A detailed microphysical model has been used to simulate polar stratospheric clouds (PSC) formed in mountain leewaves over northern Scandinavia and observed in a balloonborne multi-instrument flight on 25 January 2000. The measurements show cloud layers of large solid particles with nitric acid trihydrate (NAT) compositions at relatively high temperatures and layers containing liquid particles with supercooled ternary solution compositions at very low temperatures. The same PSC particle layers have been observed several times during the 2 1/2 h flight, offering a nearly Lagrangian picture of the particle evolution. The applied PSC model describes homogeneous freezing of ice below the ice frost point and diffusion-limited nonequilibrium and size-dependent growth and composition of liquid and solid-phase particles. The microphysical box model calculations are performed on two isentropic surfaces, corresponding to different observed particle layers, using temperature histories from combined high-resolution nonhydrostatic mesoscale and synoptic-scale model analyses of the meteorological conditions characterized by strong mountain leewaves. The calculated particle composition, physical phase, and particle size distributions are compared with the in situ measurements of the same particle properties. It appears that homogeneous freezing of ice in liquid solutions a few degrees below the ice frost point and subsequent release of NAT at higher temperatures might explain the characteristics of the observed solid PSC particles.
If you can't find a tool you're looking for, please click the link at the top of the page to go "Back to old version". We'll be adding more features regularly and your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Polar stratospheric clouds are known to appear either as liquid micrometer-sized droplets or as solid-phase particles or mixtures of both. The cloud particles play a well known role for stratospheric ozone depletion in polar regions by serving as sites for heterogeneous chemical activation of halogen compounds, converting these species into potentially ozone destroying radicals [World Meteorological Organization (WMO), 1999].
 The distinction between liquid and solid phase of the particles is important because the heterogeneous activation depends on the chemical composition and phase of the particles. More importantly, only solid type PSC particles can grow to sizes with significant fall speeds. Sedimentation of PSC particles leads to an irreversible downward transport of nitric acid of which the particles are composed. This denitrification process may prolong the lifetime of reactive chlorine and thereby enhance the chemical ozone depletion.
 PSC particles were originally classified according to their occurrence above (type 1) or below (type 2) the water ice frost point (Tice) [Poole and McCormick, 1988]. Type 1 PSCs were later sub-classified into nonspherical solid particles (type 1a) and spherical liquid particles (type 1b) [Toon et al., 1990]. Early measurements revealed that PSC particles contain nitric acid [Fahey et al., 1989]. The chemical composition of liquid particles was associated with supercooled ternary solutions (STS; H2SO4/HNO3/H2O) [Tabazadeh et al., 1994; Carslaw et al., 1994], which later has been verified directly by in situ measurements [Schreiner et al., 1999]. Nitric acid trihydrate (NAT) is the stable HNO3 hydrate under stratospheric conditions [Hanson and Mauersberger, 1988], and in situ measurements have for the first time shown that solid PSC particles can be composed of NAT [Voigt et al., 2000].
 In the Arctic, denitrification is usually observed without concurrent dehydration [Fahey et al., 1990]. This observation suggests that the denitrification cannot be explained by falling ice particles with inclusions of nitric acid. Denitrification can also not be explained by liquid PSC particles. These particles do not have a nucleation barrier for their growth and the available nitric acid is distributed on all particles in the background size distribution during condensation, thus preventing the formation of large liquid particles with significant fall speeds. Instead denitrification appears to be caused by a selective, but yet unknown, nucleation mechanism responsible for the formation of a small number of large solid particles as first observed in the 1999/2000 Arctic winter stratosphere [Fahey et al., 2001]. Such particles, with radii of 5–10 μm, could induce a substantial denitrification.
 Solid PSC particle formation remains difficult to explain [Tolbert and Toon, 2001]. Ice freezes out of liquid ternary solutions at temperatures several degrees below the ice frost point [Chang et al., 1999]. On synoptic scales in the Arctic such low temperatures occur infrequently whereas solid particles are often observed in large-scale PSCs [Toon et al., 2000]. Large-scale temperature histories indicate that solid PSCs have spent several days close to or below the condensation temperature of NAT (TNAT) [Tabazadeh et al., 1996; Larsen et al., 1997]. It has been suggested that stable NAT particles may form from numerous small particles in a metastable water-rich HNO3/H2O phase [Tabazadeh and Toon, 1996]. Also a slow freezing process, generating nitric acid hydrate particles above Tice, [Salcedo et al., 2001; Tabazadeh et al., 2001], could play an important role.
 Complicated phase transitions have been observed in mountain leewaves, resulting in the formation of solid PSC particles and explained by detailed nucleation processes [Carslaw et al., 1998; Wirth et al., 1999]. It has also been suggested that nonequilibrium conditions in leewave induced temperature fluctuations may drive the smaller particles to compositions with high HNO3 concentrations which may favor a selective freezing into nitric acid dihydrate (NAD) [Tsias et al., 1997]. Solid PSC particles formed in mountain leewaves may have a long lifetime if temperatures stay below TNAT. It remains to be investigated in more detail if particles of this origin could be a source of solid PSCs on synoptic scales in the Arctic as indicated by model calculations [Carslaw et al., 1999].
 In this study we investigate the formation of solid PSC particles in mountain leewaves, using a microphysical PSC model that is constrained by a comprehensive set of simultaneous in situ measurements of the chemical composition and size distributions of PSC particles obtained in the Arctic 1999/2000 winter. Although freezing processes above TNAT are not excluded, the aim is to investigate if a relatively simple homogeneous freezing process of ice out of STS particles at temperatures below the ice frost point [Koop et al., 2000], followed by the release of NAT after ice evaporation at higher temperatures, could explain the observations. Being of primary importance, we start by presenting the measurements and, in particular, those properties of the observed particles, which are important for the model to represent. Then follows a survey of the conditions under which the particles could have formed in terms of the temperature histories that the particles have experienced. The microphysical model is introduced and the model results are finally discussed in comparison with the measurements.
 During the balloon flight on 25 January 2000 from Kiruna in Sweden, PSC particles were measured by a suite of instruments, providing a comprehensive set of simultaneously observed chemical and physical properties of the particles together with a characterization of the gas phase. The measurements include chemical compositions and size distributions of the particles, aerosol backscatter and depolarization (physical phase), number concentrations of condensation nuclei, size distributions of the background aerosol, which originally served as sites for the PSC formation, gas-phase water vapor concentrations, and temperatures. The measurements and instrumentation have been described in detail in the accompanying paper by Schreiner et al.  together with an overall characterization of the meteorological conditions during the flight.
 The balloon made three ascent/descent maneuvers through the PSC layers during the 2.3 h flight in the PSC layers. Estimates from mesoscale model air parcel trajectories in comparison with the course of the balloon indicate that the vertical wind shear within the altitude of the PSCs was low. Therefore, we can assume that the balloon probed the same air parcels within a PSC, in succession 4–5 times, as the air parcels moved downwind on different potential temperature surfaces. In this study we shall concentrate on solid PSC particles observed 5 times at potential temperatures around 508 K and liquid particles observed 4 times around 530 K.
 The particles in the upper layer (530 K) were observed on 25 January 2000 at 21.11, 21.39, 2190, and 2204 UT decimal hours. During the first ascent/descent pair of passes, the particles were observed at very low temperatures (187–191 K) with a chemical composition consistent with STS (H2O:HNO3 molar ratios around 4), low depolarization and color indices, all indicating the bulk of the particles to be liquid PSCs, [cf. Schreiner et al., 2002, Figures 4 and 5]. The cumulated particle size distributions at these two passes are given as black and red dots in Figure 1a, measured by optical particle counter (OPC). The figure shows the number concentration of particles with radii larger than r, N(>r). The curves in the figure illustrate the results of microphysical simulations, which will be discussed later in the paper. It should be noticed that the cloud at the first pair of passes contains relatively high concentrations of small particles with radii up to 0.3 μm and a mode in the size distribution (indicated by the arrow) with larger particles (up to 1.25 μm) at low concentrations around 10−3 cm−3. The particles represent a total condensed volume around 0.1–0.6 μm3 cm−3 (±40%). During the second pair of ascent/descent passes of this layer, the temperature has increased to 191–195 K. The small STS particles have evaporated whereas the large mode remains nearly unchanged, as indicated by green and blue dots in panel a in Figure 1.
 The particles in the lower layer (508 K) were observed 5 times at 21.03, 21.53, 21.78, 22.14, and 2250 UT decimal hours. Except for the first pass, the temperatures were relatively high (190–194 K), and in all cases the chemical measurement showed NAT compositions of the particles (H2O:HNO3 molar ratios around 3.0) and depolarization and color indices were high, [cf. Schreiner et al., 2002, Figures 4 and 5], all indicating the presence of solid PSC particles. Dots with different color-coding in Figure 1b show the size distributions at the five passes of this layer. The size distributions correspond to a condensed volume of 1–3 μm3 cm−3(±40%). A number of features should be noticed in comparison with the size distribution in the upper layer. The concentration of small particles is high even though temperatures are much higher than the first passes in the upper layer while the concentrations of large particles are comparable. The size distributions are nearly unchanged during the five passes of the 508 K layer and there are indications of a number of modes in the size distributions as shown by the arrows. In the following, potential formation processes of the two different particle layers will be described in more detail.
3. Meteorological Conditions
 The flight took place on the inside edge of the polar vortex. As described in greater detail by Schreiner et al.  and Dornbrack et al. , the meteorological situation prior to and during the flight was characterized by strong mountain waves. On 25 January mountain waves developed as a result of strong westerly ground winds forced over the Scandinavian mountains with perturbations penetrating into the stratosphere, resulting in strong adiabatic temperature oscillations in stratospheric air parcels passing the mountain region northwest of Kiruna.
 Short-term mesoscale and long-term synoptic-scale isentropic air parcel trajectories, based on the MM5 mesoscale meteorological model and ECMWF analyses, respectively, have been calculated backward from the time and locations around the last measurements obtained by the balloon. Figure 2 combines a mesoscale analysis for the 5 h prior to observation with a synoptic-scale analysis for the 66 h prior to observation to provide a detailed picture of air parcel temperature histories for the two cases considered 530 K (upper panels) and 508 K (lower panels). The mesoscale trajectories (right-hand side panels) go back to slightly after 1700 UT on 25 January when the air parcels were located west of the Scandinavian mountains. The synoptic-scale trajectories (left-hand side panels) go back to 23 January 0000 UT (plotted as time −48 h) with the air parcels over northeastern Siberia (70°N, 130°E). Figure 2 also shows the ice frost point temperatures at the measured water vapor concentrations (blue curves), the NAT condensation temperatures (assuming 7, 10, 12, and 14 ppbv HNO3 for the lower to the upper green curves), and the SAT melting temperature (orange curves). Red dots show the measured temperatures in the altitude ranges 505–510 K (lower panel) and 527.5–532.5 K (upper panel).
 Although the synoptic-scale trajectories indicate some temperature oscillations passing the mountains around 1700 UT, the mesoscale model shows much larger temperature amplitudes while the phase of the oscillations are nearly the same in the two model results. Compared to the observations, the mesoscale model tends to overestimate the temperatures in the early part of the flight (closest to the mountains) while the measured temperatures at the end of the flight are well simulated.
 It should be noticed that the air parcels experienced temperatures above the SAT melting temperatures less than three days before observations and that the temperatures dropped and stayed below TNAT for less than 24 h thereafter. The lowest temperatures were several degrees below the ice frost point passing over the peak of the mountains between 1900 and 2000 UT and possibly were close to Tice around 1500 UT. Based on this temperature history we assume that the observed solid particles formed during the passage of the mountains. Considering that the solid particles are relatively small, compared to “NAT rocks” [Fahey et al., 2001], and they are observed at upper and middle PSC altitudes, it is not likely that these particles have fallen in from above.
 As input to the microphysical simulations the synoptic-scale temperatures before 1700 UT have been combined with the mesoscale model temperature thereafter. Corrections have been applied for the deviation between measured and modeled temperatures together with some corrections at the lowest temperatures as explained below.
4. Microphysical Model
 A detailed microphysical box model [Larsen, 2000] has been used to simulate the PSC formation as observed during the balloon flight. The model is driven by isentropic temperature histories from the synoptic and mesoscale trajectories and initialized in consistency with all observations.
 The model simulates the changes in size distribution, chemical composition, and physical phase of an ensemble of liquid particles, frozen sulfate aerosol particles, solid-phase PSC particles, and ice. The liquid particles include sulfate aerosol turning into STS particles at low temperatures, type 1b PSC. The frozen sulfate aerosol particles are assumed to be composed of sulfuric acid tetrahydrate (SAT), the type 1a assumed to be composed of NAT with inclusions of SAT, and the ice to contain inclusions of SAT and NAT. The model incorporates a number of microphysical processes. These processes include: (1) homogeneous freezing of ice, NAT (NAD) in STS, (2) nucleation of NAT by vapor deposition of HNO3 on pre-activated SAT, and of ice by vapor deposition of H2O on NAT, () dissolution of SAT at low temperatures in the presence of HNO3 in the gas phase, (4) condensation and evaporation of HNO3 and H2O to and from STS, NAT, and ice particles, and (5) mass balance calculations of these species between gas and condensed phase.
 The structure of the model is illustrated in Figure 3. All condensation/evaporation processes involving STS, NAT, and ice are calculated by the basic vapor diffusion equation [Pruppacher and Klett, 1980], applying a full kinetic approach [Fuchs and Sutugin, 1971; Toon et al., 1989] and taking into account the Kelvin effect, exchange of latent heat, and effects from nonsphericity of solid particles. Vapor pressures over STS, NAT, and ice are taken from Luo et al. , Hanson and Mauersberger , and Marti and Mauersberger , respectively. Liquid sulfate aerosol particles (left-hand side of the figure) take up HNO3 and H2O at decreasing temperatures, turning into fully developed STS particles a few degrees above the ice frost point Tice. During the fast changing temperature conditions in mountain leewaves, the particles will typically not be in equilibrium with the gas phase, and the composition of the liquid particles will depend on their radius.
 At 3–4 K below Tice, freezing of ice out of STS become possible. The model applies the homogeneous ice freezing rates, derived from experimental data, as given by Koop et al. . The much slower homogeneous freezing of NAT (or NAD) out of solution at temperatures between TNAT and Tice [Salcedo et al., 2001; Tabazadeh et al., 2001] is included in the model, but has a negligible effect due to the short duration at temperature below TNAT in the cases studied here (cf. Figure 2). The probability of freezing is proportional to the volume of the particles, and the freezing rate for ice and NAT are illustrated in Figure 4, calculated for different particle sizes. The left-hand side of Figure 4 shows the ice freezing rates. Assuming a critical freezing rate of 1 s−1, complete freezing occurs in a very narrow temperature interval of 0.1–0.2 K at temperatures roughly 3 K below Tice at an H2O/ice saturation ratio around 1.7. In comparison with ice freezing, the NAT homogeneous freezing rates (right-hand side in Figure 4) peak a few K above Tice at HNO3/NAT saturation ratios around 30 with much lower freezing rates. The NAT freezing rates and saturation ratios shown in Figure 4 have been calculated for equilibrium STS compositions as function of temperature, taking account of the uptake of HNO3 and H2O in the solution. It should be noticed that to simulate a partial ice freezing of only the larger particles in a size distribution in a lee wave situation, temperature histories would be needed with an accuracy of a tenth of a degree, which is unrealistic in any meteorological model. It should also be noted that once a fraction of the large-size end of the distribution is frozen into ice particles, fast condensation of water vapor to these particles will decrease the partial water pressure towards equilibrium with ice and thereby require even lower temperatures to freeze the remaining smaller particles. This gas-phase depletion of H2O will cause a substantial broadening of the temperature interval in which freezing of the full range of particles may take place. Therefore accurate cooling rates are also required.
 During freezing it is assumed that each sulfate molecule combines with four water molecules and each nitric acid molecule combines with three water molecules forming SAT and NAT inclusions in the ice particle. Remaining water molecules freeze into ice. Referring to Figure 3, solid type particles holding ice are classified as type 2 PSC particles; solid type particles without ice but still holding NAT are classified as type 1a PSC particles, whereas solid type particles without ice and NAT are classified as SAT particles.
 In comparison with previous microphysical studies of PSC formation in mountain lee waves [Carslaw et al., 1998; Wirth et al., 1999; Tsias et al., 1999, hereinafter referred to as the Mainz model] it should be noticed that we assume heterogeneous nucleation of NAT in all frozen particles, and that ice freezing starts from the large-size end of the liquid particle size distribution. The volume effect in the homogeneous ice freezing process explains the partial NAT nucleation in our model, which therefore becomes very sensitive to the temperature history. In the Mainz model, ice is assumed to form in all particles when temperatures are sufficiently low (≈4K below Tice), but heterogeneous nucleation of NAT only takes place in a fraction of the frozen particles. In the Mainz model NAT nucleation occurs on those ice particles that grow from the smallest droplets, since a liquid layer, surrounding the ice, preferentially breaks on these particles, and the ice becomes exposed to the gas phase. This assumption is motivated by laboratory experiments showing slow nucleation of NAT out of solution containing ice [Koop et al., 1997]. The number of NAT particles assumed (2%–60%) shows to be a critical parameter in terms of explaining the optical characteristics of the PSCs in the Mainz model, and the cooling rate has the largest influence on NAT nucleation although the amount of H2SO4 could play a minor role under background aerosol conditions [Wirth et al., 1999].
 Heterogeneous nucleation of NAT or ice by HNO3 or H2O vapor deposition could take place when cooling SAT particles below TNAT, or cooling NAT particles below Tice, respectively (cf. Figure 3). Although theoretical and laboratory investigations have shown that SAT is not suited for NAT nucleation [MacKenzie et al., 1995; Iraci et al., 1995], other laboratory studies have shown that NAT may nucleate onto pre−activated SAT at HNO3/NAT saturation ratios around 7–13 [Zhang et al., 1996]. In the present model configuration SAT particles have previously been in contact with NAT and thus could be assumed to be pre-activated. The model applies standard heterogeneous nucleation theory [Pruppacher and Klett, 1980], assuming a “compatibility parameter,” m, to describe the ability of nucleation of one crystal structure upon the surface of another crystal. For NAT nucleation upon SAT, m = 0.76 has been adopted which is the upper limit as derived from deposition experiments [Iraci et al., 1995], resulting in nucleation at the above mentioned saturation ratios for pre-activated SAT [Zhang et al., 1996]. Ice nucleation on NAT should be more favorable, and m = 0.95 is used for this process. Dissolution of SAT particles when HNO3 is present in the gas phase [Koop and Carslaw, 1996; Iraci et al., 1998] is included in the model. This process, which is assumed to take place when HNO3/NAT saturation exceeds 15, turning SAT into STS or NAT particles, may compete with the NAT nucleation on pre-activated SAT. Finally, it is assumed that SAT particles will melt into liquid sulfate aerosol particles at temperature higher than 210–215 K [Middlebrook et al., 1993]. Calculation of particle fall speed is included in the model, but sedimentation has a negligible effect in the short-duration processes studied here.
 The model applies “Lagrangian” particle growth in radius space, meaning that radius, chemical composition, and physical phase are calculated for 500 individual particles, each representing a fixed number of particles per kilogram of air taken from the full range of the initial size distribution. The advantage of the Lagrangian approach is that numerical diffusion between size bins during condensation/evaporation is avoided. The disadvantage is that all particles represented by a given size bin must be assumed to freeze when the calculated freezing rate for that bin exceeds a critical value. In reality there will be a small range of size bins where only a fraction of the particles, represented by these bins, will freeze. It is not a serious problem for the fast process of ice freezing in STS solution although small effects can be seen in the size distributions as explained below. In order to simulate slow nucleation processes, a “Eulerian” particles growth model would be more suited [Voigt et al., 2002].
5. Microphysical Simulations
 The model is initialized at −42 UT hours, where the particles at temperatures above SAT melting (cf. Figure 2) could be assumed to be liquid sulfate aerosols in equilibrium with the ambient water vapor. The initial size distributions are estimated from unimodal and bimodal lognormal fits of the size distribution measurements obtained by OPC with a heated inlet [Schreiner et al., 2002]. For the upper layer at 530 K we have used a unimodal lognormal parameter set
and for the lower layer at 508 K, a bimodal parameter set with
where Nt is the total number concentration, rm the median radius, and σ the geometric standard deviation. Gas-phase H2O concentrations have been initialized in agreement with the hygrometer measurements at the isentropic surfaces, using 6.0 and 5.7 parts per million by volume (ppmv) for the upper and lower isentropic surfaces. Unfortunately, simultaneous in situ measurements of gas-phase nitric acid are not available. The initial HNO3 concentrations have been adjusted such that within the clouds calculated total particle volumes are in agreement with the observations. This lead to 7 and 14 parts per billion by volume (ppbv) in the upper and lower layer, respectively. This result is a bit surprising since such high gradients in HNO3 concentrations are not expected over this limited altitude range. One explanation could be that some vertical redistribution of HNO3 by sedimentation of large NAT particles has taken place before the start of the simulations. Sensitivity studies were performed with less severe gradients in HNO3 between the layers and the results are not significantly different.
 Results of microphysical simulation of the particle evolution in the upper layer at 530 K potential temperature appear in Figure 5. Only the last part of the particle evolution from 1500 UT on the day of observation, shortly before the air parcels enter the mountain region, is shown. The results are compared with in situ measurements obtained in the altitude range 527.5–532.5 K potential temperature. Calculated size distributions at the time of the four passes of the 530 K level are shown as solid curves in comparison with the measurements in Figure 1.
 The mass spectrometer measurements clearly show that the bulk of the particles have an STS composition, which is in agreement with the simulation results, showing that the majority of the particles are liquid with STS compositions during the first pass of the upper layer (cf. Figure 5, lower-right and middle-right panels). A second mode in the size distribution with particles larger than 0.5 μm in low number concentrations (≈10−3 cm−3) and not contributing significantly to the total volume (cf. Figure 1) could be explained by the model as being NAT particles. This interpretation is consistent with the observation that these particles remain with nearly the same size distribution in the second pair of passes of the upper layer at higher temperatures close to TNAT where the small STS particles have evaporated in agreement with the simulation. Furthermore, it is not likely that the large mode is composed of liquid particles since this interpretation would also require a large mode to be present in the background sulfate aerosol distribution, which is not confirmed by the heated inlet OPC measurements. As explained in section 3, it is most likely that the large particles formed during the passage over the Norwegian mountains shortly before observation.
 In the simulation (cf. Figure 5, upper right) particles remain in the liquid phase, growing into fully developed liquid PSC particles until shortly after 1900 UT where the mesoscale temperatures drop a few degrees below the ice frost point. As discussed above, temperatures would be needed with an accuracy of less than 0.1 K for an exact simulation of a partial freezing of only the largest sizes among the liquid particles. Here we have adjusted the minimum temperature at 1900 UT hours to obtain the best resulting size distribution of the large-size mode. At this point in the simulation the largest STS particles freeze into ice particles, which grow to sizes up to 10 μm (upper-right panel in Figure 5) with ice volumes up to 6 μm3 cm−3 (middle-right), while a small fraction of the gas-phase H2O is condensing on the particles (lower-left). This thin ice PSC cloud exists for about 1.5 h. Around 2060 UT, temperatures have increased and the ice has evaporated, leaving behind a liquid PSC with a small number of NAT particles which slowly increase their volume as the remaining liquid particles evaporate and HNO3 is transferred to the NAT particles. The evolution of the size distribution during the evaporation process after 2000 UT would also require a more accurate temperature history than can be provided by mesoscale models (cf. Figure 2), but the simulated change in size distributions is not inconsistent with the measurements (middle-left).
 It should be noticed that this case is an example of the development of a liquid PSC, immediately turning into an ice PSC when sufficiently low temperatures are reached. Upon heating, the liquid PSC reappear and a tail of solid PSC particles emerge out of the liquid cloud later in the process. Similar cloud evolution has previously been observed in mountain lee waves and simulated by Carslaw et al.  and Wirth et al. . The present case also represents an example of a potential way to generate large NAT particles. If temperatures had remained below TNAT after the last observation in the upper layer, the small number of NAT particles could have grown to large sizes, gaining significant fall speed and potentially contributing to denitrification through sedimentation.
 Both the Mainz-model [Wirth et al., 1999; Tsias et al., 1999] and the model applied here show that NAT particles develop into a relatively narrow size distribution. The main difference between the two models lies in the representation of the liquid particles in coexistence with NAT particles at low temperatures as observed in the first pair of passes at the 530 K level. In the present model NAT particles nucleate out of the large-size end of the size distribution, leaving the small-size end nearly unaffected in good agreement with the observations. In the Mainz-model, the NAT particles are nucleated out of the small-size end of the size distribution and only low concentrations of liquid particles exist with sizes smaller than the NAT particles, depending on the assumed fraction of the ice particles, which serve as sites for NAT nucleation.
 The present model shows that a gap in size develops between the small liquid and the larger NAT particles (cf. upper right panel in Figure 5 and the flat region between 0.2 and 2 μm in the modeled size distribution in Figure 1a). This feature is an artifact using the Lagrangian approach. In reality there should be some bins in this size range where only a fraction of the liquid particles freeze which, however, is not possible to represent in a Lagrangian approach. This partial freezing is probably the reason for the slope in the observed size distribution, which is only represented by a step in the calculated distribution.
 According to the scenario represented by the model, the observation (at relatively high temperatures) of a large-size mode in the size distribution, as seen in the second pass of the 530 K layer, is a sign of a previous freezing process. This process leaves the solid particles in a nearly fixed size distribution that changes rather slowly as long as temperatures remain between TNAT and Tice. This is actually the situation for the five observations in sequence at the 508 K level. The particles have a clear NAT composition, and there are indications of at least two, perhaps three, modes of medium and large-sized particles in the distribution that do not change significantly between TNAT and Tice (cf. Figure 1). The mesoscale temperature histories indicate that at least two freezing processes could have taken place while the air parcels passed the mountains (cf. Figure 2).
 Results from the simulated particle evolution at the 508 K level are shown in Figure 6 in the same format as Figure 5. Again the minimum temperatures have been adjusted to obtain the best agreement between calculated and measured size distributions. Here, this adjustment gives rise to freezing of a much larger fraction of the liquid particles than in the upper layer. In the first freezing process shortly after 1600 UT, the liquid PSC nearly instantaneously freezes into a dense ice PSC. After evaporation of the ice, liquid PSC particles reappear with a mixture of a small concentration of solid NAT particles. During the subsequent heating the liquid particles evaporate and HNO3 vapor is transferred to the NAT particles, which grow to become the dominant particle type. The particles undergo a very slow change in their size distribution in the time interval where the observations are being performed with chemical compositions and total volumes in good agreement with the observations. Detailed comparison between the observed and modeled size distributions appear in Figure 1. The two modes in the modeled size distributions are the result of the two freezing processes in sequence and are consistent with the observations. In fact, the best fit to the observations is obtained by assuming that the synoptic temperatures drop sufficiently below the ice frost point in the temperature minimum between 1500 and 1700 UT. Results consistent with the observations could also be obtained, assuming the two freezing processes to take place at 1900 and 1970 UT. Again a more accurate temperature history would be required to represent the fine-scale structures in the distribution.
 For the first time, a comprehensive set of in situ measurements of PSC particle properties and a characterization of the gas phase have been obtained in a balloon flight in a mountain leewave situation [Schreiner et al., 2002]. The measurements showed clearly different types of PSC particles at different altitudes. Due to the ascent/descent maneuvers of the balloon and the low vertical wind shear in the PSC, the temporal evolution of the cloud particles could be followed for a period of time. The air parcel temperatures show that most likely the PSC particles were formed just shortly before being observed. This data set and the meteorological conditions have offered a unique opportunity for model simulations to be performed for interpretation of the particle formation processes. The evolution of two clearly distinct types of PSC particles, composed respectively of STS and NAT, has been investigated by a detailed microphysical model based on air parcel temperature histories from combined synoptic and mesoscale trajectory calculations.
 The same freezing and other microphysical processes have been used to describe the evolution of the two different types of particles in good agreement with all simultaneous measurements of the chemical compositions and size distributions. It appears that a relatively simple process of homogeneous freezing of ice in STS solution [Koop et al., 2000], followed by heterogeneous nucleation of NAT, offers a good explanation of the observed characteristics of the solid particles.
 The homogeneous freezing is strongly dependent on temperature, and accurate temperature histories beyond the capability of any meteorological model are required to perform a detailed simulation of the particle evolution. Therefore a firm proof that the homogeneous ice freezing processes have determined the properties of the solid PSC as observed cannot be given. However, a consistent picture does appear that in the upper layer with dominantly liquid STS particles, a partial freezing of the large-size end of the size distribution must have taken place, leaving behind a small mixture of solid particle which remain nearly unchanged up to temperatures close to TNAT. It seems that the freezing process almost fixes the size distributions of the solid particles as long as temperatures remain below TNAT, at least on timescales as observed here. The same fixation of the solid type size distributions can also be noticed in the lower layer with NAT particles. Nearly all particles are frozen and two or three modes of medium and large particles indicate that the particles have experienced a number of consecutive freezing processes as simulated fairly accurately.
 The model described here and that used by Carslaw et al.  and Wirth et al.  represent essentially the same PSC particle evolution; however, differences occur in the process of NAT nucleation, and these differences may be difficult to resolve. The Mainz model of nucleation assumes ice will freeze in all liquid particles when temperatures are sufficiently low; however only a fraction of the ice particles will nucleate NAT. In this model NAT nucleation occurs only from the breaking of a liquid film surrounding the ice. This assumption implies that the fraction of ice particles which serve as particles for NAT nucleation arise preferentially from the small-particle end of the size distribution. In our model we assume that NAT forms in all frozen particles, i.e. the number of ice particles equals the number of NAT particles downstream after ice evaporation. The partial freezing in our model arises from the fact that the largest particles will freeze first, and this implies a very strong dependence on the particle's temperature history.
 As noted by Carslaw et al.  understanding NAT nucleation requires in situ measurements during the nucleation process. The measurements by Carslaw et al.  and Wirth et al.  approached this ideal with lidar observations of air parcels in which ice and NAT nucleation had occurred. These measurements indicated that the number of ice particles is significantly larger than the number of NAT particles downstream, observations that are consistent with the Mainz model. While our measurements did not capture the nucleation process, they do provide detailed in situ measurements of particle size, phase, and composition down wind of the nucleation process. These observations in the upper layer during the first pair of passes suggest that the NAT particles were nucleated from the large-particle end of the size distribution, i.e. the number of NAT particles is consistent with the concentration of the larger particles in the background sulfuric acid size distribution. This observation is consistent with the assumptions concerning NAT nucleation within our model.
 Processes such as NAT or NAD freezing out of STS solution at temperatures above Tice [Salcedo et al., 2001; Tabazadeh et al., 2001] may lead to the formation of large NAT particles on a synoptic scale. These longer-duration processes could not be investigated from the present data set. Rather, our observations and the microphysical simulations clearly demonstrate that NAT can form in mountain leewave cooling events. The exact temperature histories play a dominant role in determining their size distributions ranging from small mixtures of solid particles inside a dominantly liquid PSC to a fully developed solid PSC with clear indications of previous freezing processes evident in the particle size distributions. If only a small fraction of the liquid particles freeze into ice, leaving behind a small number of NAT particles mixed into a dominantly liquid cloud after ice evaporation, as observed in the upper layer, these NAT particles may grow to large sizes and contribute to denitrification, provided that temperature remains below TNAT.
 The balloon flight was performed as part of the European-American SOLVE/THESEO 2000 campaign in winter 1999/2000. We appreciate the excellent work of the balloon operators from CNES. This work has been supported by the Commission of the European Union through the Environment and Climate Programme (contracts ENV4-CT97-0523-PSC ANALYSIS and EVK2-CT-2000-00095-CIPA). T.D, C.K., J.M.R and N.T.K were supported by the U.S. National Science Foundation.