Evidence of significant large-scale impacts of boreal

fires on ozone levels in the midlatitude Northern

Hemisphere free troposphere

K. Lapina,1 R. E. Honrath,1 R. C. Owen,1 M. Val Martın,1 and G. Pfister2

Received 2 February 2006; revised 21 March 2006; accepted 18 April 2006; published 27 May 2006.

[1] Summertime observations of O3 and CO made at thePICO-NARE station during 2001, 2003, and 2004 are usedto assess the impact of boreal forest fires on the distributionof O3 mixing ratios in the midlatitude Northern Hemisphere(NH) lower free troposphere (FT). Backward trajectorieswere used to select measurements impacted by outflowfrom high-latitude regions. Measurements during theseperiods were segregated into two subsets: those obtainedduring periods with and without apparent significantupwind fire emissions. Periods affected by fire emissionswere identified based on enhanced CO levels confirmed byglobal simulations of fire emissions transport. During fire-impacted periods, O3 was shifted toward higher mixingratios, with medians significantly higher than in periodswithout detectable upwind fire impacts. This implies asignificant impact of boreal wildfires on midlatitude lowerFT background O3 during summer. Predicted future increasesin boreal wildfires may therefore affect summertime O3

levels over large regions. Citation: Lapina, K., R. E. Honrath,

R. C. Owen, M. Val Martın, and G. Pfister (2006), Evidence of

significant large-scale impacts of boreal fires on ozone levels in the

midlatitude Northern Hemisphere free troposphere, Geophys. Res.

Lett., 33, L10815, doi:10.1029/2006GL025878.

1. Introduction

[2] Boreal wildfires are known to significantly impacttropospheric composition [e.g., Van der Werf et al., 2004;DeBell et al., 2004]. CO emissions have gained consider-able attention in recent years, as the resulting impacts ontropospheric background CO levels are significant [Novelliet al., 2003; Edwards et al., 2004]. Recent studies have alsoidentified increased mean summertime O3 at boundary layer(BL) sites in northwest N. America and at Mace Head,Ireland, during years of exceptionally large area burned[Jaffe et al., 2004; Simmonds et al., 2005], suggesting thatboreal fires may also significantly impact background O3.However, loss of ozone from the BL may mute this signal atBL sites [DeBell et al., 2004], and the large-scale impact ofboreal fires on FT O3 levels remains poorly characterized.[3] O3 plays a central role in tropospheric chemistry as a

major source of hydroxyl radical. O3 is phytotoxic, harmfulto health and an important greenhouse gas [Houghton et al.,2001]. Tropospheric O3 production in the NH midlatitudes

is dominated by anthropogenic sources, which impacttropospheric O3 over large regions [e.g., Chandra et al.,2004]. However, northerly regions can be a substantialsource of the O3 precursors nitrogen oxides (NOx, NO +NO2), CO and volatile organic compounds [Goode et al.,2000] during the boreal fire season (May–September).Quantifying the resulting impact of boreal fires on tropo-spheric O3 is a challenging task, due to the large degree ofvariability in the type of fire and fuel [Kasischke et al.,2005], which leads to variability and uncertainty in emis-sions of NOx, a limiting factor for O3 production. As aresult, the magnitude of O3 production from boreal fires ishighly variable and uncertain, with the limited number ofstudies available reporting O3 enhancements ranging fromhigh [e.g., Bertschi and Jaffe, 2005; Honrath et al., 2004;Law et al., 2005], to very low [e.g., Tanimoto et al., 2000]in several-days-old fire plumes.[4] Most studies of O3 in aged fire plumes are based on

observations obtained during the sampling of individualplumes and therefore characterize a specific combination offire type, meteorological conditions and plume age, or alimited number of such combinations. In contrast, this workutilizes multi-year (2001, 2003 and 2004) measurements ata site far downwind from the boreal fire region. Thus, theresulting dataset includes observations of fire plumes char-acteristic of multiple sources that have aged and dispersedduring 5–15 days of transport [Honrath et al., 2004]. Inaddition, this work utilizes FT measurements, a key dis-tinction as BL O3 loss may obscure the true magnitude ofboreal fire O3 impacts. It therefore provides a sampling ofthe large-scale impacts of boreal fires on FT O3 levels.

2. Methods

[5] To characterize the impact of boreal fires on O3

mixing ratios in regions well downwind, we combineanalysis of observations made at the PICO-NARE stationwith backward trajectories and the results of global simu-lations of fire emissions transport and chemistry.

2.1. Measurements

[6] The PICO-NARE station is located on the summitcaldera of Pico mountain in the Azores Islands (Portugal) inthe central North Atlantic Ocean (38�N, 28�W). This stationis frequently impacted by air from high-latitude regions,often without downwind transport over anthropogenicsource regions [Honrath et al., 2004]. It is therefore wellsuited to study the outflow from N. American and Siberianboreal wildfires. Station altitude (2225 m) is well above themarine boundary layer (MBL). To ensure that the measure-ments analyzed here are characteristic of the FT, we have

GEOPHYSICAL RESEARCH LETTERS, VOL. 33, L10815, doi:10.1029/2006GL025878, 2006

1Department of Civil and Environmental Engineering, MichiganTechnological University, Houghton, Michigan, USA.

2Atmospheric Chemistry Division, National Center for AtmosphericResearch, Boulder, Colorado, USA.

Copyright 2006 by the American Geophysical Union.0094-8276/06/2006GL025878

L10815 1 of 4

excluded from this analysis all periods potentially affectedby upslope flow of MBL air [Kleissl et al., 2005]. (Thisscreening removed 30% of the measurements, but did notaffect the results significantly.)[7] CO was determined using a modified non-dispersive

infrared absorption instrument (Thermo Environmental, Inc.,Model 48C-TL), calibrated daily with standards referenced tothe NOAA CMDL standard [Novelli et al., 2003]. O3 wasdetermined with commercial ultraviolet absorption instru-ments (Thermo Environmental Instruments Inc., Franklin,Massachusetts; Model 49C) referenced to the NOAA CMDLnetwork ozone standard (S. Oltmans, NOAA/ESRL, personalcommunications, 2001, 2004). As a result of instrumentdamage due to water ingestion following heavy icing theprevious winter, no O3 data are available for summer2002. Additional details on the instruments and calibrationmethods are provided elsewhere [Honrath et al., 2004; Owenet al., 2006]. Data were recorded as one-minute averages, andwere further averaged to obtain the one-hour averages used inthis work.

2.2. Transport Analysis

[8] To identify air mass transport pathways, we calculatedhourly backward trajectories using the HYSPLIT model(R. Draxler and G. Rolph, HYSPLIT4 (HYbrid Single-Particle Lagrangian Integrated Trajectory) model, NOAAAir Resources Laboratory, Silver Spring, Maryland, 2003,available at http://www.arl.noaa.gov/ready/hysplit4. html).The model uses hourly wind vectors interpolated from6-hourly National Weather Service National Center forEnvironmental Prediction FNL output [Stunder, 1997]. Foreach hour, a set of six trajectories was initiated: one termi-nating at the station, four terminating at end points separatedby 1� of latitude and longitude from the station, and oneterminating directly below the station, at a height of 2000 m.Trajectories were run for 10 days backward in time, withtrajectory locations recorded every hour.[9] We selected time periods potentially affected by

outflow from boreal regions by picking hours when oneor more of the trajectories passed over Alaska or Canadanorth of 50�N. These flow periods are referred to asnorthern N. American periods (abbreviated as NNA peri-

ods). They include multiple intervals of time impacted by N.American or Siberian fires, which last from hours to days[Honrath et al., 2004].

2.3. MOZART Simulations

[10] The global chemical transport model MOZART(Model for OZone And Related chemical Tracers) wasused to simulate the impact of emissions from the 2004 N.American boreal fires on CO at the PICO-NARE station.MOZART was driven by 6-hourly meteorological fieldsfrom the National Centers for Environmental Prediction(NCEP) dataset. The spatial resolution is approximately2.8� � 2.8� with 28 vertical levels between the surface and2 hPa. The chemical time step of the model is 20 minutes.The MOZART simulation results presented in this workare mixing ratios averaged over 2-hour windows andinterpolated to the location and pressure of the PICO-NARE observatory. These simulations utilized updated N.American boreal fire emissions that were optimized usingMOPITT (Measurements Of Pollution In The Tropo-sphere) CO columns as described by Pfister et al.[2005]; CO emissions were injected between 0 and 9 km.To infer the magnitude of the fire impact at the PICO-NARE station, we use the MOZART-simulated ratio ofCO fire tracer (CO that was emitted from the N. Americanboreal fires) to total CO mixing ratio at the PICO-NAREstation, ([COf]/[CO])MOZART. A second model run, inwhich N. American boreal fire emissions were turnedoff, was also conducted. Simulation results for O3 fromthis run were used to assess the importance of processesother than boreal fire emissions on O3 levels.

3. Results

[11] Figure 1a shows the distribution of summer COmixing ratios for the years of 2001, 2003 and 2004. In2003 and 2004 CO exhibited maximum levels nearly twiceas high as in 2001 (Table 1). High mixing ratios were alsomore frequent. Median values increased from 67 ppbv in2001 to 108 and 87 ppbv in 2003 and 2004, respectively.We have shown previously [Honrath et al., 2004] that2001 was a low-fire year, while the highest CO levels in2003 were the result of extreme Siberian fires that year.These fires consumed the largest area burned in more thanten years [Jaffe et al., 2004]. 2004 was also a high-fireyear, with the largest fire season on record in Alaska[Pfister et al., 2005]. To confirm these fires as the primarycause of the high CO that year, we have inspected thebackward trajectories for CO observations exceeding the70th percentile of the 2004 summer measurements (98ppbv). At least 75% of these cases were consistent withtransport from the region of active fires in Alaska and/orwestern Canada based on MODIS (MODerate resolutionImaging Spectroradiometer) fire counts, with haze in theupwind region clearly visible in MODIS true-color images.Thus, the considerably higher CO mixing ratios observedin both 2003 and 2004 were the result of emissions fromlarge fires transported to the Azores region in NNA flow.This conclusion is in accordance with recent works show-ing that boreal wildfires strongly affect the NH summerCO background during years of high fire activity [Novelliet al., 2003; Edwards et al., 2004; Van der Werf et al.,2004; Kasischke et al., 2005].

Figure 1. Distributions of (a) CO and (b) O3, including allmeasurements for each summer. A small number of COobservations greater than the maximum value shown areincluded in the rightmost bin.

L10815 LAPINA ET AL.: IMPACT OF BOREAL FIRES ON OZONE L10815

2 of 4

[12] The distributions of summer O3 mixing ratios werealso shifted toward higher levels in 2003 and 2004 relative to2001 (Figure 1b). Median O3 increased from 31 ppbv in2001 to 40 ppbv in 2003 and 41 ppbv in 2004 (Table 1).Given the prior observations of O3 enhancements in indi-vidual boreal fire plumes noted above and the presence ofelevated levels of nitrogen oxides in fire plumes sampled atthe PICO-NARE station during 2004 [Val Martın et al.,2005], O3 production in the fire plumes may have contrib-uted to the higher O3 levels in 2003 and 2004. To evaluatethis, we now test the hypothesis that boreal fire emissionsresult in significantly increased O3 levels in the highly agedfire plumes sampled at the PICO-NARE station.[13] To assess the impact of boreal fires on O3 levels, we

analyze only those measurements obtained during NNAperiods. This minimizes the potential impact of latitudinalO3 gradients on the results. We selected from the full set ofNNA periods those apparently impacted by fire emissions(termed ‘‘fire’’ air masses) and those with clean backgroundconditions (referred to as ‘‘non-fire’’ air masses). Fire peri-ods were identified based on the occurrence of significantlyenhanced CO levels, confirmed by simulations of fireemissions transport. Non-fire periods were identified basedon the occurrence of low CO levels and absent or minimalmodel-simulated fire impacts. The 40th and 60th percentilesof CO measurements in NNA flow periods (Table 1) wereused for the non-fire and fire CO cutoff values, respectively.For 2004, we required ([COf]/[CO])MOZART > 0.1 for the fireperiods; no intervals with ([COf]/[CO])MOZART > 0.09 wereincluded within non-fire periods. For 2001 and 2003, weused the results of previous analyses [Honrath et al., 2004],in which the NRL Aerosol Analysis and Prediction Systemmodel was used to identify PICO-NARE measurementspotentially impacted by upwind boreal fires. Fire periodswere required to be within those previously identifiedintervals, and non-fire periods were required to excludethose previously identified intervals.[14] Histograms of O3 for fire and non-fire periods in each

year are shown in Figure 2. For each year, the two distribu-tions are clearly different. Table 1 presents the medians,

means and ranges of each distribution. To confirm that theO3 distributions in the fire air masses were significantlydifferent from the O3 distributions during the non-fire peri-ods, we performed a nonparametric Wilcoxon Sum-rank test.We also performed a two-sample t-test to test for differencesbetween the means of two distributions. These two tests gaveconsistent results at the 0.01 level of significance, indicatingthat the distributions were significantly different, with signif-icantly higher mean values in the fire subset.[15] The differences between the median O3 values in the

non-fire and fire subsets each year (DMedian) are shown inTable 1. To estimate the contribution to DMedian resultingfrom latitudinal or altitudinal O3 gradients not caused byfires, we repeated the analysis with the following difference.Instead of using measured O3 mixing ratios, we used [O3]simulated at Pico in the 2004 MOZART simulation thatexcluded N. American boreal fire emissions. The resultingDMedian was 3 ppbv—14% of the actual value (22 ppbv:Table 1). To assess potential impacts of anthropogenic O3

production, we repeated the original analysis but excludedperiods with trajectories passing over N. America south of48�N. DMedian changed by <1 ppbv. (Both reanalyses wereconducted for 2004, the only year with sufficient data andMOZART simulations.) These results support the hypothe-sis that a majority of the observed fire minus non-fire O3

differences is the result of boreal wildfire emissions.

Table 1. Statistics of CO and O3 Distributions by Yeara

Data Subset

CO O3

N Mean Median Min Max N Mean Median Min Max DMedianb

2001All datac 606 70 67 47 111 593 34 31 14 65NNAc 187 82 85 57 111 184 42 44 14 65Non-fire, NNAd 50 71 69 61 82 50 32 32 14 59Fire, NNAd 25 97 96 88 111 25 53 55 37 65 24

2003All datac 442 104 108 54 201 259 40 40 12 78NNAc 263 114 114 56 201 153 45 46 13 78Non-fire, NNAd 37 80 79 56 108 37 35 30 13 69Fire, NNAd 33 130 130 117 142 33 52 53 39 59 23

2004All datac 1362 90 87 51 233 1332 41 41 13 79NNAc 852 97 95 56 233 842 45 46 15 79Non-fire, NNAd 200 76 76 56 90 200 33 30 15 69Fire, NNAd 248 122 114 99 233 248 51 52 32 74 22aAll concentrations are in ppbv. N is number of hourly average data points.bDifference between medians of the fire and non-fire O3 subsets.cMeasurements with either CO or O3 are included. See text for description of ‘‘NNA’’.dNon-fire and fire subsets were selected using the 40th and 60th percentiles of the ‘‘NNA’’ CO distribution, MOZART

simulations, and previously identified fire periods (see text). These cutoff values are equal to the non-fire maximum and the fireminimum, respectively. Measurements during all 1-hr periods with simultaneous CO and O3 are included.

Figure 2. O3 frequency distribution in Northern N.American (NNA) flow, fire and non-fire subsets. The areaunder the non-fire O3 distributions is shaded.

L10815 LAPINA ET AL.: IMPACT OF BOREAL FIRES ON OZONE L10815

3 of 4

[16] We discussed above the effect of high fire activity in2003 and 2004 on CO. Here, we see evidence of fire impactsin 2001 as well. This is the result of a relatively smallnumber of 2001 events in which Siberian fire emissionsreached the Azores, as noted previously [Honrath et al.,2004]. These 2001 events had little effect on the overall COand O3 distributions, but nevertheless provide additionalevidence of fire-enhanced O3.[17] It is possible to obtain ozone enhancement ratios by

dividing the values of DMedian (corrected by the MOZART-based estimate of 2004 non-fire contribution) by the differ-ence between median CO in the fire and non-fire subsets. For2004, the ratio is 0.5. This value is relatively large, but is inthe range of the highest values measured in aged boreal fireplumes by the sources cited in the Introduction. However,further work is required to explain the mechanism leading tothese high ozone enhancements.

4. Conclusions

[18] By separating multi-year summertime O3 observa-tions impacted by outflow from high-latitude regions intotwo subsets composed of fire-affected and non-fire (rela-tively clean) periods, we found that O3 levels in the NorthAtlantic lower FT are significantly increased when borealfire impacts are present. O3 production from boreal fireprecursors was therefore at least partially responsible forsignificant shifts in the O3 distributions toward highermixing ratios during the high-fire-activity summers of2003 and 2004, when median O3 values were 9–10 ppbvgreater than during the summer of 2001 (a low-fire year).Given the long distance from the fires in northwestern N.America and Siberia to the Azores where our measurementswere made, these results imply that boreal fires affectbackground O3 levels over a very large region of the NHmidlatitudes.[19] Surface temperatures in the boreal regions have risen

more rapidly than global average temperatures over recentdecades [Hassol, 2004], and Global Circulation Modelsimulations predict that boreal climate warming will leadto increased boreal fire danger in future decades [Stocks etal., 1998]. Our results imply that increased boreal firemagnitudes would lead to (or may have already led to) anincrease in the summertime O3 background over largeregions of the NH, providing a climate forcing feedback(as O3 is a greenhouse gas) and negatively affecting theability of downwind nations to meet O3 air quality standards.

[20] Acknowledgments. We thank J. Strane for assistance in air flowanalyses and M. Dziobak and P. Fialho for their roles in successful PICO-NARE station operation. NOAA Air Resources Laboratory provided theHYSPLIT transport model. The MODIS data used in this study wereacquired as part of NASA’s Earth Science Enterprise and obtained from theGoddard Distributed Active Archive Center. This work was supported byNOAA, Office of Global Programs, grants NA16GP1658, NA86GP0325and NA03OAR4310002, and by National Science Foundation grants ATM-0215843, ATM-0535486 and INT-0110397.

ReferencesBertschi, I. T., and D. A. Jaffe (2005), Long-range transport of ozone,carbon monoxide, and aerosols to the NE Pacific troposphere duringthe summer of 2003: Observations of smoke plumes from Asian borealfires, J. Geophys. Res., 110, D05303, doi:10.1029/2004JD005135.

Chandra, S., J. R. Ziemke, X. Tie, and G. Brasseur (2004), Elevated ozone inthe troposphere over the Atlantic and Pacific oceans in the Northern Hemi-sphere, Geophys. Res. Lett., 31, L23102, doi:10.1029/2004GL020821.

DeBell, L. J., R. W. Talbot, J. E. Dibb, J. W. Munger, E. V. Fischer, and S. E.Frolking (2004), A major regional air pollution event in the northeasternUnited States caused by extensive forest fires in Quebec, Canada,J. Geophys. Res., 109, D19305, doi:10.1029/2004JD004840.

Edwards, D. P., et al. (2004), Observations of carbon monoxide and aero-sols from the Terra satellite: Northern Hemisphere variability, J. Geophys.Res., 109, D24202, doi:10.1029/2004JD004727.

Goode, J. G., R. J. Yokelson, D. E. Ward, R. A. Susott, R. E. Babbitt, M. A.Davies, and W. M. Hao (2000), Measurements of excess O3, CO, CH4,C2H4, C2H2, HCN, NO, NH3, HCOOH, CH3COOH, HCHO, andCH3OH in 1997 Alaskan biomass burning plumes by airborne Fouriertransform infrared spectroscopy (AFTIR), J. Geophys. Res., 105(D17),22,147–22,166.

Hassol, S. J. (2004), Impacts of a Warming Arctic: Arctic Climate ImpactAssessment, Cambridge Univ. Press, New York.

Honrath, R. E., R. C. Owen, M. Val Martn, J. S. Reid, K. Lapina, P. Fialho,M. P. Dziobak, J. Kleissl, and D. L. Westphal (2004), Regional andhemispheric impacts of anthropogenic and biomass burning emissionson summertime CO and O3 in the North Atlantic lower free troposphere,J. Geophys. Res., 109, D24310, doi:10.1029/2004JD005147.

Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden,X. Dai, K. Maskell, and C. A. Johnson (2001), Climate Change 2001:The Scientific Basis, Cambridge Univ. Press, New York.

Jaffe, D., I. Bertschi, L. Jaegle, P. Novelli, J. S. Reid, H. Tanimoto,R. Vingarzan, and D. L.Westphal (2004), Long-range transport of Siberianbiomass burning emissions and impact on surface ozone in western NorthAmerica, Geophys. Res. Lett., 31, L16106, doi:10.1029/2004GL020093.

Kasischke, E. S., E. Hyer, P. Novelli, L. Bruhwiler, N. French, A. Sukhinin,J. Hewson, and B. Stocks (2005), Influences of boreal fire emissions onNorthern Hemisphere atmospheric carbon and carbon monoxide, GlobalBiogeochem. Cycles, 19, GB1012, doi:10.1029/2004GB002300.

Kleissl, J., M. P. Dziobak, and R. E. Honrath (2005), The influence oforographic flows on PICO-NARE trace-gas measurements, Eos Trans.AGU, 86(52), Fall Meet. Suppl., Abstract A51D-0110.

Law, K., et al. (2005), Evidence for long-range transport of North Americananthropogenic and wildfire emissions to Europe from airborne andground based LIDAR measurements during European ITOP (IGACLagrangian 2K4, ICARTT), Eos Trans. AGU, 86(52), Fall Meet. Suppl.,Abstract A41D-01.

Novelli, P. C., K. A. Masarie, P. M. Lang, B. D. Hall, R. C. Myers, and J. W.Elkins (2003), Reanalysis of tropospheric CO trends: Effects of the1997–1998 wildfires, J. Geophys. Res., 108(D15), 4464, doi:10.1029/2002JD003031.

Owen, R. C., et al. (2006), An analysis of transport mechanisms of NorthAmerican emissions to the central North Atlantic, J. Geophys. Res.,doi:10.1029/2006JD007062, in press.

Pfister, G., P. G. Hess, K. Emmons, J.-F. Lamarque, C. Wiedinmyer, D. P.Edwards, G. Petron, J. C. Gille, and G. W. Sachse (2005), QuantifyingCO emissions from the 2004 Alaskan wildfires using MOPITT CO data,Geophys. Res. Lett., 32, L11809, doi:10.1029/2005GL022995.

Simmonds, P., A. Manning, R. Derwent, P. Ciais, M. Ramonet, V. Kazan,and D. Ryall (2005), A burning question: Can recent growth rate anoma-lies in the greenhouse gases be attributed to large-scale biomass burningevents?, Atmos. Environ., 39, 2513–2517, doi:10.1016/j.atmosenv.2005.02.018.

Stocks, B., et al. (1998), Climate change and forest fire potential in Russianand Canadian boreal forests, Clim. Change, 38, 1–13.

Stunder, B. (1997), NCEP model output—FNL archive data, Tech. Rep.TD-6141, NOAA Air Resour. Lab., Silver Spring, Md. (Available athttp://www.arl.noaa.gov/ready-bin/fnl.pl.)

Tanimoto, H., Y. Kajii, J. Hirokawa, H. Akimoto, and N. P. Minko (2000),The atmospheric impact of boreal forest fires in far eastern Siberia on theseasonal variation of carbon monoxide: Observations at Rishiri, a north-ern remote island in Japan, Geophys. Res. Lett., 27(24), 4073–4076.

Val Martın, M., R. Honrath, R. Owen, J. Kleissl, P. Fialho, G. Pfister, andK. Lapina (2005), Large enhancements of NOx over the central NorthAtlantic lower free troposphere resulting from boreal wildfires: Observa-tions at the PICO-NARE station during summer 2004, Eos Trans. AGU,86(52), Fall Meet. Suppl., Abstract A41D-02.

Van der Werf, G. R., J. T. Randerson, G. J. Collatz, L. Giglio, P. S.Kasibhatla, A. F. Arellano Jr., S. C. Olsen, and E. S. Kasischke(2004), Continental-scale partitioning of fire emissions during the 1997to 2001 El Nino/La Nina period, Science, 303, 73–76.

�����������������������R. E. Honrath, K. Lapina, R. C. Owen, and M. Val Martın, Department of

Civil and Environmental Engineering, Michigan Technological University,Houghton, MI 49931-1295, USA. (reh@mtu.edu; klapina@mtu.edu)G. Pfister, Atmospheric Chemistry Division, National Center for

Atmospheric Research, Boulder, CO 80307-3000, USA.

L10815 LAPINA ET AL.: IMPACT OF BOREAL FIRES ON OZONE L10815

4 of 4