Reconstructing ozone chemistry from Asian wild fires using models , satellite and aircraft measurements during the ARCTAS campaign

R. Dupont, B. Pierce, J. Worden, J. Hair, M. Fenn, P. Hamer, M. Natarajan, T. Schaack, A. Lenzen, E. Apel, J. Dibb, G. Diskin, G. Huey, A. Weinheimer, and D. Knapp Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, USA NOAA/NESDIS/STAR, Madison, WI, USA NASA Langley Research Center, Hampton, VA, USA Science Systems and Applications, Inc, Hampton, VA, USA Space Science and Engineering Center, University of Wisconsin, Madison, WI, USA National Center for Atmospheric Research, Boulder, CO, USA University of New Hampshire – EOS, Durham, NH, USA

Satellite data have allowed monitoring of the large numbers of seasonal fires over Siberia/northeast Asia (Cahoon et al., 1994;Tanimoto et al., 2000;Kajii et al., 2002;Kato et al., 2002;Soja et al., 2004) and over the Southeast Asian subcontinent (Christopher et al., 1998;Goloub and Arino, 2000). Consequently, Asian fires can represent 10% of global Carbon emissions (Van der Werf et al., 2006). Those fires play an important role in total tropospheric CO concentrations and interannual variability in the Northern Hemisphere (Novelli et al., 2003;Edwards et al., 2004;Kasischke et al., 2005;Pfister et al., 2005;Nedelec et al., 2005;Fisher et al., 2010). 15 Aircraft and ground-based measurements have shown that Asian fires can influence O 3 concentrations over North America Morris et al., 2006), the Arctic (Warnecke et al., 2008) and even Europe (Simmonds et al., 2005). For example, Bertschi and Jaffe (2005) have shown that Siberian fires have caused three ozone pollution events over the coast of Washington state in 2002. However, the impact of 20 fires on tropospheric O 3 concentrations is challenging to estimate. Indeed, O 3 production is highly variable inside boreal fire plumes because of the influence of clouds and aerosols on photochemical production as well as highly variable PAN and NO x concentrations and the cycling between these species (Mauzerall et al., 1996;Lapina et al., 2006;Val Martin et al., 2006;Real et al., 2007;Verma et al., 2009). 25 During spring 2008, fires burned throughout Asia and Siberia. Smoke from these fires was observed across the Pacific by satellite and aircraft measurements during the ARCTAS/ARCPAC campaign. In this paper we use fire counts from the MODIS satellite instrument, CO and O 3 profiles from Tropospheric Emission Spectrometer (TES), in situ measurements from ARCTAS aircraft data and the Real-time Air Quality Modeling System (RAQMS) to understand the factors influencing O 3 production and transport from these Asian fires and their impact on middle and high latitude tropospheric chemistry.

Spring ARCTAS 2008
The NASA ARCTAS field campaign consisted of two phases; a spring and summer deployment and was part of the international IPY/POLARCAT arctic field program for atmospheric composition (http://zardoz.nilu.no/ ∼ andreas/POLARCAT/), which involved a consortium of countries (e.g., United States, Canada, Germany, France, Norway, 10 Russia...) several aircraft, surface, and shipbased measurement platforms.
The Spring phase of the ARCTAS campaign took place as 3-week aircraft deployments in April 2008. It involved the NASA DC8 as an airborne platform for detailed atmospheric composition and two smaller aircraft to focus more closely on aerosols and radiation. As a complement to aircraft measurement NASA's polarorbiting satel- 15 lites (Terra and the Atrain) observe the longrange transport of pollution and its seasonal accumulation in the Arctic, enabling a better understanding of pollution sources, transport pathways, and radiative impacts of boreal smoke plumes.
ARCTAS has four major scientific themes, including the impact of Boreal forest fires on atmospheric composition, and long-range transport of pollution to the Arctic includ-20 ing tropospheric ozone. As an atmospheric receptor of long-range transport of pollutants from the northern midlatitudes continents (Shaw, 1995;Quinn et al., 2007), the Arctic is increasingly beset by emissions from massive forest fires in boreal Eurasia (Hao and Ward, 1993;Cahoon et al., 1994;Tanimoto at al., 2000;Fromm et al., 2000;Dlugokenchy et al., 2001;Conard et al., 2002;Kasischke and Bruhwiler, 2003). However, there remain large uncertainties regarding the transport pathways and the relative 26755 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | contributions of different source regions to arctic pollution (Staudt et al., 2001;Heald et al., 2003a;Liu et al., 2003;Liang et al., 2004;Koch and Hansen, 2005;Stohl, 2006;Shindell et al., 2008;Turquety et al., 2008;Shindell et al., 2008;Yashiro et al., 2009;Fisher et al., 2010;Jacob et al., 2010). Integration of satellite and aircraft observations with models through ARCTAS provides a means to address this issue (Eckhardt 5 et al., 2003;Klonecki et al., 2003;Koch and Hansen, 2005;Stohl, 2006;Shindell et al., 2008). Spacebased observations of CO and ozone from the Tropospheric Emission Spectrometer (TES) are of particular interest for characterizing synoptic transport events, the seasonal buildup of pollution during winter and spring and ventilation of the Arctic to midlatitudes in spring and summer. 10 Another major focus of ARCTAS is to evaluate chemical transport model (CTM) simulations of sourcereceptor relationships for pollutants in the Arctic which is a challenge because of the complexity of the transport patterns involved, the paucity of meteorological data, the stratification of the atmosphere, and uncertain chemistry and surface interactions. During April 2008, MODIS fire count (Fig. 1) shows large and relatively intense wildfires taking place in Asia, particularly in Kazakhstan, Siberia and Thailand. This agrees with previous observations by the Advanced Very High Resolution Radiometer (AVHRR) over Siberia and North Asia (Cahoon et al., 1994;Tanimoto et al., 2000;Kajii et al., 20 2002; Kato et al., 2002) and Southeast Asia (Christopher et al., 1998;Goloub and Arino, 2000). On the 19 April, during flight 11 between Fairbanks (Alaska) and Palmdale (California), the DC8 aircraft sampled several biomass-burning plumes, as identified based on in situ measurements (Fig. 2). The DC8 took off around 19:00 GMT (11:00 ADT) and landed approximately at 24:51 UTC (05:51 p.m. PDT) and remained along the OMI/TES track until around 22:51 UTC. The aircraft flew between 2 and 9 km to sample several biomass burning plumes. Using the DIfferential Absorption Lidar (DIAL) measurements, the altitude and thickness of the plumes were detected and 26756 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | allowed vertical profiling through them measuring chemical and aerosol composition with the in situ sensors aboard the aircraft.

DIfferential Absorption Lidar (DIAL)
During DC8 flight 11, the DIfferential Absorption Lidar (DIAL) instrument measured elevated concentrations of O 3 and aerosols in what appeared to be two biomass burning 5 plumes (Figs. 3 and 4). Figure 3 shows Aerosol Scattering Ratio (ASR) profiles measured by DIAL during DC8 flight 11 as a function of UTC time, latitude and longitude. The ASR is defined as the ratio of the attenuated aerosol backscatter laser light to the calculated molecular backscatter expected from an aerosol free atmosphere (Browell, 2001). 10 Values of ASR (591 nm) ranging from 0.1 to 5.0 and showing high concentrations of aerosols were observed in the lower and middle troposphere (2 to 6 km) all along the flight track. Forward model trajectories from the wildfire locations (Sect. 3.1) indicate that this plume originated primarily from the Kazakhstan and Baikal fires with contributions from anthropogenic emissions. In the same way, aerosols were detected 15 in the upper troposphere (8 to 12 km) between 45 and 55 • of latitude North. Forward model trajectories from the wildfire locations (Sect. 3.1) indicate that this plume came from the Thailand fires. Both middle and upper tropospheric plumes showed enhanced ozone concentrations relative to the background as seen in Fig. 4. In the same figure, O 3 concentrations higher than 125 ppbv are observed in the upper troposphere around 20 56 • of latitude north (20:00 UTC time). This is the result of a stratospheric air mass descent and is not due to wildfires.

In situ measurements of plume chemical composition
The NASA DC8 aircraft flew through two parts of the lower-middle tropospheric fire plume between 55 • N and 50 • N (20:00 to 21:00 UTC) and around 40 • N (around plumes. In situ measurements showed O 3 concentrations of as high as 140 ppbv (Fig. 6). Unfortunately, no in situ sampling of the high tropospheric plume was performed. Figure 5 shows CO (blue dots), SO 2 (green dots) and BC (red dots) concentrations along the DC8 flight track as a function of UTC time. Lower/middle tropospheric 10 biomass burning plume, originated from Kazakhstan, was sampled in two sections between 20:00 and 21:10 UTC, and around 23:00 UTC. Both parts show high CO, SO 2 and BC levels respectively up to 400 ppbv, 125 pptv and 250 particles per cm 3 in the northern part (55 • N, 20:00-21:10 UTC) and 220 ppbv, 450 pptv and 220 particles per cm 3 in southern part (40 • N, 23:00 UTC). Those high levels of CO (Roths and Harris, 15 1996;Jaffe et al., 1997;Yurganov et al., 1998;Ponchanart et al., 2003;Fisher et al., 2010) and aerosols (Haywood and Boucher, 2000;Zhao et al., 2002;Eck et al., 2003;Myhre et al., 2003;Massie et al., 2004;Abel et al., 2005;Reid et al., 2005a, b;Forster et al., 2007) observed by in situ data point to the biomass burning origin of the plumes. Moreover, SO 2 measurements are significantly above background levels in both part of 20 the lower/middle tropospheric plume; these enhanced amounts suggest mixing of the biomass burning plume with anthropogenic pollution. Figure 6 shows higher concentrations of PAN (up to 2500 pptv) (blue dots) in the northern part of the plume. PAN is an important source and sink of NO x with higher temperatures leading to NO x production from PAN and lower temperatures leading 25 to PAN production from NO x . PAN can be produced from NO x present in biomass burning plumes (Mauzerall et al., 1998;Real et al., 2007). Then, concentrations of NO x and PAN will depend on plume temperature (Singh and Hanst, 1981). The plume has been transported in the middle troposphere and at high latitude (50-55 • N) where temperatures are low and PAN is stable leading to high PAN concentrations. Figure 6 shows O 3 (green dots) and NO x (red dots) concentrations along the flight track as a function of UTC time. Both parts of the plume show high concentrations of NO x (90-100 pptv in average) and also high level of O 3 with mean concentration of 90 ppbv that is 30 to 40 ppbv above background level. 5 In summary, the aircraft data suggest that the observed plumes in the middle and lower troposphere could originate from biomass burning and were subsequently mixed with anthropogenic pollution. We use these aircraft data along with forward wildfire trajectories and CO and ozone profile measurements from TES as well as the RAQMS model to confirm the biomass burning origin of these plumes and explain the ozone 10 distributions observed during NASA/DC8 flight 11 on the 19 April 2008. We identify plume sources and reconstruct the chemical evolution of these plumes in Sect. 3 using the TES satellite data and RAQMS model and estimate the contribution to ozone from photochemistry and exchange from the stratosphere. 15 TES is an infrared Fourier transform spectrometer that measures the thermal emission of the Earth's surface and atmosphere over the spectral range 650-2250 cm −1 . It was designed to provide simultaneous vertical profile retrievals of tropospheric O 3 , CO and other trace gases on a global basis (Beer et al., 2001;Beer, 2006). The nadir footprint is 5.3 km across the spacecraft ground track and 8.5 km along track for the 16-detector 20 average (Beer et al., 2001). TES has two basic science operating modes: Global Survey and Special Observations. Global Surveys are conducted every other day while special observations are taken as needed in between Global Surveys. We used global survey observations of TES O 3 and CO obtained between 7 April and 19 April 2008 with a nadir sampling of 1.6

Overview of TES
• spacing along the ground track.

25
The analysis presented here utilizes TES version 003 data (Osterman et al., 2007). An overview of the TES retrieval algorithm and error estimation are discussed by Bowman et al. (2006)

and the characterization of errors and vertical information for 26759
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | individual TES profiles are discussed by  and Kulawik et al. (2006). The vertical resolution of TES nadir O 3 retrievals is about 6 km for cloud-free scenes, with sensitivity to both the lower and upper troposphere Bowman et al., 2006). To date, TES tropospheric O 3 validation has been conducted through comparisons with ozonesondes (Worden et al., 2007) and lidar (Richards et al., 2008). 5 These validation studies show that TES O 3 estimates are typically biased high in the upper troposphere by approximately 10%. Nassar et al. (2008) shows that TES O 3 is biased high by 3-10 ppb in the upper troposphere.

Real-time Air Quality Modeling System: overview
Chemical and aerosol analyses from the Real-Time Air Quality Modeling System 10 (RAQMS) and ensemble wild fire trajectories are used to examine the different processes influencing the evolution of trace gases (e.g., O 3 and CO) within fire plumes prior to sampling these plumes by the DC8. RAQMS is a unified (stratosphere/troposphere), online (meteorological, chemical, and aerosol) modeling system which has been developed for assimilating satellite observations of atmospheric chem- 15 ical composition and providing real-time predictions of trace gas and aerosol distributions (Pierce et al., 2003(Pierce et al., , 2007Kittaka et al., 2004). The chemical formulation follows a family approach with partitioning on the basis of photochemical equilibrium approximations. The non-methane hydrocarbon (NMHC) chemical scheme is based on the carbon bond lumped structure approach (Pierce et al., 2007). Photolytic rates 20 are calculated using the FAST-JX code, an updated version of FAST-J2 code (Bian et al., 2003). The RAQMS aerosol model incorporates online aerosol modules from GOCART (Chin et al., 2002(Chin et al., , 2003. Seven aerosol species (SO 4 , hydrophobic and hydrophilic organic carbon (OC), and black carbon (BC), dust, sea-salt) are transported. RAQMS biomass burning emissions use twice daily ecosystem/severity based emission estimates coupled with Moderate-Resolution Imaging Spectroradiometer (MODIS) Rapid Response fire detections (Al-Saadi et al., 2008). Total direct carbon emissions are calculated as the product of area burned and the ecosystem-and severity-specific 26760 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | carbon consumption estimates. Ecosystem-dependent carbon consumption databases for three classes of fire severity (low, medium, and high) are considered. Fire weather severity is estimated using the US Forest Service Haines Index, which considers atmospheric moisture and thermal stability (Haines, 1988). Emissions of other species are determined by combining published emission ratios for different ecosystems (Cofer et 5 al., 1991;Andreae and Merlet, 2001). The RAQMS chemical analysis used in the current study is from a retrospective 4-month (February-May 2008) 2 × 2 • assimilation that includes assimilation of cloud cleared OMI total column O 3 measurements and stratospheric O 3 profiles from the Microwave Limb Sounder (MLS) on the NASA Aura satellite. MODIS Aerosol Optical Depth (AOD) from instruments onboard the Terra and Aqua satellites (Remer et al., 2005;Davies et al., 2004) was also assimilated. A Mie code based look-up table of speciated aerosol mass extinction coefficients and relative humidity dependent hygroscopic growth factors is used to convert the predicted aerosol mass to speciated extinction, which is integrated vertically to obtain a first guess AOD for assimilation. 15 The masses of all aerosol species are adjusted within each model layer on the basis of the total AOD analysis increment and the relative contribution of each aerosol species to the total layer extinction. The resulting RAQMS aerosol analysis is in good agreement with April 2008 global Aeronet measurements (r = 0.7, bias=0.05). During the chemical and aerosol assimilation cycle the RAQMS meteorological forecasts are 20 reinitialized from NOAA Global Forecasting System (GFS) analyses at 6 h intervals.

Source identification of observed plumes
As discussed earlier, significant wild fires occurred in western Asia (Kazakhstan), eastern Asia (Siberia) and southeastern Asia (Thailand). Satellite fire counts over Russia were more than two times the April average in April 2008 (Fisher et al., 2010). Kaza- 25 khstan and Siberian fires are mainly due to burning of mixed coniferous and deciduous forest and croplands as observed by one-degree land cover derived from AVHRR.

RAQMS trajectory analysis
We use RAQMS chemical and aerosol analyses in conjunction with ensemble trajectory analysis (Pierce et al., 2009;Verma et al., 2009) to follow the chemical evolu-5 tion of ensemble trajectories initialized at daily wildfire locations. The resulting Lagrangian photochemical histories of the wildfire plumes can be used to relate emissions from these biomass burning regions to the observed plumes. First, we look at the lower/middle tropospheric plume identified in Sect. 2.2.1 and detailed in Sect. 2.2.2. RAQMS forward wildfire trajectory analysis (left graph of Fig. 8) shows that the plume 10 originates from Kazakhstan on the 7 April. This day was marked by a sudden increase of Kazakhstan fire CO emissions from less than 0.5 to more than 1.5 TgC day −1 . After initialization, ensemble trajectories are transported eastward, in the middle troposphere, over North Asia, across the Pacific via the Bering Sea and then south of Alaska where they are intersected by the NASA/DC8 flight track on the 19 April 2008, 15 12.5 days after initialization. During transport over Asia, the RAQMS model indicates that the wildfire plume was enhanced by Siberian biomass burning emissions with increase in CO and, carbonaneous aerosol (BCOC) (not shown in the figures). Moreover, the plume subsequently mixed with anthropogenic emissions over the Bering Sea, as indicated by increased SO 4 concentrations in RAQMS that is consistent with aircraft Several studies (Bey et al., 2001;Jacob et al., 2003) have already shown that CO from biomass burning in Southeast Asia is transported over the Pacific by the particularly strong Asian outflow in spring. The frequent cyclogenesis off the coast of East Asia transport pollution to the free troposphere (Stohl, 2001;Liu et al., 2003;Miyazaki et al., 2003) where air masses are isolated from the surface. Then, plumes are trans-5 ported to northern high latitudes and divided into northern and southern branches. The plume that originated from Thailand was transported over China and the RAQMS model suggested no enhancement by anthropogenic emissions. Figure 9 shows a comparison between DIAL Aerosol Scattering Ratio (ASR) at the bottom and RAQMS carbonaceous aerosol mixing ratio from the baseline simu-10 lation at the top. The RAQMS baseline carbonaceous aerosol distribution along the DC8 flight curtain agrees spatially with the DIAL measurements and shows a major carbon aerosol enhancement in the lower troposphere (2-6 km) (up to 4.6 ppbv) and lower aerosol amounts in the upper troposphere (8-12 km) (up to 2.4 ppbv). RAQMS gives a realistic estimation of ozone concentrations in the troposphere but underesti- 15 mates DIAL lower tropospheric ozone for the northern part of the flight (RAQMS max O 3 = 55 ppbv, DIAL max O 3 = 105 ppbv) (not shown in the figures). We use RAQMS sensitivity studies to quantify the impact of Asian biomass burning plumes on aerosols, O 3 and CO sampled by the DC8. 20 We conducted a series of RAQMS simulations (without aerosol or ozone assimilation) where we restricted April 2008 wildfire emissions to within either the Kazakhstan, Siberian, or Thailand regions. Differences between the baseline RAQMS simulation and simulations with the restricted wildfire emissions are then used to infer which fires were primarily responsible for the aerosol and ozone enhancements observed by the 25 DC8. Figure 10 shows RAQMS Carbonaceous aerosol (CC), CO and O 3 enhancements relative to background due to Asian fire emissions (Kazakhstan, Siberia and Thailand combined) and DIAL O 3 profiles along the DC8 flight 11. Based on this model 26763 We also determined the impact of each fire on species concentrations along the profile but don't show the results. Kazakhstan wildfires account for the majority of lower tropospheric wildfire signatures with peak CC, CO and O 3 enhancements respectively of 3.4 ppbv, 33 ppbv and 5.5 ppbv. However, Siberian fires account for a small fraction 10 of lower tropospheric wildfire signatures mostly below 2 km and at northern (60 • N) part of DC8 flight. As seen before in Sect.

Ozone origin in Asian fire plumes
Using TES and RAQMS CO and O 3 profiles we observe and interpret CO and O 3 concentrations during transport of the Kazakhstan/Siberian and the Thailand plumes from emission to flight location. We also discuss the impact of fire emissions and stratospheric-tropospheric exchanges on O 3 concentrations.
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Kazakhstan and Siberian fires
We next examine the Kazakhstan plume CO and O 3 concentrations at different locations during transport across the Pacific using the TES data and RAQMS chemical analyses. RAQMS ensemble trajectory analysis is used to determine which TES profiles sampled the wildfire plumes and we use TES and RAQMS CO and O 3 profiles 5 to characterize the plume chemical and dynamical evolution. Table 1 shows dates, locations and UTC time of each TES wildfire plume profile used during the study. Ozone and CO concentrations are examined in three locations as shown in Fig. 11a through Fig. 11c. First, we observe O 3 and CO concentrations in the plume on 7 April 2008 (YYYY/MM/DD) over source region (Kazakhstan). Then, we quantify  Fig. 11a. The bottom left panel shows these same RAQMS concentrations after the TES averaging kernel and constraint vector have been applied to account for the TES vertical sensitivity and retrieval bias (e.g., Jones et al., 2003;Pierce et al., 2009;Bowman et al., 2002Bowman et al., , 2006Bowman et al., , 2009. The CO/O 3 scatter 20 plots are color-coded by the predicted CO enhancement associated with the Kazakhstan wildfires, as determined by the differences between the RAQMS baseline and no Kazakhstan wildfire emission simulations. Color-coding helps to indicate the impact of Kazakhstan fire emissions on RAQMS CO and O 3 , and the same value are applied to TES observations. In all graphs, green to red points show high Kazakhstan fire 25 influence on RAQMS CO concentrations. Comparison of the bottom left panel of Fig. 11a to the upper left panel indicates that the RAQMS model estimates for maximum CO mixing ratios (∼200 ppbv) agrees 26765 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | well with the TES data after accounting for the vertical resolution and a priori of the TES data. However, comparison of the bottom left panel to the bottom right panel indicates that CO concentrations in the plume could be much larger (∼700 ppbv) than observed by TES. The blue points indicate low fire influence on RAQMS CO concentrations. Comparison between the TES and the adjusted RAQMS model ozone (with TES 5 operator) for these points indicates that RAQMS has lower background ozone in this region relative to TES. The negative correlation between O 3 and CO suggests that this air may have originated from the stratosphere. Therefore, for TES observations over Kazakhstan source region (Fig. 11a), elevated O 3 is mainly due to stratospheric enhancements and wildfire emissions do not result in significant O 3 enhancements within 10 the plume. At this stage, O 3 concentrations in the plume are up to 60 ppbv, 10 ppbv over background level. This can be due to the fact that the plume sampling occurs right after plume emission and does not allow a sufficient time for O 3 precursors to react in the troposphere and produce O 3 . After emission, the Kazakhstan wildfire plume was transported eastward in the mid- 15 dle troposphere (∼ 500hPa) over Asia. Before reaching the Pacific, the Kazakhstan plume encountered fire emissions from Siberia (Baikal) (Fig. 11b) leading to additional CO enhancements (up to 600 ppbv) that were not associated with the initial Kazakhstan emissions. TES and the adjusted RAQMS model agree in the estimation of maximum CO concentration (respectively 220 ppbv and 240 ppbv), but the adjusted 20 RAQMS model is higher than TES CO concentrations by more than 70 ppbv. In the same way, the adjusted RAQMS model underestimates maximum O 3 concentrations observed by TES by more than 50 ppbv. In comparison, in Fig. 12, in situ measurements of O 3 onboard of the DC8 aircraft show higher concentrations with O 3 ranging from 60 to 120 ppbv with collocated CO over 150 ppbv (pink dots).
As suggested by the aircraft in situ SO 2 measurements (Sect. 2.2.2) and confirmed by the RAQMS analysis of the Lagrangian chemical evolution of the Kazakhstan wildfire plume, the plume mixed with Asian anthropogenic emissions. The plume was then transported in the middle troposphere over the Pacific via the Bering Sea and reached flight location around the 18 April (Fig. 11c). Both TES and the adjusted RAQMS model show maximum CO concentrations (around 225 ppbv) and O 3 concentrations (max O 3 = 60 ppbv, min O 3 = 40 ppbv). However, lower CO concentrations are still overestimated by the adjusted RAQMS model by more than 60 ppbv and TES shows a low O 3 event, down to 20 ppbv of O 3 , which is not captured by the adjusted 5 RAQMS model. In situ measurements within the wildfire plume from flight 11 give O 3 concentrations ranging from 80 to 100 ppbv in the plume. Figure 12 shows that the RAQMS model significantly underestimates the dynamic range of in situ CO and O 3 (blue dots) along the DC8 flight track. RAQMS underestimates maximum in situ CO concentrations within the wildfire plume (yellow dots) and also underestimates ozone 10 mixing ratios within the stratospheric intrusion sampled by the DC8 and observed in Fig. 3. Comparisons between RAQMS and in situ CO also suggests that RAQMS overestimates background CO by approximately 60 ppbv, which is consistent with the comparisons with TES in regions which were not strongly influenced by the Kazakhstan wildfire plume. Because of averaging across the tropopause, Investigation of TES and 15 MODIS cloud data show that there is high cloud density over the region where the DC8 sample the main part of the plume. As a result, TES shows little sensitivity in O 3 between 50 and 60 • N.

Thailand fires
Using the same approach as with the Kazakhstan fires, we examined the Thailand 20 plume CO and O 3 concentrations at different locations during transport from the source and across the Pacific where the Thailand plume was eventually observed by DIAL on the DC8. Table 2 shows dates, locations and UTC time of each TES profiles used during the study.
In order to understand the plume's chemical and dynamical evolution, we use  As observed in the Kazakhstan/Siberian plume study, RAQMS CO downwind of source region (Fig. 13a) is higher than TES by almost a factor of 6 prior to applying the TES averaging kernel (AK) and a priori constraint. This discrepancy is due to the low sensitivity of the TES estimates to CO in the upper troposphere . However, after application the TES AK constraint to the RAQMS results, both 15 ranges of variability of CO are in good agreement. As opposed to CO, the TES sensitivity is sufficient to resolve ozone variations in the middle and upper troposphere (e.g. Liu et al., 2009;. Consequently, the TES and RAQMS ozone distributions are in better agreement prior to applying the TES AK. Two days after plume emission (Fig. 13a) Fig. 14 show d O 3 and d CO concentration perturbations due the Thailand fires predicted by the RAQMS model. It allows a better representation of the plume position and elevation, and gives perturbations due to fires up to 60 ppbv of O 3 and 300 ppbv of CO. In Fig. 13a

Conclusions
Using satellite ozone and CO tropospheric profile estimates from TES, and in situ and lidar data from the ARCTAS campaign, we explored the impact of Asian Boreal and Southeast Asian fires on tropospheric chemistry, long-range transport of associated pollutant over the Pacific and the chemical and dynamical processes influencing ozone 20 concentrations.
On 19 April 2008, Differential Absorption Lidar (DIAL) onboard of NASA/DC8 aircraft sampled two major biomass burning plumes ashore of western North America. Real time Air Quality Modeling System (RAQMS) ensemble wildfire trajectory analysis showed that those plumes originated from Kazakhstan and Thailand where 25 large wildfires were taking place with monthly mean CO emissions up to three times higher than usual April means. In situ measurements onboard the aircraft showed high concentrations of CO, PAN, and BC consistent with biomass burning along with signatures of anthropogenic emissions (SO 2 ) and significant O 3 enhancements (up to 100 ppbv). RAQMS sensitivity studies allowed us to assess the relative impact of Asian fires on CO, O 3 and CC concentrations along DIAL profiles. It showed that Thailand fires were responsible for almost the entire biomass burning signature in the upper tro-5 posphere (8 to 12 km); whereas Kazakhstan and Siberian fires were responsible for the majority of the fire signature in the lower middle troposphere (2 to 6 km).
We examined CO and O 3 concentrations within these two distinct plumes using TES and RAQMS data and characterized the chemical evolution along their respective transport pathways. Over the Kazakhstan fire source region, the majority of O 3 ob-10 served by TES was due to stratospheric enhancement. Upon transport eastward in the middle troposphere (500 hPa) over Asia, the plume went over the Siberian fire and was exposed to additional CO and O 3 precursor enhancements. Finally, the plume crossed the Pacific over the Bering Sea and reached the flight location on the 19 April. Along the flight track, TES and RAQMS observe similar O 3 concentrations ranging from 40 15 to 60 ppbv. However, in situ measurements from flight 11 measured O 3 concentrations ranging from 80 to 100 ppbv in the plume; this increased ozone enhancement could be due to localized ozone production within narrow filaments of the wildfire plume that are not observed by TES due to its relatively broad weighting functions and not captured by RAQMS due to it's relatively coarse vertical and horizontal resolution. The relatively good agreement between RAQMS predictions and TES observations during the wildfire plume transport allows us to estimate how much ozone these fires produce. We find that on a monthly mean scale, the springtime Kazakhstan and Siberian fires increase ozone concentrations by 3 DU over source region and 1 DU over North America and that the Thailand fires increase ozone concentrations by ap-5 proximately 10 DU over source region and 4 DU over North America. However, this estimate likely underestimates local ozone production based on comparisons between RAQMS and insitu DC8 measurements within the plume.
Considering this seasonal trend of biomass burning pollutant transport from Asia, it is important to characterize these events in order to understand their impact on air quality and climate. Recently, increase of CO by biomass burning in boreal area, and Southeast Asian fire emissions have been identified as affecting human health (Maynard and Waller, 1999) and perturbing atmospheric chemistry at regional and even global scales (Logan et al., 1981;Roths and Harris, 1996;Novelli et al., 1998;Bowman et al., 2009;. Increases in population in Southeast Asia tend to increase burning of postharvest agricultural waste that is an important type of biomass burning in this region of the world (Crutzen and Andreae, 1990;Nguyen et al., 1994). Increased temperatures from global warming also tend to increase drought and decrease snow cover in Northern Asia leading to increases in fuel quantity and a longer fire season (Euskirchen et al., 2007). 20 gations of fire, smoke, and carbon monoxide during April 1994 MAPS mission: Case studies over tropical Asia, J. Geophys. Res., 103, 19327-19336, 1998 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Variability of biomass burning aerosol optical characteristics in southern Africa during the SAFARI 2000 dry season campaign and a comparison of single scattering albedo estimates from radiometric measurements, J. Geophys. Res., 108, 8477, doi:10.1029/2002JD002321, 2003 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Estimation of area burned and emissions of pollutants by advanced very high resolution radiometer satellite data, J. Geophys. Res.,107(D24) Kato et al., 2002) and Southeast Asia (Christopher et al., 1998;Goloub and Arino, 2000).
On the 19th of April, during flight 11 between Fairbanks (Alaska) and Palmdale   temperatures leading to NO x production from PAN and lower temperatures leading to PAN production from NO x . PAN can be produced from NO x present in biomass burning plumes (Mauzerall et al., 1998;Real et al., 2007). Then, concentrations of NO x and PAN will depend on plume temperature (Singh and Hanst, 1981). The plume has been transported in the middle troposphere and at high latitude (50-55°N) where temperatures are low and PAN is stable leading to high PAN concentrations. Figure 6 shows O 3 (green dots) and NO x (red dots) concentrations along the flight track as a function of UTC time. Both parts of the plume show high concentrations of NO x (90-100 pptv in average) and also high level of O 3 with mean concentration of 90 ppbv that is 30 to 40 ppbv above background level.
In summary, the aircraft data suggest that the observed plumes in the middle and lower troposphere could originate from biomass burning and were subsequently mixed with anthropogenic pollution. We use these aircraft data along with forward wildfire trajectories