and Physics A case study of pyro-convection using ...lidar.ssec.wisc.edu/papers/damoah/damoah_pyro_convection.pdf · R. Damoah et al.: Pyro-convection 177 Fig. 5. (a) MODIS ﬁre

Atmos Chem Phys 6 173ndash185 2006wwwatmos-chem-physorgacp6173SRef-ID 1680-7324acp2006-6-173European Geosciences Union

AtmosphericChemistry

and Physics

A case study of pyro-convection using transport model and remotesensing data

R Damoah1 N Spichtinger1 R Servranckx2 M Fromm3 E W Eloranta4 I A Razenkov4 P James5 M Shulski6C Forster7 and A Stohl7

1Department of Ecology Technical University of Munich Freising Germany2Canadian Meteorological Centre Montreal Canada3Naval Research Laboratory Washington DC USA4University of Wisconsin-Madison Madison WI USA5Hadley Centre for Climate Prediction and Research Exeter UK6Alaska Climate Research Center Fairbanks AK USA7Norwegian Institute for Air Research Kjeller Norway

Received 8 July 2005 ndash Published in Atmos Chem Phys Discuss 18 August 2005Revised 9 November 2005 ndash Accepted 13 December 2005 ndash Published 26 January 2006

Abstract Summer 2004 saw severe forest fires in Alaskaand the Yukon Territory that were mostly triggered by light-ning strikes The area burned (gt27times106 ha) in the year 2004was the highest on record to date in Alaska Pollutant emis-sions from the fires lead to violation of federal standards forair quality in Fairbanks

This paper studies deep convection events that occurredin the burning regions at the end of June 2004 The con-vection was likely enhanced by the strong forest fire activ-ity (so-called pyro-convection) and penetrated into the lowerstratosphere up to about 3 km above the tropopause Emis-sions from the fires did not only perturb the UTLS locallybut also regionally POAM data at the approximate loca-tion of Edmonton (535 N 1135 W) show that the UTLSaerosol extinction was enhanced by a factor of 4 relative tounperturbed conditions Simulations with the particle disper-sion model FLEXPART with the deep convective transportscheme turned on showed transport of forest fire emissionsinto the stratosphere in qualitatively good agreement withthe enhancements seen in the POAM data A correspond-ing simulation with the deep convection scheme turned offdid not result in such deep vertical transport Lidar mea-surements at Wisconsin on 30 June also show the presenceof substantial aerosol loading in the UTLS up to about13 km In fact the FLEXPART results suggest that thisaerosol plume originated from the Yukon Territory on 25June

Correspondence toR Damoah(damoahforsttu-muenchende)

1 Introduction

Emissions of aerosols and trace gases like carbon monox-ide nitrogen oxides and non-methane hydrocarbons fromforest fires play an important role for atmospheric chemistry(Crutzen 1973 Crutzen and Andreae 1990) and radiationtransmission in the atmosphere (Robock 1991 Iacobellis etal 1999) Due to long-range transport of forest fire emis-sions they affect not only local surroundings but also in-fluence the atmosphere on regional (Tanimoto et al 2000Kato et al 2002) and even continental (Wotawa and Trainer2000) and hemispheric scales (Forster et al 2001 Wotawaet al 2001 Damoah et al 2004 Spichtinger et al 2004Yurganov et al 2004 Novelli et al 2003)

It is well known that volcanic eruptions can produce theexplosive force to inject material from the surface deep intothe stratosphere (Graf et al 1999) Only recently how-ever it was realized that deep convection triggered or en-hanced by forest fires (so-called pyro-convection) can alsopenetrate deeply into the stratosphere leading to the deposi-tion of gaseous and particulate products of the burning in theupper troposphere and lower stratosphere (UTLS) (Frommand Servranckx 2003 Fromm et al 2005) The mecha-nisms causing this deep convection are not yet entirely clearbut it appears that normal deep convection that occurs regu-larly in the boreal region in summer-time under favourablemeteorological conditions can be greatly enhanced by theheat released by the fire as well as by microphysical and ra-diative processes due to the large amounts of aerosols emit-ted These aerosols reduce cloud droplet numbers suppressprecipitation formation and enhance the significance of the

copy 2006 Author(s) This work is licensed under a Creative Commons License

174 R Damoah et al Pyro-convection

Fig 1 June daily temperatures (green) for(a) Anchorage(b) Juneau(c) Nome and(d) Fairbanks respectively The horizontal red and bluelines indicate record high and low temperatures respectively (see fromhttpclimategialaskaedu)

20

Fig 2 The annual area burned in Alaska during the fire seasonsbetween the years 1997 and 2004 (seehttpclimategialaskaedu)

ice phase (hence also enhance the amount of latent heatingavailable at high altitudes) (Andreae et al 2004) Similarprocesses have been observed to be effective in tropical pyro-convection (Andreae et al 2004) Furthermore the radiationabsorption by black carbon particles emitted by the fires andthe associated heating of the atmosphere above the cloud topscould lead to further lifting after the initial convective injec-tion into the UTLS region (Westphal and Toon 1991 Her-ring and Hobbs 1994 Trentmann et al 2002) While the rel-ative importance of these different mechanisms is not clearseveral recent papers have presented dedicated case studiesof pyro-convection events (Fromm and Servranckx 2003Fromm et al 2005) or cases that can likely be explained bypyro-convection although the authors did not explicitly sug-gest pyro-convection to be the transport mechanism (Liveseyet al 2004 Jost et al 2004) Some of the observations (egJost et al 2004) even show that the emission products can betransported to levels in the stratosphere with potential tem-peratures above 380 K a region that is commonly referred toas the stratospheric overworld

Atmos Chem Phys 6 173ndash185 2006 wwwatmos-chem-physorgacp6173

R Damoah et al Pyro-convection 175

Fig 3 ECMWF (European Centre for Medium-Range Weather Forecasts) Geopotential Height in gpdm (geopotential decametres) at 500 hPaon (a) 20 June 2004 at 0000 UTC(b) 21 June 2004 at 0000 UTC(c) 21 June 2004 at 1200 UTC(d) 22 June 2004 at 1200 UTC(e) 25June 2004 at 0000 UTC and(f) 26 June 2004 at 0000 UTC

In June 2004 dry conditions above normal temperaturesand strong lightning activity caused severe forest fires inAlaska and the Canadian Yukon Territory Air quality wassignificantly degraded in large regions by these fires Forinstance particulate matter smaller than 25microm (PM-25)of about 1000microgm3 were measured in Fairbanks exceed-ing EPArsquos threshold (300microgm3) by a factor of more thanthree (Rozell 2004) Also in Fairbanks measured CO of92 ppm nearly violated the Clean Air Act (95 ppm) whereastypical mid-summer CO background mixing ratios are onlyabout 80 ppb Emissions from the burning were also trans-ported over long distances High enhancements of aerosolscattering were observed at Cheeka Peak (483 N 1246 W480 m asl) (personal communication D Jaffe) and several

research aircraft were guided repeatedly into the fire plumesover eastern North America the Atlantic and over Europeduring the ICARTT (International Consortium for Atmo-spheric Research on Transport and Transformation) cam-paign and measured large enhancements of fire tracers in thelower and middle troposphere In fact Cammas et al (2006)1

in a paper in preparation linked an enhanced CO measure-ment taking during a commercial airliner cruising at about11 km over the North Atlantic to biomass burning Using

1Cammas J-P Brioude J Chaboureau J-P Duron J MariC Mascart P Hedelec P Smit H Volz-Thomas A Stohl Aand Fromm M Stratospheric injection of biomass fire smoke fol-lowed by long-range transport a MOZAIC case study in prepara-tion 2006

wwwatmos-chem-physorgacp6173 Atmos Chem Phys 6 173ndash185 2006


Fig 4 (a) ECMWF analyses of mean sea level pressure (colored contour lines) 2-m temperatures at selected locations (colored buttons)and 250 hPa winds (arrows) for 23 June at 0000 UTC plotted over a topography map of Alaska and western Canada Shaded areas showthe convective precipitation from a 12-h forecast from the 0000 UTC analysis(b) Brightness temperatures from the NOAA-12 satelliteat channel 4 (107microm) on 23 June at 0500 UTC and(c) on 24 June at 0100 UTC The black rectangle and white circle in (b) indicateapproximately the area shown in Fig 5b and location of the radiosonde (Fig 7) respectively

convective mass fluxes from a mesoscale model they showthat deep convection was responsible for the high CO con-centrations in the stratosphere

In this paper we focus on a particular episode of burning inJune 2004 when pyro-convection transported fire emissionsinto the UTLS Section 2 describes model calculations usedto study the transport of the plume and Sect 3 gives the me-teorological conditions in the observed area during the eventModel results satellite and lidar measurements during theevent are reported in Sect 4 and finally summary and con-clusions are presented in Sect 5

2 Model description

For studying the transport phenomena associated with thepollution plumes we used the Lagrangian particle dispersionmodel FLEXPART (Stohl et al 1998 2005) to simulate thetransport of a CO tracer CO tracer was used because CO hasa relatively long life time that ranges from 1 month (in thetropics) to 4 months (in the mid-latitudes) (Seinfeld and Pan-dis 1998) and because we want to focus here on the transportonly and do not consider chemical processes

FLEXPART simulates long-range transport diffusion dryand wet deposition and radioactive decay of air pollutantsreleased from point line or volume sources It treats ad-vection and turbulent diffusion by calculating the trajectories



Fig 5 (a)MODIS fire hot spots for 23 June 2004 The black rectangle is approximately the area shown in panel (b)(b) MODIS Aquasatellite image over Alaska and the Yukon Territory on 22 June at about 2200 UTC Red spots show active fires whereas blue-grey colorsindicate smoke and the whitish colors show snow ice and clouds

of a multitude of particles Stochastic fluctuations obtainedby solving Langevin equations (Stohl and Thomson 1999)are superimposed on the grid-scale winds from global me-teorological datasets to represent transport by turbulent ed-dies which are not resolved Global data sets also do notresolve individual convective cells although they reproducesome of the large-scale effects of convection ThereforeFLEXPART has recently been equipped with a convectionscheme (Emanuel and Zivkovic-Rothman 1999) to accountfor sub grid-scale convective transport The scheme requiresthe grid-scale temperature and humidity fields as input andcalculates as a by-product a displacement matrix for eachhorizontal grid box The matrix provides the necessary in-formation for the particle redistribution which is performedon the same vertical levels as the input data The convec-tion scheme is called every FLEXPART time step which is900sims The temperature and specific humiditiy are read infrom the meteorological data and linearly interpolated to thecurrent time

FLEXPART can be driven by meteorological analysis dataeither from the European Centre for Medium-Range WeatherForecasts (ECMWF 1995) or from the Global Forecast Sys-tem (GFS) of the National Center for Environmental Predic-tion (NCEP) Due to the significantly higher vertical resolu-tion of the ECMWF data and the higher intrinsic horizontalresolution of the operational ECMWF forecast model com-pared to the GFS model we use ECMWF data in this studyThese data have 60 hybrid model levels in the vertical andwere retrieved on a global 1times1 latitude-longitude grid Sixhourly analyses were supplemented by 3-h forecast step datato provide a 3-hourly temporal resolution An output grid

with a 1times2 latitudelongitude resolution a vertical spacingof 1000 m and an output interval of 3 h was employed

In the tracer simulations 5times106 particles were released tocalculate the transport of CO emissions from fires in Alaskaand the Yukon Territory For our simulation we consid-ered the 2-week period from 17 to 30 June 2004 duringwhich the strongest deep convection and transport event oc-curred In order to describe the regional and daily variationsof the fires the MODIS (httpmapngdcnoaagovwebsitefiredetectsviewerhtm) hot spot data were counted daily on a1

times1 grid For every 3-day period the maximum daily num-ber of hot spots in a grid cell was taken in order to minimizethe effect of missing hot spot data in the presence of cloudsThe daily area burned in North America on a per-province (inCanada) and per-fire-state (in the United States) basis takenfrom a web page at the Center for International Disaster In-formation (httpwwwcidiorgwildfire) was distributed tothose cells in the respective province containing the hot spotsThen a constant CO emission factor of 4500 kg per hectareof forest burned which is similar to recent estimates basedon emissions from the Canadian Northwest Territories (Coferet al 1998) was applied to release CO on a 1

times1 gridThe altitude at which fire emissions are effectively releasedinto the atmosphere varies from day to day and is actuallynot known Lacking this information we released the COtracer into the lowest 3 km of the model atmosphere which isconsistent with other fire simulations (Damoah et al 2004Spichtinger et al 2004) Transport to high altitudes must beproduced by the model itself either as a result of resolvedvertical transport or by the convection scheme It must benoted that all the processes related to the fire itself are not



Fig 6 Picture of a pyro-cumulonimbus located at 58 N 126 Won 27 June at 2100 UTC The picture was taken out of the cockpitof a commercial airliner cruising at about 10 km (Picture courtesyof Noriyuki Todo of Japan Airlines)

treated in the ECMWF model To some extent the effects ofthe fires (eg a destabilization of the atmosphere) might havebeen indirectly introduced into the analyses through the as-similation of observations (eg temperature data from satel-lites and radiosondes in the vicinity of the fires) Howeverthese effects are rather local and it is not clear whether ob-servations that were affected by the fires indeed entered theECMWF data assimilation It is thus likely that convectivetransport over the fires is underestimated by FLEXPART

3 Meteorological conditions in the burning area

It is known that climatologically maximum daily tempera-tures are related to the number of lightning strikes and thenumber of lightning strikes is in turn related to the num-ber of started fires (Rorig and Ferguson 1999) During theyear 2004 most of Alaska experienced the warmest summeron record In June temperatures were warmer than normalacross the State (Alaska Climate Research Center seehttpwwwclimategialaskaedu) Anomalies of the monthlymean temperature ranged from 25C to 4C In Anchor-

age (lat 61 N long 150 W) southern Alaska the monthlyaverage temperature in June was about 14C 25C abovenormal The highest temperature of the month (28C) oc-curred on 26 June (Fig 1a) More than 65 of the dailymaximum temperatures were above normal The conditionsin Juneau (lat 58 N long 134 W) were extreme a string of7 days (18ndash24 June) had record highs (red asterisks Fig 1b)The highest measured temperature was above 29C Nome(lat 64 N long 165 W) also registered three record high tem-peratures (Fig 1c) The monthly mean temperature in Fair-banks (lat 65 N long 148 W) was almost 195C 4C abovenormal

Between 14 and 15 June about 17 000 cloud-groundflashes were reported to have hit Alaska within 24 h thehighest number in 25 years (AICC 2004) Of the 707 firesreported in 2004 almost 300 were triggered by lightningstrikes (seehttpfireakblmgov) Being the third driestsummer on record (Rozell 2004) the fires also fanned bysurface winds of about 9 ms (Rozell 2004) on average inJune consumed large areas in Alaska and its surroundingsFigure 2 shows the annual area burned in Alaska from 1997to 2004 In 2004 707 fires burned a new record of about27times106 ha The old record of 2times106 ha was registered in1957

During the second half of June 2004 anti-cyclonic con-ditions prevailed over Alaska and north-western Canada(Fig 3a) However on 21 June a mid-tropospheric low-pressure system developed west of the Canadian west coastweakening the anticyclone over Alaska and north-westernCanada over the next two days (Figs 3b c d) Then the an-ticyclone strengthened again (Figs 3e f) and remained untilthe end of June The low pressure system caused only a slightdecrease in surface pressure and did not bring a major changein surface temperatures (see colored buttons in Fig 4a andtemperature time series in Fig 1) but colder temperaturesassociated with the mid-tropospheric cut-off low triggeredstrong convection as shown in an infrared satellite image on23 June at 0500 UTC (Fig 4b) Several centers of convec-tive activity can be seen with the coldest cloud top bright-ness temperatures (ltminus52C) located over the Yukon Terri-tory and British Columbia On the following days even lowerbrightness temperatures (ltminus60C) were observed (Fig 4c)As we shall show below the convective activity was strongestduring the afternoon and evening hours and weakened overnight This daily cycle was repeated several days after 20June The northern cluster of convection seen in Figs 4band c was responsible for the upward transport of forest fireemissions into the UTLS

4 Transport of fire emissions into the upper troposphereand lower stratosphere

Figures 5a and b show the distribution of active fires andsmoke on 23 and 22 June respectively from NASArsquos



Fig 7 Potential temperature profile derived from radiosonde temperature data at Fairbanks Alaska(a) on 23 June at 0000 UTC and(b) on24 June at 1200 UTC The horizontal orange line indicates the dynamical tropopause height calculated from ECMWF data

Moderate Resolution Imaging Spectroradiometer (MODIS)aboard the Terra and Aqua satellites (Kaufman et al 20031998 Justice et al 1996) It indicates intense burning inPingo and Winter Trail along the Yukon River (upper left ofFig 5b) The Billy Creek and Porcupine fires south-east ofthe Tanana River sent piles of smoke (blue-grey) into the at-mosphere Another cluster of hot spots can be seen in YukonTerritory through to Juneau and in the north-west of BritishColumbia (Fig 5a) These fires had been burning for weeksand also continued to burn into July

By comparing Figs 4b 4c and 5 it can be seen that thenorthern clusters of deep convection were located almostexactly over the region with the strongest burning activityWhile it is impossible to say to what extent the fires en-hanced the convection the strong correlation between firelocations and the coldest cloud tops is striking such thata causal relationship is suggested In support of this viewFig 6 shows a picture taken out of the window of a com-mercial aircraft on 27 June within the region The pictureshows a pyro-cumulonimbus cloud with cloud tops above10 km Three things about this image are striking Firstly thecloudrsquos brownish color which is reminiscent of pictures ofvolcanic eruptions and is caused by the huge mass of aerosolstransported upward by the cloud from the forest fires seento be burning underneath Secondly the exact co-locationof the convective cloud and the forest fire and itrsquos consider-able depth The cloud top is obviously much higher thanthe tops of the convective clouds seen in the backgroundThirdly despite the maturity and depth of the convection thecloud doesnrsquot seem to precipitate This supports our sugges-tion that in addition to the heat and water vapor (ie latentheat) released by the fire microphysical mechanisms similarto those reported by Andreae et al (2004) suppressed pre-cipitation and thus allowed the latent heat of freezing to bereleased at high altitude

Based on radiosonde potential temperature measurementsfrom Fairbanks (Fig 7) taken approximately within thewhite circle indicated in Fig 4b the coldest cloud tops werelocated at around 11 to 12 km According to the soundingsthe thermal tropopause height was at approximately 11 km on23 June at 0000 UTC and at 12 km on 24 June at 1200 UTCThe dynamical tropopause (orange lines in Fig 7) taken fromthe ECMWF analysis was located somewhat lower at about105 km on 23 June (the dynamical tropopause is the altitudewhere the potential vorticity exceeds the value of 2 potentialvorticity units) The radiosonde temperature profiles showa strong gradient and no significant structure in the free tro-posphere but a very sharp tropopause These features arecharacteristic of an environment shaped by deep convection(Salby 1996) Note that the sounding from 24 June wastaken during night hours and shows a very shallow surfaceinversion The cloud top height is consistent with the simu-lated transport of the CO emissions from the fires The fireswere located east of Fairbanks and a vertical section of theFLEXPART CO tracer at 64 N with convection turned on(Fig 8a) and off (Fig 8b) on 24th at 1200 UTC confirmsthe injection of fire emissions into the Alaskan lower strato-sphere due to convection at around 138 W In the simulationwith convection turned on forest fire tracer can be found upto about 12 km altitude The top of the tracer cloud at about12 km is in good agreement with the coldest cloud top tem-peratures (Fig 7) No injection into the stratosphere takesplace in the simulation without convection showing that theuse of the convection scheme in FLEXPART brings a betteragreement between measurements and model results

Figure 9 shows FLEXPART vertical cross sections along64 N cutting through the region with the strongest burningactivity at around 140 W It demonstrates how the convec-tion weakens during the night and morning hours and howthe forest fire emission injected into the free troposphere and



Fig 8 Vertical sections through the FLEXPART CO tracer at 64 latitude on(a f) 23 June at 0900 UTC and(b g) 2100 UTC(c h) on24 June at 0000 UTC(d i) 0300 UTC and(e j) 0600 UTC Panels (andashe) show simulations with convection turned on panels(fndashj) showsimulations with convection turned off The black solid line is the dynamical tropopause height (height of the 2PVU surface) determinedfrom ECMWF operational data



Fig 9 TOMS aerosol index over North America on(a) 23 and(b) 24 June 2004 FLEXPART 9ndash15 km CO tracer columns over NorthAmerica with convection turned on at(c) 23 June 0300 UTC(d) 24 June 0300 UTC(e)and(f) Same as (c) and (d) respectively but withconvection turned off The orange and green rectangles shown in Fig 10f indicate the locations of the POAM and SAGE measurementsrespectively shown in Fig 11a The red rectangle shows the location of the lidar measurements shown in Fig 12

lower stratosphere during the day are transported eastwards(Figs 9a b) In the afternoon hours of the following day theconvection gets active again (Fig 9c) and transports forestfire emissions to regions above the tropopause within a cou-ple of hours (Figs 9d e) This cycle is repeated every dayin the period from 20 June until the end of June and forestfire emissions were repeatedly injected into the stratosphereduring this period Figures 9fndashj show the same as Figs 9andashebut for a FLEXPART simulation with the convective parame-terization turned off No transport into the stratosphere takesplace reflecting again the importance of the representation ofconvection in transport models

41 UTLS observations of forest fire smoke

The aerosol index derived from data provided by the To-tal Ozone Mapping Spectrometer (TOMS) instrument onboard the Earth Probe satellite (seehttptomsgsfcnasagovfordetails) gives information on the spatial distribution ofUV-absorbing tropospheric aerosols (Hsu et al 1996) On23 and 24 June the TOMS aerosol index (Figs 10a and brespectively) shows filaments of enhanced values that stretchfrom Fairbanks and the Yukon Territory through Edmontonto Wisconsin Distinct maxima of more than 45 are observedover Fairbanks and the Yukon Territory on both days Theselocations correspond remarkably well with the locations ofthe very cold cloud tops in Figs 4b and c The sensitivity



Fig 10 (a) POAM III (orange) and SAGE II (green) aerosol extinction profiles at 1microm approximately over Edmonton on 24 June at0400 UTC The location of the profiles is shown in panel (f) The unperturbed black profile is the POAM aerosol extinction over Quebec at0000 UTC(b) Vertical section through the FLEXPART CO tracer at 54 latitude on 24 June at 0300 UTC with convection turned on Theblack solid line is the dynamical tropopause height (height of the 2PVU surface) determined from ECMWF operational data

of the TOMS aerosol index to the presence of absorbingaerosols increases strongly with altitude Thus the maxi-mum values of the TOMS aerosol index are likely due toaerosols located at high altitudes which have passed throughthe deep convective clouds

The features of enhanced TOMS aerosol index on 23 and24 June are in remarkable agreement with the FLEXPARTCO tracer located between 9 and 15 km (Figs 10c and d) inthe simulation with the convection parameterization turnedon On 23 June at 0300 UTC (Fig 10c) the simulated COtracer shows a similar filament over the Yukon Territory andBritish Columbia while on 24 June at 0300 UTC (Fig 10d)the maxima are at eastern Fairbanks Yukon Territory andEdmonton Figures 10e and f show the corresponding re-sults but with the convection parameterization turned offThe very low CO tracer in this simulation shows again thatthe upward transport in the simulation with convection turnedon was indeed due to the sub-grid-scale convection parame-terization In fact the average of the tracer between 9 and15 km with convection relative to without convection showsa nearly nine-fold enhancement on both days The feature inthe TOMS image which is close to the Beaufort Sea on 23and over the sea on 24 (red rectangle) is not visible in theFLEXPART 9ndash15 km column (Figs 10c and d) Howeverinspection of the 0ndash15 km column (not shown) indeed showsCO tracer over the Beaufort Sea Thus this feature may belocated at lower (lt9 km) altitude

Other instruments that testify to the perturbation of theUTLS by the Alaskan fires are the Polar Ozone andAerosol Measurement (POAM III) aboard the SPOT-4 satel-lite (Lucke et al 1999) and the Stratospheric Aerosol andGas Experiment (SAGE II) aboard the Earth Radiation Bud-

get satellite (Mauldin et al 1985) Figure 11a shows en-hanced aerosol extinction at 1microm wavelength in the UTLSfrom SAGE II (green) and POAM III (orange) over Ed-monton (green and orange rectangles in Fig 10f) on earlyhours of 24 June The SAGE profile shows a strong en-hancement between 10 and 12 km the ECMWF dynamicaltropopause (Fig 11b) over the location is at about 10ndash11 kmThe POAM profile over Edmonton shows an even strongerenhancement above 12 km but no values below This termi-nation of POAM data at such a high altitude is what is re-ferred to as ldquoHigh Zminrdquo (Fromm et al 1999) This nor-mally indicates that optically sufficiently thick clouds ob-scure the sun from the POAM sun tracker and thus mea-surements are available only above the cloud An investiga-tion of GEOS infrared imagery however revealed that therewere no high clouds in the region Visible imagery showedsome greyish features whose color suggests these were notnormal water or ice clouds In fact backward trajectories(not shown) from the enhanced POAM profile travel back tothe fire source in Alaska Thus it seems that the aerosol layerwas optically sufficiently thick to terminate the POAM suntracker measurements Another POAM profile over Quebecat 54 N 66 W (black line in Fig 11a) outside the regionof the Alaskan fire emission plume shows an unperturbedprofile for reference The two POAM profiles (orange andblack profiles in Fig 11a) above 12 km show an approxi-mately four-fold enhancement of the aerosol extinction of theperturbed (orange) relative to the unperturbed profile (black)Although the FLEXPART CO tracer across 54 N on 24 Juneat 0600 UTC (Fig 11b) shows the main part of the forestfire plume west of Edmonton there is clear evidence of for-est fire CO plume well above (1ndash2 km) the local tropopause



Fig 11 Lidar time-height plot of aerosol backscatter cross section over Madison (red rectangle in Fig 10f) on(a) 26 June from 1200 UTCto 2400 UTC(b) 28 June from 1200 UTC to 2400 UTC and(c) 30 June 2004 from 0100 UTC to 1200 UTC The vertical resolution is75 m FLEXPART time-height plots of the CO tracer from the simulation with convection turned on at the lidar location on(d) 26(e)28 (f)30 June for the same time periods as the corresponding lidar data plots (andashc)(g) (h) and(i) Same as (d) (e) and (f) respectively but withconvection switched off

at Edmonton This underlines that the enhanced POAM andSAGE extinction in the UTLS were likely due to a pyro-convective blow up

42 Lidar observations at Madison

During the last four days of June 2004 the High SpectralResolution Lidar (HSRL) (Eloranta 2005) at the Universityof Wisconsin (red rectangle in Fig 10f) used for long-termcloud studies (431 N 894 W) registered unusually highvalues of the aerosol backscatter cross section in the uppertroposphere (Fig 12) The strong signal was presumablysmoke that was transported from Alaska and the Yukon Ter-ritory Figures 12andashc show observed aerosol backscatter on26 28 and 30 June respectively (note that the first two pan-els show the last 12 h of the day while the last panel showsthe first 12 h of the day) It is obvious that there is a consistentaerosol layer build up high above the atmospheric boundarylayer at altitudes up to 13 km

The observed patterns of the aerosol backscatter through-out this period agree well with the simulated FLEXPARTCO tracer patterns from the fires both with convection on(Figs 12dndashf) and off (Figs 12gndashi) Although FLEXPART

tends to underestimate the plume height by about 2ndash3 km theoverall features are well produced It is evident from the sim-ulations that convection did not alter the CO tracer distribu-tion over Madison significantly during 26 and 28 June How-ever during the last 2 h of 26 of June traces of the CO liftedinto the UTLS by pyro-convection is evident above 9 kmThe strongest aerosol backscatter was observed on 30 Junewith the pyro-convective CO plume seen at 6 to 13 km Infact according to a FLEXPART movie showing maps of thetotal tracer column (not shown) this thick plume over Madi-son was advected north-eastwards from the Yukon Territoryon 25 June 24 h later it has taken a south-eastern direc-tion from south of Victoria Island (69ndash73 N 100ndash115 W)where it was stagnant for some hours before it headed to Wis-consin where it engulfed most of the upper troposphere overMadison and reached about 13 km

5 Summary and conclusions

In this paper we have used satellite and lidar data and theLagrangian transport model FLEXPART to study a pyro-covection event that occurred in Alaska and the Yukon



Territory in June 2004 Although pyro-convection is not in-cluded in the model and that the vertical transport of theforest fire CO is due to other processes (ie regular deepconvection) there is no generally accepted definition forpyro-convection at least for now so one cannot clearly drawa line between regular deep convection and pyro-convectionBut pyro-convection seems likely from the observation

After a mostly dry period of anticyclonic conditions overAlaska with unusually high temperatures a cyclone devel-oped west of the Canadian coast on 21 June It weakened theanticyclone over Western Canada and Alaska and triggeredstrong convection Several deep convective centers with verylow cloud top brightness temperatures were observed Thecluster of the coldest cloud top temperatures (belowminus60C)was located almost exactly over the region with the strongestfire activity in the boreal forest suggesting that the fires mayhave enhanced the convection A picture taken from a com-mercial airliner a few days later over another fire also showsthe pyro-convection over the fire to be much higher thanthe convective clouds in the environment The picture alsoclearly shows how smoky the pyro-convection was

POAM and SAGE extinction profiles that were availableon 24 June one day after one of the pyro-convection eventsshowed that there was enhanced aerosol extinction up toabout 13 km whereas the local tropopause was at about11 km Lidar measurements at Madison Wisconsin showedstrongly enhanced aerosol backscatter at up to about 13 kmaltitude on 30 June about 1 week after the strongest pyro-convection

The Lagrangian particle dispersion model FLEXPARTwas used to establish relationships between the different ob-servations of the fire plume These simulations could indeedtrace back the POAM and SAGE observations as well as thelidar observations to the pyro-convection events in the secondhalf of June 2004 However the simulations were only suc-cessful when the deep convection parameterization of FLEX-PART was switched on Without the convection parameteri-zation the deep transport of the fire emissions into the strato-sphere could not be simulated at all Even with the convec-tion parameterization turned on the maximum altitude of thesimulated pollution plume over the lidar location was some-what lower than observed This is likely because the convec-tion parameterization does not account for processes relateddirectly to the fire However the ECMWF data used to driveFLEXPART likely benefited from the assimilation of close-by (only about 557 km from the coldest cloud top tempera-tures) radiosonde data that clearly showed the signatures ofthe deep convection taking place In other cases where thereare no such close-by radiosonde data available tracer simula-tions of pyro-convection may be much less successful Thisdemonstrates the need to include information about fire ac-tivity in numerical weather forecast models and observationassimilation schemes

AcknowledgementsWe acknowledge ECMWF and the GermanWeather Service for permitting access to the ECMWF archivesWe appreciate the provision of data via the internet by the scienceteams at the Alaska climate research centre including those ofMODIS TOMS POAM SAGE and lidar group in Madison Weare thankful to N Todo of Japan Airlines International Co Ltd forhis image of a pyro-cb This study was funded by the EuropeanCommission in the European Research Framework 5 programas part of the PARTS (no EVK2CT200100112) project and theGerman Federal Ministry for Education and Research in theFramework of the Atmospheric Research 2000 program as part ofNOXTRAM (07 ATF05)

Edited by H Wernli

References

Alaska Interagency Coordination Center (AICC) Report (seehttpfireakblmgov) Review of the 2004 fire season 2004

Andreae M O Rosenfeld D Artaxo P Costa A A Frank GP Longo K M Silva-Dias M A F Smoking rain clouds overthe Amazon Science 303 1337ndash1342 2004

Cofer W R Winstead E L Stocks B J Goldammer J G andCahoon D R Crown fire emissions of CO2 CO H2 CH4 andTNMHC from dense jack pine boreal forest fire Geophys ResLett 25 3919ndash3922 1998

Crutzen P J A discussion of the chemistry of some minor con-stituents in the stratosphere and troposphere Pure Appl Geo-phys 106ndash108 1385ndash1399 1973

Crutzen P J and Andreae M O Biomass burning in the trop-ics impact on atmospheric chemistry and biogeochemical cy-cles Science 250 1669ndash1678 1990

Damoah R Spichtinger N Forster C James P Mattis IWandinger U Beirle S Wagner T and Stohl A Aroundthe world in 17 days ndash hemispheric-scale transport of forest firesmoke from Russia in May 2003 Atmos Chem Phys 4 1311ndash1321 2004SRef-ID 1680-7324acp2004-4-1311

ECMWF User Guide to ECMWF Products 21 Meteorol BullM32 ECMWF Reading UK 1995

Eloranta E W High Spectral Resolution Lidar in Lidar Range-Resolved Optical Remote Sensing of the Atmosphere edited byWeitkamp K Springer Series in Optical Sciences Springer Ver-lag New York 143ndash163 2005

Emanuel K A and Zivkovic-Rothman M Development and eval-uation of a convection scheme for use in climate models J At-mos Sci 56 1766ndash1782 1999

Forster C Wandinger U Wotawa G James P Mattis I Al-thausen D Simmonds P OrsquoDoherty S Jennings S KleefeldC Schnieder J Trickl T Kreipl S Jager H and Stohl ATransport of boreal forest fire emissions from Canada to EuropeJ Geophys Res 106 22 887ndash22 906 2001

Fromm M Bevilacqua R Hornstein J Shettle E Hoppel Kand Lumpe J D An analysis of Polar Ozone and Aerosol Mea-surement (POAM) II Arctic polar stratospheric cloud Observa-tions 1993ndash1996 J Geophys Res 104 24 341ndash24 357 1999

Fromm M D and Servranckx R Transport of forest fire smokeabove the tropopause by supercell convection Geophys ResLett 30 1542 doi10292002GL016820 2003



Fromm M Bevilacqua R Servranckx R Rosen J Thayer PJ Herman J and Larko D Pyro-cumulonimbus injection ofsmoke to the stratosphere observations and impact of a superblowup in northwestern Canada on 3ndash4 August 1998 J Geo-phys Res 110 D08205 doi1010292004JD005350 2005

Graf H Herzog M Oberhuber J M and Textor C The effectof environmental conditions on volcanic plume rise J GeophysRes 104 24 309ndash24 320 1999

Herring J A and Hobbs P V Radiatively driven dynamics ofthe plume from the 1991 Kuwait oil fires J Geophys Res 9918 809ndash18 826 1994

Hsu N C Herman J Gleason J Torres O and Seftor CSatellite detection of smoke aerosols over a snowice surface byTOMS Geophys Res Lett 26 1165ndash1168 1999

Iacobellis S F Frouin R and Somerville R C J Direct climateforcing by biomass-burning aerosols impact of correlationsbetween controlling variables J Geophys Res 104 12 031ndash12 045 1999

Jost H J Drdla K Stohl A Pfister L Loewenstein M LopezJ P Hudson P K Murphy D M Cziczo D J Fromm MBui T P Dean-Day J Gerbig C Mahoney M J Richard EC Spichtinger N Pittman J V Weinstock E M Wilson JC and Xueref I In-situ observations of mid-latitude forest fireplumes deep in the stratosphere Geophys Res Lett 31 11 101doi1010292003GL019253 2004

Justice C O Kendall J D Dowty P R and Scholes R Satel-lite remote sensing of fires during the SAFARI campaign usingNOAA advanced very high radiometer data J Geophys Res101 23 851ndash23 863 1996

Kato S Pochanart P Hirokawa J Kajii Y Akimoto H OzakiY Obi K Katsuno T Streets D G and Minko N P Theinfluence of Siberian forest fires on carbon monoxide concentra-tions at Happo Japan Atmos Environ 36 385ndash390 2002

Kaufman Y J Justice C O Flynn L P Kendall J D PrinsE M Giglio L Ward D E Menzel W P and Setzer A WPotential global fire monitoring from EOS-MODIS J GeophysRes 103 32 215ndash32 238 1998

Kaufman Y J Ichoku C Giglio L Korontzi S Chu D AHao W M Li R-R and Justice C O Fire and smoke ob-served from Earth Observing System MODIS instrument ndash prod-ucts validation and operational use Int J Remote Sensing 241765ndash1781 2003

Livesey N J Fromm M D Waters J W Manney G L San-tee M L and Read W G Enhancement in lower stratosphericCH3CH observed by the Upper Atmosphere Research SatelliteMicrowave Limb Sounder following boreal forest fires J Geo-phys Res 109 D06308 doi1010292003JD004055 2004

Lucke R Korwan D Bevilacqua R Hornstein J Shettle EChen D Daehler M Lumpe J Fromm M Debrestian DNeff B SquireM Konig-Langglo and Davis J The PolarOzone and Aerosol Measurement (POAM) III instrument andearly validation results J Geophys Res 104 18 785ndash18 7991999

Mauldin L Zuan N McCormic M Guy J and Vaughn WStratospheric Aerosol and Gas Expriment II instrument A func-tional description Opt Eng 24 307ndash312 1985

Novelli P C Masarie K A Land P M Hall B D MyersR C and Elkins J W Reanalysis of tropospheric CO trendsEffects on the 1997-1998 wildfires J Geophys Res 108 4464

doi1010292002JD003031 2003Robock A Surface cooling due to forest fire smoke J Geophys

Res 96 20 869ndash20 878 1991Rorig M L and Ferguson S A Characteristics of lightning and

wildland fire ignition in the Pacific Northwest J Appl Meteor38 1565ndash1575 1999

Rozell N Smoked pike on menu for Yukon Flats scientists AlaskaScience Forum 1710 2004

Salby M L Fundamentals of Atmospheric Physics InternationalGeophysics Series 622 pp Academic Press 1996

Seinfeld J H and Pandis S N Atmospheric Chemistry andPhysics 1326 pp John Wiley Inc New York 1998

Spichtinger N Damoah R Eckhard S Forster C James PBeirle T Wagner T Novelli P C and Stohl A Boreal forestfires in 1997 and 1998 A seasonal comparison using transportmodel simulations and measurement data Atmos Chem Phys4 1857ndash1868 2004SRef-ID 1680-7324acp2004-4-1857

Stohl A Hittenberger M and Wotawa G Validation of theLagrangian particle dispersion model FLEXPART against largescale tracer experiment data Atmos Environ 32 4245ndash42641998

Stohl A and Thomson D J A density correction for Lagrangianparticle dispersion models Boundary-Layer Meteorol 90 155ndash167 1999

Stohl A Forster C Frank A Seibert P and Wotawa G TheLagragian particle dispersion model FLEXPART version 62 At-mos Chem Phys 5 2461ndash2474 2005SRef-ID 1680-7324acp2005-5-2461

Tanimoto H Kajii Y Hirokawa J Akimoto H and Minko NP The atmospheric impact of boreal forest fires in far easternSiberia on the seasonal variation of carbon monoxide observa-tions at Rishiri a northern remote island in Japan Geophys ResLett 27 4073ndash4076 2000

Trentmann J Andreae M O Graf H-F Hobbs P V OttmarR D and Trautmann T Simulation of a biomass-burningplume Comparison of model results with observations J Geo-phys Res 107 4013 doi1010292001JD000410 2002

Westphal D and Toon O B The short-term temperature responseto smoke from oil fires J Geophys Res Lett 96 22 379ndash22 400 1991

Wotawa G and Trainer M The influence of Canadian forest fireson pollutant concentrations in the United States Science 288324ndash328 2000

Wotawa G Novelli P C Trainer M and Granier C Inter-annual variability of summertime CO concentrations in theNorthern Hemisphere explained by boreal forest fires in NorthAmerica and Russia Geophys Res Lett 28 4575ndash4578 2001

Yurganov N L Blumenstock T Grechko I E Hase F HyerJ E Kasischke E S Koike M Kondo Y Kramer I Le-ung F-Y Mahieu E Mellqvist J Notholt J Novelli C PRinsland P C Scheel E H Schulz A Strandberg A Suss-mann R Tanimoto H Velazco V Zander R and Zhao YA quantitative assessment of the 1998 carbon monoxide emis-sions anomaly in the Northern Hemisphere based on total col-umn and surface concentration measurements J Geophys Res109 D15305 doi1010292004JD004559 2004



Fig 1 June daily temperatures (green) for(a) Anchorage(b) Juneau(c) Nome and(d) Fairbanks respectively The horizontal red and bluelines indicate record high and low temperatures respectively (see fromhttpclimategialaskaedu)

20

Fig 2 The annual area burned in Alaska during the fire seasonsbetween the years 1997 and 2004 (seehttpclimategialaskaedu)

ice phase (hence also enhance the amount of latent heatingavailable at high altitudes) (Andreae et al 2004) Similarprocesses have been observed to be effective in tropical pyro-convection (Andreae et al 2004) Furthermore the radiationabsorption by black carbon particles emitted by the fires andthe associated heating of the atmosphere above the cloud topscould lead to further lifting after the initial convective injec-tion into the UTLS region (Westphal and Toon 1991 Her-ring and Hobbs 1994 Trentmann et al 2002) While the rel-ative importance of these different mechanisms is not clearseveral recent papers have presented dedicated case studiesof pyro-convection events (Fromm and Servranckx 2003Fromm et al 2005) or cases that can likely be explained bypyro-convection although the authors did not explicitly sug-gest pyro-convection to be the transport mechanism (Liveseyet al 2004 Jost et al 2004) Some of the observations (egJost et al 2004) even show that the emission products can betransported to levels in the stratosphere with potential tem-peratures above 380 K a region that is commonly referred toas the stratospheric overworld













2 Model description































































Edited by H Wernli

References

























































2 Model description































































Edited by H Wernli

References


















































2 Model description































































Edited by H Wernli

References









































































































Edited by H Wernli

References
































































































Edited by H Wernli

References





















































































Edited by H Wernli

References














































































Edited by H Wernli

References











































































Edited by H Wernli

References





































































Edited by H Wernli

References






























































Edited by H Wernli

References




















































Edited by H Wernli

References














































































Documents

and Physics A case study of pyro-convection using ...lidar.ssec.wisc.edu/papers/damoah/damoah_pyro_convection.pdf · R. Damoah et al.: Pyro-convection 177 Fig. 5. (a) MODIS ﬁre