Impact of the 2002 Canadian Forest
Fires on Particulate Matter
Air Quality in Baltimore City
AMIR SAPKOTA,
†
J. MOREL SYMONS,
†
JAN KLEISSL,
‡
LU WANG,
†
MARC B. PARLANGE,
‡
JOHN ONDOV,
§
PATRICK N. BREYSSE,
†
GREGORY B. DIETTE,
|
PEYTON A. EGGLESTON,
|
AND
TIMOTHY J. BUCKLEY*
,†
Johns Hopkins University Bloomberg School of Public Health,
Johns Hopkins University Whiting School of Engineering, and
Johns Hopkins Hospital, Baltimore, Maryland 21205, and
University of Maryland, College Park, Maryland 20742
With increasing evidence of adverse health effects
associated with particulate matter (PM), the exposure
impact of natural sources, such as forest fires, has substantial
public health relevance. In addition to the threat to
nearby communities, pollutants released from forest fires
can travel thousands of kilometers to heavily populated urban
areas. There was a dramatic increase in forest fire
activity in the province of Quebec, Canada, during July
2002. The transport of PM released from these forest fires
was examined using a combination of a moderate-
resolution imaging spectroradiometer satellite image, back-
trajectories using a hybrid single-particle Lagrangian
integrated trajectory, and local light detection and ranging
measurements. Time- and size-resolved PM was evaluated
at three ambient and four indoor measurement sites
using a combination of direct reading instruments (laser, time-
of-flight aerosol spectrometer, nephelometer, and an
oscillating microbalance). The transport and monitoring
results consistently identified a forest fire related PM episode
in Baltimore that occurred the first weekend of July
2002 and resulted in as much as a 30-fold increase in
ambient fine PM. On the basis of tapered element oscillating
microbalance measurements, the 24 h PM
2.5
concentration
reached 86 µg/m
3
on July 7, 2002, exceeding the 24 h
national ambient air quality standard. The episode was
primarily comprised of particles less than 2.5 µmin
aerodynamic diameter, highlighting the preferential transport
of the fraction of PM that is of greatest health concern.
Penetration of the ambient episode indoors was efficient
(median indoor-to-outdoor ratio 0.91) such that the high
ambient levels were similarly experienced indoors. These
results are significant in demonstrating the impact of a
natural source thousands of kilometers away on ambient
levels of and potential exposures to air pollution within an
urban center. This research highlights the significance
of transboundary air pollution and the need for studies
that assess the public health impacts associated with such
sources and transport processes.
Introduction
Forest fires are a significant source of air pollution. Every
year, fires consume millions of hectares of forest in North
America, generating vast air pollution emissions that pose
significant potential for wide-scale human exposure. Biomass
emissions from large-scale forest fires include particulate
matter (PM), carbon monoxide (CO), polycyclic aromatic
hydrocarbons (PAHs), aldehydes, and semivolatile and
volatile organic compounds (VOCs) (1). Plumes of hot gases
containing airborne pollutants ascend from forest fires into
the mixed layer. These plumes can then detrain from the
mixed layer and be advected over long distances in the free
troposphere by prevailing winds (2, 3). Transport of air
pollutants in the free troposphere is more efficient than in
the mixed layer because of fewer loss mechanisms and
stronger winds (4). Previous research has shown that this
long-range transport can affect pollutant concentrations
thousands of kilometers from the source (5-10). Wotawa
and Trainer (11) reported that extensive forest fires in Canada
during 1995 influenced concentrations of CO and ozone (O
3
)
in the southeastern United States. Building on these findings,
McKeen and colleagues (12) showed that transport of nitrogen
oxide (NOx) led to elevated O
3
levels over eastern U.S. urban
centers. In addition to gaseous pollutants, several investiga-
tors (13-15) have reported high concentrations of PM in
different areas of Southeast Asia during the Indonesian forest
fires of 1997. Though these studies have shown biomass
burning to be a major source of ambient PM, to date, there
are no published findings that have evaluated the impact on
the indoor environment in population centers that are located
thousands of kilometers away.
Historic events characterized by unusually high levels of
air pollution provide evidence that increased concentrations
of airborne pollutants adversely impact human health (16-
18). In considering the health impact of air pollutants such
as PM, penetration of particles of outdoor origin into the
indoor environment is a key consideration because the indoor
environment is the dominant venue for exposures to air
pollutants due to the large fraction of time spent indoors
(19). Penetration is determined by a combination of particle
type and size and building and ventilation characteristics
(20).
The current study examines the impact of a significant
source event (Canadian forest fires) on the PM levels in an
urban center (Baltimore) located thousands of kilometers
from the source. The PM exposure continuum is examined
including the source, fate, and transport, ambient concen-
trations, the size distribution of particles comprising the
episode, and the indoor concentrations of PM within homes
of susceptible individuals, i.e., asthmatic children. This
examination illustrates the global significance of biogenic
air pollution sources and the meteorological conditions that
determine transport and impact on ambient and indoor
environments.
Methods
The impact of the July 2002 Canadian forest fires on PM air
quality in Baltimore was assessed across the exposure
continuum using an array of methods. Characterization of
the source term was based on data available from the
* Corresponding author address: Department of Environmental
Health Sciences, Johns Hopkins Bloomberg School of Public Health,
615 N. Wolfe St. (Room E6614), Baltimore, MD 21205; phone: (410)-
614-5750; fax: (410)955-9334; e-mail: tbuckley@jhsph.edu.
†
Johns Hopkins University Bloomberg School of Public Health.
‡
Johns Hopkins University Whiting School of Engineering.
§
University of Maryland.
|
Johns Hopkins Hospital.
Environ. Sci. Technol.
2005,
39,
24-32
24
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 1, 2005 10.1021/es035311z CCC: $30.25 2005 American Chemical Society
Published on Web 11/17/2004
Canadian Ministry of Forestry including reports tracking the
number of hectares of forest burned per month. The transport
of the smoke plume from the forest fires to Baltimore was
illustrated using satellite imagery and hybrid single-particle
Lagrangian integrated trajectory (HYSPLIT) modeling. The
existence of the smoke plume above Baltimore before the
peak of the episode and its reentrainment into the mixed
layer during the episode was determined using light detection
and ranging (LIDAR). The effect of this long-range transport
on the ambient PM air quality in Baltimore was determined
using monitors at three locations within the city. Finally, the
overall impact of the episode on the indoor environment
was assessed by measurement of PM within children’s
bedrooms.
The fate and transport of the forest fire plume from Quebec
to Baltimore was assessed using back-trajectory calculations
from HYSPLIT. This model employs archived Eta data
assimilation system (EDAS) weather model outputs to trace
the origins of an air parcel at a specific latitude, longitude,
and height over the continental United States. This model
is well suited and validated for assessing long-range air
pollution transport (21-24). Sets of back-trajectories with
end points over Baltimore (39.29° N, 76.55° W) were
computed every 6 h with end heights of 100, 500, and 1000
m above ground level (AGL). Each trajectory was computed
in 1 h increments for a 72 h period. These data were then
imported into the software package MATLAB (The Math-
Works, Inc., Natick, MA) and plotted over the U.S. carto-
graphic boundary files from the 2000 census. The plume’s
fate and transport were further assessed qualitatively through
publicly available satellite images acquired with a moderate-
resolution imaging spectroradiometer (MODIS).
The time-resolved vertical aerosol profile was measured
at site A1 (Figure 1) using the Johns Hopkins University
miniature elastic backscatter LIDAR (JHU-MEBL) as de-
scribed by Pahlow et al. (3). The LIDAR instrument was
operated in vertical upward pointing mode on a daily basis
during the EPA Baltimore Supersite summer intensive
sampling period (see www.jhu.edu/∼dogee/mbp/super-
site2001/for details on the measurements and data visual-
ization). A laser emits light pulses at a wavelength of 1.06 µm
parallel to the axis of a 26 cm, f/10, Cassegrain telescope.
In the lower troposphere, laser light is scattered mainly by
aerosols, whose size is on the order of the wavelength of the
laser light. During an ideal backscatter event, the laser light
is reflected 180° downward and is collected by the telescope.
Reflected light passes through an interference filter and a
lens that focuses the energy ona3mmdiameter, IR-enhanced
silicon avalanche photodiode. Both the travel time of the
light emitted from the laser to its collection and the intensity
of the backscatter radiation are digitized. In this way a vertical
profile of relative aerosol backscatter is obtained at a vertical
resolution of 3.6 m, a range of 6 km, and a time step of 5 s.
The strong vertical mixing in the daytime mixed layer (ML)
and the weak mixing above the temperature inversion on
top of the ML lead to an accumulation of aerosol, heat, and
moisture within the ML and a strong drop in these quantities
in the free troposphere. Due to this vertical gradient in aerosol
backscatter across the ML top, the ML height can be
determined with LIDAR.
Time- and size-resolved outdoor PM was measured at
two different locations (see Figure 1) using two different
methods. In central Baltimore (site A2), 5 min measurements
were made with a laser particle counter (Climet CI-500, Climet
FIGURE 1. Baltimore City map showing the locations of ambient and residential indoor monitoring sites. (A1) Baltimore Supersite ambient
monitoring station. Measurements include those with the aerodynamic particle sizer and the tapered element oscillating microbalance,
LIDAR, and meteorological monitoring. (A2) Baltimore Traffic Study monitoring station. Indoor and outdoor particle counts are made using
a switching manifold and a Climet-CI500 laser particle counter. (A3) Ambient monitoring station. The PM concentration was measured
with an MIE DataRam. (I-1-I-4) Indoor measurements made within homes (child’s bedroom) participating in a study of childhood asthma
in the urban environment. Measurements were made using an MIE DataRam.
VOL. 39, NO. 1, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
25
Instrument Co., Redlands, CA). The CI-500 was connected
to an isokinetic port of a stainless steel sampling manifold.
The manifold and Climet were located on the second floor
of a three-story row home. The CI-500 registers counts based
on light scattered by particles passing through a laser beam.
The amount of light scatter is directly proportional to the
size of the particle, which is then converted to an electrical
pulse by a photodetector. The amplitude of the pulse is
compared to a reference voltage, and the particle generating
the pulse is assigned to preset size bins of 0.3-0.5, 0.5-1.0,
1.0-2.5, 2.5-5, and 5-10 µm, respectively. The instrument
was operated at 2.8 L/min and was factory calibrated within
one year of use. The second monitoring site was the Baltimore
Supersite (site A1) located approximately 4 mi to the east of
the first location on the southeastern edge of the city.
Measurements at this location were made using both an
aerodynamic particle sizer (APS; TSI, Inc., Shoreview, MN)
and tapered element oscillating microbalance (TEOM; Rup-
precht & Patashnick Co., Inc., Albany, NY). The APS was
programmed to provide integrated particle counts every 5
min over 52 size bins between 0.452 and 20 µm. The TEOM
was equipped with a 2.5 µm sampling inlet and was operated
at 30 °C, providing mass concentration results in 30 min
intervals.
PM concentrations were measured in the bedrooms of
four children participating in a study of childhood asthma
in the urban environment and on the rooftop of a three-
story row home (site A3) centrally located in the neighborhood
where the children lived. Within each home, PM was
measured over 72 h using a portable direct reading
nephelometer (MIE pDR1000s, ThermoElectron, Franklin,
MA). The four study participants were 6-12 years of age and
lived in East Baltimore. Three of the four children were
asthmatic. Indoor source activity (e.g., smoking, cooking,
and vacuuming) and ventilation (windows, air conditioning)
during the monitoring period were obtained from a daily
questionnaire completed by an adult respondent within each
home. Details of this study and baseline exposure levels of
the entire cohort are described elsewhere (25).
The nephelometer uses a pulsed emitting diode producing
near-infrared light at 880 nm that is scattered by particles
entering the sensing cell. The intensity of light scattered by
the passing particles is proportional to the particle concen-
tration. The nephelometer’s inlet is designed to passively
sample particles of e10 µm. The manufacturer specifies a
measurement range from 0.1 to 10 µm; however, it has been
shown that particles in the range of 0.3-2 µm are prefer-
entially detected (26, 27). The nephelometers were factory
calibrated using SAE Fine (ISO Fine) test dust characterized
as having a mass median diameter of 2-3 µm, a bulk density
of 2.6-2.65 g/cm
3
, and a refractive index of 1.54 (Powder
Technology, Inc., Schofield, WI). Before being deployed to
the field, the nephelometers were zeroed against clean air
obtained using a HEPA filter according to the manufacturer’s
operation manual.
Meteorological variables were measured at site A1 using
a variety of instruments with sensors positioned on an 11 m
tower. In particular, incoming solar radiation was measured
with an Eppley Pyranometer at 11 m, the temperature and
relative humidity were logged from a Vaisala HMP45AC probe
at 4.9 m, rain was collected by a Campbell Scientific TE525MM
tipping bucket rain gauge, and the wind speed and turbulence
were measured by a Campbell Scientific three-component
anemometer-thermometer CSAT3 at 9 m.
Results
The impact of the Canadian forest fires on Baltimore PM
concentrations was assessed by examining the source activity,
fate and transport of the plume, resulting ambient PM
concentrations, and, ultimately, its infiltration into the homes
and bedrooms of asthmatic children. The source activity was
evaluated on the basis of the Canadian Forest Service estimate
of hectares of forest combusted in the Canadian province of
Quebec during the period of interest. For the forest fire season
of 2002, there was more than a 100-fold increase in hectares
of forest burned during the month of July compared to other
months (Figure 2). Specifically, during the first week of July,
there were 85 active wildfires reported in the province of
Quebec (28).
The transport of the plume across the eastern seaboard
was evaluated both qualitatively and quantitatively. A
qualitative evaluation was provided by a satellite image
(MODIS) taken on July 7 at 1035 EDT showing a coherent
smoke plume extending from the forest fires in Quebec to
the U.S. northeastern seaboard (Figure 3A).
A quantitative assessment of the plume transport was
provided using HYSPLIT. Multiple trajectories were run
representing a range of heights (100, 500, and 1000 m) and
times (every 6 h); however, the results were consistent in
showing air parcels contributing to the pollution episode
originated in the Quebec region, whereas those before and
after the episode originated elsewhere. For clarity of illustra-
tion, a single plot per day, capturing the conditions before,
during, and after the episode, is given in Figure 3B. The
trajectories shown are for 1300 EDT and represent a height
FIGURE 2. Hectares of forest burned in Quebec during the 2002 fire season.
26
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 1, 2005
of 500 m AGL. The selection of the 500 m height is justified
because the model is less accurate at lower altitudes due to
incomplete resolution of near-surface frictional and turbulent
effects. In addition, air pollution transport is better repre-
sented at heights above the inversion layer, especially during
the summer when strong nocturnal inversions occur (29).
The selection of 1300 EDT as an end time for the trajectories
is justified because atmospheric mixing is well established.
Back-trajectories are shown for the 5 days including and
surrounding the PM episode. The back-trajectories clearly
show that the air mass arriving in Baltimore on July 6-8
originated from the area of Quebec where the forest fires
were burning. In contrast, on July 5 and 9, the arriving air
mass originated from western Canada and the U.S. Midwest,
respectively.
The local plume was observed aloft using LIDAR mea-
surements. LIDAR provides the relative aerosol concentration
as a function of time and height z above ground level (Figure
4). Shortly after 0900 EDT on July 7, the forest fire smoke
plume was situated at a height between 1400 and 1800 m.
The LIDAR verifies that the smoke cloud shown by MODIS
images (recorded at 1035 EDT) existed at heights between
1100 and 1800 m above ground level before entrainment
into the mixed layer. The daytime turbulent mixed layer can
be identified due to the homogeneous vertical mixing within
it. The mixed layer height increased from 600 m at 1000 EDT
to approximately 1000 m at 1100 EDT that day. Until that
time, the aerosol concentration in the mixed layer was lower
than in the smoke layer aloft. Shortly after 1100 EDT, the
mixed layer reached the height of the smoke plume. Large
FIGURE 3. (A) MODIS satellite image taken on July 7, 2002, 1035 EDT. The red circles mark areas of high forest fire activity. MODIS satellite
image courtesy of the Land Rapid Response Team, NASA/GSFC, Greenbelt, MD. (B) Back-trajectories of air parcels ending at Baltimore,
MD, from the HYSPLIT model. The trajectory end times are shown in the legend, and the end height is 500 m AGL. Each segment of the
trajectory lines indicates a 12 h interval.
FIGURE 4. Time profile of range-corrected relative aerosol backscatter (Pr
2
) in arbitrary units from ground-based LIDAR measurements
over Baltimore on July 7, 2002 measured at site A1. Clean air scatters less light and is shown in blue, whereas the strongly scattering
smoke-laden air is shown in red (adapted from Pahlow et al.).
VOL. 39, NO. 1, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
27
eddies infringed the lower interface of the smoke plume and
started to mix the smoke-laden air with surface air beginning
at 1130 EDT. This process increased in intensity during the
afternoon. Thus, heavily polluted forest fire smoke entered
the mixed layer from aloft, resulting in elevated PM levels at
ground level. After 1800 EDT the mixing became less intense.
The ML height slowly diminished to around 650 m and the
forest fire plume moved upward to >1000 m at 2030 EDT.
Pahlow et al. (3) and Adam et al. (30) provide further details
on the mixed layer dynamics during this event.
The entrainment of the smoke plume into the mixed layer
evident from the LIDAR measurements coincided with a
dramatic increase in ground-level PM concentrations at both
ambient monitoring sites (sites A1 and A2), clearly defining
the forest fire pollution episode. Figure 5 shows the increase
in PM concentration (at site A1) across various particle size
ranges during the episode compared to the background. The
background was estimated by the 4 week average (2 weeks
before and after the episode) PM concentration matched by
day and time. Consistent with LIDAR observations previously
described, the peak of the forest fire episode in Baltimore
was observed on July 7 between 1100 and 1800 EDT. Much
of the increase in PM concentration was attributable to
particles less than 2.5 µm in size. During the peak of the
episode, there was as much as a 30-fold increase in PM
concentration in the 0.8-0.9 µm size range.
A similar episode-related particle concentration profile
was observed at site A2 on the basis of laser particle count
measurements. Figure 6 presents the outdoor particle count
concentration profiles for July 6 and 7 relative to the average
concentrations for all other weekends in June and July 2002.
This figure shows a marked increase in the concentration of
smaller particles during the peak of the episode with little
or no change in PM concentration for particles of g2.5 µm,
similar to that observed at site A1 measured with the APS.
The beginning of the forest fire episode is clearly identified
by the elevated levels of PM concentration over the back-
ground starting at 1600 EDT on July 6, reaching maximum
levels on July 7 between 1100 and 1800 EDT.
An episode-related increase in ambient PM (site A3) was
observed within the bedrooms of children (n ) 4, sites I-1-
I-4) participating in a study of asthma and air quality (Figure
7A). Consistent with the ambient measurements, the peak
in PM concentrations inside the homes occurred between
1100 and 1800 EDT on July 7, 2002. The indoor measurements
show a high level of correspondence to the associated outdoor
central site, A3, where PM was measured using the same
method. The median indoor-to-outdoor ratio during the
episode was 0.91 (quartiles of 0.74 and 1.05), indicating
efficient particle penetration. The daily questionnaire re-
vealed some smoking activity in two homes during the
measurement period (I-1 and I-4) and incense burning in a
third home (I-3). Further, all homes reported at least two
windows open on July 7. No health data were collected during
the period of monitoring. The PM
2.5
concentration profile at
site A1 measured with a TEOM (Figure 7B) agrees closely
with the bedroom and rooftop measurements, but differs in
absolute value. The maximum PM concentrations measured
at sites A1 (PM
2.5
TEOM), A3 (PM
10
MIE), and I-1-I-4 (PM
10
MIE) were 199, 645, and 590 µg/m
3
, respectively. On the basis
of the TEOM measurements, the 24 h integrated PM
2.5
concentration on July 7 was 86 µg/m
3
, exceeding the 24 h
average national ambient air quality standard (NAAQS; 65
µg/m
3
). The 24 h PM
2.5
concentration remained elevated on
July 8 and 9 (56 and 44 µg/m
3
, respectively), before decreasing
to 25 µg/m
3
on July 10.
Meteorology during the episode was characterized by a
long period without precipitation (June 28-July 9) and strong
northerly winds. Local wind directions oscillated around a
northern flow before the episode with average wind speeds
FIGURE 5. Three-dimensional plot showing the time profile of the particle mass concentration ratio (episode:background) by size as
measured with the aerodynamic particle sizer at site A1. The background level is given by the mean mass concentration on the same
weekend days preceding and following the episode.
28
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 1, 2005
