Multi-scale model analysis of boundary layer ozone over East Asia

This study employs the regional Community Multiscale Air Quality (CMAQ) model to examine seasonal and diurnal variations of boundary layer ozone (O 3) over East Asia. We evaluate the response of model simulations of boundary layer O3 to the choice of chemical mechanisms, meteorological fields, boundary conditions, and model resolutions. Data obtained from surface stations, aircraft measurements, and satellites are used to advance understanding of O3 chemistry and mechanisms over East Asia and evaluate how well the model represents the observed features. Satellite measurements and model simulations of summertime rainfall are used to assess the impact of the Asian monsoon on O3 production. Our results suggest that summertime O3 over Central Eastern China is highly sensitive to cloud cover and monsoonal rainfall over this region. Thus, accurate simulation of the East Asia summer monsoon is critical to model analysis of atmospheric chemistry over China. Examination of hourly summertime O 3 mixing ratios from sites in Japan confirms the important role of diurnal boundary layer fluctuations in controlling ground-level O 3. By comparing five different model configurations with observations at six sites, the specific mechanisms responsible for model behavior are identified and discussed. In particular, vertical mixing, urban chemistry, and dry deposition depending on boundary layer height strongly affect model ability to capture observed behavior. Central Eastern China appears to be the most sensitive region in our study to the choice of chemical mechanisms. Evaluation with TRACE-P aircraft measurements reveals that neither the CB4 nor the SAPRC99 mechCorrespondence to: M. Lin (mlin26@wisc.edu) anisms consistently capture observed behavior of key photochemical oxidants in springtime. However, our analysis finds that SAPRC99 performs somewhat better in simulating mixing ratios of H2O2 (hydrogen peroxide) and PAN (peroxyacetyl nitrate) at flight altitudes below 1 km. The high level of uncertainty associated with O 3 production in Central Eastern China poses a major problem for regional air quality management. This highly polluted, densely populated region would greatly benefit from comprehensive air quality monitoring and the development of model chemical mechanisms appropriate to this unique atmospheric environment.


Introduction
Ozone (O 3 ) is a secondary pollutant produced in the troposphere by photochemical oxidation of volatile organic compounds (VOCs) and carbon monoxide (CO) in the presence of nitrogen oxides (NO x =NO+NO 2 ).Ozone may also be transported from the stratosphere to the troposphere.Ground-level O 3 is a major component of urban smog that poses a significant risk for public health, and O 3 throughout the troposphere is an important greenhouse gas (Fishman et al., 1979).Ozonesonde observations in Japan (Naja and Akimoto, 2004) and Hong Kong (Liu et al., 2002) show that the seasonal cycle of O 3 in the boundary layer has a broad summer minimum at lower latitudes of the Asian Pacific Rim, in contrast to summer maximums observed at regionally polluted sites in North America and Europe (Jacob, 1999).Recent studies over Eastern China report that surface O 3 exhibits a narrow peak in early summer (May or June) and a sharp drop in July and August, based on measurements Published by Copernicus Publications on behalf of the European Geosciences Union.
taken in downwind of Beijing (Ding et al., 2008;Wang et al., 2008;Lin et al., 2008c), at three mountaintop sites (Li et al., 2007), and at a rural site near Shanghai (Xu et al., 2008).Previous studies have suggested that the East Asia monsoon is responsible for the observed summer minimum of boundary layer O 3 along the west Pacific coast (Liu et al., 2002;He et al., 2008, and references therein).However, a distinctly different seasonal pattern of surface O 3 with a broad summertime maximum during May-August is observed at Mt. Waliguan (3.8 km above sea level), located on the northeastern edge of the Tibetan Plateau.Meteorological simulations (Ding and Wang, 2006) and regional model analysis from tagged emission sources (Ma et al., 2002(Ma et al., , 2005) ) suggest that the episodic elevated surface O 3 mixing ratios at Waliguan are mostly caused by the downward transport of stratospheric air, rather than transport of anthropogenic pollution from Eastern China (Zhu et al., 2004).Current understanding of the spatial and temporal variations of tropospheric O 3 in China is far from complete.More detailed model analysis is needed to interpret the observed features of tropospheric O 3 over this rapidly growing economic region.
Accurate prediction of tropospheric O 3 presents a particular challenge in chemical transport models (CTMs) due to the complex physical and chemical processes occurring from global to local scales.A number of regional CTMs have been employed to study episodic chemical transport and transformations of Asian pollutants in springtime (e.g.Carmichael et al., 2003c;Zhang et al., 2003;Wang et al., 2006), and the seasonal cycle of surface O 3 in Eastern China (Li et al., 2007) and Japan (Yamaji et al., 2006).Results from the MICS-Asia regional model intercomparison study found that O 3 predictions for July over Central Eastern China differ by ∼20 ppbv among seven regional CTMs (Han et al., 2007).This divergence in model estimates suggests that important questions remain on the key mechanisms controlling the O 3 budget over East Asia.Chemical mechanisms have been found to substantially impact model predictions of O 3 over North America, and the effects of NO x and VOCs emissions controls vary depending on chemistry (e.g., Sarwar et al., 2008;Luecken et al., 2007;Arnold and Dennis, 2006).However, the response of local and regional O 3 in Asia to choice and implementation of chemical mechanisms has not been evaluated yet.Several studies have addressed the significant impact of the Asian monsoon on O 3 production over Eastern China (He et al., 2008;Wang et al., 2008), but evaluation of how well meteorological models can reproduce monsoon rainfall and its implication on O 3 prediction is insufficient.
This study employs the regional Community Multiscale Air Quality (CMAQ v. 4.6) model (Byun and Ching, 1999;Byun and Schere, 2006) to examine seasonal and diurnal variations of boundary layer O 3 over East Asia.The CMAQ model is driven by MM5 and by WRF meteorological fields.Chemical boundary conditions for CMAQ are derived from the global Model for Ozone and Related Tracers (MOZART v. 2.4) (Horowitz et al., 2003).This is a contemporary model set-up, and there are not many results presented yet for East Asia.The purpose of this paper is to provide a first step in evaluating the response of model simulations of local and regional O 3 in Asia to the choice of chemical mechanisms, meteorological fields, boundary conditions, and model resolutions.From this analysis, we examine key mechanisms controlling boundary layer O 3 over East Asia, discuss the reasons for discrepancies between model results and observations, and make recommendations regarding optimal regional-scale model configurations for future studies of East Asian O 3 prediction.Section 2 gives an overview of observational data and model configurations.Discussion of O 3 seasonality is presented in Sect.3. We first examine the impacts of global pollution inflow, the Asian monsoon, and photochemistry on the seasonal cycle of surface O 3 at ground-based stations in Siberia, Japan, China, and Southeast Asia.Satellite data and model simulations of precipitation and tropospheric NO 2 columns are used to help interpret seasonal behavior of boundary layer O 3 .Then, with a focus on the spring and summer seasons, Sect. 4 analyzes major photochemical products using two widely used chemical schemes and compares model results with aircraft measurements.Section 5 presents the diurnal variability of summertime ground-level O 3 .We examine the impacts of boundary layer mixing, local chemistry, and dry deposition processes on the formation of ground-level O 3 .

Meteorological fields
We tested the CMAQ model with three sets of meteorological fields for 2001.Table 1 gives horizontal scales, vertical layers, large-scale meteorological fields and physical parameterizations employed for the meteorological simulations.CMAQ was initially driven with meteorological fields from MM5 on an 81×81 km 2 primary domain over East Asia and on a 27×27 km 2 nested domain over Northeast Asia (hereafter referred to as MM5-CMAQ) (Fig. 1a).Vertical layer collapsing was performed to generate eight-layer meteorological fields for CMAQ.Similar configurations of MM5-CMAQ were successfully employed to study longrange transport of acidifying substances over East Asia, and results were presented in Lin et al. (2008a) for model evaluation and in Lin et al. (2008b) for estimating source-receptor relationships.The reader is referred to Lin et al. (2008a) for a detailed description of emissions data and its processing.To examine if vertical layer collapsing would undermine model performance for O 3 predictions, we employ the Weather Research and Forecasting (WRF) model with higher vertical resolution as an alternative meteorological model for CMAQ (hereafter referred to as WRF-CMAQ).The WRF-CMAQ simulation includes 29 vertical layers extending from 8 m to 20 km above ground, with eleven layers in the lowest 2 km.The depth of the first WRF-CMAQ layer is set to 17 m.The vertical resolution of WRF-CMAQ should be fine enough to simulate surface inversion and land-surface interactions.WRF-CMAQ is run for March and summer months (JJA) only, on a 27×27 km 2 domain shown in Fig. 1b.Both MM5 and WRF simulations apply three-dimensional grid nudging towards global analysis.The impacts of meteorological fields on O 3 predictions are discussed in both Sect. 3 and 5.

Gas phase chemistry
In the interest of evaluating model sensitivities to photochemical schemes, we have tested two widely used chemical mechanisms, the Carbon Bond IV (CB4) mechanism (Gery et al., 1989) and the Statewide Air Pollution Research Center (SAPRC99) mechanism (Carter, 2000).The speciation of VOCs to the mechanism-dependent species of CB4 and SAPRC99 is based on a new emission processing model as described in Lin et al. (2008a).CB4 is a lumped-structure condensed mechanism in which organic species are categorized according to reactions of similar carbon bonds (C-C, C=C, C-CHO etc.).The CB4 mechanism in CMAQ contains 36 species and 93 reactions including 11 photolytic reactions.Compared with CB4, SAPRC99 includes more detailed organic chemistry, explicit organic peroxy radicals, and more complete organic intermediates, and provides better representation of peroxides for low NO x conditions.The SAPRC99 mechanism has assignments for 400 types of VOCs, and can be used to estimate reactivities for 550 VOC categories (Carter, 2000).A total of 24 model species are used to represent the reactive organic product species: 11 are explicit, and 13 represent groups of similar oxidation reactivity and emission magnitudes using the lumped molecule approach.The SAPRC99 mechanism in CMAQ includes 72 species and 214 reactions including 30 photolytic reactions.

Boundary conditions
Treatment of lateral and top boundary conditions (BCs) of chemical species is one of the major factors affecting regional model results, especially for long-lived species such as O 3 , PAN, and CO etc.Most regional models do not apply chemical top BCs, nor does the CMAQ model.Thus, we restrict our comments here to the lateral BCs.CMAQ was initially developed for regulatory purposes, where groundlevel pollution is of greatest concern, and numerous evaluations with ground-based measurements have been carried out.For most studies focusing on boundary layer pollution, vertical layer collapsing in the upper troposphere is generally performed to alleviate computational costs (e.g.Arnold and Dennis, 2006;Lin et al., 2008a;Han et al., 2007;Fan et al., 2005).The coarse vertical resolution in the upper troposphere might not resolve the tropopause.tests with and without stratospheric O 3 imported into the WRF-CMAQ domain were carried out, and results are compared with observations from a DC-8 flight designated to examine the influence of stratospheric intrusion.Unless mentioned in the text, other simulations of WRF-CMAQ do not import stratospheric O 3 at the domain boundaries.
In addition to MOZART BC, we have tested several other options, including monthly mean measurement-adjusted MOZART BC and hourly varying BC from the coarse CMAQ simulation.The simulation of MOZART is driven with NCEP reanalysis meteorology for 2000-2002.The evaluation by Holloway et al. (2007) showed that MOZART tends to overpredict monthly averaged surface O 3 over Japan, except in springtime.The overprediction is particularly significant, up to 40%, in summertime.The summertime overprediction is likely due to the coarse horizontal scale implemented by MOZART, which has limited ability in simulating cloud cover and thus enhances O 3 production due to underpredicted cloud cover.In order to reduce the uncertainties introduced in the import of MOZART-derived BCs, we adjusted the concentration of O 3 at the domain boundaries to correct for MOZART bias over Japan as evaluated in Holloway et al. (2007).The reduction of O 3 inflow, which varies on a monthly basis from 0 to 40%, is plotted in Fig. 2a.Results of MM5-CMAQ simulations with original and adjusted MOZART BCs are compared in Section 3.1.Similar adjust-ment of MOZART BC is applied for the WRF-CMAQ simulation.Monthly mean BC provides a seasonal perspective, but does not resolve possible episodic signals.The MM5-CMAQ simulation with 27 km resolution uses hourly varying BC extracted from the 81 km simulation of CMAQ to examine how the temporal variation of BCs can influence prediction of O 3 variability.

Observational data
Ground-based measurement data for O 3 and NO x in Japan obtained under the EANET (Acid Deposition Monitoring Network in Asia) monitoring program (EANET, 2002(EANET, -2008)), were made available for this study.The NASA TRACE-P (Transport and Chemical Evolution over the Pacific) research campaign from February-April 2001 (Jacob et al., 2003), provides a rich dataset for evaluating model results.We use the 5-min merged data sets of the DC-8 and P-3B flights to examine the difference between the CB4 and SAPRC99 chemical mechanisms in simulating key photochemical products.Figure 1 shows locations of groundbased sites and TRACE-P flight paths.Compared with Japan, there are few long-term measurement sites for O 3 in other Asian countries.

Seasonal cycle of surface ozone
We performed three CMAQ simulations at 81 km horizontal scale with different chemical mechanisms and global BCs, and compared seasonal variations of surface O 3 predictions with ground-based measurements.Measurement sites are classified into three groups representing different seasonal patterns at higher latitudes (Fig. 2a), middle latitudes (Fig. 2b) and lower latitudes (Fig. 2c) in the study domain.The correlation coefficient (R), mean bias (MB), and root mean square error (RMSE) are chosen to measure model performance.Evaluation with TRACE-P measurements shows that WRF-CMAQ reproduces well the magnitude and variation of O 3 at flight altitudes below 3 km (Sect.4.2), however, generally underpredicts O 3 at altitudes above 3 km.Carmichael et al. (2003c) and Han et al. (2007) found a similar pattern with the lowest correlation coefficient for O 3 above 3 km, using other regional models not applying top BCs of chemical ) & *+,,-./0 1 2 3 2 0 0 3 4 ) 5 6
As the need to use global model BC increases, along with the need to apply CMAQ for regions frequently impacted by stratospheric intrusions, it is becoming apparent that chemical top BCs should be implemented in future versions of CMAQ and other regional models.
Compared with the remote island sites, local photochemical production and transport of regionally polluted air tend to have a greater impact on the O 3 budget at the inland Chinese sites.Observed surface O 3 at the rural site Lin'an and at three mountaintop sites (Huang, Tai and Hua) exhibit late spring/early summer maximums and drop to low values in July and August.Compared with the observed seasonal trend of surface O 3 , MM5-CMAQ/81 km with CB4 chemistry shows correlations of R=0.7∼0.8 at these sites in Eastern China.The observed peak in late spring and sharp drop in July of surface O 3 at Mt. Huang are well reproduced, but surface O 3 levels in August at all Eastern China sites are overpredicted by MM5-CMAQ/81 km.Especially at Mt. Tai, Mt.Hua and Beijing, model simulations with both gas phase chemistry of CB4 and SAPRC99 do not capture the decrease in observed O 3 in July and August relative to June.Overprediction of July and August O 3 over Central Eastern China from the MM5-CMAQ/81km simulation shown in this study is consistent with results reported for seven regional CTMs (Han et al., 2007) and 21 global CTMs (Fiore et al., 2008) in comparison to EANET data from Japan.Wang et al. (2008) also found that the GEOS-Chem global CTM overestimated July O 3 mixing ratios observed at a rural site near Beijing.

S
Observed and simulated seasonal cycle of daytime (05:00 a.m.-04:00 p.m.) ozone over Beijing.The whiskers are ± one standard deviation of measurement data.
July.Relative to MM5, the WRF simulation at 27 km horizontal scale reproduces well both the magnitude and spatial distribution of TRMM rainfall over South China.The distinct difference in simulating monsoonal rainfall between MM5 and WRF can be attributed to differences in physical parameterizations (Table 1), model resolution, and the initialization process.The MM5 simulation was initialized with the NCEP/NCAR reanalysis at the beginning of the study period and run continuously throughout the year.Considering the cumulative errors of meteorological models, we reinitialize the WRF model every 5.5 days with FNL analysis to maximize the consistency with large-scale observational data ingested in FNL.Similar reinitialization processes have been applied in other modeling studies (e.g.Tang et al., 2007;Ding and Wang, 2006).The underestimate of monsoonal rainfall in MM5 is primarily due to the cumulative errors of a long-term simulation and the inability of the model's coarseresolution to simulate cloud cover.
Clouds have significant radiative effects on the photochemistry of O 3 production near the surface (Lefer et al., 2003;Liu et al., 2006).Transmission of solar radiation is reduced below cloud level, thus photochemical production of O 3 is suppressed in the presence of increased cloudiness associated with the monsoon rainfall.Monsoon rainfall also removes soluble O 3 precursors from the atmosphere.In response to reduced cloudiness and monsoon rainfall, MM5-CMAQ produces 10-15 ppbv higher O 3 over Central Eastern China than WRF-CMAQ.The MM5-CMAQ overestimate of surface O 3 at inland Chinese sites relates primarily to the underestimate of monsoonal rainfall and cloud cover, as well as coarse vertical resolution in which strong surface inversions might not be adequately developed.With the improved meteorological fields from WRF, CMAQ simulates a 10-20 ppbv decrease in O 3 mixing ratios at Mt. Hua and Mt.Tai during June-Augsut, and WRF-CMAQ results with CB4 chemistry are even lower than the observed values (Fig. 2b).For Beijing, the decrease in O 3 mixing ratios in August relative to June is captured by WRF-CMAQ, but the magnitude of low-observed O 3 during July and August is still overpredicted (Fig. 3). Figure 4 shows that WRF predicts less than 2 mm/day rainfall in July over the Beijing surrounding region -which is approximately 50% lower than the TRMM satellite analysis.TRMM estimates 3-5 mm/day rainfall in July over Beijing, consistent with rain gauge records reported in Ding et al. (2008).The unrealistic dry and less cloudy conditions in the model promotes O 3 production over this region.Our analysis confirms that systematic overprediction of summertime O 3 found in both regional and global CTMs is due to model inability to accurately simulate cloud cover and monsoon rainfall and inadequate representation of southwesterly/southeasterly inflow of marine air masses in the global meteorological reanalysis data.
It should be noted that model results are for 2001 while measurement data at three mountaintop sites (Mt.Tai, Mt.Hua, and Mt.Huang) are for March 2004-February 2005 (Li et al., 2007), and the MOZAIC aircraft observations for Beijing are the average of data mostly collected during 1997-1998and around 2005(Ding et al., 2008)).Our analysis of NCEP/NCAR winds and geopotential heights shows that the general transport pattern in 2001 is similar to the climatology over 1980-2005 even though a slightly different pattern is found in August.Figure 5 compares the climatological pattern of NCEP/NCAR winds and geopotential fields at 850 hPa in August for 1980-2005 with those for 2001.The Western North Pacific (WNP) anticyclone plays an important role in the connection between the East Asia summer monsoon and the tropical sea surface temperature anomalies (e.g. Lee et al., 2008), and is characterized by southwesterly airflow along the east Asian coast.We find that in August 2001 the WNP anticyclone moved to the northeast as compared with the climatological pattern.2001 is also found to be a weak La Niña year.The distributions of geopotential height in Fig. 5  analysis suggests that the weakened summer monsoon over Central Eastern China in August 2001 might contribute to O 3 enhancement over this region, but the impact is likely lower than 5-10 ppbv.

Ozone June maximum over Central Eastern China
Boundary layer O 3 over Central Eastern China appears to reach its seasonal maximum in June.Modeled afternoon (12:00 p.m.-06:00 p.m.) O 3 in June shows a typical spatial pattern with low mixing ratios south of 30 • N and a region of O 3 greater than 50 ppbv stretching across northern China (Fig. 6).The high-O 3 enhancement across northern China coincides with the region of relatively low precipitation as revealed in meteorological model simulations and TRMM satellite measurements.Elevated O 3 over the Yellow Sea reflects significant contributions of northeastward offshore transport of aged and polluted air across the water (higher production of OH via O( 1 D)+H 2 O→2OH) and the effects of the marine boundary layer.The marine boundary layer suppresses ventilation of surface pollutants as well as reduces dry deposition velocities since marine boundary layer processes, in general, are less turbulent than their counterparts over land.Beijing and its surrounding region exhibit typical O 3 mixing ratios in excess of 75 ppbv -the US EPA air quality standard of 8-h O 3 .Evaluation with TRMM satellite measurements demonstrates that the WRF model generally captures intra-seasonal differences in monsoon rainfall during June, July and August 2001.With the improved meteorological fields from WRF, we find that CMAQ largely underpredicts the observed June O 3 maxima at Mt. Hua and Mt.Tai (Fig. 2b).To exam-ine if there are additional emission sources in June missing in the model, we compared model estimates of tropospheric NO 2 columns with GOME observations (Fig. 7).The comparison reveals that a relatively strong signal of GOME NO 2 columns over Central Eastern China in June is not seen in the model, even though the magnitude of GOME NO 2 in July and August over most areas is reasonably reproduced.We have also examined NO 2 satellite data from more recent instruments (SCIAMACHY, GOME-2, and OMI) and found similar patterns in other years with greater June NO 2 over Central Eastern China.The seasonal variation of atmospheric NO 2 is mainly determined by the changing sunlight hours over the year, which affect the dominant sink of NO 2 through reaction with the hydroxyl radical.NO 2 tends to have a shorter lifetime in June relative to rainy/cloudy July and August; thus the difference in lifetime can not explain the stronger signal of GOME NO 2 columns in June.We find that the spatial extents of these high NO 2 features coincide with local fire activity in June 2001 as detected by MODIS (Fig. 8).Fu et al. (2007) found a similar feature in the GOME HCHO column and suggested that the observed high HCHO columns in June are produced by crop residue burning after the harvest of winter wheat.Additional emissions from agricultural burning over Central Eastern China, combined with the pre-summer monsoon conditions, would explain the observed relatively high values of O 3 mixing ratios, GOME NO 2 and HCHO columns in June.
The 81-km run of MM5-CMAQ uses biomass burning emissions data for 2000 from the GEIA/RETRO database (http://www.aero.jussieu.fr/projet/ACCENT/RETRO.php).A recent Global Fire Emission Database (GFED) with inter-annual variations (van der Werf et al., 2006) shows that the intensity and spatial distribution of biomass burning emissions are similar between June 2001 and 2004, but weaker in June 2000.We have tested the WRF-CMAQ run using GFED fire emissions for June 2001 and found that GFED fire emissions do not fully account for the strong NO 2 signal from GOME.The magnitude of observed high O 3 at Mt. Tai and Mt.Hua in June is still underpredicted by the CMAQ simulation using GFED fire emissions.There is increasing evidence that VOC emissions over Central Eastern China are underestimated by the current bottom-up emission inventories.Fu et al. (2007) used the GOME HCHO columns to constrain reactive VOC emissions over East Asia and found that GOME-inferred VOC emissions are 25% higher for anthropogenic sources and almost 5 times higher for biomass burning emissions than the Streets et al. (2003) inventory.The GOME-inferred biogenic isoprene emissions over Eastern China are 3 times higher than the estimate by Guenther et al. (2006).A recent study by Zhao et al. (2009) showed that the regional chemical transport model (REAM) greatly underestimates the VOC species leading to PAN formation.The GEOS-Chem global model using GOME-constrained VOC emissions drives a 5-20 ppbv increase in surface O 3 over Central Eastern China relative to using the bottom-up inventories (Fu et al., 2007).Underestimates of VOCs from Eastern China likely explain why CMAQ underpredicts the June maximum of O 3 at Mt. Tai and Mt.Hua as shown in this study.Further sensitivity tests of regional model results with improved emission inventories over China and evaluation with additional ground-based measurements over this region would be extremely valuable in the future.

Spatial and seasonal variations of chemical sensitivity
Examination of modeled O 3 seasonal cycles shows that the difference in surface O 3 is negligible in wintertime between CB4 and SAPRC99 in the MM5-CMAQ/81 km simulations, but is ∼10 ppbv in summertime.This divergence in summertime O 3 predictions between two chemical schemes is particularly pronounced at regionally polluted or downwind sites at mid-latitudes (Figs.2b and 3).In contrast, photochemistry exerts a much smaller influence at coastal and island sites at lower latitudes (Fig. 2c). Figure 9 illustrates the spatial distribution of differences in simulated monthly mean surface O 3 and PAN in July between SAPRC99 and CB4.SAPRC99 and CB4 show little differences in simulating summertime O 3 south of 30-35 • N in East Asia where O 3 production tends to be in the NO x -limited regime, reflecting the high VOC/NO x ratios due to large contributions of emissions from biogenic sources in summertime.Ozone formation between 30 • N and 45 • N across a highly industrialized region in East Asia appears to be most sensitive to the choice of chemical mechanisms.The SAPRC99 mechanism produces 20% higher O 3 and 50-70% more PAN over this polluted region.The high chemical sensitivity indicates that the summertime O 3 budget over this polluted region is dominated by local processes of NO x -VOC-CO photochemistry.Carmichael et al. (2003c), using a regional model analysis of TRACE-P measurements, suggested that photochemical production of O 3 over Central Eastern China is in the VOClimited regime during springtime.Zhao et al. (2009), using simulated HCHO/NO y ratios, inferred that O 3 production over Central Eastern China is less sensitive, relative to other regions of China, to NO x emissions even though it is still in the NO x -limited regime in June.Our analysis suggests that O 3 production over Central Eastern China is also highly sensitive to VOC emissions and their reactivities, inferred from higher O 3 yields in SAPRC99 for this region.Prior studies over North America show similar results, with SAPRC99 producing higher O 3 values than CB4 (Arnold and Dennis, 2006;Faraji et al., 2008;Luecken et al., 2007;Yarwood et al., 2003).All of these prior studies have focused on O 3 yields at polluted regions where the O 3 production tends to be in the VOC-limited regime.We show the same tendencies over Central Eastern China where the summertime O 3 budget shows complex interactions among monsoonal flow, regional transport, and local photochemistry.Understanding of O 3 chemistry and mechanisms over this region is far from complete.Differences in chemical sensitivities give rise to differences in predictions of effects of VOCs and NO x emissions controls.This poses a particular challenge in choosing a mechanism that best represents regional atmospheric conditions in regulatory applications, for example, designing an emissions control policy for Central Eastern China.
Compared with measurements, MM5-CMAQ/81km with both CB4 and SAPRC99 tends to overpredict summertime O 3 at mid-latitude sites.The CB4 mechanism better reproduces the magnitude of observed summertime O 3 at these sites, however, analysis of satellite measurements and MM5 simulations of monsoonal rainfall presented in the last section, has revealed that CB4 in the MM5-CMAQ simulations actually gives the right answer for the wrong reasons.
Several reasons can explain why SAPRC99 produces higher summertime O 3 over Central Eastern China.From the emissions point of view, total VOC emissions in SAPRC99 should be higher than in CB4, which explains in part why regions in the VOC-limited regime of O 3 production are most sensitive to the choice of chemical mechanisms.SAPRC99 includes a more detailed emission representation of hydrocarbon classification for individual VOC species.Formation of O 3 in SAPRC99 is sensitive to aggregated emissions of highly reactive VOC species.Emission speciation of VOCs, however, may not take full advantage of detailed VOC categories by SAPRC99 since most VOC speciation profiles for Asia are based on the source measurements in North America or Europe.Despite the expected uncertainty in aggregated emissions for individual VOC species, VOC species estimates by Streets et al. (2003) have held up remarkably well in subsequent model and observation comparisons (e.g.Carmichael et al., 2003b).Natural sources of reactive VOCs (especially isoprene and terpenes) are also poorly known.SAPRC99 includes both isoprene and terpene photo-oxidation reactions, while CB4 includes isoprene only and the isoprene chemistry in CB4 is much coarser than in SAPRC99 (Zaveri and Peters, 1999).Isoprene emissions, primarily from oaks and other deciduous trees, exhibit strong seasonal and diurnal variability depending on temperature, solar radiation, Leaf Area Index and plant functional type.Biogenic isoprene plays an important role in contributing to summertime O 3 in many areas.A sensitivity test of CMAQ with SAPRC99 was carried out by removing biogenic emissions, and we find that surface O 3 mixing ratios over Central Eastern China and downwind areas decrease by 10-15 ppbv on average.The results suggest that the uncertainty in esti-mating natural VOC fluxes greatly affects summertime O 3 predictions.Online calculation of biogenic emissions depending on meteorological conditions is recommended for future studies.
From a chemistry point of view, the newer mechanism SAPRC99 has a better characterization of radical recycling than CB4.With inclusion of radical recycling processes, compounds that were previously considered termination products can re-form reactive species.For example, the rate of the main radical termination process, the reaction of the hydroxyl radical with NO 2 (OH+NO 2 →HNO 3 ) is greater in CB4 relative to SAPRC99.In summertime, CB4 produces 10-30% higher nitric acid (HNO 3 ) that is quickly dry deposited and thus permanently removes a significant fraction of NO x from the system which otherwise may have been recycled.In addition to the radical recycling process, Faraji et al. (2008) identified aromatic reactions as an important source of differences in O 3 formation between CB4 and SAPRC99.Evaluation of modeled key photochemical species with measurements would provide further insights to photochemical mechanisms in CB4 and SAPRC99.Summertime is found to be the most sensitive season in our study to the choice of chemical mechanisms, however, there is no observational data available to evaluate model results for summertime.

Evaluation with TRACE-P measurements
Measurements of key photochemical species during the TRACE-P campaign for March 2001 provide a unique opportunity to evaluate regional models for Asia.Model results from the WRF-CMAQ simulations were interpolated to the aircraft location and time.Performances of CB4 and SAPRC99 mechanisms in simulating main photochemical oxidants (O 3 , PAN, and H 2 O 2 ) and key radicals (OH and HO 2 ) are compared.Results are shown in Fig. 10 with scatter plots of all DC-8 and P-3B flights in the study region, and in Figs.11-12 for comparison along the DC-8 flight paths on 7 and 27 March.
SAPRC99 and CB4 show little difference in simulating springtime O 3 .Overall both mechanisms reproduce well measured O 3 levels at flight sections below 3 km with correlations of R≥0.75.We discussed the discrepancy above 3 km in Sect.3.1.In many flight segments, modeled OH from both mechanisms closely tracked the observed values.However, CB4 tends to predict higher OH levels at some flight sections, such as the DC-8 flight at 02:00 to 04:00 UTC on 7 March (Fig. 11).SAPRC99 shows stronger variability in the OH level along the flight path, reflecting the faster production and termination processes of radicals.Large differences in PAN and H 2 O 2 mixing ratios are found between CB4 and SAPRC99 and are discussed below.
UiCure 897 B/FT,+#</& /C /D<"+6"7 ,&7 H?bIB!>X T+"7#-;"7 <T"-#"< ,0/&U ;@" 5BIe ß#U@; T,;@ /& !,+-@ 4J9 :"" b#U9)D C/+ ;@" ß#U@; T,;@9 the DC-8 flight path on 7 March illustrates that observed and measured PAN distributions clearly show outflow extending to altitudes of 6∼8 km as the DC-8 flew through a cold front in the period of 04:00∼05:00 UTC.The elevated PAN then serves as a reservoir and carrier of NO x in the middle and upper troposphere, and may be transported over long distances leading to enhanced O 3 production away from primary NO x sources by releasing NO 2 when the air masses warm up (Moxim et al., 1996).Compared with all TRACE-P measurements, CB4 shows a correlation of R=0.71 and a regression slope of S=0.39.SAPRC99 gives a correlation of R=0.45 and a regression slope of S=0.67.CMAQ captured remarkably well the variation of PAN levels along the flight track.However, both SAPRC99 and CB4 tends to overpredict low-observed PAN (<0.5 ppbv) at flight altitudes near 3 km (Fig. 11), likely due to errors in the thermal decomposition rate of PAN.SAPRC99 shows somewhat better ability to capture high-observed PAN levels below 1 km with mean bias (MB) of −0.05 ppbv in contrast to MB=−0.19 ppbv for CB4.This can be traced back to the different treatment of PAN and its analogues in CB4 and SAPRC99.SAPRC99 considers PAN, peroxy propionyl nitrate and other higher alkyl PAN analogues (PAN2), PAN analogues formed from aromatic aldehydes (PBZN), and a PAN analogue formed from methacrolein (MPAN).CB4 considers just the PAN.The rate of PAN decomposition is also higher in SAPRC99 even though the rate of PAN formation is similar between CB4 and SAPRC99.
Hydrogen peroxide (H 2 O 2 ) and the hydroperoxide radical (HO 2 ) are photochemical products that serve as precursors of odd-oxygen as well as reservoirs of odd-hydrogen radicals.Comparison with TRACE-P measurements reveals that CB4 systematically overpredicts H 2 O 2 mixing ratios with a regression slope of S=1.21 (Fig. 10).Examination of observed and modeled results along DC-8 flight 7 (Fig. 11) found that CB4 and SAPRC99 give similar predictions at altitudes above 3 km.Below 3 km, however, the overprediction of H 2 O 2 in CB4 is up to a factor of 3. The difference in H 2 O 2 and HO 2 values are relevant.CMAQ with both mechanisms tracked the observed HO 2 levels remarkably well, and CB4 tends to produce slightly higher HO 2 radicals than SAPRC99.Self-reaction of HO 2 is the only important gas-phase source of H 2 O 2 in the troposphere.CB4 uses a 62% higher conversion rate for HO 2 to H 2 O 2 reaction than SAPRC99 (Luecken et al., 2007).In consequence, SAPRC99 produces 20-50% less H 2 O 2 in summertime over Central Eastern China and downwind areas than CB4.Our analysis for the sringtime suggests that the conversion rate for the HO 2 to H 2 O 2 reaction is too high in CB4.Previous sensitivity studies over North America have suggested that SAPRC99 tends to produce higher radical concentrations leading to enhanced O 3 production (e.g.Luecken et al., 2007), but our results show that SAPRC99 produces lower OH and HO 2 radicals in springtime compared with CB4.
The analysis of photochemical sensitivity in this study has demonstrated that the choice of chemical mechanisms can cause a large difference in simulating key oxidants over East Asia.Neither the CB4 nor SAPRC99 mechanisms consistently capture the observed behavior of PAN and H 2 O 2 in springtime.Understanding the causes of these differences is needed to give us more confidence, and a better understanding of the uncertainties in current models.Use of detailed mechanisms is necessary for: 1) large releases of precursors, 2) reactivity-based alternative simulations, and 3) multi-day effects.But model results need to be interpreted with care because they may not necessarily be the best representation of the actual processes.

Diurnal variation of ground-level ozone
Previous studies examining variations of boundary layer O 3 over East Asia generally used monthly or daily mean data (e.g.Liu et al., 2002;Yamaji et al., 2006).The present section of this study uses hourly data to examine the role of chemistry evolution, vertical mixing and dry deposition processes on diurnal behavior of ground-level O 3 .Model results of hourly O 3 and NO x mixing ratios with five different configurations are compared with EANET measurements taken in Japan (Fig. 13).

Resolution dependence of chemical processes
The difference between CB4 and SAPRC99 is further examined by comparing both 81-km and 27-km simulations of MM5-CMAQ with hourly measurement data from EANET.Table 3 shows the overall statistics of five model scenarios.Comparison of hourly O 3 predictions between CB4 and SAPRC99 shows an intriguing feature: with 81 km resolution, CB4 systematically improves statistical scores over SAPRC99 in reproducing the magnitude of observed O 3 , but with 27 km resolution, SAPRC99 shows a better ability to capture observed 1-h maximums of O 3 mixing ratios.For example, 9-11 July at Sado-seki and some days at Ijira exhibit this phenomena.In general, fine-grid predictions show better agreement against observations than coarse-grid predictions for minimum and maximum O 3 mixing ratios on observed low-O 3 days.For example, during 17-26 July at Sado-seki, the average difference between the 81-km and 27-km simulations of MM5-CMAQ amounts to 40 ppbv.At the Oki site MM5-CMAQ with 27-km resolution (hereafter referred to as MM5-CMAQ/27 km) presents a correlation of R=0.42 and RMSE=15 ppbv in contrast to R=0.2 and RMSE=23 ppbv in the MM5-CMAQ/81,km simulation.Examination of hourly mixing ratios reveals that MM5-CMAQ/81km's overprediction of monthly mean O 3 at Oki (Fig. 2b) is primarily due to large biases on low-O 3 days.Increased spatial resolution improves model results regardless of chemical mechanisms.Our results, going from hemispheric-scale (81 km) to regional-scale (27 km), are consistent with Arnold and Dennis (2006) who show similar improvement from regional (32 km) to urban (8 km and 2 km) scales.Since we are using global model BCs and then downscaling the 81 km results to 27 km, the results indicate the important resolutiondependence of physical and chemical processes.

Boundary layer mixing and local chemistry
Planetary boundary layer (PBL) and land-surface processes have critical implications for air quality simulations.The temporal evolution of O 3 mixing ratios at ground level is controlled strongly by the diurnal variation of the atmospheric boundary layer.This section analyzes the influence of PBL height on O 3 and NO x mixing ratios, and deposition processes.Variations of ground-level O 3 and NO x mixing ratios, PBL height and 10 m wind speed on an hourly basis are analyzed for a rural site (Ijira -Fig.13a), a mountain site (Happo -Fig.13b), and two remote sites (Sado-seki -Fig.13c and Yusuhara -Fig.13d).Observations at urban sites are directly influenced by local emissions which might not be well represented by the model, thus are not discussed here.

Rural sites
Ijira is a rural EANET site located near the city of Nagoya in central Japan.Intensified diurnal variations of NO x and O 3 mixing ratios were observed at Ijira (Fig. 13a and Fig. 14 $""" $@"" %""" %@"" &""" AOLB*+G/ &0E; @0E; ?0E; #0E; $$0E; $&0E; $@0E; $?0E; $#0E; %$0E; %&0E; %@0E; %?0E; %#0E; " @ $" $@ %" %@ P4A6$"*+GQ</ " @ $" $@ %" %@ P4A6$"*+GQ</ " @ $" $@ %" %@ P4A6$"*+GQ</ by the YSU scheme from WRF in part promotes 40 ppbv of daytime O 3 mixing ratios during 26-27 July even though predicted temperature at Ijira from WRF is slightly lower than that from MM5.The decrease in PBL height and wind speed, as well as more concentrated O 3 precursors in the surface layer, explain why WRF-CMAQ/27 km better reproduces observed daytime O 3 maximums during 26-27 July than MM5-CMAQ/27km.The WRF-CMAQ/27 km simulation significantly improves nighttime O 3 predictions at Ijira.EANET observations represent air pollutants measured a few meters above ground.Thus WRF-CMAQ, with a 17 m deep of surface layer, better represents the observed vertical level.The WRF-CMAQ simulated mixing ratios of O 3 at 180 m are also shown in Fig. 14.The difference in O 3 mixing ratios between the 180-m level and 8-m surface level reveals that the vertical gradient in O 3 mixing ratios reaches its maximum at night.After sunrise, ground-level O 3 becomes equal to that at the 180-m level.This profile suggests that the nighttime accumulation of NO x in the stable PBL and deposition processes have significant impacts on ground-level O 3 removal at night.Radiative cooling at night leads to the formation of a stable surface layer near the ground.Above the stable surface layer and under the upper-level inversion, the characteristics of the atmosphere are relatively uniform; this layer is called the nocturnal residual layer.If elevated NO emission sources such as tall industrial smokestacks are not present, O 3 mixing ratios in the residual layer remain high, since deposition and other removal processes that occur near the surface are absent aloft.This can be observed at three remote sites (Fig. 13b-13d).At the rural site Ijira, however, relatively high mixing ratios of NO x were observed.Thus ground-level O 3 mixing ratios at Ijira decrease after sunset to very low values at nighttime due to NO titration (NO+O 3 →NO 2 +O 2 ) and deposition processes, and increase dramatically in the morning, reaching a maximum value in the afternoon (01:00-02:00 p.m.).The correlation coefficient between O 3 and NO x is −0.84 for model results and −0.27 for observations.The negative correlation between O 3 and NO x suggests that O 3 formation at rural Ijira is in the VOC-limited regime.We find that when nighttime O 3 is depleted by NO x titration, simulated NO x levels at night are much higher in WRF-CMAQ than the observed values at Ijira.The major sink of NO x at night is reaction with O 3 followed by hydrolysis of N 2 O 5 on $""" $@"" %""" AOLB*+G/ &0E; @0E; ?0E; #0E; $$0E; $&0E; $@0E; $?0E; $#0E; %$0E; %&0E; %@0E; %?0E; %#0E; " @ $" $@ %" %@ P4A6$"*+GQ</ " @ $" $@ %" %@ P4A6$"*+GQ</ " @ $" $@ %" %@ P4A6$"*+GQ</ aerosols.Heterogeneous pathways of N 2 O 5 hydrolysis is a weak part of current atmospheric chemistry models.Inadequate treatment of N 2 O 5 hydrolysis would in part contribute to the overestimate of nighttime NO x levels.Another possible cause is the fact that EANET observed aged air masses at Ijira, with most of the emitted NO x converted to other forms by the time that the air masses were sampled.
The large rate of change in nighttime O 3 among five model simulations clearly illustrates the contribution of NO x emissions intensity and deposition processes to O 3 destruction at ground level.The artificial dilution of NO x emissions in the coarse horizontal and/or vertical resolution of MM5-CMAQ results in reduced depletion of O 3 through NO x titration at night.The intensity of NO x titration is also influenced by PBL heights and mixing.For example, the difference in nighttime O 3 levels between MM5-CMAQ/81km and MM5-CMAQ/27 km simulations amounts to 40 ppbv during 2-3 July and 21-23 July.During these periods, the MM5/27km PBL height is ∼500 m lower at nighttime, which causes more NO x to be trapped near the surface and provokes a higher depletion of O 3 in CMAQ.Impacts of PBL heights on O 3 depletion at nighttime can also be examined in the sharp vari-ations of NO x mixing ratios on a few days as highlighted in Fig. 14.The correlation between PBL heights and NO x in the WRF-CMAQ simulation is −0.85.When the PBL height is substantially lower than 250 m, NO x increases can reach 20 ppbv, which causes the rapid depletion of O 3 at ground level.

Remote sites
In contrast to the rural site of Ijira, an obvious diurnal pattern of surface O 3 is not observed at Mt. Happo, Sado-seki or Yusuhara located in relatively clean, remote areas.In remote areas outside the impact of urban pollution plumes, horizontal advection and downward vertical mixing processes are the dominant mechanisms for the build-up of ground-level O 3 .All five simulations capture well the day-to-day variability in the observed data at all three remote sites.Day-to-day variability of daytime O 3 at Happo is shown to have a correlation of R=0.5 with changes of surface wind speeds.High O 3 episodes during 7-10 and 19-22 July are correlated with low winds under stagnant conditions.Although a similar diurnal pattern of boundary layer heights is found at Ijira and at Mt. Happo, O 3 mixing ratios at Mt. Happo show little diurnal variation.On 7 and 10 July, the highest observed O 3 mixing ratios are found at midnight, and this behavior is not reproduced by the models.Such behav-ior is characteristic of mountain stations located above the nocturnal inversion that isolates the site from the effects of surface deposition.The 27-km horizontal scale used in this study is still too coarse to resolve mountain topography and the associated local topographical circulation.The model, due to unrealistic removal of O 3 through dry deposition, generally underestimates nighttime O 3 at Mt. Happo.It would be valuable in the future to examine if a high spatial (1 km) resolution model can reproduce O 3 diurnal behavior at highaltitude sites.
The Sado-seki site is located on Sado island on the west coast of Japan.All model simulations successfully reproduce an elevated-O 3 episode on 10 July associated with stagnant conditions (Fig. 13c).The WRF model predicts large diurnal variations of PBL heights, temperature and latent heat flux reflecting land-surface interactions.Both MM5 simulations predict almost no diurnal variations of these variables.The likely reason for this remarkable difference is that both MM5 81km and 27km grids for Sado-seki are not identified as land in the model.The WRF-CMAQ prediction better reproduces the observed low-O 3 days during 14-22 July at Sado-seki, which greatly improves statistical scores shown in Table .3. Evolution of the boundary layer at Yusuhara is quite different from that at the other three sites (Fig. 13d).The PBL height at Yusuhara remains lower than 500 m, except during 7-10 and 25-27 July when O 3 mixing ratios also exhibit higher values than on other days.The positive correlation between ground-level O 3 and PBL height indicates the influence of vertical transport of pollutants aloft down into the surface layer.During 25-27 July, photochemical production, as indicated by high NO x mixing ratios, partly explains the high levels of surface O 3 in this period.Four MM5-CMAQ simulations systematically overpredict the observed O 3 during 27-28 July, while the WRF-CMAQ prediction reproduces the observed values well.We find that surface wind speed during 27-28 July from WRF is approximately 3 m/s higher than that from MM5, which helps the ventilation of ground-level O 3 at Yusuhara from the boundary layer.The MM5-CMAQ overprediction of O 3 during these two days is likely caused by errors in surface winds from MM5.
Examination of hourly summertime O 3 mixing ratios from sites in Japan confirms the important role of diurnal boundary layer fluctuations in controlling ground-level O 3 .By comparing five different model configurations with observations at six sites, the specific mechanisms responsible for model behavior are identified and discussed.In particular, vertical mixing, NO x titration, and dry deposition depending on PBL height strongly affect model ability to capture observed behavior.Lower PBL height enhances the build-up of groundlevel O 3 during daytime, but causes a more complete depletion of O 3 by NO x titration at night.A large vertical gradient of boundary layer O 3 is found over rural and urban areas.Our analysis demonstrates clear benefits of using a highresolution model in evaluating transport and chemical transformation of Asian pollutants.The WRF-CMAQ simulation using 27 km horizontal scale and 29 vertical layers gives the best statistical scores in evaluating the day-to-day and diurnal variations of ground-level O 3 .Future studies should ensure that model vertical resolution is adequate to resolve near surface chemical processes and land-surface interactions.Surface layer depth less than 20 m is recommended to better represent measured conditions of ground-based stations.
Central Eastern China appears to be the most sensitive region in our study to the choice of chemical mechanisms.Evaluation with TRACE-P aircraft measurements reveals that neither the CB4 nor the SAPRC99 mechanisms consistently capture observed behavior of key photochemical oxidants in springtime.However, our analysis finds that SAPRC99 performs somewhat better in simulating mixing ratios of H 2 O 2 and PAN at flight altitudes below 1 km.The CB4 mechanism overpredicts H 2 O 2 by a factor of two, which is caused by higher self-reaction rates of HO 2 .SAPRC99 predicts 50% higher PAN, because more PAN analogues and a lower decomposition rate of PAN are considered in SAPRC99 than in CB4.The overall higher load of VOC, explicit organic peroxy radicals, and inclusion of radical recycling processes explain why SAPRC99 predicts 10-20 ppbv higher O 3 during summertime for Central Eastern China and downwind areas.Our findings suggest that future model studies over this region should carefully examine chemistry sensitivities, especially in regulatory applications -such as the design of emission control policies.Between CB4 and SAPRC99, the more detailed SAPRC99 mechanism is suggested for future studies examining high-O 3 episodes, the impacts of large releases of precursors, and multi-day effects.Improvements of regional atmospheric chemistry models will require efforts in reducing uncertainty in emissions, meteorology, photochemistry, and global pollution inflow.The high level of uncertainty associated with O 3 production in Central Eastern China poses a major problem for regional air quality management.This highly polluted, densely populated region would greatly benefit from comprehensive air quality monitoring and the development of model chemical mechanisms appropriate to this unique atmospheric environment.

Fig. 1 .
Fig. 1.Model domains: (a) MM5-CMAQ domains (D1 and D2) overlayed with ground-based measurement sites.(b) WRF-CMAQ domain (D3) overlayed with intensity of NO x emissions (moles/km 2 /h) and flight paths of NASA DC-8 on 7 March (red), 27 March (blue), and other flights (black) during the TRACE-P campaign Three ground-based measurement sites (Mondy, Rishiri, and Tappi) close to the northern boundary of the model domain are chosen to diagnose the impacts of global pollution inflow.The comparison between two SAPRC99 simulations with original and controlled MOZART BC demonstrates that adjusting O 3 inflow from the global model reduces the mean bias by ∼5 ppbv on seasonal predictions at these three sites.CMAQ predictions at Mondy with adjusted O 3 inflow show root mean square error (RMSE) of 1.7 ppbv as compared to RMSE=7.4 ppbv with the original MOZART BC.Adjusting the inflow of O 3 pollution exerts a greater influence on wintertime O 3 predictions, reflecting longer lifetime of O 3 in the wintertime atmosphere.Using measurement-adjusted O 3 inflow tends to improve CMAQ's performance in simulating wintertime surface O 3 in Northern (Rishiri and Tappi) and Central Japan (Oki), as compared with measurements.

Fig. 2a .
Fig. 2a.Seasonal variations of observed and predicted surface ozone from the MM5-CMAQ simulations with 81 km resolution at (a) high-latitude sites, (b) mid-latitude sites, and (c) low-latitude sites.The dotted line shows the percentage reduction of ozone mixing ratios imported from the MOZART global model at the domain boundaries.The reduction percentage corresponds to the bias between MOZART predictions and measurements over Japan.The whiskers are ± one standard deviation of measurement data for available sites.

Fig. 2b .
Fig. 2b.Continued.Blue crosses represent model results from the WRF-CMAQ simulation with the CB4 chemistry, which is run for March and June-August 2001 only.

Fig. 4 .
Fig. 4. Comparison of observed and modeled precipitation in June and July 2001.

DFig. 7 .
Fig. 7. Tropospheric NO 2 column in 10 14 molecules cm −2 from GOME (left panel) and the WRF-CMAQ simulation (right panel) for June and August of 2001.The CMAQ simulation uses anthropogenic emissions from Streets et al. (2003) and fire emissions from van der Werf et al. (2006).The crosses indicate locations of measurement sites shown in Fig. 1.

Fig. 9 .
Fig. 9. Differences in monthly mean surface O 3 and PAN in July 2001 between SAPRC99 and CB4 predictions.

Fig. 12 .
Fig. 12.Comparison of observed and WRF-CMAQ predicted species along the DC-8 flight path on 27 March.See Fig. 1b for the flight path.

Fig. 13a .
Fig. 13a.Hourly time series of observed and predicted mixing ratios of NO x and O 3 in July 2001 (a) at the rural site Ijira, (b) the remote mountain site Happo, (c) the remote sites Sado-seki, and (d) Yusuhara.The upper panel shows predicted PBL height (left axis) and wind speed at 10 m (right axis).
tions of boundary layer O 3 and key photochemical oxidants to the choice of chemical mechanisms, meteorological fields, boundary conditions, and model resolutions.Although measurement data are relatively scarce in Asia, we have used available ground-based measurements, aircraft observations obtained from the TRACE-P campaign and satellite data to advance understanding of O 3 chemistry and mechanisms over East Asia and evaluate how well the model represents the observed features.

Fig. 14 .
Fig. 14.Diurnal variations of O 3 , NO x , PBL height, and dry deposition at the rural site Ijira during 20-27 July 2001.Episodes with large impacts of PBL height on O 3 mixing ratios are highlighted in gray.

Table 1 .
Model configurations Data for initial twelve hours are not used by CMAQ to allow the spin up of clouds and other climate processes.b WRF-CMAQ uses dynamic hourly BC for March, and monthly mean BC for JJA.

Table 2 .
Ground-based measurement data used in this study.
a Most MOZAIC data were collected during 1997-1998 and around 2005.No data are available for 2001.b Data in 2001 available for March-May only.

Table 3 .
Statistics for hourly mixing ratios of ground-level O 3 in July 2001 a Unit for MB and RMSE is ppbv.