The aircraft-based 2002 Intercontinental Transport and Chemical Transformation experiment intercepted and chemically analyzed pollution plumes transported from Asia to the western United States. The research flight on 10–11 May 2002 detected mixing between polluted and stratospheric air at midtropospheric levels above the California coast. This study uses a Lagrangian domain-filling trajectory technique to illustrate that this event was the result of mixing between two warm conveyor belts (WCB) containing Asian pollution and the remnants of a deep tropopause fold from a downstream midlatitude cyclone (referred to as the stratospheric component of a dry airstream or SCDA). Advection of the trajectory particles shows how the SCDA decayed over 7.5 days. One component dispersed into a downstream WCB, while another component descended into the lower troposphere and became entrained by an upwind WCB. After 7.5 days of transport 22% of the SCDA mass was transported into the troposphere. The portions of the SCDA that penetrated to the lowest altitudes had the greatest likelihood of being transported into the troposphere. For example, over 90% of the SCDA at altitudes below the 600 hPa level was transported to the troposphere, but none of the mass at the 200 hPa level was exchanged. More than half of the exchange occurred during the first 48 hours as the deepest portions of the tropopause fold decayed over the Pacific. The rest of the exchange occurred over the following 5.5 days as the remnants of the SCDA sheared apart along the edge of the stratospheric polar vortex and became entrained into subsequent tropopause folds and vortex breakaway features. Stratosphere to troposphere exchange resulted in the transport of 0.5 Tg of stratospheric ozone to the troposphere during the 7.5 day study period. Roughly half of the SCDA particles that entered the troposphere dispersed into the upwind and downwind WCBs.
 Researchers have known for years that tropopause folds are an integral component of the dry airstream (DA) of extratropical cyclones [Reed, 1955; Browning, 1997; Carlson, 1998]. The DA, also known as the dry intrusion, is a coherent airstream that descends from the upper troposphere and lower stratosphere on the polar side of extratropical cyclones. The DA descends isentropically into the middle and lower troposphere, west of the surface cold front. The descent is generally in an equatorward direction, although portions can advect poleward in the same direction as the cyclones's warm conveyor belt (WCB). The common occurrence of STT in DAs has been documented by Johnson and Viezee  who detected stratospheric ozone intrusions in virtually every upper level trough that they sampled over the central United States in the spring and fall of 1978. Similarly, analysis of daily ozonesondes launched from Virginia and Indiana during the monthlong AEROCE study in the spring of 1996 showed that stratospheric ozone intrusions are highly characteristic of DAs [Cooper et al., 1998]. Understanding of the relationship between tropopause folds and dry airstreams has been furthered by the studies of Cooper et al. [2001, 2002a, 2002b] who analyzed the trace gas composition of DAs and other cyclone airstreams (warm conveyor belt, cold conveyor belt and post cold front airstream) to create a conceptual model of the typical trace gas signatures of midlatitude cyclones.
 The recent literature contains many case studies of the evolution of tropopause folds, and stratospheric intrusions within cut-off lows, either in terms of just the meteorological processes, or also in terms of the transfer of trace gases from the stratosphere to the troposphere [Price and Vaughan, 1993; Lamarque and Hess, 1994; Cox et al., 1997; Wernli and Davies, 1997; Stohl and Trickl, 1999; Wirth and Egger, 1999; Gouget et al., 2000; Vaughan et al., 2001; Zanis et al., 2003]. Other studies have reported observations of exchanged stratospheric ozone immediately adjacent to WCBs containing anthropogenic pollution and describe the meteorological process that brought these air masses together [Prados et al., 1999; Parrish et al., 2000; Cooper et al., 2001; Fischer et al., 2002]. Also, a recent study by Esler et al.  gives detailed descriptions of mixing between stratospheric intrusions and WCBs in cold frontal regions on timescales shorter than 2 days above the United Kingdom during spring. The present study builds on these earlier findings and is unique because it not only considers the multiday evolution and decay of a tropopause fold, but for the first time explores the extent to which tropopause fold remnants disperse into adjacent WCBs over several days and across much of the Northern Hemisphere. This Lagrangian-based analysis illustrates how a tropopause fold is transported within a midlatitude cyclone in the North Pacific Ocean during May 2002. The fold is followed over 7.5 days as it decays and disperses into regions occupied by its companion WCB and two upwind WCBs. Finally, it is shown that the modeled transport of these airstreams clearly explains aircraft-based observations of mixing in the midtroposphere between polluted air within the remnants of a WCB and stratospheric air within a tropopause fold.
 This analysis presents results from the Intercontinental Transport and Chemical Transformation (ITCT 2K2) experiment, an aircraft-based study conducted from Monterey, California, during April and May 2002. ITCT 2K2 was designed to intercept and chemically analyze pollution plumes that had traveled from Asia across the North Pacific Ocean to the western United States. In addition to the flights conducted from Monterey, aircraft measurements were made above Washington State, surface measurements were made at the coastal sites of Cheeka Peak Observatory, Washington (48.3°N, 124.6°W, 480 m), and Trinidad Head, California (41.1°N, 124.2°W, 107 m), and daily ozonesondes were launched from Trinidad Head. The measurement and modeling techniques from the experiment are described in section 2. The results from this analysis are presented in section 3, and are discussed in section 4.
 Three-dimensional forward and backward trajectory ensembles were calculated at the NOAA Aeronomy Laboratory using the u, v and w wind fields from the NOAA NCEP Final Analyses (FNL). The global FNL data were downloaded from the National Center for Atmospheric Research data archive, available every 6-hours with a horizontal grid spacing of 1° × 1°, and 21 vertical levels between 1000 and 100 hPa. The wind field data were interpolated onto a terrain-following sigma coordinate system with horizontal grid spacing of 1° × 1°, and 22 vertical levels between the surface and 100 hPa. The three-dimensional trajectories were calculated using a linear interpolation scheme in space and time. The trajectory calculation routine has been written as a forward and backward trajectory model, called FABtraj, and its verification is discussed by Cooper et al. .
 Error in trajectory calculations [Draxler, 1991; Stohl et al., 1995] is the result of 1) the inability of any model-generated set of wind fields to accurately reproduce the horizontal and vertical motions of the atmosphere, 2) the failure of the temporal and spatial output of the wind fields to resolve all atmospheric motions important for air mass transport, and 3) errors arising from the spatial and temporal interpolation of gridded wind field data. The 1° × 1° FNL analyses used in this study, do not resolve all atmospheric boundary layer (ABL) processes or the vertical motions of individual convective cells in the free troposphere. However, they capture well synoptic scale subsidence and the vertical motions associated with slantwise ascent within midlatitude cyclones, such as the vertical lifting of air through WCBs.
 Quantifying the error in trajectory calculations can be difficult as the true trajectory path is not usually known. Several studies have compared calculated trajectory paths to the known paths of inert chemical trace gases, smoke and dust plumes and constant volume balloons. These studies are reviewed by Stohl  who concludes that three-dimensional trajectories are more accurate than other trajectory types, such as isentropic, however the position errors associated with three-dimensional trajectories are typically 20% of the travel distance, and can be much more under critical flow situations. Recent studies have identified two additional sources of trajectory uncertainty resulting from air parcel filamentation [Stohl et al., 2002] and inconsistencies between wind field analyses [Stohl et al., 2004].
 In this study trajectories are used to illustrate the transport and decay of midlatitude cyclone airstreams. Three warm conveyor belts and the stratospheric component of a dry airstream are modeled by filling their spatial domains with trajectory ensembles at a mature phase of their life cycle, following the method of Cooper et al. . In this study the WCBs were defined using the broad definition of a WCB as the general region of ascent beneath and within the WCB cloud shield (see Cooper et al.  for a discussion of the WCB definition). The spatial domains of the WCBs were determined with 3-dimensional humidity, wind and temperature fields from the same FNL analyses used to calculate the trajectories. An air parcel was classified as belonging to a WCB if it had a relative humidity value greater than 80%, isentropic potential vorticity (IPV) less than 1.5 potential vorticity units (PVU), and was under the region of the cloud shield visible in the infrared satellite imagery. Gradients in potential temperature were used to discriminate between the WCBs and near-by humid air masses.
 IPV is a quasi-conservative dynamic tracer of stratospheric air and can be used to define the tropopause. Previous definitions of the tropopause typically range between 1 PVU [Shapiro, 1980] and 2 PVU [Appenzeller et al., 1996; Lamarque et al., 1996] but have been stated as high as 4 PVU (see discussion by Cox et al. ). A current approach, and one that is adopted in this study, is to consider the region of the atmosphere between 1 and 2 PVU as a mixing zone between the stratosphere and troposphere [Stohl, 2001]. The stratospheric component of the DA (hereinafter referred to as SCDA) is defined as the portion of the cyclone with IPV > 1.5 PVU, extending as high as the top of the WCB (12 km above sea level). The tropospheric component of the DA is not considered in this study. The SCDA contains portions of the lowermost stratosphere that remain in the stratosphere throughout the study period and a very deep tropopause fold that eventually mixes into the troposphere.
 The spatial domains of the airstreams were filled with trajectory particles initialized on a regular grid with horizontal spacing of 22 km and vertical spacing of 25 hPa. The motions of the trajectory ensembles were inspected for 24 hours to ensure that the trajectory particles were flowing as a coherent air mass. Outlying particles that sheared away from the airstream within the first 24 hours were removed from the analysis.
2.2. Satellite Imagery
 Hourly infrared satellite imagery from the Japanese GMS-5 and the NOAA GOES-West geostationary satellites were downloaded and archived via near real-time Internet delivery during the ITCT 2K2 field campaign. Both satellites are located 42,168 km from the center of the Earth and positioned above the equator, with GMS-5 above 140° east longitude and GOES-West above 135° west longitude. In general terms these satellite images depict the temperature of the Earth as seen by the imagers on board the satellites, which measure the intensity of the radiation emitted by the Earth between 10.2 and 11.2 μm. McIDAS software was used to remap and combine the temperature values from these two satellites to create an equidistant cylindrical projection of the entire Pacific basin at 5 km horizontal resolution. In this paper the images are colored such that yellows and reds represent the cold upper-level cloud tops, greens represent the tops of the warmer midlevel and low-level clouds, and blues are the much warmer surface of the Earth.
 A useful tool for discerning between air masses of polar and subtropical origin in the midtroposphere to upper-troposphere is the newly developed altered water vapor product [Wimmers et al., 2003]. These images are produced by correcting the GOES satellite 6.7 μm water vapor channel for temperature and zenith angle bias [Moody et al., 1999; Wimmers and Moody, 2001]. The result is a depiction of specific humidity, rather than relative humidity [Soden and Bretherton, 1993], in the midtroposphere to upper troposphere. Because the standard GOES water vapor product yields a depiction of relative humidity in the midtroposphere to upper troposphere, dry air masses of polar origin can appear moist due to their very cold temperatures. The moist appearance of these dry polar air masses can make it difficult to distinguish them from the warmer, moister subtropical air masses. In contrast the altered water vapor product's depiction of specific humidity allows the cold and dry air masses of polar origin to be clearly distinguished from the moist and warmer subtropical air masses.
2.3. Aircraft Measurements
 During ITCT 2K2 the NOAA WP-3D aircraft, based in Monterey, was equipped with a suite of instruments to measure a wide range of chemical species and aerosol particles along the west coast of the United States. In this study we report measurements of ozone and CO. These species were measured as 1 second averages, which corresponds to approximately 100 m horizontal resolution and better than 10 m vertical resolution during ascents and descents. O3 was measured by a NO-O3 chemiluminescence instrument similar to that described by Ridley et al. . During the pre-flight period the chemiluminescence instrument was calibrated with a commercial UV absorption instrument (TECO Model 49) in prepared mixtures of O3 in zero air. During flight both instruments measured ambient O3. Ryerson et al.  give more details of the O3 measurement. The total uncertainty of the O3 measurements was ±(0.4 ppbv +2%). CO was measured by a vacuum UV fluorescence instrument similar to that described by Gerbig et al. . Holloway et al.  describe the operation of the instrument including the calibration and zeroing procedures. The one-sigma precision of the results was better than 1 ppbv, and the accuracy was within 2.5%.
 During ITCT 2K2, daily ozonesondes were launched from Trinidad Head, a coastal site in northern California (41.1°N, 124.2°W, 107 m). The balloon-borne ozonesondes were equipped with the widely used and tested electrochemical concentration cell (ECC) sensor [Komhyr, 1969; Komhyr et al., 1995]. Personnel training and instruments were provided by the Ozone and Water Vapor Group of the Climate Monitoring and Diagnostic Laboratory of the National Oceanic and Atmospheric Administration (NOAA). See Oltmans et al. , for an explanation of the equipment and techniques employed during this and many other studies. The ozonesondes produced vertical profiles of ozone, temperature and frost point between the surface and approximately 35 km above sea level (a.s.l.). The data were partitioned into 0.25 km vertical layers, and reported as layer averages.
2.5. FLEXPART Simulated Ozone
 Using the FLEXPART Lagrangian particle dispersion model [Stohl et al., 1998; Stohl and Thomson, 1999], we calculated the average ozone mixing ratios of stratospheric ozone transported into the troposphere along the North American west coast during ITCT 2K2 (0000 UTC, 20 April 2002, to 0000 UTC, 21 May 2002). FLEXPART simulates the transport and dispersion of non-reactive tracers based on operational data from the European Centre for Medium-Range Weather Forecasts [European Centre for Medium-Range Weather Forecasts (ECMWF), 1995], available every 3 hours (analyses at 0000, 0600, 1200, 1800 UTC; 3-hour forecasts at 0300, 0900, 1500, 2100 UTC) with 1° × 1° horizontal grid spacing on 60 vertical levels. Boundary layer turbulence is parameterized by solving Langevin equations [Stohl and Thomson, 1999], and the tracer concentrations on a three-dimensional grid are determined by applying a kernel method. FLEXPART has been used to simulate stratospheric intrusions over the Alpine region [Stohl et al., 2000] and to identify the stratospheric influence along flight tracks of aircraft measurement campaigns [Stohl et al., 2003b].
 Here we simulate the transport of two stratospheric tracers, the first simply represents stratospheric air, and the second represents stratospheric ozone. Neither simulation includes ozone photochemistry. The model domain extends from 80°E to 80°W and from 10°N to 90°N. Approximately 5 million particles were initialized with uniform distribution throughout the stratosphere (defined as the regions of the atmosphere with IPV greater than 1.6 PVU), such that particles at higher altitudes (where the density is lower) carry less mass than particles at lower altitudes. Particles are terminated at the out-flowing boundary and are generated at the in-flowing boundary according to the initialization procedure. The stratospheric air tracer is initialized with a value of 1 in the stratosphere and 0 in the troposphere. Stratospheric ozone was initialized using the relationship ozone(ppbv) = S (ppbv/PVU) × PV(PVU) [Danielsen, 1968; Beekmann et al., 1994]. S varies with season having a maximum in spring and a minimum in fall. Here, we use S = 70 ppbv/PVU which is comparable to the springtime values reported by Ancellet et al. . Every particle is tagged with a time flag, which is zero in the stratosphere and starts to count as soon as the particle enters the troposphere. If a particle moves back to the stratosphere the clock is reset.
2.6. MOZART-2 Model Description
 MOZART-2 (Model of Ozone and Related Chemical Tracers, version 2) is a global chemical transport model designed to simulate the distribution of tropospheric ozone and its precursors [Horowitz et al., 2003]. The model simulates the concentrations of 63 chemical species from the surface up to the middle stratosphere. In this study, the model is driven with meteorological inputs from the NCEP Aviation (AVN) model analyses, which have a resolution of T170 (approximately 0.7° latitude × 0.7° longitude) with 42 vertical (sigma) levels extending up to 2 hPa. The meteorological fields are averaged to a horizontal resolution of 1.9° latitude × 1.9° longitude, the resolution at which MOZART-2 is run. A timestep of 15 minutes is used for all chemistry and transport processes.
 Stratospheric concentrations of ozone are constrained by relaxation toward zonally and monthly averaged values from observed ozone climatologies from Logan  (for O3 below 100 mb) and HALOE [Randel et al., 1998] (for O3 above 100 mb). This relaxation is performed from the local thermal tropopause (defined by a lapse rate of 2 K km−1) to the model top at each timestep, with a relaxation time constant of 10 days.
 A supplemental animation is available from the JGR-Atmospheres server. Viewing the animation is not necessary for understanding the results presented in the manuscript, but doing so would allow the reader to see more clearly how a stratospheric intrusion decays and disperses into adjacent WCB's. The animation, corresponding to Figure 6, is in QuickTime format and the QuickTime Player can be freely downloaded from http://www.apple.com/quicktime.
3.1. Relevance of the Case Study
 The ITCT 2K2 flight on 10 May 2002 was the eighth of thirteen research flights by the NOAA WP-3D, with this particular flight designed to intercept a deep stratospheric intrusion and a plume of Asian air pollution immediately to the west of the intrusion, which was forecast by the FLEXPART particle dispersion model (see Forster et al.  for a description of the model and its performance during the study). As forecast, the polluted air mass was located in the midtroposphere, immediately adjacent to stratospheric air, and arrived at the west coast of the United States in the wake of another major pollution episode reported in detail by Cooper et al. . As will be shown, the observed event on 10 May is the result of mixing between airstream remnants of two consecutive midlatitude cyclones, both of which transported polluted air from Asia to North America.
 The CO mixing ratios encountered were much greater than expected, with 1-second average values reaching as high as 290 ppbv. This is a major pollution episode considering that this value is slightly greater than the maximum CO measurement from the midtroposphere or upper troposphere during the spring 1996 field study of the North Atlantic Regional Experiment (NARE), conducted over the eastern United States and western North Atlantic Ocean. The flight lasted from 1806 UTC, 10 May, until 0128 UTC, 11 May 2002, traversing the flight track shown in Figure 1. The aircraft intercepted enhanced CO (>175 ppbv) in three locations, and comparison of Figures 1b and 1c shows that in each case the CO was adjacent to a region of enhanced IPV (>1.5 PVU). This region of high IPV is the western edge of a tropopause fold attached to the upper level trough shown in Figure 2 and demarcated by the IPV gradient. The fold extends downward and westward into the midtroposphere and lower troposphere just west of California.
Figure 3 shows a scatterplot of 1-second average ozone and CO measurements from the flight, colored according to potential temperature. The key feature in this diagram is that the data points with enhanced CO (>200 ppbv) have the same potential temperature (approximately 308 K) as many of the data points that were measured in the tropopause fold (ozone > 120 ppbv). Because these adjacent air masses have similar potential temperatures there is no thermodynamic barrier to mixing and the string of data points that connect the polluted and stratospheric air masses show that mixing clearly occurred.
Figure 4 shows minute average ozone mixing ratios from the flight plotted against IPV. The IPV values were obtained by spatially and temporally interpolating IPV from the gridded FNL meteorological fields onto the flight track. The data distribution in Figure 4 is quite typical of the free troposphere and lowermost stratosphere. IPV values less than 1 PVU indicate air parcels within the troposphere and as a result there is no relationship between ozone and IPV. Air parcels with IPV values between 1 and 2 PVU are typically in the mixing zone between the troposphere and stratosphere. Accordingly, these data points have a positive slope and ozone ranges from values typical of the free troposphere to those typical of the lowermost stratosphere. Air parcels with IPV greater than 2 are typically of stratospheric origin and these data points exhibit a positive ozone/IPV slope. However, there is some uncertainty in the IPV values because they were interpolated from wind fields with grid spacing of 1° × 1° in the horizontal, 50 hPa in the vertical and temporal availability of 6 hours. For example, the highly polluted air (CO > 175 ppbv) shows IPV values between 1.2 and 2.4 PVU. Because these data points were immediately adjacent to a tropopause fold their exact location could not be accurately resolved by the IPV fields and some of the data points were assigned stratospheric IPV values (>2 PVU).
3.2. Meteorological Scenario
 The chemical and meteorological features intercepted by the 10 May flight are the result of interactions between two midlatitude cyclones. Describing the inter-mingling of airstreams between the two cyclones is a challenge but many graphics are provided to make the processes as clear as possible. The reader is strongly encouraged to download the supplemental animation as running this movie is the best way to grasp the transport processes.
 The first cyclone is shown in Figures 5a and 5d above the central North Pacific Ocean at 1200 UTC, 3 May 2002. The WCB of this cyclone (hereinafter referred to as WCB1) transported a major pollution plume from Asia to North America. This event was intercepted by the fifth ITCT 2K2 research flight on 5 May 2002, and the transport mechanism is described in detail by Cooper et al. . The location of the stratospheric component of the cyclone's dry airstream is also shown in Figures 5a and 5d. Because the cyclone was at a mature phase of its life cycle, this time was chosen to initialize the trajectory ensembles within the domains of WCB1 (62,000 trajectories) and the SCDA (91,000 trajectories). Note that the SCDA contains a tropopause fold that extends down to 2 km a.s.l. (Figure 6a). Playing the movie shows that over the first 48 hours WCB1 moves toward North America, with the portions in the lower troposphere marked by rising motion. As discussed by Cooper et al.  WCB1 decays as it passes over North America with only a small portion passing through the lower troposphere. Meanwhile, during the same 48-hour period, the SCDA shears into three components. The first component consists of the upper portions of the airstream near WCB1 that travel with the same upper level flow as the warm conveyor belt and are advected above and alongside WCB1. The second component represents the bulk of the SCDA and moves poleward, while the third component is the portion of the SCDA within the deep tropopause fold. Most of the trajectory particles in the fold remain over the western North Pacific and descend clockwise into the surface anticyclone west of the midlatitude cyclone exactly in the manner described by Danielsen , and form a tube-like feature, similar to that described by Bithell et al. .
 The movie and Figure 6a also show particles associated with two other warm conveyor belts, WCB2 (46,000 red particles) and WCB3 (101,000 blue particles). The domains of these WCBs are defined at 0000 UTC, 7 May, when WCB2 was at a very late stage of its lifecycle and WCB3 was at a very early stage of its lifecycle (the location of the WCBs at 1200 UTC, 7 May, is shown in Figures 5b and 5e). As a result WCB3 is characterized by much stronger uplift and as will be shown, eventually overruns WCB2. These WCBs had not yet formed at the time the domains of the SCDA and WCB1 were defined. The WCB2 and WCB3 particles shown in Figure 6a and in the movie between 1200 UTC, 3 May, and 1800 UTC, 6 May, are back trajectory particles that were run from the domains of the WCBs at 0000 UTC, 7 May. By showing these back trajectory particles we illustrate how WCB2 and WCB3 formed. In Figure 6a the particles that will form WCB2 and WCB3 are spread over the western Pacific and eastern and southeastern Asia. Many are in the lower troposphere and rise as they flow into the WCBs, while a large proportion are in the midtroposphere and upper troposphere. While WCBs are typically thought of as originating in the lower troposphere, Cooper et al.  showed that when WCBs are broadly defined as the general region of ascent beneath and within the WCB cloud shield the airstream also entrains air masses from the midtroposphere and upper troposphere. Playing the movie shows that as WCB2 and WCB3 form they entrain air from the lower troposphere of Asia (likely polluted) and WCB2 entrains stratospheric air from the portion of the SCDA that descended into the lower troposphere (also shown in Figure 6b).
 By 1200 UTC, 7 May, 96 hours after the start of the simulation, the three components of the SCDA have proceeded along their initial pathways, described above (Figure 6b). The first component has continued to head east and the particles are clearly dispersing into the remnants of WCB1. The second component continues poleward, and much of the third component is entrained by WCB2.
 Between 1200 UTC, 7 May, and the end of the simulation at 0000 UTC, 11 May, many of the remnants of the SCDA become entrained into several weather systems that produce a high degree of vertical transport. For example at 1200 UTC, 9 May, some of the SCDA particles that descended into the lower troposphere of the western Pacific are lofted by a new midlatitude cyclone forming east of Japan. Conversely, some particles that remained in the stratosphere are descending through a tropopause fold just west of the Great Lakes, and other particles are descending as they disperse into the remnants of WCB1 above the North Atlantic. Meanwhile the younger and more energetic WCB3 ascends above the remnants of the older WCB2 and the two airstreams flow into the upper level ridge immediately west of the upper level trough that contains many of the SCDA particles (compare Figures 5b, 5e, and 6b to Figures 5c, 5f, and 6c).
 By the end of the simulation at 0000 UTC, 11 May, remnants of the SCDA and WCB1 have become thoroughly inter-mingled over the North Atlantic. Above California many SCDA particles have descended through the new tropopause fold intercepted by the NOAA WP-3D aircraft during the 10–11 May ITCT flight. Furthermore, the particles from WCB2 and WCB3 are immediately adjacent to this tropopause fold and are inter-mingling with the particles from the SCDA. Remnants of WCB2 and WCB3 are visible as cloud features south of Alaska in Figure 5c. However, no cloud is present in the region of the flight track as the WCBs evaporated due to moist-adiabatic warming associated with descending motion on the eastern side of the upper level ridge. The locations of WCB2 and WCB3 in relation to the SCDA particles and the fact that these WCBs largely originated in the lower troposphere of eastern Asia explains how polluted air was sampled next to stratospheric air on 10–11 May. The excellent agreement between the position of the trajectory particles and the features intercepted by the aircraft strongly supports the accuracy of the trajectory simulation.
 Additional support is provided by qualitatively comparing the location of the SCDA particles to the 1° × 1.25° degree daily plots of total ozone, remotely sensed by the Earth Probe TOMS instrument. Figure 7 shows total ozone on 3, 7 and 11 May with contours of the SCDA particle density overlain. The Earth Probe TOMS instrument does not resolve fine scale features resulting from tropospheric ozone variation so only the SCDA particles with IPV > 1 PVU are contoured. Figure 7a shows the ozone enhancements in the upper level trough associated with the SCDA just east of Japan on 3 May. Four days later the SCDA particles have advected along the edge of the stratospheric polar vortex demarcated by the TOMS ozone gradient (Figure 7b) and the IPV and moisture gradients shown in the altered water vapor imagery (Figure 5e). Many of the particles are associated with ozone features that have sheared away from the polar vortex over the eastern Pacific and western North Atlantic (Figure 7b). By 11 May the SCDA particles are still within the boundary of the polar vortex or associated with the vortex break-away features (Figure 7c).
3.3. Decay and Fate of the SCDA
Figure 8 shows the locations of the SCDA trajectory particles at 0000 UTC, 11 May, after 7.5 days of transport. Most of the particles remained in the stratosphere (67%), while 22% entered the troposphere and 12% were in the mixing zone between the troposphere and stratosphere (1 <= IPV < 2). Each particle represents an equal mass of air so these percentages also represent the SCDA mass in each of the three regions of the atmosphere. During this time period the SCDA remnants circled the Northern Hemisphere, and the particles that entered the troposphere were dispersed throughout the lower, middle, and upper troposphere, with less than 1% passing through the lower troposphere (<3 km) of the United States and southern Canada. A particularly interesting region is the North American west coast. The particles in the stratosphere extend southward to southern California and are within the newly formed tropopause fold sampled by the NOAA WP-3D on 10–11 May. These particles were originally in the SCDA component that traveled poleward but later headed south after they became entrained into a newly formed upper level trough. Many of these particles descended through the tropopause fold and were in the process of being transported into the troposphere, as indicated by the particles in the mixing zone along the western edge of the fold. Of all the particles in the troposphere immediately west of the 11 May tropopause fold, only a few had been exchanged during this particular event. Most entered the troposphere several days earlier over the central Pacific and came out of the original tropopause fold that was defined at 1200 UTC, 3 May. These particles were subsequently entrained into WCB2, and transported eastward until the remnants of WCB2 abutted against the 11 May tropopause fold. This scenario illustrates the complexity of airstream interactions, namely that air associated with a decaying tropopause fold can descend into the lower troposphere, become entrained into the following WCB, rise into the middle and upper troposphere and then inter-mingle with remnants of the original SCDA which enter the troposphere through a newly formed tropopause fold.
Figure 9 shows when and where the SCDA trajectory particles entered the troposphere since 1200 UTC, 3 May. Of all the particles that entered the troposphere, 57% entered within the first 48 hours (Figure 10). This exchange occurred over the central Pacific, mainly between 8 and 2 km a.s.l., during the decay of the initial tropopause fold. Over the next 5.5 days particles continued to enter the troposphere at a fairly steady rate (Figure 10) through a variety of pathways. Some were exchanged as a portion of the SCDA retreated poleward (particles north of the Bering Strait in Figure 9), some were exchanged when the SCDA dispersed into WCB1 over the central North Atlantic, while others entered the troposphere through downstream tropopause folds indicated in Figure 9 by the particles above California, northern Russia, central Asia and south of Greenland. STT can also result from radiational cooling of the lowermost stratosphere.
 The likelihood that a particle entered either the mixing zone or the troposphere was dependent upon its altitude at the time the particles were initialized at 1200 UTC, 3 May. Figure 11 shows the percentage of the SCDA trajectory particles at each level of initialization that were located within the stratosphere, mixing zone or troposphere after 7.5 days of transport. Of the particles that were initialized at pressures greater than 600 hPa, more than 90% were transported to the troposphere with the remainder located in the mixing zone. As pressure decreases from 600 to 200 hPa the percentage of particles that are transported to the troposphere steadily decreases and the percentage that remains in the stratosphere increases, while the percentage of particles in the mixing zone reaches a maximum of 36% at 400 hPa.
Figure 12a shows the percentage of particles in the three regions of the atmosphere at every 6-hour time step during the 7.5 day simulation. The figure shows that most of the particles that enter the troposphere do so within the first 48 hours. The fact that the percentage of particles in the mixing zone only decreases from 19% to 12% over 7.5 days suggests that transport between the troposphere and stratosphere is a continuously active process. Figures 12b and 12c show the number of particles in the three regions of the atmosphere as a percentage of all the particles initialized at pressures below and above 500 hPa, respectively. At pressures less than 500 hPa the trajectory particles enter the troposphere at a steady rate. Most of these particles are at an altitude above the main tropopause fold that is thoroughly dispersed into the troposphere, as shown in Figure 11. Instead, most of the upper portions of the SCDA remain in the stratosphere, but as the airstream shears and decays as it is advected along the edge of the polar vortex some of the particles become entrained in downstream disturbances and are subsequently transported to the troposphere. At pressures greater than 500 hPa most of the particles transported to the troposphere are exchanged within the first 48 hours. After 48 hours nearly all of the particles are either in the troposphere or the mixing zone. The percentage of particles in the troposphere fluctuates between 75% and 85% as particles move back and forth between the troposphere and the mixing zone. A small percentage of the particles in the mixing zone are transported into the stratosphere.
 Next we estimate the amount of ozone within the SCDA and the amount transported into the troposphere. This was achieved by taking the common approach of deriving a linear relationship between ozone and IPV. The ozone/IPV relationship varies seasonally and by latitude due to the seasonal cycle of ozone in the lower stratosphere [Fahey and Ravishankara, 1999]. Fortunately, the daily ozonesondes launched from Trinidad Head, California, during the ITCT 2K2 experiment provide a high quality concurrent data set for the midlatitude North Pacific Ocean. IPV calculated from the FNL meteorological fields was interpolated onto the ozonesonde profiles. A straight line fit through the ozone and IPV data resulted in the following relationship: ozone (ppbv) = −15 + 61(IPV) (r-squared = 0.66). The uncertainly in the slope of the regression line at the 95% confidence interval is ±7%, and the uncertainty of the ozonesonde measurements in the lower stratosphere is ±5%. This yields a total error of ±9% for the ozone/IPV relationship, for average conditions when one is considering many data points, such that the relatively large residuals (standard deviation of 68 ppbv) cancel out. From this relationship we estimate that the SCDA contained 5.2 Tg of ozone at 1200 UTC, 3 May. After 7.5 days, 0.5 Tg of ozone were transported to the troposphere, with another 0.4 Tg of ozone in the mixing zone. Therefore after 7.5 days of transport 10% of the SCDA ozone was located in the troposphere, while 22% of the overall SCDA mass had been exchanged. The reason that the ozone percentage is less than the mass percentage is due to the fact that the SCDA particles with the greatest probability of entering the troposphere were in the lower portions of the airstream. These particles also had lower IPV values and therefore lower ozone mixing ratios.
3.4. Dispersion of the SCDA Into WCBs
 In section 3.2 we described the processes by which the SCDA decayed over time and dispersed into the three WCBs. In this section we quantify the extent of the dispersion. We assume that each trajectory particle represents a fixed parcel of air (22 km × 22 km × 25 hPa) that does not deform during transport, even though in reality a parcel of air elongates into filaments as it experiences turbulent mixing and wind shear [Stohl et al., 2002]. Owing to the discrete volume and coherence of the air parcels this analysis cannot diagnose true mixing, which requires the exchange of air mass characteristics [Fairlie et al., 1999]. Instead we diagnose the degree to which airstreams disperse through the same region at the same time, and then if they meet certain criteria, assume that the air parcels are close enough to experiencing large scale mixing. The Northern Hemisphere (NH) was first divided into blocks with equal mass, between the surface and 100 hPa. Then the number of blocks containing trajectory particles from the SCDA, WCB1, WCB2 and WCB3 at each 6-hour timestep was totaled. Given that the airstreams were distinct and separate entities at the time their domains were defined we consider particles from separate airstreams to be close enough to experience large-scale mixing if they are located within the same block at any time after the airstream domains were defined. Because the results of this exercise are dependent upon the mass of air contained within each box we use two different box dimensions on scales similar to the grid spacing of the NCEP wind fields: (1) 222 km × 222 km × 50 hPa and (2) 111 × 111 × 50 hPa. These totals were recorded separately for particles in the stratosphere, mixing zone and troposphere. For example, a SCDA particle with a stratospheric IPV value and a WCB particle with a mixing zone IPV value in a block that straddles the stratosphere and the mixing zone are not classified as having been mixed together.
Figure 13 shows the percentage of the NH (divided into 222 km × 222 km × 50 hPa blocks) containing particles from the SCDA. At the time the domain of the SCDA was defined, it occupied 1% of the NH. After 7.5 days of transport the SCDA remnants were spread through 7% of the NH. At the end of the simulation 67% of the particles were still in the stratosphere but were only spread though 2.5% of the NH. In contrast only 12% of the particles were in the mixing zone but they were spread though 1.7% of the NH, and 22% of the particles were in the troposphere but they were spread through 4% of the NH. The stratospheric particles experienced less dispersion due to the stable conditions within the stratosphere and weaker vertical motions compared to the mixing zone and the troposphere. The sum of the percentages for the individual stratosphere, mixing zone and troposphere regions in Figure 13 is slightly greater than the percentage for all SCDA particles. This is because it is possible for a single block to contain particles from two or even three of the atmospheric regions, and these particular blocks will be attributed to more than one of the atmospheric regions.
Figure 14 shows the percent of the total SCDA mass that dispersed into WCB1 and the combined airstreams of WCB2 and WCB3, which eventually merged (WCB2&3), based on analyses with 222 × 222 km and 111 × 111 km boxes. In all cases the percentage of the SCDA that combined with the WCBs is greater for the 222 × 222 km boxes. This is because a larger box increases the likelihood that two trajectory particles from different airstreams will occupy the same box. Below we present results from both analyses to give a range of estimates of the extent of the co-dispersion between the SCDA and the WCBs.
 After 7.5 days of transport 10–21% of the total SCDA mass co-dispersed with WCB1 particles in the stratosphere. Similarly, 6–9% of the mass dispersed into WCB1 in the troposphere and 3–5% dispersed into WCB1 in the mixing zone. Relatively few WCB1 particles were actually transported into the stratosphere. In places, the SCDA particles were very dense close to the tropopause so a block containing 100 SCDA particles but just one WCB1 particle skews the degree of dispersion between the SCDA and WCB1 in the stratosphere. The percentage of the SCDA that dispersed into WCB2&3 is much less than for WCB1 due to greater spatial and temporal separation between these particular airstreams: 0–1% in the stratosphere, 3–5% in the troposphere, and 0–1% in the mixing zone.
 Of all the SCDA particles that entered the troposphere, 28–43% dispersed into WCB1 and 15–22% dispersed into WCB2&3. This means that a total of 43–65% of the SCDA mass in the troposphere dispersed into WCBs, located either upwind or downwind. Similarly, a total of 25–44% of the SCDA mass in the mixing zone also dispersed into WCBs.
4. Discussion and Conclusions
 It is clear that once air from a tropopause fold has been irreversibly transported into the troposphere it will eventually become well mixed with the surrounding air masses. However, the most interesting result from this case study is that roughly half (43–65%) of the SCDA particles that entered the troposphere dispersed into the upwind and downwind WCBs. This process is an important one to consider when interpreting trace gas measurements made within WCBs. For example, a data point in Figure 3 shows 120 ppbv of ozone associated with a polluted air mass of 200 ppbv of CO. Without considering the mixing with stratospheric air one may assume that the high ozone mixing ratios are entirely the result of photochemical ozone production, when in fact a large percent is of recent stratospheric origin. It seems likely that photochemical ozone production played a role in producing the data points that show 86 ppbv of ozone associated with 290 ppbv of CO, but even when CO is this high we still can't rule out the possibility that some stratospheric air was mixed into this air mass after it left the Asian atmospheric boundary layer and raised the ozone substantially.
 Consideration of the mixing between stratospheric intrusions and WCBs is especially important along the west coast of North America, the most likely region in the northern hemisphere to experience deep stratospheric intrusions [Sprenger and Wernli, 2003]. To illustrate this point further, the amount of stratospheric ozone within the troposphere was simulated during the ITCT study period (20 April to 20 May 2003) using the FLEXPART particle dispersion model. Figure 15 shows the mean ozone mixing ratios of stratospheric ozone that has resided in the troposphere between one and ten days, along the west coast of North America. This simulation only considers STT that occurred between 80°E and 80°W. Above 5 km average ozone of stratospheric origin is typically greater than 20 ppbv at midlatitudes and often greater than 40 ppbv. At lower latitudes average ozone of stratospheric origin is greater than 60 ppbv in the upper troposphere, the result of the highly frequent shallow tropopause folds that occur near the subtropical tropopause [Sprenger et al., 2003]. A striking feature is the extension of stratospheric ozone into the lower troposphere near 40°N, resembling the shape of a tropopause fold. This average feature is the result of frequent tropopause folding along the west coast observed by both the WP-3D aircraft and the ozonesondes launched from Trinidad Head, with one stratospheric intrusion producing nearly 140 ppbv of ozone at 2.5 km altitude.
 In section 3.3 we estimated that the decay of the SCDA resulted in the transport of 0.5 Tg of ozone to the troposphere over 7.5 days. Of this 0.5 Tg, 0.3 Tg was exchanged over the Pacific within the first 48 hours as the deepest portion of the tropopause fold decayed. The remaining ozone was exchanged as the SCDA sheared apart along the edge of the stratospheric polar vortex, which entrained the airstream into other tropopause folds and polar vortex breakaway features. If we extrapolate this amount of exchanged ozone (0.3 Tg) over the entire spring we can estimate the contribution of deep STT to the total springtime STT ozone flux. However, this exercise can only provide a very rough estimate because there is a large degree of uncertainty in the annual global flux of ozone from the stratosphere to the troposphere. For example, Hauglustaine et al.  report the global STT ozone flux from several chemical transport models with values ranging between 391 and 846 Tg per year. A recent intercomparison of nine different models and methodologies showed that while the models and methods can yield similar temporal and spatial evolution of a STT event the mass exchange varied by a factor of four [Meloen et al., 2003]. Furthermore the frequency of tropopause folding varies seasonally and from day to day. STT has a broad maximum from December to April and a minimum in August and September [Sprenger and Wernli, 2003], and deep tropopause folds like the one in this study are most common in winter but rare during summer [Sprenger et al., 2003]. Limiting our estimate to the springtime Northern Hemisphere, we assume that on average 1 deep tropopause fold forms each day and that the amount of ozone exchanged by their decay is 0.3 Tg per tropopause fold. This results in 28 Tg of ozone transported into the troposphere during the March–May period. Using a chemical transport model, Kentarchos and Roelofs  estimate STT injects 140 Tg of ozone into the springtime Northern Hemisphere troposphere. Comparing our deep tropopause fold ozone flux estimate to that of Kentarchos and Roelofs , we conclude that deep tropopause folds could comprise 20% of the springtime STT ozone flux. However, the net annual Northern Hemisphere stratosphere-to troposphere ozone flux (437 Tg O3 yr−1) calculated by Kentarchos and Roelofs  is at the high end of published estimates, therefore if smaller estimates of net springtime STT are considered, deep tropospause folds would contribute more than 20% of the springtime STT ozone flux. A one-year trajectory study by Wernli and Bourqui  estimates that deep STT is responsible for 15–20% of the total springtime STT mass exchange in the Northern Hemisphere (compare their Figures 3a and 11a). However, these two estimates are not directly comparable as Wernli and Bourqui  estimate the atmospheric mass exchange and not the ozone exchange, in addition they define the tropopause as the 2 PVU surface and require all trajectories to descend below the 700 hPa surface, while the present study defines the tropopause as the 1.5 PVU surface and does not require that all trajectories cross the 700 hPa surface. The remaining 80% of the springtime STT ozone flux would be accounted for by the far more numerous medium and shallow tropopause folds [Sprenger et al., 2003], the erosion of cut-off lows, gravity wave breaking and the decay of stratospheric streamers through shearing and radiative processes.
 One further consideration of this event is the question of its impact on the surface of North America. The supplemental movie shows that as the SCDA and WCB1 decay they pass over North America mainly in the middle and upper troposphere with very few particles reaching the surface. However, a domain-filling trajectory analysis of the deep tropopause fold that formed over California on 10 and 11 May 2002 (not shown), revealed that a substantial portion descended through a surface anticylone into the lower troposphere above southern Louisiana and southeastern Texas on 14–15 May 2002. Archived maps of ozone mixing ratios obtained from the EPA-AIRNow website (http://www.epa.gov/airnow) show elevated ozone (61–79 ppbv) across broad regions of eastern Texas and Louisiana on 14–16 May 2002. This ozone episode occurred beneath the western side of a surface anticyclone, which not only provides ideal conditions for photochemical ozone production, but also allows for the descent of stratospheric air to the surface [Davies and Schuepbach, 1994; Cooper and Moody, 2000; Kentarchos et al., 2000]. We examined this event with the MOZART-2 chemical transport model [Horowitz et al., 2003]. While a full discussion of the analysis is beyond the scope of this paper we report here that the model indicated an ozone peak of 75 ppbv above the south-central United States between 700–800 hPa at 0000 UTC, 14 May, 83% of which was stratospheric in origin. Between 800 and 1000 hPa the average ozone mixing ratio on 14 May was 65 ppbv with 43% of stratospheric origin. Furthermore, given the close proximity of the Asian pollution to the stratospheric intrusion on 10 May, it is feasible that the pollution would follow the same transport pathway and also impact the surface of the United States.
 While this study has established a clear link between the hemispheric-scale inter-mingling of stratospheric intrusions and WCBs and described the processes by which it occurs, it does have several limitations. First, trajectories were initialized at just one instance in the life cycles of the airstreams. This study does not consider the air parcels that exited the airstreams at earlier stages, nor does it consider air parcels that enter the airstreams at later stages, and therefore does not explore the inter-mingling between these non-modeled portions of the airstreams. Second, defining the tropopause is difficult as there is no sharp transition between the troposphere and stratosphere. In this paper we treated the 1–2 PVU region as a mixing zone between the troposphere and stratosphere and demarcated the SCDA with an IPV value on 1.5 PVU. We could have chosen a more conservative value of 2 PVU, but this would have excluded the stratospheric particles that had recently entered the mixing zone. Conversely a boundary of 1 PVU would have included too high a proportion of mass with recent tropospheric origin. So defining the SCDA with the 1.5 PVU surface appears to be a good compromise. Finally we are not certain how representative this case study would be for other regions of the globe and during other seasons, or for smaller, less energetic midlatitude cyclones. Despite the limitations the analysis produced several important results.
 1. The observation of mixing between polluted and stratospheric air over the west coast of North America was shown to be the result of inter-mingling between two WCBs containing Asian pollution and the remnants of the SCDA of the downwind midlatitude cyclone.
 2. The SCDA decayed into three main components. The first component headed east and dispersed into the cyclone's WCB. The second component headed poleward, much of which remained in the stratosphere but over the simulation period portions transported along the edge of the polar stratospheric vortex were entrained into the troposphere. The third component constituted the lower portions of the tropopause fold. This component descended into the lower troposphere through the anticyclone west of the cyclone, and was subsequently entrained by the upstream WCB.
 3. After 7.5 days of transport 22% of the SCDA mass was transported into the troposphere. The portions of the SCDA that penetrated to the lowest altitudes had the greatest likelihood of being transported into the troposphere. For example, over 90% of the SCDA at altitudes below the 600 hPa level was transported to the troposphere, but none of the mass at the 200 hPa level was exchanged.
 4. STT resulted in the transport of 0.5 Tg of stratospheric ozone to the troposphere during the 7.5 day study period.
 5. More than half (57%) of the ozone exchange occurred during the first 48 hours as the deepest portions of the tropopause fold decayed over the Pacific. The rest of the exchange occurred over the following 5.5 days as the remnants of the SCDA sheared apart along the edge of the stratospheric polar vortex and became entrained into subsequent tropopause folds and vortex breakaway features. Portions of the SCDA may also have entered the troposphere via radiational cooling processes.
 6. Roughly half of the SCDA particles that entered the troposphere subsequently dispersed into the upwind and downwind WCBs.
 As mentioned above this study has examined the dispersion and decay of the stratospheric component of a single dry airstream. Future studies can build on the present results by examining other deep tropopause folds to see if similar dispersion and inter-mingling processes occur. While this event contained a very deep tropopause fold it was not unusual for this time of year. However, shallower tropopause folds are far more common and it would be important to see if these events experience the same degree of inter-mingling between stratospheric intrusions and WCBs. A box-model study by Esler et al.  suggests that the mixing of stratospheric ozone and tropospheric water vapor leads to enhanced hydroxyl radical concentrations compared to background tropospheric and stratospheric values, which increases the oxidation of CO, methane and higher hydrocarbons. Therefore chemical transport models should be run to determine the effect of freshly entrained stratospheric air on the chemical transformation processes within WCBs as they export pollution from the continents and transport it around the globe.
 We thank the Data Support Section of NCAR's Scientific Computing Division for making the NCEP FNL analyses available for download. GMS-5 satellite images were provided by the Space Science and Engineering Center, University of Wisconsin-Madison. GOES satellite imagery and AVN gridded meteorological fields were provided by UNIDATA Internet delivery and displayed using McIDAS software. We thank the Ozone Processing Team of NASA/Goddard Space Flight Center for providing the Earth Probe TOMS data. Finally, we thank three anonymous referees for their very helpful comments and suggestions.