Quantifying Pollution Inflow and Outflow over East Asia in Spring with Regional and Global Models

Understanding the exchange processes between the atmospheric boundary layer and the free troposphere is crucial for estimating hemispheric transport of air pollution. Most studies of hemispheric air pollution transport have taken a large-scale perspective using global chemical transport models with fairly coarse spatial and temporal resolutions. In support of United Nations Task Force on Hemi-spheric Transport of Air Pollution (TF HTAP; www.htap. org), this study employs two high-resolution atmospheric chemistry models (WRF-Chem and CMAQ; 36×36 km) driven with chemical boundary conditions from a global model (MOZART; 1.9×1.9 •) to examine the role of fine-scale transport and chemistry processes in controlling pollution export and import over the Asian continent in spring (March 2001). Our analysis indicates the importance of rapid venting through deep convection that develops along the leading edge of frontal system convergence bands, which are not adequately resolved in either of two global models compared with TRACE-P aircraft observations during a frontal event. Both regional model simulations and observations show that frontal outflows of CO, O 3 and PAN can extend to the upper troposphere (6–9 km). Pollution plumes in the global MOZART model are typically diluted and insufficiently lofted to higher altitudes where they can undergo more efficient transport in stronger winds. We use sensitivity simulations that perturb chemical boundary conditions in the CMAQ regional model to estimate that the O 3 production over East Asia (EA) driven by PAN decomposition contributes 20% of the spatial averaged total O 3 response to European (EU) emission perturbations in March, and occasionally contributes approximately 50% of the total O 3 response in subsiding plumes at mountain observatories (at approximately 2 km altitude). The response to decomposing PAN of EU origin is strongly affected by the O 3 formation chemical regimes, which vary with the model chemical mechanism and NO x /VOC emissions. Our high-resolution models demonstrate a large spatial variability (by up to a factor of 6) in the response of local O 3 to 20% reductions in EU anthropogenic O 3 precursor emissions. The response in the highly populated Asian megacities is 40– 50% lower in our high-resolution models than the global model, suggesting that the source-receptor relationships inferred from the global coarse-resolution models likely overestimate health impacts associated with intercontinental O 3 transport. Our results highlight the important roles of rapid convective transport, orographic forcing, urban photochem-istry and heterogeneous boundary layer processes in controlling intercontinental transport; these processes …


Introduction
Pollution export and import from local to global scales is a major concern because of wide-ranging consequences for human health, ecosystems, visibility degradation, changes in radiative forcing, the hydrological cycle, and tropospheric oxidizing capacity.Findings from numerous observational and modeling studies indicate that atmospheric composition throughout the Northern Hemisphere is affected by emissions from upwind continents (e.g.Jaffe et al., 1999, Lelieveld, et al., 2002;Stohl et al., 2002;Fiore et al., 2002;Akimoto, 2003).Wind speeds generally increase with Published by Copernicus Publications on behalf of the European Geosciences Union.height, causing pollutants at higher altitudes to be transported rapidly.Chemical species are relatively long-lived in the free troposphere due to weakened chemical loss and the absence of surface deposition.Therefore, hemispheric transport of atmospheric constituents cannot be understood without greater knowledge of exchange processes between the atmospheric boundary layer and the free troposphere.A number of key processes controlling vertical mass transport have been characterized, including the importance of midlatitude frontal passages (e.g.Cooper et al., 2002Cooper et al., , 2004)), deep convection (Lelieveld and Crutzen, 1994;Lawrence et al., 2003;Hess, 2005), and orographic forcing over complex terrains (Henne et al., 2004;Chen et al., 2009;Ding et al., 2009).The large-scale processes have become better understood in recent years, through integrated analysis of intensive measurement campaigns, satellite data, and ground-based observation networks with models (e.g.Jacob et al., 2003;Heald et al., 2003;Fehsenfeld et al., 2006;Singh et al., 2006Singh et al., ., 2009)).However, the importance of fine-scale mixing processes, such as by rapid deep convection, dry convection, mountain-valley breezes, and entrainment from the free troposphere into the atmospheric boundary layer, has not been well documented, and these transport mechanisms are often not resolved in global model simulations with a typical resolution of 2 • ×2 • .A number of previous studies suggested that the response of vertical transport and of the distribution of ozone (O 3 ) and its precursors varies in different models (e.g.Doherty et al., 2005;Lawrence et al., 2003;Kiley et al., 2003;Zhao et al., 2009).Convective systems encompass small-scale fair weather cumuli, active thunderstorms, and meso-scale convective systems (Cotton et al., 1995).The corresponding lifetime of these systems increases with their size from minutes to about half a day.Therefore, the representation of convective transport and associated clouds and precipitation processes need to be parameterized in numerical models and is typically sensitive to both the model's spatial and temporal resolution.
Anthropogenic emissions over Asia have experienced a rapid increase in recent decades (Richter et al., 2005;Zhang et al., 2007).Impacts of Asian emissions on the global environment have been examined through observational and model studies focusing on key export and transpacific processes (e.g.Liu et al., 2003;Carmichael et al., 2003;Cooper et al., 2004), seasonal and episodic variability (e.g.Yienger et al., 2000;Stohl et al., 2002), and impacts over North America (e.g.Fiore et al., 2002;Zhang et al., 2008;Singh et al., 2009;Reidmiller et al., 2009) and Europe (e.g., Stohl et al., 2007;Fiedler et al., 2009).Much less attention has been paid to the import of foreign emissions to Asia.A few studies using global tropospheric chemistry models estimated that European impacts on Asian surface O 3 range from 0.5-5 ppbv for various receptor areas (e.g.Wild et al., 2004;Holloway et al., 2007;Fiore et al., 2009;Jonson et al., 2009).In addition to these model estimates, signatures of European air masses have been detected at a remote mountain site in east Siberia (Pochanart et al., 2003).Compared to pollution export from East Asia and North America, rapid vertical exchange associated with frontal systems is weaker and less frequent over continental Europe (Stohl et al., 2002).The main transport pathway of European pollutants to East Asia occurs predominantly through the boundary layer (Wild et al., 2004), where winds are slower and chemical reactions and surface deposition may lead to additional pollutant loss.Thus, the representation of boundary layer and land surface processes is essential for estimating the import of European pollutants and subsequent impacts on surface air quality over East Asia.
It is not well understood how the key processes in tracer transport, oxidant formation and loss interact at finer temporal and spatial scales.To date, most studies of hemispheric air pollution transport have taken a large-scale perspective: using global chemical transport models (CTMs), focusing on synoptic-scale export events, and evaluating intercontinental source-receptor (S-R) relationships for relatively large regions.Under the United Nations Economic Commission for Europe (UNECE) Convention on Long Range Transboundary Air Pollution (LRTAP), the Task Force on Hemispheric Transport of Air Pollution (TF HTAP; www.htap.org)was established to advance the understanding of hemispheric transport of air pollutants in the Northern Hemisphere.EMEP (European Monitoring and Evaluation Programme) is a scientifically based and policy driven programme under the LRTAP Convention for international co-operation to solve transboundary air pollution problems.Under the umbrella of EMEP and TF HTAP, a set of global and hemispheric CTMs was employed to quantify intercontinental S-R relationships for O 3 (Fiore et al., 2009;Reidmiller et al., 2009;Jonson et al., 2009), oxidized nitrogen deposition (Sanderson et al., 2008), Arctic pollution (Shindell et al., 2008), aerosols, mercury, and persistent organic pollutants (TF HTAP, 2007).Most pertinent to our study are the findings that 8-15% of emitted NO x is transported over 1000 km from the source region boundaries (Sanderson et al., 2008), sub-continental scale variability in O 3 response to foreign emission changes (Reidmiller et al., 2009) and the wide range in model estimates for EU influence on EA O 3 in spring (over a factor of 2) (Fiore et al., 2009).Continental-scale average estimates of the surface O 3 response to foreign emission changes from the global coarse-resolution models have been used to assess the impacts of intercontinental transport on human health, mortality and crop productivity (Ellingsen et al., 2008;Casper-Anenberg et al., 2009), and will inform global air pollution policy.Thus, it is essential to understand the limitation and uncertainties in estimating S-R relationships using these coarsely resolved models (TF HTAP model resolution ranged from 1 to 5 degree horizontal).
While global CTMs provide an essential framework for estimating intercontinental scale S-R relationships, they often utilize fairly coarse spatial and temporal resolutions and simplified physical and chemical parameterizations because of computational limitations.Higher resolution regional-scale CTMs generally better simulate regional air pollution meteorology including mixing depth, wind speed, cloud and precipitation patterns, and hence improve the prediction of tracer transport and chemical evolution (e.g., Lin et al., 2008;Lin et al., 2009).In addition, significant uncertainty remains regarding the representation of non-linear urban chemistry relevant to air quality in these global-scale models.Emmerson and Evans (2009) compared an explicit chemical mechanism containing ∼5600 species and ∼13 500 reactions with six simplified tropospheric chemistry schemes utilized in the global models.They found significant uncertainties in the chemical schemes including the treatment of N 2 O 5 hydrolysis, PAN formation and sink, isoprene chemistry, and NO 3 nighttime chemistry.Therefore, it is important to evaluate the sensitivity of pollution export, chemical evolution and eventual mixing into surface air over remote regions to model physics and resolution, and to determine how estimated S-R relationships vary within a region (e.g.Lin et al., 2008b;Reidmiller et al., 2009).
We employ two high-resolution regional-scale atmospheric chemistry models -the online-coupled Weather Research and Forecasting model with Chemistry (WRF-Chem v3.0) (Grell et al., 2005) and the Community Multiscale Air Quality model (CMAQ v4.6) (Byun and Schere, 2006) -to examine the export of Asian pollutants to the global atmosphere and the import of European pollution to East Asia.Both models solve for gas-phase and heterogeneous atmospheric chemical processes in a regional, three-dimensional, Eulerian framework, and they represent the two most widely used state-of-the-science models of their kind.To realistically quantify pollution import at the regional boundaries, we apply dynamic boundary conditions of relatively long-lived species derived from the global Model for Ozone and Related Tracers (MOZART v2) (Horowitz et al., 2003).We evaluate MOZART, WRF-Chem, and CMAQ with aircraft observations; examine pollution export processes and the sensitivity of local surface O 3 responses to hemispheric transport.Intercomparison of three different regional and global models in representing pollution import and export permits a greater understanding of how global and regional processes interact, and it allows us to evaluate the role of physical parameterizations vs. resolution in contributing to model uncertainties.
The paper is structured as follows.The models and data used are briefly introduced in Sect.2.1.We focus our analysis and discussion on three questions: (1) how well did the models represent the observations (Sect.2.2) (2) how do the regional and global models differ in simulating the surfaceto-free troposphere exchange (Sect.3)? (3) What is the spatial variability of Asian O 3 response to European emission perturbations and the role of direct import of European O 3 vs.production driven by PAN decomposition (Sect.4)?We then discuss the connection between these regional-to-urban scale processes and hemispheric transport of air pollution, as well as the uncertainty in intercontinental S-R relationships estimated with global models.

Meteorological fields and emissions
The WRF regional weather forecasting model is run at 36 km×36 km, with eleven vertical layers in the lowest 2 km and remaining eighteen layers extending to 20 km.The online chemistry component of WRF employs the CBM-Z scheme for gas-phase chemical reactions, the MOSAIC scheme with 4 size bins for aerosol parameterizations, and the Fast-J photolysis scheme coupled with hydrometeors, aerosols and convective parameterizations (Fast et al., 2006 and references therein).The reader is referred to Lin et al. (2009) for a detailed description of the physical options applied in the WRF meteorological predictions.Atmospheric chemical reactions and advection are calculated offline in CMAQ, driven with archived hourly meteorology fields from WRF (hereafter WRF-CMAQ).We choose the SAPRC99/AERO3 scheme for the chemistry of trace gases and aerosols in CMAQ.Regional meteorology should be considered very similar between the two high-resolution air quality models since the feedback of the chemistry to meteorology is not turned on in the WRF-Chem simulations presented here.The key difference between WRF-Chem and WRF-CMAQ lies in the calculation of photolysis rates, gasphase chemistry, and aerosol dynamics.
The MOZART global model is run at ∼1.9 • ×1.9 • horizontal scale, with 28 levels in the vertical.MOZART is one of the global models participating in the HTAP emission perturbation experiments, which carried out a 20% reduction in anthropogenic emissions from East Asia (EA, 95 (Fiore et al., 2009).We applied MOZART simulations with base case emissions and with 20% reductions in anthropogenic emissions of O 3 precursors from EU (scenario SR6EU).Each regional model was run twice, with temporally varying (hourly for CMAQ, 6 hourly for WRF-Chem) chemical boundary conditions from the MOZART base and SR6EU simulations, respectively.Boundary conditions consisted of 12 species: O 3 , carbon monoxide (CO), peroxyacetyl nitrate (PAN), ethane, propane, acetone, nitrogen oxides (NO x ), sulfur dioxide (SO 2 ), and speciated aerosols.Ethane, propane, and acetone were selected to represent relatively long-lived non-methane volatile organic compounds (NMVOCs).We evaluate the response in O 3 concentrations over EA due to changes in EU pollutants imported into our regional model domain.This approach builds on previous regional model studies by applying boundary conditions from global models to estimate the impacts of foreign emissions (Carmichael et al., 2007;Lin et al., 2008b).To assess the relative influence of direct transport of EU O 3 vs.local production driven by decomposing PAN of EU origin, we conducted an additional simulation for CMAQ using boundary conditions of PAN and NO x from the MOZART base run and other chemical species from the MOZART SR6EU scenario.Differences in EA O 3 concentrations between the above sensitivity simulation and the simulation with all boundary species from the MOZART SR6EU scenario represent the contribution of European PAN to EA O 3 responses.The results are discussed in Sect. 4.
The regional and global models differ in the treatment of boundary layer dynamics and vertical transport processes as well as the spatial and temporal resolutions of driving meteorology.The meteorological fields for 2001 for driving MOZART were taken from the NCEP/NCAR reanalysis and were interpolated from a 6-h time structure to the 20-min time steps of the simulations.The meteorological fields for driving chemistry and transport in the regional models are predicted using the WRF model with 3min dynamic time steps and with initial and boundary conditions interpolated from the NCEP Final Analysis (FNL, 6 hourly, 1 • ×1 • ).Four-dimensional data assimilation was implemented to force the simulated meteorology toward the FNL analysis, and the model was reinitialized every five days to reduce the accumulative errors as discussed in Lin et al. (2009).We choose the Yonsei University (YSU) nonlocal boundary layer vertical diffusion scheme for planetary boundary layer (PBL) parameterizations (Hong et al., 2006) in WRF.The YSU PBL scheme includes an explicit treatment of the entrainment process of heat and momentum at the top of PBL, which resolves two major problems in other PBL schemes: too much mixing with strong wind shear, and too little mixing in the convection-dominated PBL (Hong et al., 2006).Such explicit treatment of the entrainment processes was not implemented in the boundary layer scheme of Holtslag and Boville (1993) employed in the MOZART model.Inadequate treatment of entrainment processes at the PBL top may affect the budget of pollution exchange between the boundary layer and the free troposphere.Convective parameterizations in WRF employ an improved version of the Grell-Devenyi ensemble cumulus scheme (Grell et al., 2002) -Grell-3d, which allows for subsidence in neighboring columns.The MOZART model diagnoses convective mass fluxes using the shallow and midlevel convective transport of Hack (1994) and deep convection scheme of Zhang and McFarlane (1995).
All models employ biomass-burning emissions from the Global Fire Emission Database (GFED v2) (van der Werf et al., 2006).Both CMAQ and WRF-Chem adopt fossil fuel emissions from Streets et al. (2003), updated for the year 2001 as described in Carmichael et al. (2007).Fossil fuel emissions in MOZART are based on EDGAR v2.0 and so representative of the early 1990s (Horowitz et al., 2003).Biomass burning emissions in MOZART are distributed vertically up to 4 km with 70% of these emissions occurring below 2 km, other emissions for all three models are constrained in the surface model layer.Comparing the global and regional model inventories, we find that the largescale patterns of major O 3 precursors are similar between the two inventories, with the regional inventory estimating higher emissions from urban areas and power plants (Supplementary Fig. S1, http://www.atmos-chem-phys.net/10/4221/2010/acp-10-4221-2010-supplement.pdf).Total emissions of NO x and CO in the regional emission inventory are ∼15% and ∼5% higher, respectively (Table 1).Total emissions of NMVOCs are close between MOZART and CMAQ, but are ∼50% lower in WRF-Chem.WRF-Chem employs an online biogenic VOC emissions module based on Simpson et al. (1995) and Guenther et al. (1994), and we find that calculated isoprene concentrations over Southeast Asia in WRF-Chem are lower by a factor of 2-5 than CMAQ (Supplementary Fig. S2, http://www.atmos-chem-phys.net/10/4221/2010/acp-10-4221-2010-supplement.pdf), which uses offline biogenic VOC emissions from the POET database for 2000 (Granier et al., 2005;Oliver et al., 2003).Anthropogenic VOC emissions are also approximately 15% lower in WRF-Chem since the CBM-Z chemical scheme considers a smaller set of VOC species than the SAPRC99 mechanism used in CMAQ.Discrepancies in VOC emissions and speciation in the models partly contribute to the bias of O 3 predictions and responses to emission perturbations, and will be discussed in the remaining sections.

Overall model evaluation
Evaluating models with observations is essential for establishing credibility in all applications of model results.Our study focuses on March 2001 since intercontinental transport is generally strongest in the Northern Hemisphere spring season (Yienger et al., 2000;Wild et al., 2004;Stohl et al., 2002).Focusing on March 2001 also gives us an opportunity to evaluate the representation of pollutant export and chemical evolution in regional and global models against intensive measurements obtained from the TRACE-P (Transport and Chemical Evolution over the Pacific) campaign (Jacob et al., 2003).Whereas CMAQ has been well evaluated in prior studies examining chemistry and transport over East Asia (e.g.Uno et al., 2003;Zhang et al., 2004;Lin et al., 2008a 2009), few studies to date have presented WRF-Chem results for this part of the world (Tie et al., 2009;Matsui et al., 2009).Here, we give an overall evaluation of WRF-CMAQ, WRF-Chem, and MOZART against TRACE-P aircraft measurements, and a forthcoming manuscript (hereafter referred to as Lin et al., 2010) will expand the intercomparison and evaluation of WRF-Chem and WRF-CMAQ over East Asia.
The models were sampled every five minutes along the flight tracks.Figure 1 shows the scatter plots of observed and modeled major tracer species (CO and ethane) and photochemical oxidants (O 3 and PAN) at altitudes below 2 km, 2-4 km and above 4 km.Both ethane and CO react relatively slowly in the gas phase and are not efficiently removed by wet processes, thus their distributions reflect mainly the source distribution and transport pathways.All models reproduce well the variations of observed ethane and CO along the flight tracks with correlation coefficients of 0.6-0.8,suggesting that large-scale transport processes were well www.atmos-chem-phys.net/10/4221/2010/Atmos.Chem.Phys., 10, 4221-4239, 2010 captured in the models.Mixing ratios of CO are underestimated by a factor of two below 2 km, consistent with prior studies suggesting that CO emissions from China are higher than assumed in the TRACE-P inventory (Streets et al., 2003;Carmichael et al., 2003;Streets et al., 2006 and references therein).WRF-Chem gives the greatest correlation coefficients with observations of CO and ethane among all three models, indicating the improved ability of the online-coupled climate-chemistry model to resolve more sharply the temporal variation of regional meteorology and associated chemical evolution.Section 3 will further discuss the mechanisms explaining model behavior, in particular the large differences between regional and global models in simulating vertical transport of chemical species.Larger differences are found in the models' simulation of the main photochemical products, O 3 and PAN (Fig. 1b), than in CO and ethane (Fig. 1a).This divergence in model estimates suggests that important questions remain regarding the complex photochemical processes involving NO x and VOC precursors.The two regional models give higher regression coefficients than the MOZART global model, as would be expected, given the ability of CMAQ and WRF-Chem to resolve more sharply the spatial gradients in the short-lived O 3 and PAN precursors.We find that MOZART generally underestimates the production of boundary layer O 3 and PAN in springtime, although the HTAP global model intercomparison found that most models, including MOZART, overestimate summertime O 3 (Fiore et al., 2009).Comparing the two regional models, we find that CMAQ gives better statistical scores for O 3 below 2 km, while WRF-Chem better simulates observed O 3 and PAN from 2-4 km.For near surface O 3 below 2 km, CMAQ shows correlation coefficient of R=0.89 and mean bias of MB=1.8 ppbv as compared to R=0.85 and MB=−3.2 ppbv for WRF-Chem.WRF-Chem's underestimate of boundary layer O 3 likely reflects lower VOC emissions from both anthropogenic and biogenic sources included in the simulation (Table 1).
Figure 1b indicates significant variations in the calculated concentration of PAN among the models.The magnitude and variation of PAN mixing ratios along the flight tracks are generally well reproduced in the WRF-Chem model with correlation coefficients of 0.6-0.7 and a negative mean bias <0.1 ppbv.MOZART underestimates PAN at all altitude levels.CMAQ predicts the highest concentrations of PAN, and tends to overestimate the observed PAN levels between 2-4 km altitudes.We find that the positive bias of PAN in CMAQ tends to occur in the low-level postfrontal outflow sampled during TRACE-P and often correlates with O 3 underestimates in the same airstreams (figure not shown).The negative correlation between PAN and O 3 suggests that CMAQ's underestimate of O 3 in the postfrontal outflow is likely due to the excessive NO x uptake by PAN.Lin et al. (2009) reported that CMAQ with the SAPRC99 chemical mechanism predicted 50% higher concentrations of surface PAN over polluted regions than with the Carbon Bond IV (CBIV) mechanism.The SAPRC99 mechanism includes peroxy propionyl nitrate (PPN) and other higher alkyl PAN analogues in addition to PAN, and has explicit treatments of major PAN precursors -acetaldehyde, glyoxal, methylglyoxal, and separate ketone species (ACET, MEK, MVK).Our evaluation with TRACE-P aircraft measurements reveals that SAPRC99 somewhat agrees better than the carbon-bond mechanism in simulating the higher PAN mixing ratios in polluted surface air masses, but overestimates PAN at cleaner conditions above 2km, consistent with the evaluation with INTEX-B aircraft measurements over the eastern Pacific presented by Adhikary et al. (2010).The errors may come from PAN-forming VOC emissions (including biogenic isoprene) and speciation, the thermal decomposition rate, the PAN+OH sink, and other rate constants.The variations in model calculated PAN will affect estimates of the long-range transport of reactive nitrogen and hence O 3 production over remote regions.This will be discussed in Sect. 4.

Export of Asian pollutants
Midlatitude cyclones tracking from west to east have been recognized as the primary mechanism for rapidly exporting air pollution from Asia and North America to downwind continents (Cooper et al., 2002(Cooper et al., , 2004;;Stohl et al., 2002).Questions remain, however, as to the amount of surface emissions that make their way to the free troposphere.Here, we examine the importance of both synoptic and fine-scale venting processes in controlling the total budget of Asian emissions exported to the free troposphere as represented in WRF-Chem, WRF-CMAQ and MOZART.Chemistry is dynamically solved online in WRF-Chem without any interpolation of meteorological fields, while both CMAQ and MOZART are driven with temporally interpolated meteorological fields (1-h to 12-min for CMAQ and 6-h to 20-min for MOZART).Therefore, we focus our analysis and discussion of export processes on the comparison of WRF-Chem and MOZART results (Figs. 2,3,4,and 6) to highlight the role of model temporal and spatial resolutions.Figure 5 and supplementary Fig. S4 (http://www.atmos-chem-phys.net/10/4221/2010/acp-10-4221-2010-supplement.pdf)present WRF-CMAQ results analogous to those of WRF-Chem, and the difference between the two regional models is also discussed below.

Episodic nature of Asian outflow
An important feature of the East Asian winter monsoon is frequent southeastward intrusion of cold air triggered by the Siberian anticyclone.Along with the propagation of cold fronts over China and Japan, a midlatitude synoptic wave sweeps northeastward over the East Asian coast.We identified a total of five major frontal events during the March 2001 study period, diagnosed by the decrease in daily mean surface temperature (Liu et al., 2003).Figure 2  MOZART and WRF-Chem calculated vertical profiles of CO and PAN zonal fluxes along 140 • E for six frontal outflow events.The comparison for other six episodes is presented in the supplementary Fig. S3 (http://www.atmos-chem-phys.net/10/4221/2010/acp-10-4221-2010-supplement.pdf).A key feature inferred from Figs. 2 and S3 is that pollution plumes in the high-resolution WRF-Chem model tend to have a multiple -layered structure and are generally more intensified in a certain region and located at higher altitudes where they can undergo more efficient transport in stronger winds.In contrast, pollution plumes in the global MOZART model are typically diluted and insufficiently lofted to higher altitudes.For example, the plumes extend to the upper troposphere (4-9 km) in WRF-Chem as compared to the middle troposphere (2-4 km) in MOZART for the episodes of 7, 20, and 26 March.WRF-Chem also predicts stronger boundary layer outflow confined within the 35-45 N latitude bands during the episodes of 18-20 March.The magnitudes of CO and PAN fluxes for some regions differ by up to a factor of two between the two models.Similar patterns are found for CO, PAN, and ethane (figure not shown), suggesting that the major differences between the models are driven by transport.The stronger outflow of PAN to the western Pacific in WRF-Chem implies that the amount of potential O 3 production over downwind regions (e.g.Western US) driven by decomposing PAN in subsiding transpacific plumes will be larger than in MOZART.Figure 3 illustrates the variation of total fluxes of CO and PAN along 140 • E that is integrated from 25-40 • N latitudes and from 4-8.5 km altitudes.The 4-8.5 km altitude band is selected to examine the relative pollutant content of air masses transported to the upper troposphere in the models.The shaded bands in Fig. 3 indicate the timing of cold frontal passages over central China, and it should be noted that the shaded area does not necessarily cover the entire outflow episode as the influence of a cold front on pollution outflow may persist over a few days.Outflow of CO and PAN exhibits a strong episodic behavior in association with the passage of midlatitude synoptic waves.Both regional and global models capture elevated frontal outflow, with the WRF-Chem regional model estimating stronger zonal fluxes in the upper troposphere (>4 km).While the global MOZART model can resolve synoptic-scale transport events, the intensity of CO and PAN fluxes is generally weaker than simulated in WRF-Chem.The enhanced lofting of Asian pollutants to higher altitudes as simulated in the WRF-Chem model occurs every 3-4 days throughout the study period.

compares
Total emissions of CO are only 5% higher in WRF-Chem than MOZART, which is not enough to explain the striking discrepancies in the vertical location of pollution plumes as illustrated in Fig. 2. While the injection heights of emissions could contribute to the discrepancy, both anthropogenic and biomass burning emissions are distributed in the surface  layer of WRF-Chem so the lofting of pollution to higher altitudes in WRF-Chem is not due to the injection heights of emissions.The greatest discrepancies between the two models are found during 5-8 March when a vigorous cold front swept over East Asia.Calculated total zonal fluxes of CO and PAN are approximately 50% higher in WRF-Chem than in MOZART for this episode.We focus below on this episode, compare model simulated vertical profiles of trace gases with observations, and examine the mechanisms explaining model behavior.

Role of rapid vertical transport
Figure 4 shows the WRF-Chem forecast of elevated CO in the free troposphere in association with the passage of the cold front on 7 March.A deep cyclone was located over northern Japan (centered at 145 • E, 40 • N), coexisting with a Siberian anticyclone and a strong convergence band extending from southwest China all the way to northern Japan.We have compared the simulated precipitation pattern in the frontal zone with the multi-satellite data from the Tropical Rainfall Measuring Mission (TRMM) (figure not shown), and their close agreement suggests that the placement, geographical extent, and intensity of this frontal event are well captured by the WRF model.Mixing ratios of CO at the 5-km level reach 200-350 ppbv within the warm conveyor belt, east of the cold front.To identify the origins of the elevated CO in the middle and upper troposphere, we parison of model simulated convective and updraft mixing of CO in association age of the cyclonic wave during 6-7 March.Shown are vertical distributions of CO uivalent potential temperature (contours) along the frontal system convergence band igure 4. look at the vertical distribution of CO along ∼25 • N, approaching the south side of the convergence zone sweeping over west Myanmar, a heavily burned area in March 2001 (Fig. S1, http://www.atmos-chem-phys.net/10/4221/2010/acp-10-4221-2010-supplement.pdf), and major population centers along the Yangtze River -Chongqing, Wuhan, and Shanghai.Results are shown in Fig. 5 for three models, and we find that the updraft mixing of CO is strongest in WRF-Chem and weakest in MOZART.WRF-Chem predicts strong CO updrafts near Chongqing (105-115 • E) extending to the upper troposphere (>4 km) and WRF-CMAQ shows a similar vertical distribution of CO, suggesting that some transport processes were not adequately captured in the global MOZART model, which resulted in a weaker lifting of CO than the two regional models.The WRF-Chem panel in Fig. 5 also shows the vertical distributions of equivalent potential temperature (θ e ), calculated using the formula from Bolton (1980).In stable conditions, θ e increases with altitude.If θ e decreases with height, convection can occur, as is the case, above Myanmar (90-100 • E) and (105-110 • E), indicating convective conditions.In lar, deep convection near Chongqing, leading to strong upward transport of CO to the upper troposphere in the regional models.The lofting of occurred within just a few hours and was rapidly pushed offshore by the strong westerly winds.
This event was sampled by the NASA DC-8 platform flying through the frontal region near the East China coast on 7 March (Fig. 4). Figure 6 compares the observed and modeled vertical distributions of CO, O 3 and PAN along the flight path.It can be seen that the observed CO distributions along the flight path clearly show outflow within the frontal zone (04:00-06:00 UTC) extending to altitudes of ∼6-8 km.Impacts of the primary airstreams that composed this cyclone, e.g.strong advection and lifting in the warm conveyor belt (03:00-06:00 UTC) and dry intrusion of upper level background air (08:00-10:00 UTC), are remarkably well simulated in the high-resolution WRF-Chem model.WRF-Chem predicted outflow of CO to the western Pacific exhibits a complex structure appearing in three vertical layers: mixing of fresh emitted emissions capped in the boundary layer below 1 km; background outflow of CO originated from the coastal region of China confined below ∼3 km over the prefrontal region; and large-scale outflow above 4 km within the frontal zone.While the global MOZART model generally captures the low-level outflow below ∼3 km, the plume in the frontal region (04:00-06:00 UTC) is diluted and appears in the wrong place, strongest in the lower free troposphere (2-4 km) rather than in the middle and upper free troposphere (6-8 km) as in WRF-Chem and in the observations.The export of CO emissions during the frontal event discussed above is associated with elevated mixing ratios of major photochemical oxidants O 3 and PAN (Fig. 6).Observed mixing ratios of O 3 in the intense CO plume exceed 60 ppbv between 4-8 km (04:00-06:00 UTC), comparable to the lowlevel outflow confined below 4 km in the prefrontal region, and this observed feature of O 3 enhancement in the frontal region is well captured in the WRF-Chem model.The frontal plume in MOZART (2-4 km, 04:00-06:00 UTC), however, does not show an O 3 enhancement as strong as observed and simulated in WRF-Chem.MOZART somewhat captures the frontal outflow of PAN, but underestimates the observed mixing ratios of PAN by 20-30%.Figure 5 illustrates a similar vertical transport pattern between WRF-CMAQ and WRF-Chem, and supplementary Figure S4 demonstrates the capability of WRF-CMAQ to capture the timing (04:00-06:00UTC) and location (4-8 km) of DC-8 observed frontal outflow of CO on 7 March.We find that CO mixing ratios in WRF-CMAQ are generally lower by ∼15% than simulated in WRF-Chem.While differences in isoprene emissions could contribute to the discrepancy, the isoprene emissions are larger in WRF-CMAQ (Fig. S2) so the excess CO in WRF-Chem is not due to CO produced during isoprene oxidation.The WRF-Chem model domain covers a larger area of biomass burning region in Southeast Asia (Figs.S1 and S2, http://www.atmos-chem-phys.net/10/4221/2010/acp-10-4221-2010-supplement.pdf).The CMAQ model domain does not include the WRF boundary grid boxes (five in total), and as a result a part of the heavily burned area at the western boundaries has been cut, which partly explains CMAQ's lower estimate of CO mixing ratios in elevated plumes affected by fire emissions.In addition, we suspect that the temporal resolution of convective processes, solved every three minutes in WRF-Chem as compared to the interpolation from a 1-h to 12-min time structure in CMAQ, will play a role in influencing the magnitude of elevated CO.Sources of CO in the layers above 4 km sampled by the TRACE-P aircrafts are mainly attributable to biomass burning emissions in Southeast Asia (Carmichael et al., 2003) and to industrial emissions along the frontal system convergence bands as illustrated in Fig. 4. All models in this study applied the same dataset for biomass burning emissions (Fig. S1).The striking bias in simulating vertical distributions of trace gases illustrated in Figures 2, 5 and 6 is primarily due to the representation of key meteorological processes in different models.Deep convection is an important mechanism for vertically transporting tropical and subtropical biomass burning emissions out of the atmospheric boundary layer into the middle and upper troposphere (Duncan et al., 2003;Hess, 2005).The occurrence of cold fronts tends to enhance subtropical and tropical deep convection because of the intense low-level convergence along their leading edge (Garreaud, 2001), as is the case simulated in the WRF model.We find consistent enhancement of deep convection embedded in rising airstreams for multiple warm conveyor belt transport events (e.g.,7,20,26 March) in March 2001 (figure not shown).MOZART diluted and displaced the plumes at the lower free troposphere for these episodes, suggesting its limited ability to resolve or properly parameterize this rapid deep convection that develops along the leading edge of the frontal system convergence bands.The CO vertical cross-sections indicate that the updraft mixing of biomass burning emissions from Southeast Asia for some episodes is likely enhanced by orographic forcing over the complex terrains in Myanmar and southwest China (figure not shown).A tracer modeling study by Lin et al. (2010) suggests that the trough (low) formed on the lee side of the Tibetan Plateau and Indochina mountains, is an important transport mechanism for uplifting biomass-burning emissions from Indochina.Other studies have reported that mountain-valley breezes affect the vertical forcing of emissions from polluted Central Eastern China up the nearby mountain slopes (e.g.Ding et al., 2009;Chen et al., 2009).
In summary, meso-scale model studies and supporting observations point out the important roles of deep convection and orographic forcing in enhancing pollution venting during warm conveyor belt lifting events.Inadequate treatment of these transport processes can lead to insufficient lofting of pollution to higher altitudes as seen in the MOZART model.The dilution of intense plumes due to numerical diffusion and coarse resolution in the global-scale models further diminishes the impacts of vigorous episodic transport events (Heald et al., 2003;Wild et al., 2004;Pfister et al., 2006;Fang et al., 2009).Enhanced lifting of surface pollutants to the free troposphere where it can undergo efficient longrange transport -as observed in aircraft studies and simulated in the regional models -might not immediately contribute to the degradation of surface air quality over downwind continents.However, pollutants are generally longlived in higher altitudes and the mixing into the background troposphere air can eventually lead to increasing levels of background atmospheric constituents over downwind continents (Fiore et al., 2002).

Uncertainty in export processes
Vertical transport and distribution of tracer species are typically sensitive to the convection schemes used in different models (e.g.Doherty et al., 2005;Lawrence et al., 2003;Kiley et al., 2003;Zhao et al., 2009).The WRF simulations in this study employed the new Grell-3d scheme for convective parameterization (Grell et al., 2002), and evaluation of chemical distributions presented here illustrates the capability of the Grell-3d convection scheme to closely simulate vertical exchanges of air masses.The convective transport schemes of Hack (1994) and Zhang and McFarlane (1995) were applied in both MOZART version 2 (Horowitz et al., 2003) and version 4 (Emmons, et al., 2010).These differences in convective parameterizations likely also contribute to the insufficient lofting of pollution plumes to higher altitudes in the MOZART model.
In addition, high-resolution meteorological fields calculated using the meso-scale models (e.g., MM5, RAMS and WRF), as opposed to the coarse reanalysis data (e.g., NCEP/NCAR, ECMWF) for driving offline global CTMs, further support the improved simulation of tracer vertical transport in the regional scale CTMs.For example, TRACE-P post-campaign analysis showed that the STEM regional model driven with RAMS meteorology (80×80 km 2 , hourly) was also able to capture the elevated CO outflow extending to the upper troposphere for the 7-March episode discussed above (Carmichael et al., 2003), while the global GEOS-Chem model driven with coarse-resolution meteorology (2 • ×2.5 • , 3 hourly) displaced the plume in the lower free troposphere similar to the distribution in MOZART (Liu et al., 2003).Recent model analysis of observations from the 2006 INTEX-B campaign over the Northeast Pacific found that most current global-scale models have limited ability to reproduce discrete Asian gas and aerosol plumes arriving at the North American West Coast (Singh et al., 2009;Dunlea et al., 2009).We suspect that this is partly due to inadequate treatment of small-scale venting over the Asian continent as illustrated in this study and excessive dilution during transpacific transport due to poor resolution in the global-scale models.
We also looked at the vertical profiles of CO at three stations along the East Asian Coast from ten HTAP global models providing vertical profiles for intercomparison and evaluation with ozonesonde observations (Jonson et al., 2010).It appears that none of the global models captured the strong upper-troposphere (6-9 km) CO outflow simulated in WRF-Chem for the 7-March episode, but there is a significant variability of CO vertical profiles among models for the other four frontal episodes in March 2001 (figure not shown).Some events suggest MOZART and WRF-Chem is on the high side of altitudes as compared to the HTAP models and others in the low side, suggesting a real need for more process oriented studies.Among the HTAP global models, MOZECH, CAM-Chem, and CHASER show relatively greater CO outflow to the free troposphere than others.Jonson et al. (2010) examined the vertical profiles of O 3 from the HTAP global models.They found a similar spread in the modeled vertical profiles of O 3 and the agreement between ozonesonde measurements and individual models tends to be at its minimum in spring and summer.The timing and location of pollution plumes may vary among models, but vertical profiles are only available at the individual stations at 12-h intervals and a detailed comparison of HTAP global models is beyond the scope of this study.
Based on available data, we are only able to evaluate model skill at capturing Asian outflow in spring, but we would expect that the fundamental limitations in global model resolution and their ability to resolve subgrid-scale surface-to-free troposphere exchange and chemical processing induce large uncertainties in estimating continental outflow in other regions and seasons.Other studies indicate similar resolution-dependant processes with importance to large-scale continental export.For example, Fang et al. (2009) reported that MOZART underestimated an observed plume with CO in excess of 500 ppbv at 6-10 km from boreal fires during the 2004 INTEX-NA field campaign over www.atmos-chem-phys.net/10/4221/2010/Atmos.Chem.Phys., 10, 4221-4239, 2010 North America by a factor of 5, likely due to excessive dilution of the fire plume or inadequate treatment of pyroconvection that injected strong boreal fire emissions into the upper troposphere and lower stratosphere (Turquety et al., 2007).
The smaller-scale venting mechanisms have greatest importance in warm seasons when large-scale stirring by synoptic systems is diminished.Kiley and Fuelberg (2006), using a meso-scale meteorological model (MM5), suggested that weak, midlatitude cyclones in summer are capable of producing vertical lifting as great or greater than much stronger cyclones over North American East Coast.For European outflow, Henne et al. (2004Henne et al. ( , 2005) ) revealed a strong influence of topographic venting on O 3 mixing ratios in the lower free troposphere over and downwind of the Alps, and orographic forcing is expected to play an important role in O 3 production and outflow on a European scale.
In addition to transport processes discussed above, there remains a large uncertainty in the outflow of reactive nitrogen (NO y ), including the fraction of individual NO y species, removal in precipitation associated with frontal lifting and convective processes, and further photochemical processing.The current generation of models captures typical correlations between tracers and NO y species in the plumes, but the estimate of the NO y amount transported from the continental boundary layer to free troposphere varies among models (Cooper et al., 2002;Parrish et al., 2004;Li et al., 2004;Sanderson et al., 2008).

European impacts over East Asia
We first discuss the spatial sensitivity of monthly mean surface O 3 responses over EA to emission changes in EU (10 • W-50 • E, 25 • N-65 • N) (Sect.4.1).Section 4.2 presents the relative contribution of direct transport of O 3 vs.production driven by PAN decomposition.Section 4.3 discusses time and vertical evolution of EU influences on tropospheric O 3 and CO over China and Japan.We select two regionally representative mountain observatories -Mount Hua (34.5 • N, 110.1 • E, 2064 m) in central China and Mount Happo (36.7 • N, 137.8 • E, 1850 m) in central Japan -to demonstrate the temporal evolution of European impacts.Regional air pollution over East Asia and source attributions have been widely examined at Mount Happo in the past decade (e.g., Carmichael et al., 1998;Wild et al., 2004;Lin et al., 2008aLin et al., , 2009;;Tanimoto et al., 2009), and at Mount Hua since the set up of regular O 3 measurements in 2004 (e.g., Li et al., 2007;He et al., 2008).
We focus our discussion of European impacts on the model results from CMAQ and MOZART because the two models emit similar quantities of NMVOC, which were found to significantly affect the response of O 3 to emission perturbations (e.g.Fiore et al., 2009).In addition, CMAQ applies hourlyvarying chemical boundary conditions, consistent with the output frequency of chemical concentrations in MOZART, while WRF-Chem employs the boundary conditions at sixhour intervals.

Spatial sensitivity of ozone responses
Figure 7 shows monthly mean O 3 responses over EA to 20% reductions in EU anthropogenic emissions of O 3 precursors.The perturbation in chemical boundary conditions of CMAQ in association with EU emission changes induces a 0.3 ppbv response in spatially averaged surface O 3 over EA.Fiore et al. (2009) reported that the model spread in the EA O 3 response to 20% EU emission reductions is largest in March, ranging from 0.1-0.7 ppbv.The CMAQ estimate falls within the range of global model estimates, but is somewhat lower (0.1 ppbv) than both MOZART and the ensemble mean (0.4 ppbv) of 15 HTAP models.
The large-scale spatial pattern of EU impacts over EA is similar between MOZART and CMAQ.Both models show 0.1-0.2ppbv stronger response of O 3 in the northwest mountainous regions than the North China Plains (30-45 • N, 115-125 • E), with the fine-scale model better resolving the sharp orographic gradients.Surface topography can enhance ventilation of surface emissions through mountainvalley wind systems as well as suppress horizontal transport across mountain ridges.Consistent with the results presented here, a few studies reported that the impacts of Asian emissions are strongest on O 3 and PM levels over the Rocky Mountains of the western United States (Jaffe et al., 2003;Reidmiller et al., 2009).Mountain top entrainment of free tropospheric air and weakened surface depositional and chemical loss are key factors contributing to enhanced responses over these regions.
We find a large local and regional variability (by up to a factor of 6) in the EA response to changes in EU O 3 and relevant precursors imported into our regional model domain (Fig. 7b), indicating the important role of surface processes, which differ from region to region.For example, the CMAQ regional model clearly demonstrates the sharp gradients of enhanced O 3 loss through surface deposition over the land, due to uptake by plants and stronger boundary layer turbulence, as opposed to over the ocean.Sharp gradients of coastal lines are not adequately resolved in MOZART, leading to a higher estimate of EU enhancements, as compared to CMAQ, on surface O 3 over the coastal regions of Eastern China, South Korea, and South Japan.Another key feature observed from the regional model results is that EU enhancements appear to be lowest (0.1-0.3 ppbv) over surface air in megacities such as Beijing, Seoul and Tokyo.In these urban areas, European O 3 imported into the regional model domain is titrated by high local emissions of NO x (Fig. S1, http://www.atmos-chem-phys.net/10/4221/2010/acp-10-4221-2010-supplement.pdf).The global model estimate is 0.4-0.6 ppbv, nearly double than the estimate by the regional model.The large difference over the megacities between the global and regional model results  Anenberg et al., 2009), as 50% of the world's population lives in cities where complex urban photochemistry may fundamentally alter the local impacts of imported pollution, and urban areas are not resolved in the HTAP global models.

The role of PAN and local chemical regimes
Foreign enhancement on local O 3 budgets can occur either through direct transport of O 3 and/or through production driven by PAN decomposition.PAN acts as an effective reservoir for NO x , and its thermal decomposition can lead to O 3 production over remote areas far away from emission source regions (Moxim et al., 1996).We conducted sensitivity simulations that perturb PAN boundary conditions in the CMAQ model, and estimate that the contribution of NO x released from PAN decomposition accounts for nearly 20% of the spatially averaged total O 3 response over East Asia to EU emission reductions (Fig. 7c).To our knowledge, this is the first time that the influence of foreign PAN on local O 3 budgets has been estimated for a region.The contribution of PAN driving O 3 production can reach up to 50% in subsiding plumes in association with cold frontal passages over central China (Mt.Hua -2064 m) and central Japan (Mt.Happo -1850 m) (Fig. 8).Consistent with our model results, recent aircraft studies have detected strong O 3 enhancements driven by PAN decomposition in subsiding plumes containing large amount of PAN formed over upwind continents (e.g.Heald et al., 2003;Real et al., 2007;Zhang et al., 2008).The significant role of PAN in driving O 3 production over remote regions as illustrated in our study indicates that the variation in PAN concentrations simulated in the different models will strongly affect the assessment of intercontinental S-R relationships of O 3 .Emmerson and Evans (2009) found that the PAN concentration calculated by six chemical schemes employed in global models varies by up to a factor of five.Our evaluation with TRACE-P aircraft measurements shows that there also exist large uncertainties in regional models' PAN simulations.The MOZART simulation presented in this study underestimates Asian outflow of PAN by 20-30% (Figs.1b and 6).Fiore et al. (2009) reported a wide range in the NA and EU O 3 response to a 20% decrease in EA emissions, and the response in MOZART is ∼0.1 ppbv lower than the ensemble mean of 15 HTAP models for March.While a variety of factors can contribute to the lower response of O 3 in MOZART, the underestimate of Asian PAN outflow reported here may also play a role.
The chemical regime where the subsiding PAN decomposes determines the O 3 production efficiency and varies with local VOC reactivity and the VOC-to-NO x ratio.Analysis of aircraft measurements and model sensitivity results suggest that O 3 formation over Northeast Asia is in the NO x -saturated regime in spring (Carmichael et al., 2003).Thus, increasing NO x from European sources has a relatively weaker influence on local O 3 production in a model with a lower VOC/NO x emissions ratio (because of increased termination from OH+NO 2 rather than OH propagation occurring by OH+VOC (Seinfeld and Pandis, 2006)).For example, the WRF-Chem model using the Carbon Bond Mechanism of Zaveri et al. (1999) (CBM-Z) has nearly 15% lower anthropogenic NMVOC emissions and 50% lower from biogenic sources, as compared to either CMAQ or MOZART.This difference likely explains the absence in WRF-Chem of the O 3 enhancements simulated by CMAQ and MOZART during the cold surges of 3-4 and 6-9 March at Mount Hua as well    as 10-12 March at Mount Happo (Fig. 9).In these subsiding plumes, as much as approximately 50% of O 3 enhancements simulated in CMAQ are driven by foreign PAN decomposition (Fig. 8).The CO enhancements are on the same magnitude among the models (Supplementary Figure S6), further supporting that the difference in O 3 enhancements during cold surges is mainly due to the variation in O 3 production efficiency of European NO x rather than in transport.We conclude that the strength of the local surface O 3 response to reductions in foreign emissions is very sensitive to the chemical regime as represented in the model chemical mechanism.

Time and vertical evolution of European air masses over East Asia
Figure 8a shows that European influence at Mount Hua is generally strongest (up to 10% of total O 3 ) on regionally clean days and weakest (<3%) on regionally polluted days, consistent with prior studies for North America (e.g., Fiore et al., 2002) S6, http://www.atmos-chem-phys.net/10/4221/2010/acp-10-4221-2010-supplement.pdf).Regional-scale models (CMAQ and WRF-Chem), by employing chemical boundary conditions from the global model (MOZART), also capture major events of European CO enhancements over China and Japan in association with cold surges.European impacts over Northern China are primarily episodic, and we found five events with 30-50 ppbv enhancements of CO and 3-5 ppbv of O 3 occurring in the atmospheric boundary layer during a single month of March 2001.In contrast, Yienger et al. (2000) estimated that only 3-5 of Asian pollution events directly impacts the atmospheric boundary layer along the US West Coast during a typical February-May period.European air masses exert a greater influence over Japan as compared to Northern China.Due to the influence of southeastward flow over Japan following trans-Eurasia transport in spring (Wild et al., 2004), European emissions also contribute approximately 20-30 ppbv of background CO at central Japan in addition to the episodic impacts of cold surges.
The relationship between boundary layer depths and tropospheric O 3 responses is demonstrated in Fig. 9.A relatively weaker response of O 3 is found in the nocturnal boundary layer for most cases, where O 3 is efficiently removed by surface deposition.During daytime, a mixed layer of vigorous turbulence grows in depth, capped by a statically stable entrainment zone of intermittent turbulence in which small-scale exchanges of chemical species may occur.The three models compared in this study show very different small-scale patterns in the vertical distribution of EU O 3 impacts.While the difference in emissions and chemical regimes discussed previously are most important during cold surges, the representation of entrainment processes in the models may play an important role during other times.Our understanding of intercontinental transport impacts on surface air quality would greatly benefit from further studies of the mixing from the free troposphere into the boundary layer using intensive three-dimensional field measurements and high-resolution models.The California Air Resources Board (CARB), the National Oceanic and Atmospheric Administration (NOAA) and the California Energy Commission (CEC) are proposing a joint field study of atmospheric processes over California and the eastern Pacific coastal region in 2010 (http://www.esrl.noaa.gov/csd/calnex/).Threedimensional sampling of O 3 and aerosols from lidar, aircraft and research vessel from this study will provide a unique opportunity to examine the entrainment of free tropospheric air into the atmospheric boundary layer and associated influences of orographic flow over the western United States.

Conclusions
This study employs two state-of-the-science regional atmospheric chemistry models (WRF-Chem and CMAQ) coupled with a global chemical transport model (MOZART) to examine how chemical transport, oxidant formation and loss interact at finer temporal and spatial scales, and access the sensitivity to model physics.The importance of synoptic to urban scale processes on the rapid export of Asian pollutants and the import of European pollutants over East Asia are evaluated using meteorology for March 2001.While our work was motivated by an interest in the role of regionalscale atmospheric processes and model resolution, it is critical to consider these findings in light of model uncertainties and structure (online vs. offline).Evaluation with intensive aircraft measurements from the TRACE-P field campaign shows that both regional and global models are able to capture the large-scale processes controlling pollution outflow, with the regional models better resolving the fine-scale variability of regional air pollution meteorology and associated chemical processing.There are significant variations in the calculated PAN concentration among models.The ability of the individual model to accurately simulate PAN will affect the long-range transport of reactive nitrogen and hence hemispheric impacts of O 3 pollution.This issue should be further addressed in future studies.
We identified significant differences between regional and global models in simulating vertical mixing of trace gases to the free troposphere and subsequent continental outflow.Our analysis indicates the importance of rapid venting through deep convection that develops along the leading edge of frontal system convergence bands, which are not adequately resolved in either of two global models compared with TRACE-P aircraft observations.Both regional models show that elevated CO, O 3 and PAN extend to the upper troposphere (6-9 km) during a vigorous frontal event.Stronger Asian outflow to the western Pacific and layered structure are simulated in the regional models for multiple events.Given that both WRF-Chem and WRF-CMAQ better match the TRACE-P observations of outflow events, we infer from Figures 2 and S3 that pollution plumes in the global MOZART model are typically diluted and insufficiently lofted to higher altitudes where they can undergo more efficient transport in stronger winds.The difference between regional and global models in simulating tracer vertical transport is much larger than the variation between the online (WRF-Chem) www.atmos-chem-phys.net/10/4221/2010/Atmos.Chem.Phys., 10, 4221-4239, 2010 vs. offline (CMAQ) regional models, suggesting that the limitations in the global models are induced by potential errors in convective parameterizations as well as coarse temporal and spatial resolution.Research needs to be undertaken to further improve the parameterizations of key export processes in large-scale climate-chemistry models, in particular deep convection and the associated cloud and precipitation processes, orographic forcing, and heterogeneous boundary layer processes.
We use sensitivity simulations that perturb chemical boundary conditions in the CMAQ regional model to estimate that the O 3 production over East Asia driven by PAN decomposition contributes 20% of the spatial averaged total O 3 response to European (EU) emission perturbations in March, and occasionally contributes approximately 50% of the total O 3 response in subsiding plumes at mountain observatories (at approximately 2 km altitude).The response to decomposing PAN of EU origin is strongly affected by the O 3 formation chemical regimes, which vary with the model chemical mechanism and NO x /VOC emissions.Our high-resolution models estimate a 40-50% weaker responses of O 3 in the highly populated Asian megacities to European emission perturbations, as compared to global model estimates.This suggests that the studies using the sourcereceptor relationships inferred from coarse-resolution global models likely overestimate health impacts associated with intercontinental O 3 transport.
These results imply an important role for dynamic downscaling using regional climate-chemistry models in evaluating global pollution transport, both to advance our understanding of atmospheric processes, and to inform decisionmaking on air quality management.Future global highresolution or meso-scale model analysis for a hemispheric domain (e.g.trans-Pacific or trans-Atlantic) should provide further insights into how the export and import processes interact, and help to narrow the uncertainty of intercontinental source-receptor relationships.Major issues that should be addressed in future research include maintaining consistency between regional and global model physics, evaluating the role of physical parameterizations vs. resolution, and developing two-way nesting approaches.In light of the strengths and limitations of each modeling framework, there is a need for further analysis of regional processes affecting global transport and further development of models to improve skill in all key processes.Additional analysis in areas with more extensive measurements of atmospheric constituents at the surface and in the free troposphere would also advance understanding of the hemispheric transport influence on surface air quality.

Figure 1a .Figure 1b .
Figure 1a.Comparison of measured (5-min merged) and modeled CO and ethane along the TRACE-P flight tracks in March 2001.Scatter plots are shown for flight sections at altitudes below 2 km (left panel), 2-4 km (middle panel), and above 4 km (right panel).Data points with stratospheric influence (O 3 /CO>1.25)are omitted.Correlation coefficient (R), mean bias (MB, in ppbv) and root mean square error (RMSE, in ppbv) are calculated for each model: MOZART (red), WRF-CMAQ (blue), and WRF-Chem (green).

Figure 3 .
Figure 3.Comparison of MOZART and WRF-Chem calculated zonal fluxes of CO and PAN along 140 • E that is integrated over 4-8.5 km altitudes and 25 • -40 • N latitudes.The shaded bands indicate the timing of cold frontal passages over central China in March 2001.

Fig. 3 .
Fig. 3. Comparison of MOZART and WRF-Chem calculated zonal fluxes of CO and PAN along 140 • E that is integrated over 4-8.5 km altitudes and 25 • -40 • N latitudes.The shaded bands indicate the timing of cold frontal passages over central China in March 2001.

Figure 4 .
Figure4.WRF-Chem simulated CO mixing ratios (filled) and horizontal fluxes (bars) at the 5-km level superimposed with sea level pressure (contours) at 04:00 UTC on 7 March.The red horizontal line approximates the frontal system convergence band, along which a vertical cross-section is shown in Figure5.The thick black line labeled with UTC denotes the flight track of the NASA DC8 for which the chemical distributions are illustrated in Figure6.

Fig. 4 .
Fig.4.WRF-Chem simulated CO mixing ratios (filled) and horizontal fluxes (bars) at the 5-km level superimposed with sea level pressure (contours) at 04:00 UTC on 7 March.The red horizontal line approximates the frontal system convergence band, along which a vertical cross-section is shown in Fig.5.The thick black line labeled with UTC denotes the flight track of the NASA DC8 for which chemical distributions are illustrated in Fig.6.

Fig. 5 .
Fig. 5. Comparison of model simulated convective and updraft mixing of CO in association with the passage of the cyclonic wave during 6-7 March.Shown are vertical distributions of CO (filled) and equivalent potential temperature (contours) along the frontal system convergence band indicated in Fig. 4. The hatched areas indicate that surface pressure is below the bottom pressure level used for vertical interpolation.

Figure 6 .
Figure 6.Observed (1-min merged) and predicted vertical profiles of trace gases along the DC-8 flight track on 7 March during the TRACE-P campaign.The black line denotes the flight path also shown in Figure 4 with corresponding UTC labels.White lines indicate boundary layer heights.

Fig. 6 .
Fig. 6.Observed (1-min merged) and predicted vertical profiles of trace gases along the DC-8 flight track on 7 March during the TRACE-P campaign.The black line denotes the flight path also shown in Fig. 4 with corresponding UTC labels.White lines indicate boundary layer heights.
Figure 8a.Day-to-day variability of calculated total O3 at Mount Hua in central China and its response to a 20% reduction in European emissions.Model results are sampled at the pressure altitude of the mountain sites.The model results are sampled at the pressure altitude of the mountain sites.The dash purple line represents the contribution of O3 production driven by PAN decomposition.The shaded bands indicate the timing of cold frontal passages.

Fig. 8 .
Fig. 8. (a) Day-to-day variability of calculated total O 3 at Mount Hua in central China and Mount Happo in central Japan, and the response to a 20% emission reduction in Europe.WRF-Chem calculated total O 3 and the observations available from the EANET are also shown for Mount Happo.The model results are sampled at the pressure altitude of the mountain sites.The dash purple line represents the contribution of O 3 production driven by PAN decomposition.The shaded bands indicate the timing of cold frontal passages.

Figure 9 .
Figure 9.Time and vertical evolution of O3 responses over China (Mt.Hua) and Japan (Mt.Happo) in association with 20% perturbations in European emissions.The black line represents boundary layer depths.The hatched areas indicate local surface pressure below the bottom pressure level used for vertical interpolation.

Fig. 9 .
Fig. 9. Time and vertical evolution of O 3 response over China and Japan in association with a 20% perturbation in European emissions.The black line represents boundary layer depths.

Table 1 .
Total emissions of ozone precursors over East Asia (15-50 • N, 95-160 • E) in March in the individual models.