Synoptic-scale transport of reactive nitrogen over the western Pacific in spring

[ 1 ] Pathways of synoptic-scale uplifted transport of pollutants from East Asia and their effects on chemical distributions of NO y species are investigated based on a subset of the aircraft data obtained during the NASA Transport and Chemical Evolution over the Pacific (TRACE-P) experiment, conducted in February–April 2001. Meteorological and chemical analyses indicate that 73% of the uplifted transport was associated with warm conveyor belts (WCBs) and convective outflow (COF), which transported air masses strongly impacted by biomass burning over Southeast Asia. The rest (27%) of the uplifted air masses originated over coastal regions of northeast China, where fossil fuel combustion was a dominant source of pollutants. Both WCB associated with a midlatitude cyclone and COF associated with a stationary front over southeast China are examined in detail for the April 4 case. During the TRACE-P period, low NO x (= NO + NO 2 )/NO y ratios in the WCB and COF indicate that a significant part of the NO x was oxidized to nitric acid (HNO 3 ) and peroxyacetyl nitrate (PAN) during transport. Low HNO 3 /NO y ratios in the WCB and COF airstreams indicate that a large amount of HNO 3 was removed during transport on timescales within 1–3 days. PAN was found to be the dominant form of NO y in air masses transported by the WCB and COF, likely due to the production of PAN in regions of biomass burning and industrial emissions, as well as due to the rapid removal of HNO 3 during transport. For emissions that were transported to the free troposphere by WCBs and COF, about 10–20% of the NO y remained after transport to the free troposphere, and 30% of the NO y surviving in the boundary layer in limited cases. The results indicate that the WCB and COF provide both an efficient sink for HNO 3 and an efficient mechanism for the transport of PAN from the boundary layer to the free troposphere over the western Pacific.


Introduction
[2] Midlatitude cyclones have been recognized to play a key role in redistributing trace species and affecting their chemical evolution in the troposphere. Airstreams associated with midlatitude cyclones, such as Warm Conveyor Belts (WCBs) and Dry Airstreams (DAs), are of great importance. The WCB is an ascending airstream ahead of a surface cold front, traveling into the middle to upper troposphere [e.g., Browning and Roberts, 1994], which transports trace species from the boundary layer to the free troposphere.
[3] Recently, characteristics of these airstream types associated with midlatitude cyclones have been studied, especially from a viewpoint of the effects of transport on trace species. Bethan et al. [1998] interpreted chemical and meteorological data in the context of the airstreams within cyclones over the eastern North Atlantic Ocean (NAO) in spring. They concluded that the conveyor belts within cyclones are capable of effectively transporting boundary layer air to the free troposphere. Kowol-Santen et al. [2001] showed that up to $70% of carbon monoxide (CO) in the boundary layer can be transported to the free troposphere in the region of WCBs, using a mesoscale model. The chemical characteristics of CO, ozone (O 3 ), and total reactive nitrogen (NO y ) were studied more systematically for these airstreams associated with midlatitude cyclones over the NAO, based on aircraft measurements [Cooper et al., 2002a[Cooper et al., , 2002b.
[4] The chemical compositions in the airstreams are controlled mainly by two factors: the geographic distribution of sources and the cyclone pathway. For example, air masses in WCBs vary from fairly clean [Grant et al., 2000] to highly polluted [Stohl and Trickl, 1999], depending on their origins. East Asia also experiences midlatitude cyclones [e.g., Chen et al., 1991Chen et al., , 1992. In fact, Stohl [2001] has shown that WCB air masses originate most frequently in the boundary layer over the warm water pools in East Asia, similarly to northeast America, based on a large number of trajectory calculations.
[5] The export of NO y from the boundary layer to the free troposphere is a key parameter of the global O 3 budget. Liang et al. [1998] showed that the influence of NO y exported from the U.S. boundary layer on the global O 3 budget is greater than the influence of direct O 3 transport, by using a three-dimensional (3-D) chemical transport model (CTM). Considering the large emissions of reactive nitrogen over the Asian continent [Streets et al., 2003], transport of NO y from the continental boundary layer over Asia to the western Pacific should be of great importance. In 3-D CTM studies, Bey et al. [2001] found that frontal lifting plays a critical role in controlling the chemical characteristics of the Asian outflow in February -March. In fact, cumulus convection associated with a cold front over northeast China has been shown with aircraft measurements to transport highly polluted air to the upper troposphere over the Japan Sea in April [Miyazaki et al., 2002]. There are still, however, large uncertainties in our knowledge concerning the emission rates, chemical processes, and transport processes of NO y species [Stohl et al., 2002].
[6] In this study, the chemical data obtained during the NASA Transport and Chemical Evolution over the Pacific (TRACE-P) experiment, conducted in February -April 2001, are used to investigate the chemical evolution of NO y associated with upward transport due to convection and midlatitude cyclones over the western Pacific in spring. A major objective of this study is to examine pathways of uplifted airstreams from the Asian continent and to investigate the role of upward transport in controlling the distributions of NO y species. For this purpose, we focus on uplifted patterns of transport from the boundary layer over the continent. In addition, the chemical evolution of NO y species is examined in detail within airstreams in the developing cyclones and convective outflow, based on aircraft measurements.

Data Description
[7] During TRACE-P, a variety of chemical species and meteorological parameters were measured on board the NASA P-3B and DC-8 aircraft. The instrumentation of the TRACE-P mission is described in an accompanying overview paper [Jacob et al., 2003]. Measurement data and the instrumentation used in this study are summarized in Table 1. In this study, CO was used as a long-lived tracer in combination with trajectories for identifying airstreams. One-minute merged data were used for the P-3B, and the measured NO y was used as total reactive nitrogen. For the DC-8 measurements, NO y is defined as the sum of NO + NO 2 + PAN + HNO 3 . Uncertainty associated with using the sum of these species as a surrogate for the total NO y was estimated to be 14% from the median values and accuracy of each measurement. Alkyl nitrates and peroxypropionyl nitrate (PPN) constituted only 2-3% when they are included in the sum of the NO y species for the DC-8 data. Since HNO 3 was measured on the DC-8 with a temporal resolution of 2 min, HNO 3 merged data were used for the DC-8 NO y analysis. Table 2 summarizes the relationship between the measured NO y and the sum of NO + NO 2 + PAN + HNO 3 for the P-3B. The data used here were obtained over 122°-170°E, 22°-42°N, which is the study area described in section 3.2. On average, 86% of NO y at 2 -6 km is accounted for as NO x , PAN, and HNO 3 . Although NO y was not measured at 7 -12 km during TRACE-P, Kondo et al. [1997a] and Singh et al. [1998] reported general agreement of the measured and the sum of NO y with a small deficit (10 -15%) at 6-12 km over the western Pacific based on data from PEM-West B, conducted in February -March 1994. The deficit of NO y in this study can be explained mainly by alkyl nitrates [Simpson et al., 2003;Talbot et al., 2003] and aerosol nitrate (NO 3 À ) [Koike et al., 2003]. We do not include these data in this analysis because the time resolution of these species (typically 5 -15 min) would make it difficult to discuss the NO y in a selected air mass.
[8] In order to select vertically uplifted patterns of lagrangian transport from the Asian continent, 5-day back trajectories were used. The trajectories were calculated along the flight tracks of the P-3B and DC-8 at 5-min intervals by the Florida State University (FSU) Kinematic Trajectory Model [Fuelberg et al., 1996]. The calculation used the European Centre for Medium-Range Weather Forecasts (ECMWF) data . The data set has a 1°Â 1°horizontal resolution and 61 sigma levels in the vertical, supplied at 6-hourly intervals throughout the TRACE-P period. Some limitations of the trajectory calculations are described in previous works [e.g., Stohl et al., 1995]. Geostationary Meteorological Satellite-5 (GMS-5) infrared (IR) images were used for the identification of airstreams associated with cyclones. Images of average rainfall rates, using the Ferriday algorithm [Ferriday and Avery, 1994] on Special Sensor Microwave/Imager (SSM/I) satellite data, are available at the NOAA website http:// www.etl.noaa.gov/satres/archive.html.

Transport Pathways From the Asian Continent
During TRACE-P

Meteorological Conditions in February-April
[9] Brief highlights of the meteorological conditions in February -April during TRACE-P are described here, which are key points for this study. Meteorological conditions concerning atmospheric transport during TRACE-P are summarized by Fuelberg et al. [2003] in more detail. Figure 1 displays the time-averaged sea level pressure and rainfall rates derived from the SSM/I instrument in February -April 2001. Stationary fronts prevail in the convergence zones over central and southern China in early spring, the season that corresponds to the transition period from the winter to summer monsoons in Asia [Nieuwolt, 1977]. Incursions of warmer air from the south become more frequent due to the building up of the Pacific high-pressure system, while movement of cold air southward decreases because the Siberian High weakens by March, as seen in the left panels of Figure 1. As a result, the warming causes frequent convection in Southeast Asia [Nieuwolt, 1977;Bey et al., 2001], indicated in the right panels of Figure 1. Moreover, a widespread region of precipitation over the western Pacific, as seen in Figure 1, is associated with pathways of midlatitude cyclones, as indicated by Fuelberg et al [2003]. They derived an average frequency of 3 -4 midlatitude cyclones per week over the western Pacific during TRACE-P.

Trajectory Analyses
[10] In order to examine the export pathways of pollutants, vertically uplifted patterns of lagrangian transport were selected by using 5-day back trajectories of sampled air masses. We selected air masses that originated in the boundary layer (pressure levels greater than 800 hPa; altitudes below 2 km) over the Asian continent and that were transported above the 800-hPa level prior to the aircraft measurements. An uplifted air mass is defined here as that ascending at least 1 km. It should be noted that the boundary layer altitude can extend higher or lower depending on a variety of local conditions. Therefore the boundary layer in the current analysis is approximated by this range of altitudes. Figure 2a shows a horizontal plot of the selected trajectories, and a summary of these trajectories is listed in Table 3. Variations of height are also given as functions of longitude ( Figure 2b) and time (Figure 2c). During the observational period between 7 March and 4 April, the uplifted transport pattern consists of 146 trajectories (90 from 9 DC-8 flights and 56 from 6 P-3B flights). The air masses were uplifted from the boundary layer within 1 -3 days (Table 3 and Figure 2c). The study area in this analysis is defined to be over 122°-170°E, 22°-42°N, where the intensive flights were made from Hong Kong and Yokota.
[11] Locations where the trajectories reached the 800-hPa level are plotted in Figure 3, which represent estimated origins of the air masses. The results show that 73% of the selected trajectories originated over a broad region in Southeast Asia (SA) where frequent convective activities occurred during the TRACE-P period, as revealed by abundant satellite-derived precipitation and lightning observations (see Figure 1 and also Fuelberg et al. [2003]). The SA region corresponds to the location where extensive biomass burning occurred during the observational season. This will be further examined in section 3.3. On the other hand, 27% of the selected trajectories showed the air masses originated in the northeast coastal region of China (NC) at 30°-40°N.
[12] Individual trajectories were also compared with cloud images observed by the GMS and with surface weather charts. In this study, WCB airstreams were distinguished from other uplifted transport, i.e., convection. For instance, Stohl and Trickl [1999] have defined WCBs as airstreams ascending at least 8 km in 60 hours, using coherent ensembles of trajectories (CETs) methods [Wernli and Davis, 1997] with PV fields and cloud images. In this analysis, airstreams associated with the WCB are defined as follows: first, cloud features associated with surface cold and warm fronts were examined in relation to the sampling location of the aircraft. Each back trajectory calculated along the flight track was then compared to the motion of the frontal system. If the trajectories run parallel to the surface cold front and ascend at least 1 km within 24 hours, the corresponding airstream was identified as a WCB. As a result, four individual WCBs within different cyclones were identified, consisting of 53 trajectories (36% of the uplifted trajectories). All of these WCBs originated in the SA region, with a mean ascent rate of 2.3 (±1.1) km/24 hours.
[13] Convective outflow (COF) was determined by trajectories that were uplifted over satellite-observed convective cloud regions and those identified with surface weather charts. The COF determined in this study includes convection such as that associated with a stationary front and that occurring behind a cold front. In addition, convection in advance of a cold front is also categorized as COF in this analysis. The results of this analysis show that 10 individual COFs were identified, consisting of 80 trajectories (55% of the uplifted trajectories). Seven of the 10 COFs originated in the SA region, while the remaining three originated in the NC region. The mean ascent rate for COFs was 4.7 (±2.5) km/24 hours, which is faster than that of the WCBs by a factor of 2. It should be noted that the trajectories have limitations of spatial and temporal resolution relative to the much smaller scales of vertical transport, including convection. Some air masses may ascend much more rapidly in a convective cell than indicated by the relatively coarse resolution of the ECMWF data currently used. Thus the uplifted transport presented here might be a lower limit of upward transport in the real atmosphere. The remaining trajectories identified as neither WCB nor COF consist of only 9% of the total, most of which include outflows in the lower free troposphere. It should also be noted that the fractions of WCBs and COF to the total uplifted trajectories are not statistically representative due to a bias in the sampling locations and time.

Chemical Tracers of Source Description
[14] In addition to the frequent convection over Southeast Asia, distributions of pollutant emissions during the observational period are also important. Figure 4 shows the annual emission rates of CO and NO x /CO molar ratios over Asia estimated for the year 2000 [Streets et al., 2003]. These emission rates are given at http://www.cgrer.uiowa. edu/people/carmichael/ACESS/Emission-data_main.html. CO and NO x are strongly emitted along the east and southeast coast of China (22°-44°N, 110°-122°E). Two major sources are considered in these estimates: biomass burning and industrial emissions. Extensive biomass burning occurs in Southeast Asia and northern India in the spring (February -April) [e.g., Duncan et al., 2003], while industrial emissions are largest in northeast China [e.g., Bey et al., 2001].
[15] In order to examine the signatures of the chemical tracers in the identified air masses, scatter plots among CO, CH 3 Cl, and C 2 Cl 4 are shown in Figure 5. In this analysis, CH 3 Cl and C 2 Cl 4 are used as tracers of biomass burning and fossil fuel combustion, respectively [e.g., Wang et al., 1995;Blake et al., 1996]. A tight, positive correlation between CO and CH 3 Cl (r 2 = 0.74) is apparent in the airstreams in the SA category. A slope of the regression line (dCH 3 Cl/dCO) of 0.52 (pptv/ppbv) is similar to the emission ratios of 0.50-0.57 (pptv/ppbv) [Rudolph et al., 1995;Andreae et al., 1996;Blake et al., 1996] from biomass burning. In contrast, the correlation was much poorer in the NC category (r 2 = 0.34), with lower values of CH 3 Cl (530 -620 pptv) than those in the SA category (570-700 pptv). These results are consistent with the trajectory analysis, showing that the WCB and COF air masses originated from the intense biomass burning region.
[16] In contrast, the mixing ratios of C 2 Cl 4 in the NC category (8 -15 pptv) were higher than those in the SA category (3 -10 pptv). The result indicates that the airstreams in the NC category were much more strongly influenced by fossil fuel combustion than those in the SA category. However, it is likely that the air masses influenced by fossil fuel combustion were mixed with the biomass burning plumes in the WCB and COF airstreams because CO was correlated with C 2 Cl 4 (r 2 = 0.43).

Meteorological Conditions
[17] This section describes an example of two different pathways of upward transport: WCB and COF. Figure 6 shows GMS water vapor images with flight tracks for the P-3B and DC-8 and sea level pressure at 0000 UT on 4 April. There were two frontal systems over the westernto-central Pacific on this day. A developing wave cyclone (Cyclone-A) was centered near 50°N, 150°E, with a cold front extending southward. The other system was a deeply occluded low (Cyclone-B) located just west of Alaska. The upper level trough associated with Cyclone-A (not shown) corresponds to a dry region over Japan in the middle to upper troposphere. The P-3B flew from Yokota Air Force Base (AFB) (35°N, 140°E) to Midway Island (28°N, 177°E), conducting extensive in-progress vertical sampling at 0 -6 km. The DC-8, which headed for Kona (19°N, 204°E), Hawaii, also followed the same route between Yokota and Midway at 0 -11 km. Both aircraft crossed a cloud region associated with the surface cold front at 146°E. In order to obtain good vertical coverage of air samplings over the western Pacific, the P-3B made low-altitude

Warm Conveyor Belt (WCB)
[19] Figure 8 shows the trajectories that represent a welldefined WCB airstream (Figure 7a). Air masses in the WCB were observed at 8°-20°east in longitude relative to the center of a cyclone, which were within a typical range of 0°-20°in the upper troposphere [Cooper et al., 2002b]. As is typical for WCB trajectories [e.g., Browning, 1990], the trajectories move parallel to the surface cold front, ascending toward the northeast over the western Pacific. The mean timescale for transport from the top of the boundary layer to the sampling points was estimated to be 2.6±0.6 days, which was derived from trajectories. The mean ascent rate of the WCB was 2.3±0.9 km/24 hours, estimated from the trajectories. High relative humidity of 85 to nearly 100% (not shown) is consistent with upward transport. The median CO mixing ratios in the WCB air masses were 180-200 ppbv, reflecting the strong influence of high emissions of pollutants in the SA region. The CO values were higher than the median values of 140 ppbv in the WCB in the middle to upper troposphere over the western North Atlantic Ocean in spring [Cooper et al., 2002b]. Trajectories in Figure 8 indicate that these air masses were strongly impacted by CO emissions over southeast China, including coastal regions, where the emissions of CO from biomass burning are large.

Convective Outflow (COF)
[20] Contrary to a typical WCB, the sampling points of the air masses shown in Figure 7b were located 0°-7°west in longitude relative to the center of Cyclone-A. These air masses were also polluted (CO mixing ratios > 200 ppbv) and humid (relative humidity of nearly 100%), indicating that they were also transported upward from the polluted boundary layer. The bulk of the trajectories shows mean ascending motion of the air masses in 1.3 ± 0.2 days from the boundary layer to the sampling points at 10 km, as seen in Figure 9. The ascent rate estimated from the trajectories was about 7.3 ± 1.7 km/24 hours, which was much faster than that of the WCB (2.3 ± 0.9 km/24 hours) described in section 4.2.1.
[21] Figure 10 shows the GMS water vapor image at 0200 UT on 3 April together with the location of a surface front 24 hours prior to sampling. A surface stationary front existed continuously at about 25°N over southeast China for approximately 3 days during 1 April and 4 April. Therefore the polluted air masses should have been rapidly lifted up to an altitude of 9-10 km by this stationary front and then transported cyclonically at this altitude along the eastern side of the upper-level trough located over the northern part of Japan.
[22] According to the trajectories, these air masses were found to originate in southeast China. As examined in section 3.3, correlation between CO and CH 3 Cl (not shown) also confirms that the influence of biomass burning was dominant in these air masses. Because the air masses in the WCB were also identified as originating from the region of this stationary front, as discussed in the previous section, some air masses were lifted up by the stationary frontal convection, while others were transported by the WCB as the cyclone developed.

Dry Airstreams (DA)
[23] It is well established that downward transport of stratospheric air into the troposphere occurs behind cold fronts [e.g., Danielsen, 1968;Shapiro et al., 1987]. A dry airstream is defined here as an airstream behind a surface cold front, originating in the upper troposphere or lower stratosphere at high latitude. Locations where DAs were observed at 5 -9 km on this flight (Figure 7c) correspond to a dry region in the middle to upper troposphere over Japan (30°-45°N, 130°-145°E), as seen in Figure 6. Air masses in the DAs were characterized by high O 3 of 120-200 ppbv, low CO of 80-100 ppbv, and a low relative humidity of 2% (not shown). Figure 11a illustrates a cross section of potential vorticity (PV) calculated along the flight track with the observed O 3 values by the aircraft. A tongue of PV can clearly be seen extending downward to the midtroposphere at 5 km and 139°-143°E, corresponding to the region of high O 3 values. The trajectories shown in Figure 11b confirm that the air masses in the DA were transported from high altitudes north of 55°N. Together, the PV map and TOMS total O 3 (not shown) clearly represent air masses of stratospheric origin associated with a dry intrusion.
[24] It is interesting to note that the DAs at 5-9 km were observed in the region adjacent to the air masses impacted by the COF at 10 km, as described in section 4.2.2. If these different types of air masses are mixed by turbulent mixing, the mixing ratios of trace gases such as O 3 , NO y , and H 2 O in the resulting air masses can vary significantly depending on the degree of mixing. In fact, evidence of the mixing of stratospheric air with polluted air transported from the boundary layer has also been reported previously. Prados et al. [1999] showed that air masses behind a cold front, transported to the midtroposphere over Bermuda, were a mixture of stratospheric air and polluted air convected ahead of another cold front. Parrish et al. [2000] reported that highly polluted air was located in the immediate vicinity of air of stratospheric origin in the lower troposphere within a cyclone system over the western NAO. A key point of the current analysis is that the mixing of biomass burning effluents transported by convection with stratospherically influenced air could occur in the upper troposphere in the vicinity of the frontal region.

Partitioning and Export of NO y Species
[25] In order to investigate the chemical evolution of NO y species in the air masses vertically transported over the western Pacific, partitioning of NO y is presented in this section for the WCB and COF identified in section 3. Figure 12 shows the median values of NO x , PAN, and HNO 3 mixing ratios and their ratios to the total NO y in the WCB and COF, as identified in section 3.2. The median values of the whole set of data obtained at the study area (22°-42°N and 122°-170°E) during TRACE-P are also shown in the figure for comparison. The NO x mixing ratios and NO x /NO y ratios in the WCB and COF were much lower (10 -30 pptv and 0.03) than the midlatitude median values (50 -70 pptv and 0.10-0.15), particularly at 6 -10 km. The NO x /NO y ratios in the WCB and COF were also lower than the median values of 0.10 -0.30 in the continental air masses at 6 -10 km over the western Pacific in February -March [Kondo et al., 1997a]. Similarly, the HNO 3 mixing ratios (80 ± 10 pptv) and HNO 3 /NO y ratios (0.15 ± 0.05) in the WCB and COF were lower than the midlatitude median values (110 ± 10 pptv and 0.30 ± 0.05, respectively). In contrast, the PAN mixing ratios (330 ± 70 pptv) and the PAN/NO y ratios (0.75 ± 0.10) in the WCB and COF were higher than the midlatitude median values (150 ± 50 pptv and 0.45 ± 0.05, respectively).
[26] NO x is oxidized to HNO 3 by reaction with OH during daytime on a timescale of 1 day and by nighttime formation of N 2 O 5 followed by hydrolysis on sulfate aerosols on a similar timescale [Dentener and Crutzen, 1993]. The residence time of air masses in the WCB and COF is 1 -3 days, while chemical recycling from HNO 3 to NO x occurs on a longer timescale of 10 days to several weeks [Kley et al., 1981]. Therefore NO x is not in chemical equilibrium with HNO 3 in these WCB and COF air masses. Moreover, the low HNO 3 /NO y ratios in Figure 5. Scatter plots of CO-CH 3 Cl (left) and CO-C 2 Cl 4 (right) in the uplifted air masses, which originated at SA (solid circles) and NC (crosses) region. The dashed lines represent a typical range of the slopes in the biomass burning plumes [Andreae et al., 1996].
the WCB and COF suggest that HNO 3 has been substantially removed during the vertical transport within 1 -3 days, leading to the reduction of chemical recycling from HNO 3 to NO x . In contrast, the midlatitude median values of NO x increased with altitude at 6 -12 km. This general increase of NO x with altitude has typically been observed over the western Pacific [Kondo et al., 1997a] and over other regions [Kley et al., 1981;Ridley et al., 1994], due to the longer chemical lifetime of NO x and the input of in situ sources of NO x in the upper troposphere such as lightning or aircraft emissions. The present results show that the air masses in the WCB and COF were isolated from ambient air and were less affected by such in situ sources of NO x . To summarize, the lower NO x /NO y ratios in the WCB and COF are due to both the chemical imbalance between NO x and HNO 3 and a small influence of in situ sources without significant mixing with surrounding air masses.
[27] The elevated PAN mixing ratios in the WCB and COF can be attributed to limited decomposition of the PAN produced over the biomass burning and industrial emissions region followed by the rapid upward transport of the air masses and limited mixing with surrounding air masses. Moreover, the dominance of PAN in NO y was also due to the rapid loss of HNO 3 during transport by the WCB and COF, leading to the increase in the PAN/NO y ratios. Loss of PAN during transport in the free troposphere should be very small due to the stability of PAN at low temperatures. In addition, wet deposition of PAN is negligible compared to HNO 3 because its solubility in water is small [Kames et al., 1991]. PAN chemistry has only a minimal impact on net NO x production of ±10% in the cold upper troposphere, as estimated using a 3-D CTM by Moxim et al. [1996]. However, the dominance of PAN in the WCB and COF can provide significant sources of NO x in remote regions through long-range transport followed by descent to the warmer lower troposphere, where PAN decomposes to NO x [Moxim et al., 1996].

Removal of HNO 3 in the WCB and COF
[28] In this section, we estimate the export efficiency (EE) with which NO y emitted from the surface over East Asia was transported to the sampling points by the WCB and COF. CO can serve as a good reference for estimating the NO y removal rate [Stohl et al., 2002], because CO is emitted simultaneously with NO x and has a lifetime of 1 -2 months. For the calculation of EE, the enhancement of NO y and CO over their background level should be used [Stohl et al., 2002]. In this analysis, the background mixing ratios of NO y (NO y(bkg) ) and CO (CO bkg ) were determined to be 100 pptv and 80 ppbv, respectively, which are approximately the lowest values at 2 -10 km over the study area (22°-42°N, 120°-170°E) during TRACE-P. The dNO y and dCO can be then defined as dNO y = NO y À NO y(bkg) and dCO = CO À CO bkg , respectively. The EE of NO y can be derived from the slope of dNO y /dCO ratios in each air mass by dividing it by the NO x /CO emission ratios over the source regions. Molar ratios of NO x /CO emitted over the SA and NC are estimated to be 0.034 ± 0.017 mol/mol and 0.051 ± 0.013 mol/mol, respectively, for the year 2000 ( Figure 4) [Streets et al., 2003]. Based on uncertainties in the CO (±156%) and NO x (±23%) emissions estimations in China [Streets et al., 2003], the overall uncertainty of the NO x /CO ratios was estimated to be ±158%.
[29] Figure 13a shows a scatter plot of CO and NO y in the WCB and COF air masses with reference data obtained at 2 -11 km over the study area. The slopes of the regression lines for the WCB (3.15 pptv/ppbv) and COF (SA) (5.27 pptv/ppbv) show lower values compared to the reference data points between CO and NO y over the study region. These lower slopes reflect large amounts of NO y removal, as discussed below. On the other hand, the slope of the regression line for the COF (NC) showed a larger value of 10 pptv/ppbv. The difference of the slopes between the SA and NC categories can be attributed to (1) the removal rate during transport and/or (2) the difference of emission ratios over the two regions. Figure 13b also represents these scatter plots of CO against NO y for the SA and NC with the expected fraction of the emission ratios. The EE value was estimated to be 9% in the WCB, while it was 15% (SA) and 19% (NC) in the COF. The results indicate that the WCB and COF transported 10-20% of NO y emitted over the Asian continent to the free troposphere over the western Pacific. If the estimated uncertainties in the emissions ratios are given, the EE values range within 6 -19% (WCB), 10 -31% (COF-SA), and 13-26% (COF-NC). Stohl et al. [2002] showed only 3 -5% of the NO y emissions over the United States were transported above an altitude of 3 km over the North Atlantic, utilizing the aircraft data combined with a lagrangian transport model.
[30] It should be noted that EE includes the loss of NO y in the boundary layer (z = 0 -2 km) in addition to the removal during uplifted transport of air masses into the free troposphere (z = 2 -12 km). In order to examine this process, the EE in the boundary layer near the source regions was  estimated. Air masses transported in the boundary layer within one day were selected based on trajectories calculated from the P-3B flight tracks on 7 and 9 March for the SA region and on 17 and 18 March for the NC region. The resulting number of trajectories was 17 for SA and 10 for NC. Figure 13b also includes scatter plots of CO and NO y in these air masses of the selected boundary layer (BL) outflow. For the calculation of EE in the BL, CO bkg and NO y(bkg) were determined to be 100 ppbv and 100 pptv, respectively, which are approximately the lowest values at 0 -2 km over the study area. The EE values in these BL outflows for the SA and NC categories were estimated to be 33% and 30%, respectively. These BL outflows were likely affected by a much smaller subset of the source regions than the WCB and COF air masses. Given that source regions exhibit different emissions ratios, ranging from 0.038 (Southeast Asia) to 0.060 (China) [Streets et al., 2003], the EE values in these BL outflows range over 19-33%. It should be noted, however, that these BL data were obtained only in limited cases and may not be representative. This is because concentrations of NO y in the BL air masses are highly variable depending on variable occurrences of removal processes since emissions and because the emission estimates include initial uncertainties as described above. Koike et al. [2003] have made a similar estimate for NO y of 15% (20 -40%) at 2 -7 km (0-2 km) transported from the northeast part of China to the Yellow Sea and Japan Sea during TRACE-P. These values are similar to those obtained in this study including the SA as a source region of NO y .   [31] Considering the EE value ranges of 10-20% for the WCB and COF at 2 -10 km and 30% at 0 -2 km, the results of this study suggest that there was additional NO y removal during transport by the WCB and COF, by a factor of 1.5-3. This removal of NO y , mainly due to HNO 3 , occurred most likely by precipitation associated with warm frontal clouds or convective clouds (see Figure 1) during transport. In order to examine this, temporal and vertical variations of relative humidity are shown in Figure 14, which were derived from the ECMWF data along the trajectories for the WCB and COF on 4 April described in section 4. Altitude ranges where the relative humidity reached 100% were about 3-4 km and 2 -4 km for the WCB and COF, respectively. These altitude ranges correspond to regions where precipitation had occurred. Mari et al. [2000] simulated that 70% of HNO 3 is efficiently scavenged by liquid precipitation in the lower (warm) part of the deep convective cloud below 7 km, by using a one-dimensional entraining/detraining plume model. The results of this study suggest that in addition to loss of HNO 3 in the BL, warm frontal clouds of the WCB as well as convective clouds provide an efficient sink for HNO 3 , especially at lower altitudes in the free troposphere.
[32] For the EE estimation in this analysis, there is an assumption that the dNO y /dCO ratios do not change due to dilution with constant dNO y /dCO in an assumed background air. Even if the assumed CO bkg value is 120 ppbv, which is the median value in the study area, the resulting EE values for each category increase by only 3%. This shows that the uncertainty of CO bkg values does not significantly affect the estimates of EE values. Uncertainties of the EE estimates also include the possibility of additional inputs of NO y during transport, such as lightning or aircraft emissions. However, as described in the previous section, the extremely low NO x /NO y ratios of less than 0.10 suggest that these additional inputs of NO y were absent in both WCB and COF air masses. Therefore the influence of these additional sources of NO y in the WCB and COF is considered negligible in the current analysis.

Summary and Conclusions
[33] We investigated the origins and pathways of synopticscale uplifted transport of pollutants and their effects on chemical distributions of NO y species over the western Pacific, based on systematic aircraft measurements obtained during TRACE-P, in conjunction with 5-day back trajectories. During the observational season, convection and cyclones over East Asia to the western Pacific were attributed to more frequent incursions of warmer air from the south due to the building up of Pacific high pressure, compared to decreasing southward movement of cold air because of the weakening of the Siberian High by March.
[34] Vertically uplifted patterns of lagrangian transport from the Asian continent were selected by the trajectories calculated from sampling points along the 9 DC-8 and 6 P-3B flight tracks. Of these, 73% originated over Southeast Asia, where biomass burning was most active during this season, while 27% originated over northeast China. The SA-originating trajectories were associated with warm conveyor belts within midlatitude cyclones and convective outflow (COF) over southern China. The strong influence of biomass burning was confirmed by the higher mixing ratios of CO and CH 3 Cl than the midlatitude median values, with a positive correlation of these species. On the other hand, the NC-originating trajectories were associated with convection within low-pressure systems. Air masses in the NC categories showed a large influence of fossil fuel combustion, revealed by higher mixing ratios of C 2 Cl 4 compared to those in the SA categories.
[35] WCB and COF associated with a stationary front were identified for the 4 April case. The WCB and COF were found to originate in southern China, suggesting that air masses in the biomass burning region were transported by both the stationary front and WCB as the cyclone developed. The dry airstream, identified as originating from the tropopause region from the trajectory and PV analysis, was found to be located adjacent to the COF. Therefore turbulent mixing can produce air masses with combined chemical signatures of stratospherically influenced air masses and pollutants from biomass burning.
[36] The fractions of NO x , PAN, and HNO 3 of the total NO y in the WCB and COF were found to be different from the midlatitude median values. The NO x levels were rela- Figure 13. (a) Scatter plot of CO against NO y for WCB (red), COF-SA (blue), and COF-NC (black) with linear fits for each category. The other values (gray) are data obtained at 2 -11 km over the study area shown for reference. (b) Scatter plots of CO against NO y for the SA (left) and NC (right) categories with the linear fits, as in Figure 13a. In addition, scatter plots between CO and NO y in the BL outflow are plotted (green) for both SA and NC regions. Also shown are the expected relations for the emissions assuming a molar ratio of NO y /CO of 0.034 and 0.051 for the SA and NC, respectively. The background CO (NO y ) mixing ratios at 2 -11 km are determined to be 80 ppbv (100 pptv), while those at 0 -2 km are determined to be 100 ppbv (100 pptv) over the study area.
tively low both in the WCB and COF due to the chemical imbalance between NO x and HNO 3 after the oxidation of a significant portion of the NO x , as well as due to the small influence of in situ sources in the isolated air masses. PAN was found to be the dominant form of the NO y in air masses transported by the WCB and COF. The high PAN/NO y ratio can be due to the substantial production of PAN in the boundary layer of the biomass burning and industrial region and the lower temperature in the middle to upper troposphere. The WCB and COF can transport NO x over long distances in the form of PAN to regions remote from Asian sources.
[37] The HNO 3 /NO y ratios in the WCB and COF were much lower than the midlatitude median values, indicating that HNO 3 was removed during transport within 1 -3 days. Fractions of NO y emitted from the surface over East Asia reaching the sampling points in the WCB and COF were estimated to be 10-20%. Similarly, fractions of NO y transported within the boundary layer were estimated to be 30% for the SA and NC, although the amount of data collected was quite limited for each air mass category. The results suggest that a large amount of NO y was removed during transport from the BL to the free troposphere. This was most likely due to loss of HNO 3 by precipitation during transport in the lower free troposphere in WCB and COF air masses, as well as loss of HNO 3 in the BL. A widespread region of precipitation over the east coast of the Asian continent to the North Pacific also suggests that the WCB and COF can provide an efficient sink for HNO 3 over the western Pacific.
[38] Acknowledgments. The TRACE-P mission was supported by the NASA Global Tropospheric Chemistry Program. The authors thank the crew of the NASA P-3B and DC-8 aircraft for their support. R. E. Newell and Y. Hu are acknowledged for providing PV values and a map of sea level pressure produced from ECMWF gridded data. The authors are grateful to H. Nakamura for helpful discussions on meteorology. Thanks also go to N. Oshima and Y. Morino for their assistance with the data analysis. Y. M. was supported by research fellowships of the Japan Society for the Promotion of Science for Young Scientists. Figure 14. Temporal and vertical variations of relative humidity (RH) in the air mass along the back trajectories for the (a) COF and (b) WCB on 4 April. The RH was obtained from the ECMWF data within the 6-hour interval prior to sampling.