Interactive comment on “ Constraints on Asian ozone using Aura TES , OMI and Terra MOPITT ”

The study first uses MOPITT CO and OMI NO2 to constrain the global CO and Chinese NOx emissions in the GEOS-Chem model for the year 2006. For NOx, emissions are separately scaled for anthropogenic, lightning, soil, and biomass burning sources under the assumption of a linear relationship between the individual emission sources and their impact on the NO2 column. Independent lower tropospheric retrievals from TES are then used to evaluate modeled CO and ozone over Eastern China and its outflow region separately, for the model run with the 2006 optimized emissions, as well as 2006-2010 modeled with a priori emissions. Overall the topic is interesting as is the analysis, but the paper should more clearly connect the analysis with the conclusions being drawn by more clearly framing the questions being addressed here. C7564


Introduction
Unprecedented growth in transportation, coal-fired power plants and the industrial sector in China has resulted in a substantial increase in the emissions of O 3 precursors (Lin et al., 2014a).Recent studies (Lamsal et al., 2011;Lin, 2012;Mijling et al., 2013) show a 5-10 % annual growth rate of NO x emission in China.Wang et al. (2012) found there was a 3 % annual growth rate of O 3 in Beijing in the period of Z. Jiang et al.: Constraints on Asian ozone using Aura TES, OMI and Terra MOPITT 2003-2010. East . East Asian O 3 can be transported to the surface of North America in about 2-3 weeks (Liu and Mauzerall, 2005) by midlatitude westerly winds (Liang et al., 2004(Liang et al., , 2005)), which likely resulted in an increase of background O 3 concentration in western North America by 3-7 ppbv (parts per billion by volume) during the period of 2000-2006(Zhang et al., 2008;;Brown et al., 2011).
Use of inverse (top-down) methods to better quantify the emission of NO x (Lamsal et al., 2011;Lin and McElroy, 2011;Lin, 2012;Mijling et al., 2013), VOCs (volatile organic compounds; Shim et al., 2005;Fu et al., 2007) and CO (Kopacz et al., 2010;Fortems-Cheiney et al., 2011;Gonzi et al., 2011) are needed to ensure consistency between bottomup inventories and observations.However, large discrepancies can still exist between bottom-up-and top-down-based inventories (e.g., Kopacz et al., 2010;Lin et al., 2012).In this work, we perform a multi-tracer assimilation with the GEOS-Chem (Goddard Earth Observing System Chemistry) model to evaluate the top-down estimates of O 3 precursors (NO x and CO) in eastern Asia.We firstly optimized the CO and NO x emission with MOPITT (Measurements Of Pollution In The Troposphere) CO and OMI (Ozone Monitoring Instrument) NO 2 retrievals, respectively, and then evaluate the a posteriori simulation of CO and O 3 by comparing the values with measurements from TES (Tropospheric Emission Spectrometer) in the period of December 2005-November 2006.Using the adjoint of the GEOS-Chem model (Henze et al., 2007), we then quantify source contributions (NO x , CO, VOC) to free tropospheric O 3 pollution over eastern China and the China outflow region in December 2005-November 2006.

TES CO and O 3
The TES instrument was launched on NASA's Aura spacecraft on 15 July 2004.The satellite is in a sun-synchronous polar orbit of 705 km and crosses the Equator at 01:45 and 13:45 LT (local time).With a footprint of 8 km × 5 km, TES measures radiances between 3.3 and 15.4 µm with a global coverage in 16 days (Beer et al., 2001) of observations.In the troposphere, TES O 3 profile retrievals have 1-2 degrees of freedom for signal (DOFS), and about 1 DOFS for CO.We use data from the "lite" product (http://tes.jpl.nasa.gov/data/)which reports volume mixing ratios (VMRs) on 26 pressure levels for O 3 and 14 pressure levels for CO.Using an optimal estimation approach, the TES retrievals are conducted with respect to the logarithm of the VMR.The relationship between the retrieved profiles and the true atmospheric state can be expressed as where z is the true atmospheric state (expressed as log(VMR)), z TES a is the TES a priori O 3 or CO profile, A TES is the TES averaging kernel matrix and Gε describes the retrieval error.The averaging kernel matrix represents the sensitivity of the retrieval to the actual trace gas in the atmosphere.The TES retrievals use a monthly mean profile of the trace gas from the MOZART-4 CTM (chemical transport model), averaged over a 10 • latitude × 60 • longitude, as the a priori information z TES a .According to the recommended quality control criterion, we only use CO and O 3 data with major quality flags of 1.These data have passed all major quality flags used to assess the TES data related to chi-2 tests, biases in the radiance residuals, and residual nonlinearity checks.The data with small DOFS (DOFS for CO smaller than 0.8) are dropped as the limited sensitivity reduces the robustness of calculated O 3 -CO correlations.We empirically find that the sensitivity of CO is the limiting factor in these comparisons; that is, if DOFS for CO is > 0.8 then the DOFS of O 3 is > 0.8.Recently, Verstraeten et al. (2013) evaluated TES O 3 measurements by using data from the World Ozone and Ultraviolet Radiation Data Centre (WOUDC) sites and found that there is a ∼ 7 ppb bias in the TES measurements in free troposphere, and that the magnitude is slightly larger in summer and smaller in winter.TES CO measurements were evaluated by Luo et al. (2007) using the aircraft measurements from INTEX-B (Intercontinental Chemical Transport Experiment B) campaign.They showed that TES CO VMR profiles are 0-10 % lower than the aircraft measurements in the lower and middle troposphere.

MOPITT CO
The MOPITT instrument was launched on NASA's Terra spacecraft on 18 December 1999.The satellite is in a sunsynchronous polar orbit of 705 km and crosses the Equator at 10:30 LT.With a footprint of 22 km × 22 km, MOPITT (version 6) combines TIR (thermal infrared; 4.7 µm) with the NIR (near infrared; 2.3 µm) and has better sensitivity to lower tropospheric CO over land (Worden et al., 2010).MOPITT CO retrievals are reported on 10 pressure levels (surface,900,800,700,600,500,400,300,200 and 100 hPa).Similar to the TES product, relationship between the retrieved CO profiles and the true atmospheric state can be expressed as where z is the true atmospheric state (expressed as log(VMR)), z MOP a is the MOPITT a priori CO profile, A MOP is the MOPITT averaging kernel matrix and Gε describes the retrieval error.
Same as TES, the a priori information of MOPITT retrievals is from the monthly mean profile of the MOZART-4 CTM, without the 10 • latitude × 60 • longitude average.We reject MOPITT data with CO column amounts of less than 5 × 10 17 molec cm −2 and if low clouds are observed.The nighttime data are excluded in the assimilation, due to the NIR radiances measured reflected solar radiation.The version 5 data have been evaluated recently against NOAA aircraft measurements (Deeter et al., 2013), which shows a small bias in the low and middle troposphere, but 14 % positive bias at the 200 hPa retrieval level.The new version 6 data significantly reduce the bias in the upper troposphere but magnify the positive bias at the surface level.In this work, we decide to use the new version 6 data, as we focus on the free troposphere (above 800 hPa), which is not affected by the positive bias in the retrieval at the surface level.

OMI NO 2
The OMI instrument was also launched on NASA's Aura spacecraft.The sensor has a spatial resolution of 13 km × 24 km (Levelt et al., 2006).OMI provides daily global coverage with measurements of both direct and atmosphere-backscattered sunlight in the ultraviolet-visible range from 270 to 500 nm; 405-465 nm is used to retrieve tropospheric NO 2 columns.In this study, the daily level-2 data from the KNMI (Royal Netherlands Meteorological Institute) DOMINO-2 product (Derivation of OMI tropospheric NO2; Boersma et al., 2011) are averaged to obtain monthly mean vertical column densities (VCDs) for subsequent emission inversion.The total error for the retrieved VCDs is about 30 % plus 0.7 × 10 15 molec cm −2 , and the magnitude is larger in winter than in summer (Boersma et al., 2011;Lin and McElroy, 2011).The pixels with cloud radiance fraction exceeding 50 % are removed.In order to have a better analysis of the spatial distribution of VCDs within short distance, we only use data from the 30 pixels around the swath center.The details for the data treatment are described in Lin (2012).

GEOS-Chem
The GEOS-Chem CTM (http://www.geos-chem.org) is driven by assimilated meteorological observation from the NASA Goddard Earth Observing System (GEOS-5) at the Global Modeling and data Assimilation Office.We use version v34 of the GEOS-Chem adjoint, which is based on v8-02-01 of GEOS-Chem, with relevant updates through v9-01-01.The standard GEOS-Chem chemistry mechanism includes 43 tracers, which can simulate detailed tropospheric O 3 -NO x -hydrocarbon chemistry, including the radiative and heterogeneous effects of aerosols.The GEOS-5 meteorological fields have 72 vertical levels and the lowest 31 levels are terrain following levels.In order to minimize the amount of memory required to run GEOS-Chem, the model is run with a reduced vertical resolution, in which the levels in the stratosphere are lumped together online.
The native horizontal resolution of GEOS-5 is 0.5 • in global-scale simulations.A nested simulation can be achieved by running a 0.5 • × 0.667 • resolution model within a regional domain using the boundary condition provided from a global, coarse-resolution mode (Wang et al., 2004;Chen et al., 2009).Recently, the adjoint of nested GEOS-Chem was developed by Jiang et al. (2014a).In this work, following Jiang et al. (2014a) and Mao et al. (2014), we run the model with 0.5 • × 0.667 • resolution over Asia.The boundary condition is generated with a global-scale 4 • × 5 • -resolution simulation.
The anthropogenic emission inventories are identical to those used in Jiang et al. (2013).The global anthropogenic emission inventory is EDGAR (Emission Database for Global Atmospheric Research ) version 3.2FT2000 (Olivier et al., 2001), updated by the following regional emission inventories: the INTEX-B Asia emissions inventory for 2006 (Q.Zhang et al., 2009), the Cooperative program for monitoring and evaluation of the long-range transmission of air pollutants in Europe (EMEP) inventory for Europe in 2000 (Vestreng et al., 2002), the US Environmental Protection Agency National Emission Inventory (NEI) for 2005 in North America, the Criteria Air Contaminants (CAC) inventory for Canada, and the Big Bend Regional Aerosol and Visibility Observational (BRAVO) Study Emissions Inventory for Mexico (Kuhns et al., 2003).Biomass burning emissions are from the interannual GFED3 (Global Fire Emissions Database) inventory with 3 h resolution (van der Werf et al., 2010).The biogenic emissions are from MEGAN 2.0 (Model of Emissions of Gases and Aerosols from Nature version 2; Millet et al., 2008).Figure 1 shows the anthropogenic emission of NO x and CO in Asia in June 2006.There are strong pollutant emissions in the North China Plain.The urban emission centers can also be clearly identified.The annual anthropogenic NO x emission over eastern China is 16.5 Tg (2006) and 20.7 Tg (2010), with a 5 % annual growth rate.

4-DVAR inversion for global CO emission
In this work, we evaluate the observed interannual variability of O 3 and CO and the GEOS-Chem model simulation for the period of 2006-2010, where the data density of TES measurements is higher relative to subsequent years.As the first year of this 5-year period, the relative contributions of O 3 precursors to free troposphere O 3 in 2006 will be studied in detail.The 2006 global CO emissions are optimized with a 4-DVAR (four-dimensional variational) method.The inverse method minimizes the cost function J (x) to provide an optimal estimate of the CO sources, where x is the state vector of emissions, x a is the a priori estimate, y is a vector of observed concentrations, and F (x) Z. Jiang et al.: Constraints on Asian ozone using Aura TES, OMI and Terra MOPITT is the forward model, which represents the transport of the CO emissions in the GEOS-Chem model and accounts for the vertical smoothing of the MOPITT retrieval.S and S a are the observational and a priori error covariance matrices, respectively.The first term of the cost function represents the mismatch between the simulated and observed concentrations.The second term represents the departure of the estimate from the a priori.The cost function in Eq. ( 3) is minimized by reducing the gradient, ∂J ∂x , using the adjoint of GEOS-Chem model in a 4-DVAR approach (Henze et al., 2007), which has been previously used for assimilation of CO and O 3 (Kopacz et al., 2010;Singh et al., 2011;Parrington et al., 2012;Jiang et al., 2014b).Similar to Jiang et al. (2013Jiang et al. ( , 2014b)), we produce improved initial conditions by assimilating MOPITT version 6 data, using the sequential sub-optimal Kalman filter (Parrington et al., 2008), from 1 January 2006 to 1 January 2007.The optimized initial conditions are archived at the beginning of each month.Consequently, the initial conditions for the model simulation are independent from the inverse analyses.

Regression-based inversion for China NO x emissions
The 2006 Chinese NO x emissions are optimized with a regression-based multistep method exploiting the distinctive seasonality of different sources (Lin, 2012).Neglecting horizontal transport and assuming a linear relationship between the total VCD of NO 2 and VCDs from individual sources, the predicted VCD ( p ) for a given grid can be expressed as the sum of individual emission sources, multiplied by certain scaling factors: The subscripts "a", "l", "s", and "b" indicate anthropogenic, lightning, soil and biomass burning sources of NO x , respectively.The updated emission estimates can be obtained by reducing the sum of [( r − p )/σ ] 2 across the 12 months; here r is the retrieved VCD and σ is the standard deviation.
To better represent the resolution-dependent NO x chemistry (Valin et al., 2011), the inversion was conducted with the highest resolution of GEOS-Chem.The seasonality-based inversion method also reduced the influence of potential biases in OMI NO 2 data (Lin et al., 2014b), particularly in winter.
The details for the inversion process were described in Lin (2012).

Evaluation of the model simulation and top-down estimates of O 3 precursors
In this work, we are interested in the domain of eastern China, as shown in Fig. 1, because it is the largest pollu-tant emission contributor in eastern Asia and the adjacent domain where outflow of Asian pollution is significant.Figure 2 shows the monthly regional mean O 3 and CO concentration at the free troposphere (681-383 hPa) for June, July and August for the period 2006-2010, using the GEOS-Chem model driven with a priori emission inventories.The modeled O 3 concentrations are generally within 10 % of the TES data after accounting for the approximately 7 ppb bias in the TES O 3 measurements (e.g., Worden et al., 2007;Verstraeten et al., 2013).However, the modeled CO is biased low, which is consistent with previous studies (Shindell et al., 2006;Kopacz et al., 2010;Naik et al., 2013).This low bias could be associated with a positive bias in OH, as indicated by Jiang et al. (2014b).The bias in CO can be reduced by integrating the coarse-resolution global and fine-resolution nested simulations in a two-way coupled manner, such that results from the nested model can be used to improve the global model (within the nested domain) and ultimately affect its lateral boundary conditions (via the global transport of CO and other species) (Yan et al., 2014).Another possible reason for the CO bias is that the TES CO data are biased towards polluted air parcels because of its relatively low sensitivity whereas the model captures background values as discussed in Pechony et al. (2013).Although the model is biased low, the interannual variabilities and trends of O 3 and CO are well correlated between the model and TES, indicating that changes in the modeled emissions, chemical production of ozone, and meteorology are well described (e.g., Zhang et al., 2006;Kim et al., 2013).O 3 -CO correlations can be used to constrain O 3 sources and transport (e.g., Zhang et al., 2006).Positive correlations usually indicate that a region has experienced photochemical O 3 production, whereas negative correlations may result from O 3 chemical loss or influence of stratospheric air.For example, Zhang et al. (2006)  Table 1 shows the monthly regional mean O 3 and CO correlation and slope values for the free troposphere (825-383 hPa) for June, July and August 2006-2010; the model is driven by a priori emissions.The uncertainty in the O 3 and CO concentrations are due to random errors in the TES O 3 and CO observations and natural variability (Zhang et al., 2006).For this reason, we also show the mean value over the analysis time period.The correlation and slope values of TES and GEOS-Chem are generally consistent for both domains.The positive correlation coefficients imply influence of photochemical O 3 production but also transport of nearby CO  emissions into pollution plumes (e.g., Worden et al., 2013) As in previous studies (Zhang et al., 2006;Voulgarakis et al., 2011;Kim et al., 2013), there are small differences between the simulation and observation.A possible reason for these discrepancies, particularly over the ocean, is the model transport error because transport of "clean" air from the Pacific can have substantially different chemical characteristics to Asian air.The consistency between the model and TES in the interannual variations, correlation coefficients and slopes implies that the spatiotemporal distribution of the CO emissions and oxidative chemical processes are consistent.As described in Sect.3, the 2006 global CO emissions are constrained with MOPITT data; the 2006 Chinese NO x emission are constrained with OMI data.As shown in Fig. 3, Chinese posterior anthropogenic NO x emissions in June 2006 increased by 14 % over the a priori emissions, from 1.86 to 2.11 Tg.Smaller adjustments are obtained for winter.In June 2006, the Chinese anthropogenic CO emissions increased from 17.09 to 18.93 Tg, with a mean scaling factor of 1.11.In December 2005, the Chinese posterior anthropogenic CO emission increased over the prior from 14.95 to 19.78 Tg.However, as indicated by Jiang et al. (2014b), a potential bias in OH fields could have a significant influence on the inferred CO emission estimates.By using the OH fields from a different GEOS-Chem version, they found that a posteriori CO emissions over eastern Asia in June-August 2004 are reduced by 28 % for the different OH fields.This large potential error in the CO emissions do not strongly affect our conclusions because the modeled CO concentrations, based on the model OH distributions and CO emissions, are consistent with the MOPITT data.The monthly regional mean O 3 and CO concentrations in the period of December 2005-November 2006 are shown in Fig. 4. In order to remove the influence of the initial conditions, the updated simulation is obtained by running the model from 1 September 2005, with updated inventories of NO x and CO.Both the model and data show increases of O 3 concentration from winter to spring, due to enhancement of photochemical production, and a substantial decrease in June-August, due to the effect of the East Asian monsoon (Yang et al., 2014).The CO concentration peaks in March, which is consistent with Shindell et al. (2006).The boreal spring CO maximum is associated with the accumulation of CO emissions in winter, while CO lifetime is longer (Duncan et al., 2007).The updated inventories significantly reduced the bias on the CO simulation.However, these changes in the NO x and CO emissions do not significantly change free tropospheric ozone.

Dependency of O 3 on anthropogenic and natural NO x , CO and VOCs
In this section, we will use the adjoint of the GEOS-Chem model (Henze et al., 2007) to quantify source contributions (NO x , CO, VOC) to free tropospheric O 3 pollution over eastern China and the China outflow region.The updated NO x and CO emission inventories are intended to improve the simulation.We are interested in these two domains as they have significant influence on the long-range pollution transport (e.g., L. Zhang et al., 2009).Similar to previous studies (L.Zhang et al., 2009;Bowman and Henze, 2012;Lapina et al., 2014), the analysis is based on a sensitivity calculation from an adjoint model.In this work, both transport and chemistry components are run backwards and thus provide a more computationally efficient method for a receptororiented problem than the traditional approach by perturbing emissions.
Figure 5 shows the contributions of anthropogenic NO x , lightning NO x , anthropogenic CO and biogenic isoprene on free tropospheric (819-396 hPa) O 3 over eastern China.The value can be explained as the percentage change of regional mean O 3 due to a fractional change in emissions in a particular grid.For example, assuming an unchanged chemical environment, one particular grid with contribution 0.02 % implies mean free tropospheric O 3 over eastern China will  be increased by 0.02 %, if the NO x emission in this gird is increased by 100 %.The result shows that anthropogenic NO x contributes significantly to the O 3 distribution in this region.Although the influence of lightning NO x is weaker, the larger geographical distribution of lightning NO x makes it an important source.The contribution of anthropogenic CO is mainly from China, whereas southeastern Asia is the major contributor of biogenic isoprene with a negative sensitivity.Assuming anthropogenic CO is a proxy of anthropogenic hydrocarbons and biogenic isoprene is a proxy of biogenic hydrocarbons, these sensitivity calculations indicate that China is a major source of anthropogenic hydrocarbons while southeastern Asia is the major source of biogenic hydrocarbons.As shown in Fig. 1, the North China Plain has strong NO x emissions, but their effect on O 3 is not significant.However, eastern China free tropospheric O 3 is more directly sensitive to CO emissions from the North China Plain.The contribution of CO to ozone production (third set of panels in Fig. 5) is similar to the CO emissions distribution.Discrepancies exist between the spatial distribution of the sensitivity of ozone to NO x and CO.For example, the sensitivity of ozone to NO x in the Beijing area is relatively small because there is too much NO x , thus limiting ozone production.Martin et al. (2004) showed that an increase of NO x emission over northeastern China in summer will decrease surface O 3 concentration, which is opposite to the positive sensitivity in this work.This difference could be associated with the larger concentrations of VOC levels (e.g., Zhang et al., 2014) and because the sensitivities shown in Fig. 5 depend on both surface ozone production and ozone production in the free troposphere, which is more NO x limited.
It should be reminded that the sensitivity of ozone to biogenic isoprene is highly dependent on the isoprene chemistry scheme, as indicated by Mao et al. (2013).They demonstrated that the sensitivity of surface O 3 concentration over the southeastern United States on isoprene could change sign, from negative to positive, with two different isoprene schemes.However, as shown in Fig. 5 the influence of the isoprene scheme on free tropospheric O 3 is small relative to that from NO x so that we do not expect errors in the isoprene scheme to significantly alter our conclusions.These results are consistent with Mao et al. (2013) as they show that a change in sign in the sensitivity of ozone to isoprene only affected their surface ozone concentrations by 2 ppb or less and therefore had an even smaller effect on free tropospheric ozone concentrations.
The contributions to the free tropospheric ozone in the the China outflow region are shown in Fig. 6.The O 3 distribution is more sensitive to the anthropogenic NO x emissions from the coast rather than from the inland continent.The sensitivity hot spots clearly show a northeastward movement as the season progresses, from southeastern China (June) to Korea and Japan (August), reflecting the influence of the East Asian monsoon.
To understand the seasonal variation of O 3 production efficiency, we calculated the global-scale sensitivities of anthropogenic and lightning NO x during December 2005-November 2006 with 4 • × 5 • resolution.The values of sensitivities, as shown in Fig. 7, are significantly larger than those in Figs. 5 and 6, due to the change of grid size and smaller effect from initial condition.The sensitivity of O 3 to anthropogenic NO x has a marked seasonal variation, increasing in the Northern Hemisphere from winter to the summer.Kondo et al. (2008) found the slope of eastern Asia O 3 formation to NO x is proportional to HO 2 and thus increases from winter to spring.Increased solar radiation is another reason for the high O 3 production rate in the summer.Figure 7 also highlights the effect of anthropogenic NO x from southwestern China, showing a significant effect on free troposphere O 3 over eastern China, particularly in September-November.Similar to anthropogenic NO x , the contribution of lightning NO x is maximum in the Northern Hemisphere summer, partly associated with the East Asian monsoon.The sensitivities of O 3 over eastern China and the China outflow region have similar distributions, although the China outflow O 3 is more sensitive to coastal emissions.
Table 2 shows the regional, total contributions of anthropogenic and lightning NO x , calculated by summing the sensitivities shown in Fig. 7. Assuming an unchanged chemical environment, it can be explained as the percentage change of regional mean O 3 due to a 100 % change in NO x emissions.For example, a 100 % increase of Chinese anthropogenic NO x emissions in June-August 2006 will result in 10.2 % increase of tropospheric mean O 3 over eastern China.Of course, the result of an actual 100 % change of NO x will be different because of nonlinear chemistry.Furthermore, this sensitivity depends on the modeled transport and the robustness of the chemical production of ozone.For example, if the production of ozone is too "fast" then the sensitivity of free tropospheric ozone to surface emissions is too small as too much ozone is produced in the boundary layer (where lossmechanisms dominate) versus the free troposphere.To evaluate the sensitivities further, we enhanced Chinese anthropogenic NO x emissions by 10 % uniformly as a perturbation.Using the initial conditions provided from the standard simulation, the 3-month perturbation simulations are started on 1 December 2005, 1 March 2006, 1 June 2006 and 1 September 2006, individually.The relative difference of regional mean O 3 between the perturbation and standard simulations is then multiplied by 10.As shown in Table 2, the results of the two methods are highly consistent, which demonstrates our sensitivity analysis works well.Similar to Wild et al. (2012), the consistency also confirms that 10 % NO x perturbation gives a linear O 3 response over eastern Asia.Considering the high computation efficiency, adjoint sensitivity analysis is thus a good alternative to the traditional perturbation method.
As shown in Table 2, the effect of increased Chinese anthropogenic NO x on free tropospheric O 3 is limited.Assuming an unchanged chemical environment, a 100 % increase of Chinese anthropogenic NO x , during a 3-month period, will only result in a 2.4 % increase of free tropospheric O 3 in the winter and 10.2 % in the summer, associated with the chemical environment of China, which is is more inclined to be VOC limited.Because of the long O 3 lifetime in the free troposphere, O 3 from initial conditions has a substantial influence on the distribution of ozone.A 15-month continuous perturbation simulation, started on 1 September 2005, will enhance the effect of Chinese anthropogenic NO x to 3.0 % in winter and 10.5 % in summer.
Because of the rapid growth of pollutant emissions, the transpacific transport of Asian pollutants to North America has attracted significant attention (Zhang et al., 2008;L. Zhang et al., 2009;Walker et al., 2010;Bertram et al., 2013;Lin et al., 2008Lin et al., , 2014a)).The major transport mechanisms include the northeastward export of Asian pollution to about 50 • N, and then the transport across the Pacific in midlatitude westerly winds (Liang et al., 2004(Liang et al., , 2005)).Over eastern China, the effect of anthropogenic NO x emissions from the rest of Asia (ROA) on free tropospheric O 3 is that Table 2. Regional, total contributions of anthropogenic and lightning NO x on free tropospheric (819-396 hPa) O 3 over eastern China and the China outflow region.The value can be explained as the percentage change of regional mean O 3 (eastern China, China outflow) due to 100 % increase of NO x in a particular region (China and ROA).The regions of China and the ROA are defined in Fig. 3.The perturbation values (Pt) are the relative difference between standard and perturbation simulations.The model simulation is evaluated with TES O 3 and CO observations.The modeled concentrations are underestimated for both O 3 and CO, but reproduce the O 3 (CO) interannual variation.As with previous studies (Zhang et al., 2006;Voulgarakis et al., 2011;Kim et al., 2013), the modeled O 3 -CO correlation and slope are consistent with the data.The updated inventories significantly reduce the bias relative to TES CO measurements, but the improvement on the O 3 simulation is not large (∼ 1-2 %).The good agreement between model O 3 and CO and its correlations with obser-vations from TES demonstrates the reliability of the model simulation, the chemical scheme and the updated CO inventories.
We quantified source contributions (NO x , CO, VOC) to free tropospheric O 3 pollution over eastern China and the China outflow region with a sensitivity calculation approach.Our results show anthropogenic emissions from China are the major contributor to free tropospheric O 3 over eastern Asia and the corresponding outflow region.The anthropogenic emissions from the ROA have an important influence on free tropospheric O 3 over this region.The observed seasonal variation in O 3 is due to the seasonal change in the O 3 production efficiency, related to HO 2 and solar radiation.The contributions of lightning NO x to free tropospheric O 3 from China and the ROA are small, except in June-August due to the effect of the East Asian monsoon.Finally, our result shows that China is the major contributor of anthropogenic VOCs, whereas the influence of biogenic VOCs is mainly from southeastern Asia.The Supplement related to this article is available online at doi:10.5194/acp-15-99-2015-supplement.
demonstrated that TES data can be used to examine global distribution of O 3 -CO correlations.Voulgarakis et al. (2011) found significant positive correlations in the northern Pacific during the summer of 2005-2008.Kim et al. (2013) used OMI O 3 and AIRS (Atmospheric Infrared Sounder) CO to show that the GEOS-Chem model is able to reproduce the observed O 3 -CO correlations and slopes in the western Pacific, but failed in some tropical regions due to model transport error associated with deep convection.

Figure 1 .
Figure 1.Anthropogenic emission of (a) NO x and (b) CO in June 2006 as used in GEOS-Chem.The unit is in molecules per square centimeter per second (molec cm −2 s −1 ).The black box defines the domains studied in this work.The "eastern China" domain includes the grids of Chinese mainland within the black box.The "China outflow region" are grids within the black box, excluding the Chinese mainland.

Figure 2 .
Figure 2. Monthly regional mean O 3 and CO concentration at free troposphere (681-383 hPa) in June, July and August 2006-2010.The red line is the GEOS-Chem model simulation with a priori emission inventories and the black line represents the TES measurements.The model results are smoothed with the TES averaging kernels.The TES ozone data are biased high by 7 ppbv.

Figure 3 .
Figure 3. (a) Scaling factors of anthropogenic NO x for June 2006.(b) Scaling factor of total CO emission (combustion + oxidation from biogenic VOCs) for June 2006.

Figure 4 .
Figure 4. Monthly regional mean O 3 and CO concentration at free troposphere (681-383 hPa) in the period of December 2005-November 2006.The red line is GEOS-Chem model simulation with a priori emission inventories.The blue line is model simulation with updated NO x and CO emission inventories.The black line represents TES measurements.The model results are smoothed with the TES averaging kernels.The positive bias in the TES O 3 data is larger in summer and smaller in winter.

Figure 5 .
Figure 5. Contributions of anthropogenic NO x , lightning NO x , anthropogenic CO, biogenic isoprene on free tropospheric (819-396 hPa) O 3 over eastern China derived from the adjoint of GEOS-Chem in June, July and August 2006.The contributions can be explained as the percentage of change of regional mean ozone due to a fractional change in the emissions in a particular grid assuming an unchanged chemical environment.The numbers are the total of the absolute value of precursor contributions for the whole domain shown in the figures.

Figure 6 .
Figure 6.Contributions of anthropogenic NO x , lightning NO x , anthropogenic CO, and biogenic isoprene on free tropospheric (819-396 hPa) O 3 over China outflow region derived from the adjoint of GEOS-Chem in June, July and August 2006.

Figure 7 .
Figure 7. Contributions of anthropogenic NO x and lightning NO x on free tropospheric (819-396 hPa) O 3 over eastern China and the China outflow region in December 2005-November 2006.

Table 1 .
Monthly regional mean O 3 and CO correlation and slope for the free troposphere (825-383 hPa) for June, July and August 2006-2010 for both TES and model (in parentheses).The model values are sampled at the TES measurement time and location and smoothed with the TES averaging kernels.The calculation of slope and correlation is based on individual measurements within a month.The numbers of measurements used to compute the correlations are shown in the Supplement.The far-right column is the mean value of 15 months.
Our results show that the influence of the ROA on O 3 pollution export is significant because the influence of the ROA is comparable with Chinese emissions in winter and about 50 % of the Chinese emissions in other seasons for the outflow region.The contribution of lightning NO x over China is generally small relative to anthropogenic emissions except during the summer (Table2).The effect of the ROA lightning NO x is similar to the Chinese contribution but slightly larger.