast bias: The GEOS moisture analysis. Mon. Wea. Rev., 128 , 3269–3282.
Denning, R. F., Guidero S. L. , Parks G. S. , and Gary B. L. , 1989: Instrument description of the airborne Microwave Temperature Profiler. J. Geophys. Res., 94 , 16757–16765.
Derber, J., and Coauthors, 1998: Changes to the 1998 NCEP operational MRF model analysis/forecast system. National Weather Service Technical Procedures Bull. 449, 37 pp. [Available online at http://www.nws.noaa.gov/om/tpb/indexb.htm.].
Dörnbrack, A., Leutbecher M. , Volkert H. , and Wirth M. , 1998: Mesoscale forecasts of stratospheric mountain waves. Meteor. Appl., 5 , 117–126.
Dörnbrack, A., Birner T. , Fix A. , Flentje H. , Meister A. , Schmid H. , Browell E. V. , and Mahoney M. J. , 2002: Evidence for inertia gravity waves forming polar stratospheric clouds over Scandinavia. J. Geophys. Res., 107 .8287, doi:10.1029/2001JD000452.
Dudhia, J., 1993: A nonhydrostatic version of the Penn State–NCAR Mesoscale Model: Validation tests and simulation of an Atlantic cyclone and cold front. Mon. Wea. Rev., 121 , 1493–1513.
Dutton, J. A., and Fichtl G. H. , 1969: Approximate equations of motion for liquids and gases. J. Atmos. Sci., 26 , 241–254.
Eckermann, S. D., 2002: Climatology of mountain wave-induced turbulence in the stratosphere over Central Asia: October–December 1994–2001. Naval Research Laboratory Tech. Memo. NRL/MR/7640-02-8594, 52 pp.
Eckermann, S. D., and Preusse P. , 1999: Global measurements of stratospheric mountain waves from space. Science, 286 , 1534–1537.
Eckermann, S. D., Broutman D. , and Bacmeister J. T. , 2000: Aircraft encounters with mountain wave-induced clear air turbulence: Hindcasts and operational forecasts using an improved global model. Preprints, Ninth Conf. on Aviation, Range, and Aerospace Meteorology, Orlando, FL, Amer. Meteor. Soc., 456–459.
Eckermann, S. D., Ma J. , and Broutman D. , 2004: The NRL Mountain Wave Forecast Model (MWFM). Preprints, Symp. on the 50th Anniversary of Operational Numerical Weather Prediction, University of Maryland, College Park, MD, Amer. Meteor. Soc., P2.9. [Available online at http://uap-www.nrl.navy.mil/dynamics/papers/Eckermann_P2.9-reprint.pdf].
Edouard, S., Legras B. , Lefévre F. , and Eymard R. , 1996: The effect of small-scale inhomogeneities on ozone depletion in the Arctic. Nature, 384 , 444–447.
Flentje, H., Renger W. , Wirth M. , and Lahoz W. A. , 2000: Validation of contour advection simulations with airborne lidar measurements of filaments during the Second European Stratospheric Arctic and Midlatitude Experiment (SESAME). J. Geophys. Res., 105 , 15417–15437.
Fritts, D. C., and Rastogi P. K. , 1985: Convective and dynamical instabilities due to gravity wave motions in the lower and middle atmosphere. Radio Sci., 20 , 1247–1277.
Fueglistaler, S., and Coauthors, 2003: Detailed modeling of mountain wave PSCs. Atmos. Chem. Phys., 3 , 697–712.
Gjevik, B., and Marthinsen T. , 1978: Three-dimensional lee-wave pattern. Quart. J. Roy. Meteor. Soc., 104 , 947–957.
Grell, G. A., Dudhia J. , and Stauffer D. R. , 1994: A description of the fifth-generation Penn State/NCAR mesoscale model (MM5). NCAR Note 398, 121 pp.
Hanson, D., and Mauersberger K. , 1988: Laboratory studies of the nitric acid trihydrate: Implications for the South Pole stratosphere. Geophys. Res. Lett., 15 , 855–858.
Hertzog, A., Vial F. , Dörnbrack A. , Eckermann S. D. , Knudsen B. M. , and Pommereau J-P. , 2002: In-situ observations of gravity waves and comparisons with numerical simulations during the SOLVE/THESEO 2000 campaign. J. Geophys. Res., 107 .8292, doi:10.1029/2001JD001025.
Hitchman, M. H., Buker M. L. , Tripoli G. J. , Browell E. V. , Grant W. B. , Hostetler C. , McGee T. J. , and Burris J. F. , 2003: Nonorographic generation of Arctic polar stratospheric clouds during December 1999. J. Geophys. Res., 108 .8325, doi:10.1029/2001JD001034.
Hogan, T. F., and Rosmond T. , 1991: The description of the Navy Operational Global Atmospheric Prediction System's spectral forecast model. Mon. Wea. Rev., 119 , 1786–1815.
Hood, L. L., Soukharev B. E. , Fromm M. , and McCormack J. P. , 2001: Origin of extreme ozone minima at middle to high northern latitudes. J. Geophys. Res., 106 , 20925–20940.
Hu, R-M., and Coauthors, 2002: Microphysical properties of wave polar stratospheric clouds retrieved from lidar measurements during SOLVE/THESEO 2000. J. Geophys. Res., 107 .8294, doi:10.1029/2001JD001125.
Jacob, D. J., and Coauthors, 2003: Transport and Chemical Evolution over the Pacific (TRACE-P) aircraft mission: Design, execution, and first results. J. Geophys. Res., 108 .9000, doi:10.1029/2002JD003276.
Jakob, C., and Coauthors, 2000: The IFS cycle CY21r4 made operational in October 1999. ECMWF Newsletter, Vol. 87, European Centre for Medium-Range Weather Forecasts, Reading, United Kingdom, 2–9. [Available online at http://www.ecmwf.int/publications/newsletters/list.html.].
Jiang, J. H., Eckermann S. D. , Wu D. L. , and Ma J. , 2004: A search for mountain waves in MLS stratospheric limb radiances from the winter Northern Hemisphere: Data analysis and global mountain wave modeling. J. Geophys. Res., 109 .D03107, doi:10.1029/2003JD003974.
Kanamitsu, M., 1988: Description of the NMC Global Data Assimilation and Forecast System. Wea. Forecasting, 4 , 335–342.
Knudsen, B. M., Pommereau J-P. , Garnier A. , Nunes-Pinharanda M. , Denis L. , Newman P. , Letrenne G. , and Durand M. , 2002: Accuracy of analyzed stratospheric temperatures in the winter Arctic vortex from infrared Montgolfier long-duration balloon flights, 2, Results. J. Geophys. Res., 107 .4316, doi:10.1029/2001JD001329.
Lahoz, W. A., 1999: Predictive skill of the UKMO Unified Model in the lower stratosphere. Quart. J. Roy. Meteor. Soc., 125 , 2205–2238.
Lander, J., and Hoskins B. J. , 1997: Believable scales and parameterizations in a spectral transform model. Mon. Wea. Rev., 125 , 292–303.
Lane, T. P., Reeder M. J. , Morton B. R. , and Clark T. L. , 2000: Observations and numerical modeling of mountain waves over the Southern Alps of New Zealand. Quart. J. Roy. Meteor. Soc., 126 , 2765–2788.
Lee, A. M., Carver G. D. , Chipperfield M. P. , and Pyle J. A. , 1997: Three-dimensional chemical forecasting: A methodology. J. Geophys. Res., 102 , 3905–3919.
Leutbecher, M., and Volkert H. , 2000: The propagation of mountain waves into the stratosphere: Quantitative evaluation of three-dimensional simulations. J. Atmos. Sci., 57 , 3090–3108.
Lipps, F. B., and Hemler R. S. , 1982: A scale analysis of deep moist convection and some related numerical calculations. J. Atmos. Sci., 39 , 2192–2210.
Luo, B. P., Voigt C. , Fueglistaler S. , and Peter T. , 2003: Extreme NAT supersaturations in mountain wave ice PSCs: A clue to NAT formation. J. Geophys. Res., 108 .4441, doi:10.1029/2002JD003104.
Lutman, E. R., Pyle J. A. , Chipperfield M. P. , Lary D. J. , Kilbane-Dawe J. , Waters J. W. , and Larsen N. , 1997: Three-dimensional studies of the 1991/1992 Northern Hemisphere winter using domain-filling trajectories with chemistry. J. Geophys. Res., 102 , 1479–1488.
Manney, G. L., Sabutis J. L. , Pawson S. , Santee M. L. , Naujokat B. , Swinbank R. , Gelman M. E. , and Ebisuzaki W. , 2003: Lower stratospheric temperature differences between meteorological analyses in two cold Arctic winters and their impact on polar processing studies. J. Geophys. Res., 108 .8328, doi:10.1029/2001JD001149.
Marks, C. J., and Eckermann S. D. , 1995: A three-dimensional nonhydrostatic ray-tracing model for gravity waves: Formulation and preliminary results for the middle atmosphere. J. Atmos. Sci., 52 , 1959–1984.
Marti, J., and Mauersberger K. , 1993: A survey and new measurements of ice vapor-pressure at temperatures between 170 and 250 K. Geophys. Res. Lett., 20 , 363–366.
McCormack, J. P., and Coauthors, 2004: NOGAPS-ALPHA model simulations of stratospheric ozone during the SOLVE2 campaign. Atmos. Chem. Phys., 4 , 2401–2423.
Nance, L. B., 1997: On the inclusion of compressibility effects in the Scorer parameter. J. Atmos. Sci., 54 , 362–367.
Nance, L. B., and Durran D. R. , 1994: A comparison of the accuracy of three anelastic systems and the pseudo-incompressible system. J. Atmos. Sci., 51 , 3549–3565.
Newman, P. A., and Nash E. R. , 2000: Quantifying the wave driving of the stratosphere. J. Geophys. Res., 105 , 12485–12498.
Newman, P. A., and Coauthors, 1996: Measurements of polar vortex air in the midlatitudes. J. Geophys. Res., 101 , 12879–12892.
Newman, P. A., Fahey D. W. , Brune W. H. , Kurylo M. J. , and Kawa S. R. , 1999: Photochemistry of Ozone Loss in the Arctic Region in Summer (POLARIS): Preface. J. Geophys. Res., 104 , 26481–26495.
Newman, P. A., Nash E. R. , and Rosenfield J. E. , 2001: What controls the temperature of the Arctic stratosphere during spring? J. Geophys. Res., 106 , 19999–20010.
Newman, P. A., and Coauthors, 2002: An overview of the SOLVE/THESEO 2000 campaign. J. Geophys. Res., 107 .8259, doi:10.1029/2001JD001303.
Pagan, K. L., Tabazadeh A. , Drdla K. , Hervig M. E. , Eckermann S. D. , Browell E. V. , Legg M. J. , and Foschi P. G. , 2004: Observational evidence against mountain-wave generation of ice nuclei as a prerequisite for the formation of three NAT PSCs observed in the Arctic in early December 1999. J. Geophys. Res., 109 .D04312, doi:10.1029/2003JD003846.
Parrish, D. F., and Derber J. C. , 1992: The National Meteorological Center's spectral statistical-interpolation analysis scheme. Mon. Wea. Rev., 120 , 1747–1763.
Pawson, S., and Naujokat B. , 1999: The cold winters of the middle 1990s in the northern lower stratosphere. J. Geophys. Res., 104 , 14209–14222.
Pyle, J. A., and Harris N. R. P. , 1995: Lessons for the future from coordinated European stratospheric campaigns. Phys. Chem. Earth, 20 , 5–12.
Randel, W. J., and Wu F. , 1999: Cooling of the Arctic and Antarctic polar stratospheres due to ozone depletion. J. Climate, 12 , 1467–1479.
Sawyer, J. S., 1962: Gravity waves in the atmosphere as a three-dimensional problem. Quart. J. Roy. Meteor. Soc., 88 , 412–425.
Schoeberl, M. R., 1985: The penetration of mountain waves into the middle atmosphere. J. Atmos. Sci., 42 , 2856–2864.
Schulz, A., and Coauthors, 2001: Arctic ozone loss in threshold conditions: MATCH observations in 1997/1998 and 1998/1999. J. Geophys. Res., 106 , 7495–7503.
Scott, S. G., Bui T. P. , Chan K. R. , and Bowen S. W. , 1990: The meteorological measurement system on the NASA ER-2 aircraft. J. Atmos. Oceanic Technol., 7 , 525–540.
Shepherd, T. G., 2000: The middle atmosphere. J. Atmos. Sol. Terr. Phys., 62 , 1587–1601.
Shutts, G. J., 1998: Stationary gravity-wave structure in flows with directional wind shear. Quart. J. Roy. Meteor. Soc., 124 , 1421–1442.
Skamarock, W. C., 2004: Evaluating mesoscale NWP models using kinetic energy spectra. Mon. Wea. Rev., 132 , 3019–3032.
Skeie, P., and Grønås S. , 2000: Strongly stratified easterly flows across Spitzbergen. Tellus, 52A , 473–486.
Solomon, S., 1999: Stratospheric ozone depletion: A review of concepts and history. Rev. Geophys., 37 , 275–316.
Sparling, L. C., Douglass A. R. , and Schoeberl M. R. , 1998: An estimate of the effect of unresolved structure on modeled ozone loss from aircraft observations of CIO. Geophys. Res. Lett., 25 , 305–308.
Stefanutti, L., MacKenzie A. R. , Balestri S. , Khattatov V. , Fiocco G. , Kyro E. , and Peter T. , 1999a: Airborne Polar Experiment/Polar Ozone, Lee-waves, Chemistry and Transport (APE/POLECAT): Rationale, road map and summary of measurements. J. Geophys. Res., 104 , 23941–23959.
Stefanutti, L., Sokolov L. , Balestri S. , MacKenzie A. R. , and Khattatov V. , 1999b: The M-55 Geophysica as a platform for the Airborne Polar Experiment. J. Atmos. Oceanic Technol., 16 , 1303–1312.