DECacp08.pdf

Atmos. Chem. Phys., 8, 4027–

4048

, 2008

www.atmos-chem-phys.net/8/4027/2008/

the Creative Commons Attribution 3.0 License.

Atmospheric

Chemistry

and Physics

Fast airborne aerosol size and chemistry measurements above

Mexico City and Central Mexico during the MILAGRO campaign

P. F. DeCarlo

1,2,*

, E. J. Dunlea

, J. R. Kimmel

, A. C. Aiken

1,3

, D. Sueper

, J. Crounse

, P. O. Wennberg

L. Emmons

, Y. Shinozuka

, A. Clarke

, J. Zhou

, J. Tomlinson

, D. R. Collins

, D. Knapp

, A. J. Weinheimer

D. D. Montzka

, T. Campos

, and J. L. Jimenez

1,3

Cooperative Institute for Research in Environmental Science (CIRES) University of Colorado, Boulder, CO, USA

Department of Atmospheric and Oceanic Science, University of Colorado at Boulder, Boulder, CO, USA

Department of Chemistry and Biochemistry, University of Colorado at Boulder, Boulder, CO, USA

California Institute of Technology, Pasadena, CA, USA

National Center for Atmospheric Research, Boulder, CO, USA

Department of Oceanography, University of Hawaii, USA

Department of Meteorology, Texas A&M University, College Station, TX, USA

now at: Laboratory of Atmospheric Chemistry, Paul Scherrer Institute, Switzerland

Received: 22 November 2007 – Published in Atmos. Chem. Phys. Discuss.: 20 December 2007

Revised: 20 June 2008 – Accepted: 20 June 2008 – Published: 25 July 2008

Abstract.

The concentration, size, and composition of non-

refractory submicron aerosol (NR-PM

)

was measured over

Mexico City and central Mexico with a High-Resolution

Time-of-Flight Aerosol Mass Spectrometer (HR-ToF-AMS)

onboard the NSF/NCAR C-130 aircraft as part of the MILA-

GRO field campaign. This was the first aircraft deployment

of the HR-ToF-AMS. During the campaign the instrument

performed very well, and provided 12 s data. The aerosol

mass from the AMS correlates strongly with other aerosol

measurements on board the aircraft. Organic aerosol (OA)

species dominate the NR-PM

mass. OA correlates strongly

with CO and HCN indicating that pollution (mostly sec-

ondary OA, SOA) and biomass burning (BB) are the main

OA sources. The OA to CO ratio indicates a typical value for

aged air of around 80

g m

−

(STP) ppm

−

. This is within

the range observed in outflow from the Northeastern US,

which could be due to a compensating effect between higher

BB but lower biogenic VOC emissions during this study. The

O/C atomic ratio for OA is calculated from the HR mass

spectra and shows a clear increase with photochemical age,

as SOA forms rapidly and quickly overwhelms primary ur-

ban OA, consistent with Volkamer et al. (2006) and Klein-

man et al. (2008). The stability of the OA/CO while O/C in-

creases with photochemical age implies a net loss of carbon

from the OA. BB OA is marked by signals at

m/z

60 and

Correspondence to:

J. L. Jimenez

(jose.jimenez@colorado.edu)

73, and also by a signal enhancement at large

m/z

indica-

tive of larger molecules or more resistance to fragmentation.

The main inorganic components show different spatial pat-

terns and size distributions. Sulfate is regional in nature with

clear volcanic and petrochemical/power plant sources, while

the urban area is not a major regional source for this species.

Nitrate is enhanced significantly in the urban area and imme-

diate outflow, and is strongly correlated with CO indicating

a strong urban source. The importance of nitrate decreases

with distance from the city likely due to evaporation. BB

does not appear to be a strong source of nitrate despite its

high emissions of nitrogen oxides, presumably due to low

ammonia emissions. NR-chloride often correlates with HCN

indicating a fire source, although other sources likely con-

tribute as well. This is the first aircraft study of the regional

evolution of aerosol chemistry from a tropical megacity.

1 Introduction

Aerosols are important components of the earth system.

Some of the effects of aerosols are reduction in visibility

(Watson, 2002), deterioration of human health (Pope and

Dockery, 2006), deposition of pollutants to ecosystems (Byt-

nerowicz and Fenn, 1996), and direct and indirect effects

on the radiative balance of the climate system. Currently,

aerosols and their associated direct and indirect effects con-

tribute the largest uncertainty to the radiative forcing of the

Published by Copernicus Publications on behalf of the European Geosciences Union.

4028

P. F. DeCarlo et al.: Aerosol size and chemistry measurements during MILAGRO

climate system (IPCC, 2007). Organic species account for

a large fraction of the submicron aerosol mass at most loca-

tions (Zhang et al., 2007a) and are especially poorly under-

stood. Urban areas are large sources of aerosols and aerosol

precursors. Pollution from megacities and large urban areas

is important not only for local effects on health, visibility,

and ecosystems/crops but also because of their collective in-

fluence in regional to global scale atmospheric chemistry and

radiative forcing (Lawrence et al., 2007; Madronich, 2006).

The Megacity Initiative: Local and Global Research Ob-

servations (MILAGRO) took place in and around Mexico

City during March of 2006. The Megacity Impacts on Re-

gional and Global Environment (MIRAGE-Mex) was the

component of the MILAGRO campaign under US NSF spon-

sorship, and included several aircraft platforms and ground

sites. The MILAGRO campaign was designed to study the

chemical characterization and transformation of pollutants

from the Mexico City urban area to regional scales in a

pseudo-lagrangian framework. A High-Resolution Time-of-

Flight Aerosol Mass Spectrometer (HR-ToF-AMS) (DeCarlo

et al., 2006) was deployed for the first time on an aircraft

platform onboard the National Science Foundation / National

Center for Atmospheric Research (NSF/NCAR) C-130 air-

craft.

Air pollution in Mexico City has been studied for many

years. An overview and detailed list of the studies in Mexico

City from 1960–2000 is given by Raga et al. (2001). Impor-

tant conclusions include the need for size-resolved composi-

tion measurements of PM

aerosol, as well as the need of

vertical measurements of aerosol species. More recent stud-

ies in Mexico City have added to our understanding of Mex-

ico City pollution. Salcedo et al. (2006) conclude that organ-

ics dominate PM

inside the city during the MCMA-2003

campaign, and a significant fraction of the organics are oxy-

genated organic aerosol (OOA). Sulfate showed a more re-

gional behavior while nitrate was mostly produced from local

photochemistry. Non-refractory (NR) chloride was present

in small levels with a diurnal cycle peaking in the morning,

but also showed some very large plumes that were not asso-

ciated with OA. The inorganic acids were neutralized by am-

monium most of the time, although some periods with am-

monium deficit were also observed when sulfate was high.

Volkamer et al. (2006) show that secondary organic aerosol

(SOA) from urban sources is produced rapidly and about 8

times more efficiently in the city than an SOA model pre-

dicted. Although only one case study is presented, the pa-

per states that similar conclusions were obtained for several

other days simulated. This study extended previous results

about SOA underprediction over regional scales (de Gouw

et al., 2005; Johnson et al., 2006a; Heald et al., 2005) to

shorter time scales and the urban environment. Kleinman et

al. (2008) analyzed the aerosol evolution with photochemical

age during MILAGRO over the urban area and near outflow

and confirm the results of Volkamer et al. (2006).

Wildfires and biomass burning (BB) have also been iden-

tified as important sources of particulate matter in Mexico

City. Bravo et al. (2002) found a correlation between es-

timated emissions from wildfires and both Total Suspended

Particles (TSP) and PM

for the years 1992–1999. Dur-

ing the Mexico City Metropolitan Area campaign in 2003

(MCMA-2003) biomass burning was identified as an impor-

tant source of fine particles, and especially for OA and also

mineral matter and black carbon, but with no enhancement of

nitrate, sulfate, or ammonium (Johnson et al., 2006b; Molina

et al., 2007; Salcedo et al., 2006). Recent results from the

MILAGRO campaign also suggest that biomass burning con-

tributed significantly to the gas and particle pollution in the

city basin and outflow, although the magnitude of the im-

pacts of fires is the subject of debate (Moffet et al., 2007;

Stone et al., 2008; Querol et al., 2008; Yokelson et al., 2007).

A source apportionment study using trace metals, inorganics,

EC, and OC at many locations in and around Mexico City for

total PM

and PM

did not identify biomass burning as a

source (Querol et al., 2008). However, specific tracers of this

source (e.g., WSOC, water-soluble K, levoglucosan) were

not included in this study, and thus biomass burning emis-

sions were probably mixed with other emission sources in

the source apportionment analysis (M. Viana, personal com-

munication, 2007). A different source apportionment study

of fine organic aerosol reports that BB made a highly vari-

able contribution in time accounting on average for 16% and

32% of the ambient organic carbon (OC) at the urban (T0)

and rural (T1) ground supersites respectively (Stone et al.,

2008).

C measurements indicate a significant fraction of

modern carbon in the organic aerosol measured at the T0

site, part of which likely originates from urban sources such

as trash burning and cooking (Aiken et al., 2007a; Gaffney et

al., 2008).

The meteorology of the Mexico City basin has been char-

acterized in earlier field campaigns. The meteorology of the

basin is very complex in response to synoptic, land/sea, and

orographic forcings (de Foy et al., 2006; Fast and Zhong,

1998). These field campaigns examined venting timescales

of the basin and concluded that typically the basin vents on

timescales less than one day (de Foy et al., 2006; Fast and

Zhong, 1998). Mixing heights during the MCMA-2003 cam-

paign were typically around 3000 m and vigorous vertical

mixing implied pollutants were well mixed in the bound-

ary layer during the day (de Foy et al., 2006). A recently

published study examines the basin scale wind transport dur-

ing the MILAGRO campaign, identifying six episode types

and compares the meteorological conditions in March 2006

with previous 10 years (de Foy et al., 2008). An overview

of the large-scale meterology during the MILAGRO cam-

paign is presented in Fast et al. (2007) and identifies 3 gen-

eral regimes. 1–14 March is the first regime and is char-

acterized by sunny and dry conditions. The second regime

from 14–23 March saw an increase in humidity and the de-

velopment of afternoon convection, which slowly diminished

Atmos. Chem. Phys., 8, 4027–

4048

, 2008

www.atmos-chem-phys.net/8/4027/2008/

P. F. DeCarlo et al.: Aerosol size and chemistry measurements during MILAGRO

4029

as the atmosphere became drier. The third regime (24–31

March) began with a cold surge, saw increased precipitation,

and a decrease in the frequency and intensity of fires.

The C-130 performed 12 research flights (RFs) during the

MILAGRO campaign. Flight altitudes were typically less

than 6 km above sea level (ASL) (

∼

96% of data), and ap-

proximately 97% of city data (defined in Sect. 3.2) were

taken at less than 2 km radar altitude. Vigorous turbulent

mixing in the boundary layer implies a well mixed boundary

layer over Mexico City during the afternoon. The regional

data average altitude is 4.0 km above sea level (a.s.l.) and

3.2 km radar altitude (above the ground). For this paper the

focus will be on RFs 2, 3, and 12 (8, 10 and 29 March 2006,

respectively). Data will also be presented from RFs 1, 9, 10,

and 11 from the 4, 23, 26 and 28 March 2006. RFs 2, 3, and

12 all had city and regional components in their flight pat-

terns. RF 1 had a city and regional component, but not all

instruments were working on the C-130. RF 9 was a flight to

the Yucatan Pennisula, and did not contain a city component.

RF 10 was cut short due to mechanical problems with the air-

craft, and the majority of the flight was spent at high altitude.

RF 11 was a flight designed to measure the morning transi-

tion of photochemistry and did not sample in the city basin.

During RFs 4-8 a leak in the shared inlet system did not allow

quantitative measurements of aerosol chemistry; when pos-

sible mass fractions of the measured AMS chemical species

were reported to the database, however they have not been

included in this paper.

In this paper, we present, the results of the measurements

of non-refractory submicron aerosol chemistry from the HR-

ToF-AMS in the C-130 which indicate that organic aerosols

(OA) are the largest NR chemical component, with nitrate,

sulfate, ammonium, and chloride also making significant

contributions. Different aerosol species have different spa-

tial distributions and tracer correlations, indicating influences

from different sources and processes. A large fraction of

the organic aerosol is oxidized, with an oxygen-to-carbon

atomic ratio (O/C) that increases with distance from the city

approaching a ratio of 0.9 away from the basin. Organic O/C

can be used as a qualitative “photochemical clock”. Finally

a detailed case study of RF 2 is presented to assemble all of

these results into a coherent and detailed picture of submi-

cron aerosol over and around Mexico City. Quantitative ap-

portionment of OA sources will be addressed in a subsequent

publication.

2 Methods

2.1 Instrumentation

2.1.1 HR-ToF-AMS

The HR-ToF-AMS has been described in detail previously

(DeCarlo et al., 2006). It improves upon previous versions

of the AMS (Drewnick et al., 2005; Jayne et al., 2000;

Canagaratna et al., 2007) by the use of a high-resolution

mass spectrometer that allows the determination of the el-

emental composition of most ions, while previous AMS

versions can only determine the total signal at each inte-

ger mass. In the rest of the paper we will refer to the

HR-ToF-AMS as “AMS” for brevity. The AMS sampled

non-refractory submicron aerosol (NR-PM

)

during the re-

search flights. Data were acquired in two acquisition modes

(Jimenez et al., 2003): Particle Time-of-Flight mode (PToF)

which allows for particle sizing, and Mass Spec (MS) mode,

which produces species concentrations and a complete mass

spectrum of the non-refractory submicron mass with no

size information, but with higher sensitivity than the PToF

mode. For most of the campaign data was averaged and

saved every 12 s. Data was also saved at longer time in-

tervals (

∼

30–45 s) early in the campaign, and sometimes a

“plume-mode” was utilized which consisted of 3 s save in-

tervals of MS mode only (no particle sizing). The HR-ToF-

AMS was run exclusively in the V-mode of ion flight for

the mass spectrometer, which is about a factor of 10 more

sensitive than the alternative W-mode. V-mode operation

has lower resolution than W-mode, but still maintains high

enough resolution to separate most ions of the same nominal

mass, and especially reduced vs. oxidized organic fragments

(DeCarlo et al., 2006). This instrument was installed to-

wards the rear of the C-130 (

http://mirage-mex.acd.ucar.edu/

Measurements/C130/Images/C-130

layout.png

) and shared

an inlet system with the Georgia Tech Particle into Liquid

Sampler (PILS). The inlet is described in more detail else-

where (Dunlea et al., 2008). Briefly, the inlet was mounted

in the belly of the aircraft and consisted of a near isokinetic

shallow conical diffuser into a 2.54 cm (1.0 inch) diameter

stainless steel tube with a smooth 90

◦

bend into the cabin

floor. The flow was then isokinetically subsampled into the

AMS line (8 l per min) and PILS (30 l per min). A pres-

sure controlled inlet was used just before the AMS (Bahreini

et al., 2008), to eliminate the fluctuations in aerosol sizing

and transmission efficiency due to changing pressures in the

aerodynamic lens of the instrument. The PCI used a set-

point of 350 Torr, with a 180 micron orifice upstream, and

a 150 micron orifice in the lens of the AMS. This allows for

sampling up to

∼

6.5 km without the need for flow or sizing

corrections to the raw AMS data. The transmission of par-

ticle sizes into the AMS was determined by the AMS inlet

(critical orifices and aerodynamic lens) as the losses in the

plane inlet and tubing were small for the AMS size range

(Dunlea et al., 2008). The residence time between the inlet

tip and the AMS was

∼

4 s (of which 3.2 s was in the PCI),

and the flow warmed up to cabin temperature due to ram

heating in the inlet and heat transfer (

∼

2–35

◦

C for air-

craft altitudes below 4 km). This may have led to some evap-

oration of aerosol components (Murphy et al., 2007), which

would be more important for ammonium nitrate and chloride,

less important for organics, and negligible for ammonium

www.atmos-chem-phys.net/8/4027/2008/

Atmos. Chem. Phys., 8, 4027–

4048

, 2008

4030

P. F. DeCarlo et al.: Aerosol size and chemistry measurements during MILAGRO

sulfate (Huffman et al., 2008b). Further information on AMS

data processing is given in Sect. 2.3.

2.1.2 Nephelometer

Total and submicrometer aerosol scattering coefficients were

measured at 450, 550 and 700 nm using two TSI model

3563 3-wavelength integrating nephelometers (Anderson et

al., 1996; Heintzenberg and Charlson, 1996; Anderson et al.,

2003). The submicrometer TSI nephelometer employed a 1-

m aerodynamic impactor maintained at 30 lpm by an Ali-

cat Scientific volumetric flow controller. While the measure-

ments were made every second, sample air residence time

inside the nephelometers was about 10 s. The noise over

10 s at 550 nm is estimated to be 0.3 Mm

−

(Anderson et al.,

1996). The instrument relative humidity was usually lower

than 30%. The scattering coefficients were detected over 7–

170

◦

, and corrected for 0–180

◦

using the measured wave-

length dependence as a surrogate for the particle size after

Anderson and Ogren (1998). For the submicron (PM

) scat-

tering, this angular truncation correction is typically less than

10%, and contributes negligible (

3%) uncertainty. Gas cal-

ibration results in a smaller (+/

−

1%) systematic error (An-

derson et al., 1996). The aerosol inlet for the nephelometer

was a shrouded solid diffuser maintained at isokinetic flow

5% and aspirated at about 100 lpm. The aerodynamic 50%

size cut was 5

m (McNaughton et al., 2007).

2.1.3 Optical Particle Counter (OPC)

An optical particle counter (OPC, a modified LAS-X, Parti-

cle Measurement Systems, Boulder, Colorado) measured the

dry (RH

30%) aerosol size distribution between 0.1

m and

about 10

m (Clarke, 1991). The He-Ne laser operates at

633 nm detecting light scattered by individual particles over

35–145

◦

. The particle size up to 2

m was calibrated with

polystyrene latex spheres whose refractive index is 1.59. For

calibrating the coarse mode, glass beads with a refractive in-

dex of 1.54 were also used. The data was obtained every 3 s,

but averaged over 30 s to reduce error due to low counting

statistics at about 1

m or larger.

2.1.4 Scanning Mobility Particle Sizer (SMPS)

The Texas A&M SMPS (Wang and Flagan, 1990) measured

the size distribution of particles between 0.012 and 0.67

mobility diameter once every 1.5 min. A high flow differen-

tial mobility analyzer (Stolzenburg et al., 1998) is used in the

SMPS with sample and sheath flow rates of 1.5 and 15 l per

min, respectively. The sampled aerosol was dried to below

10% RH prior to classification using a Nafion tube bundle.

The SMPS was located adjacent to, and shared an inlet with,

the nephelometer and Single particle soot photometer (SP2).

The “apparent” aerosol volume (DeCarlo et al., 2004) was

calculated from the SMPS size distributions with the assump-

tion of spherical particles. This volume would be biased high

if a significant fraction of the particles were non-spherical,

which is however not expected due to the large fraction of

secondary species.

2.1.5 HCN

The Caltech CIMS (Crounse et al., 2006) measured selected

product ions on the C-130 via reaction of the reagent ion

−

with analytes directly in air. HCN is measured by

monitoring the product ion at

m/z

112, which is the cluster

of CF

−

with HCN. The sensitivity to HCN is dependent

on the water vapor mixing ratio. Sensitivity changes due to

water vapor changes are corrected for using the dewpoint hy-

grometer water measurement from the C-130 aircraft, and a

water calibration curve that has been generated though lab-

oratory measurements. Non-water sensitivity changes are

corrected for using in-flight standard addition calibrations

of H

and HNO

(other species measured by the CIMS)

and proxied to laboratory calibrations of HCN. The detection

limit (S/N=1) for HCN for a 0.5 s integration period is bet-

ter than 15 pptv for moderate to low water vapor levels (H

mixing ratio

≤

0.004).

2.1.6 CO

The NCAR/NSF C-130 CO vacuum UV resonance fluores-

cence instrument is similar to that of Gerbig et al. (1999).

The MILAGRO data have a 3 ppbv precision, 1-s resolution,

and a typical accuracy of

10% for a 100 ppbv ambient mix-

ing ratio.

2.1.7 NO

, NO

measurement

(NO and NO

) and NO

(total reactive nitrogen) were

measured (along with O

) using the NCAR 4-channel chemi-

luminescence instrument, previously flown on the NASA

WB-57F (Ridley et al., 2004). NO

was measured via Au-

catalyzed conversion of reactive nitrogen species to NO, with

a time response of about 1 s. NO

was measured as NO fol-

lowing photolytic conversion of NO

, with a time response

of about 3 s due to the residence time in the photolysis cell.

NO was measured with an identical time response due to use

of a cell with an identical residence time. NO and NO

are

reported at 1-s. For NO, NO

, and NO

, the precision of a

1-s value is near 15 pptv. The overall estimated uncertainties

of 1-s values are

(15+7% of the mixing ratio) pptv for NO,

(15+10% of the mixing ratio) pptv for NO

(15+15% of

mixing ratio) pptv for NO

Atmos. Chem. Phys., 8, 4027–

4048

, 2008

www.atmos-chem-phys.net/8/4027/2008/

P. F. DeCarlo et al.: Aerosol size and chemistry measurements during MILAGRO

4031

2.1.8 MOZART model

A simulation of MOZART-4 (Model for Ozone and Re-

lated chemical Tracers, version 4) (Emmons et al., 2008

)

was run where tracers that represented the CO emissions

from Mexico City for each day were included. MOZART-

4 was driven with NCEP/GFS (National Centers for Envi-

ronmental Prediction Global Forecast System) meteorolog-

ical fields at a horizontal resolution of T170 (0.7

◦

). The

emissions are from the Mexico National Emissions Inventory

for 1999 (

http://www.epa.gov/ttn/chief/net/mexico.html

, as

gridded by M. Mena, U. Iowa) and the tracers include the

emissions between 18–20

◦

N and 98–100

◦

2.2 Ground supersites

The ground supersites, named T0, T1, and T2 are described

in Fast et al. (2007). Some additional information on the

ground supersites is given in Querol et al. (2008) and Stone et

al. (2008). Briefly T0 was inside the city, and T1 and T2 were

outside the city to the northeast about 30 and 63 km away

from T0 respectively. The names were chosen to indicate the

relative ages of air for pseudo-lagrangian experiments when

city air flowed to the northeast. A more detailed discussion

of these sites will be given in the forthcoming MILAGRO

overview paper (Molina et al., 2008

). Notably at the T0

supersite another HR-ToF-AMS was deployed jointly by the

Jimenez group and Aerodyne (Aiken et al., 2007a, 2008).

2.3 AMS data processing

2.3.1 AMS calibrations

Procedures for AMS calibration can be found in previous

publications (Allan et al., 2003, 2004; Jimenez et al., 2003;

Kimmel et al., 2006). The amplification factor of the mi-

crochannel plate detector (MCP) was measured every day

the instrument was in use (“single ion calibration”). Due to

the stability and reproducibility of the single ion throughout

the campaign a single average value was used for the entire

campaign. The lack of MCP degradation is likely due to the

very limited instrument use,

∼

120 h total, equivalent to only

5 days of operation in a ground-based field campaign. Ion-

ization efficiency (IE) calibrations were performed 6 times

during the campaign. Due to customs issues the calibration

equipment was not delivered until 10 days into the campaign,

thus the IE for the first 3 flights was assumed to be the same,

Emmons, L. K., Hess, P. G., Lamarque, J.-F., Pfister, G. G.,

Fillmore, D., Granier, C., Guenther, A., Kinnison, D., Laepple, T.,

Orlando, J., Tie, X., Tyndall, G., Walters, S., et al.: Impact of Mex-

ico City emissions on regional air quality from MOZART-4 simu-

lations, Atmos. Chem. Phys., in preparation, 2008.

Molina, L. T., Madronich, S., Gaffney, J. S., et al.: An

Overview of the MILAGRO Campaign: Mexico City Emissions

and Their Evolution, Atmos. Chem. Phys. Discuss., in prepara-

tion, 2008.

and was determined by a calibration made on 12 March 2006.

Over the course of the campaign the IE calibration values

varied by 20%, but varied by smaller amounts from flight

to flight. 20% is therefore an upper limit to the uncertainty

introduced by not having calibration equipment for the first

three flights. IE values for the rest of the flights were de-

termined from calibrations bracketing the flights. All AMS

data was converted to mass loading at standard temperature

and pressure (STP, 273 K and 1013.25 hPa). Measured AMS

mass was converted to volume using the assumption that

species volume is additive (Eq. 4 of DeCarlo et al. (2004)).

For this conversion the densities of the species were assumed

to be 1.78, 1.72, and 1.52 g cm

−

for ammonium sulfate,

ammonium nitrate, ammonium chloride respectively (Lide,

2007). A density of 1.27 g cm

−

was used for organics, based

on the measured value from Cross et al. (2007).

2.3.2 AMS data processing

AMS Data was processed in the Igor Pro 6.0 Soft-

ware (Wavemetrics Inc.

Lake Oswego, Oregon) us-

ing the standard ToF-AMS Data Analysis toolkit,

(“Squirrel”,

http://cires.colorado.edu/jimenez-group/

ToFAMSResources/ToFSoftware/SquirrelInfo/

The par-

ticle collection efficiency (CE) for the AMS was assumed

to be 0.5 unless there was evidence of acidic aerosol (see

Fig. 2 for comparisons of AMS results with those from

other instruments). In cases of acidic aerosol, the CE was

increased proportional to the mass fraction of sulfuric acid

(CE=1) to ammonium bisulfate (CE=0.5) (Canagaratna et

al., 2007; Quinn et al., 2006; Takegawa et al., 2005). The in-

tegration of total signals for individual ions (“high-resolution

sticks,” e.g. C

and C

m/z

43) from raw mass

spectral data was carried out using the “Pika” module of

Squirrel developed by our group (

http://cires.colorado.edu/

jimenez-group/ToFAMSResources/ToFSoftware/PikaInfo/

which implements the procedures described in DeCarlo et

al. (2006). Sticks for spectra acquired in open (particles

+ air + mass spectrometer background) and closed (only

background) modes were calculated. Particle + air signal

for each ion was determined by the difference of the open

signal and the closed signal. Air contributions to individual

ions (e.g. CO

m/z

44) were subtracted from the total

signal at that ion. Variations in the gas-phase CO

would

produce only a very small effect in the aerosol CO

ion

signal. The typical gas-phase CO

background of

∼

380 ppm

is equivalent to

∼

100 ng m

−

of organic-equivalent (org.-eq.

(Zhang et al., 2005a)) aerosol signal at

m/z

44.

This

signal is subtracted from

∼

2600 ng m

−

of org.-eq.

m/z

aerosol signal in the city and

∼

1000 ng m

−

in regional air.

Variations on the order of 40 ppm for gas-phase CO

would

change its contribution to total

z 44 by +/

−

10 ng m

−

which is within the noise of the measurement. Note that this

correction can be much more important for studies with low

OA concentrations.

www.atmos-chem-phys.net/8/4027/2008/

Atmos. Chem. Phys., 8, 4027–

4048

, 2008

4032

P. F. DeCarlo et al.: Aerosol size and chemistry measurements during MILAGRO

Fig. 1.

High-resolution ion signals at

m/z

29 and 43 for RF 3 (10

March 2006). The insets shows individual 12-s high resolution mass

spectra for

m/z

29 and 43. For

m/z

43, the oxygen-containing frag-

ment dominates the total signal in both the city and non-city case,

however the intensity of the C

ion increases in the city, indi-

cating additional presence of more reduced aerosol species such as

primary traffic and biomass burning emissions.

A correction at

m/z

29 was also necessary, as the 2 most

intense peaks CHO

and

(29.00274 and 29.00318,

respectively), cannot be separated at the resolution of the in-

strument. N

was calculated as a constant fraction of the

signal at

m/z

28 from calibrations with a particle filter,

with the remaining signal assigned to the CHO

ion. Two

other peaks, corresponding to C

and CH

m/z

were also resolved from the above two ions. Figure 1 shows

ion signals at

m/z

29 and 43 during research flight 3. Traces

of the ion signals for this flight are shown and insets show the

mass spectra for these

m/z

for one 12 s data point. The inset

for

m/z

43 shows 2 different mass spectra corresponding to 2

different portions of the flight; one in the city and one in the

outflow. From this it is clear that the C

fragment is

proportionately larger than the C

fragment, indicative of

the larger contribution of oxygenated species (likely SOA)

in the outflow and of more reduced species (such as traffic

exhaust, biomass burning, and less aged SOA) over the city.

2.3.3 Calculation of light scattering from AMS Data

AMS total size distributions from PToF mode were averaged

to a 5 minute timebase for RFs 1, 2, 3, 9, 10 11, and 12. The

total NR size distribution was converted to a number dis-

tribution vs. volume-equivalent diameter (d

dlog

)

, as-

suming spherical particles and using the bulk density of the

aerosol calculated from the chemical composition for each

5 min average, according to the conversions detailed in De-

Carlo et al. (2004). Scattering was calculated for this dis-

tribution using the routine of Bohren and Huffman (1983)

translated into Igor (C. Brock, NOAA, personal communi-

cation, 2006). We used a refractive index of 1.54, which is

the average of the refractive indices for Ammonium Sulfate

(1.55), and organics (1.53) (Hand et al., 2002; Kleinman et

al., 2007) and calculated scattering at 550 nm to compare to

the Nephelometer, which measured submicron scattering at

that wavelength.

2.3.4 Elemental analysis of organic aerosol from AMS data

Elemental analysis was performed on the high-resolution

data following the method described in Aiken et al. (2007b,

2008). Inorganic and air ions were removed so only or-

ganic ions were included in the calculation of elemental ra-

tios. Oxygen-to-carbon atomic ratios of the organic aerosol

(O/C) and hydrogen-to-carbon (H/C) were determined, as

well as the organic mass-to-organic carbon ratio (OM/OC).

The nitrogen-to-carbon ratio was calculated, however the V-

mode does not have enough resolution to reliably quantify

some N-containing ions, so the absolute N/C ratio is not re-

ported, although it is used in the OM/OC calculation. W-

mode data from the T0 supersite in Mexico City show that

organic N/C is typically more than an order of magnitude

smaller than the O/C (Aiken et al., 2008), comparable to

the N/C ratio found on the C-130. Thus errors in the re-

ported OM/OC due to imprecision in the N/C ratios derived

here should be small. Organonitrates and organosulfates can

produce nominally “inorganic” ions in the AMS (e.g. NO

, SO

)

. Ignoring these ions will result in a neg-

ative bias on the O/C and OM/OC. However the analysis of

the stoichiometric neutralization of ammonium vs. nitrate,

sulfate, and chloride (discussed below, see also Zhang et

al. (2007b)) suggests that the contribution of these types of

species (and thus the associated errors in O/C and OM/OC)

is small. To avoid the effect of noise at low OA concentra-

tions, ratios are only reported when the OA mass was larger

than 2

g m

−

3 Results and discussion

3.1 Aerosol Measurement Intercomparisons

Data were converted to the timebase of the slower measure-

ment when comparing with different instruments. The AMS

mass and calculated volume and the Nephelometer submi-

cron scattering were reported at approximately 12 s per data

point, and for comparisons Nephelometer data was interpo-

lated to the AMS timebase. The SMPS data was recorded on

Atmos. Chem. Phys., 8, 4027–

4048

, 2008

www.atmos-chem-phys.net/8/4027/2008/

P. F. DeCarlo et al.: Aerosol size and chemistry measurements during MILAGRO

4033

Fig. 2. (a)

Time series from RF 3 (10 March 2006) of AMS calculated volume, SMPS apparent volume, and PM

light scattering, all on their

native timebases.

(b)

Scatter plot of submicron light scattering as measured at 550 nm by the Nephelometer vs the AMS total mass. Data

is interpolated to the AMS timebase and the majority of data points in the plot are 12 s averages.

(c)

Scatter plot of volume calculated from

the AMS mass and composition-dependent density, against the apparent volume calculated from the SMPS distribution assuming particle

sphericity. The AMS data in this plot has been averaged to the 96 second SMPS timebase. A slope close to 1 and an

of 0.82 indicate

good general agreement.

(d)

Calculated scattering from 5-min AMS size distributions and Mie theory versus average measured scattering

by the Nephelometer over the same period. 27 data points out of 507 total points were eliminated due to low loading resulting in unphysical

size distributions from the AMS. Reported errors in slopes in plots b-d are the 1-

estimate for the fitted variables returned from the fitting

algorithm.

a 96 s timebase, for this comparison AMS data was averaged

to the SMPS timebase. The calculation of scattering from

AMS size distribution data was done on a 5 min time grid as

the size distributions have lower signal to noise ratios than

total mass concentrations and require more time integration.

Figure 2a shows a timeseries from RF 3 (10 March 2006) for

calculated AMS volume, SMPS apparent volume, and mea-

sured PM

Light Scattering. All data are shown in Fig. 2a at

their native sampling resolution, and high correlation among

the measurements is seen. Although only one flight is shown,

other flights show similar agreement. Correlations between

the different measurements for the whole campaign are dis-

cussed in the following sections. Note that data from takeoff

and landing were excluded from these comparisons since the

inlet for the AMS was in a different location on the aircraft

than the Nephelometer and SMPS, and particles generated

by the landing gear and exhaust are sampled differently for

these portions of the flights.

3.1.1 AMS vs. Nephelometer comparison

A direct comparison of Nephelometer Scattering to NR-

submicron aerosol mass for the entire MILAGRO campaign

shows high correlation (see Fig. 2b). The slope of the linear

regression (3.79 m

/g) is equivalent to the Mass Scattering

Efficiency (MSE), and is in good agreement with the range

of dry MSE values of 3.6

1.3 m

/g reported by Shinozuka

et al. (2007). Based on preliminary data, black carbon makes

up 1–3% of the submicron mass during MIRAGE (R. Sub-

ramanian, DMT, personal communication), and would make

up slightly less of the submicron volume due to the higher

density values for black carbon (Park et al., 2004) in rela-

tion to the dominant organic constituents; consequently the

AMS mass or calculated volume is not expected to be sig-

nificantly impacted by the exclusion of black carbon. Some

of the scatter can be explained by differences in the exact

size cuts of both instruments, and of sampling timebase,

www.atmos-chem-phys.net/8/4027/2008/

Atmos. Chem. Phys., 8, 4027–

4048

, 2008

4034

P. F. DeCarlo et al.: Aerosol size and chemistry measurements during MILAGRO

Fig. 3.

Average chemical composition of NR-PM

in different

regions. Mass concentrations are given in

g m

−

STP (273 K,

1 atm). Panel

(d)

has a gray background to differentiate the ground

data from the aircraft data

(a)

Average of whole MIRAGE aircraft

campaign.

(b)

Average of MIRAGE data when flying directly over

Mexico City.

(c)

Average of MIRAGE excluding all points over

Mexico City. (d) Average measured at the ground at the T0 Su-

persite, between the hours of 18:00–24:00 UTC (12:00–18:00 local

time). The time period was chosen for comparison purposes, as this

is when the C-130 typically flew over the city (see text). The area

of each pie is proportional to average mass concentration.

frequency, and different inlet locations on the aircraft. The

AMS sampled in MS mode approximately half of the time,

while the Nephelometer was sampling continuously. Short

plumes (

∼

6 s or less equivalent to approximately 600 m at

100 m/s) could be missed by the AMS or be sampled and be

non-representative of the average mass loading for the pe-

riod in question. The regression of submicron Nephelometer

scattering to AMS calculated volume gave a slope (Volume

Scattering Effciency) of 5.50 m

−

(

=0.86).

3.1.2 AMS and SMPS volume comparison

Figure 2c shows a scatter plot of AMS calculated volume vs.

SMPS apparent volume for all overlapping data during MI-

RAGE. Again, good agreement is found with a slope of 0.98

(

=0.82). Perfect agreement is not expected due to lack of

complete overlap in the measured size ranges, effects on siz-

ing of particle non-sphericity, and since SMPS is measuring

a different particle size at each point in space (time), while

the AMS does an interleaved average as discussed above. At

high mass loadings the AMS is above the fit line, which could

be due to ammonium nitrate evaporation in the SMPS (Gysel

et al., 2007), which was maintained at 40

◦

C to improve RH

control in the tandem differential differential mobility ana-

lyzer (TDMA) that was located in the same enclosure as the

SMPS.

3.1.3 Calculated scattering vs. measured scattering

Scattering calculated from AMS size-resolved composition

was compared to the averaged nephelometer measurements.

Figure 2d shows that agreement is good for most of the

flights. During flights with lower aerosol concentrations

there is more scatter, likely because the AMS size distribu-

tions are noisier for low loading (especially for larger sizes)

and this noise is amplified with the non-linearity in the scat-

tering calculation. Calculated scattering from INTEX-B was

also included in this figure to show the general agreement

with MIRAGE and shows the lower loadings sampled during

the INTEX-B campaign (Dunlea et al., 2008).

3.2 Submicron Aerosol Chemistry over and away from

Mexico City

Bulk NR-PM

composition was averaged for the whole air-

craft campaign as well as for periods when the aircraft

was over Mexico City (defined as the box between 19.814,

19.023 N, and 260.577 and 261.379). Figures 3a–c show this

information. Figure 3d shows the average aerosol composi-

tion as measured at the ground during the entire campaign

at the T0 supersite (19.48973 N,

−

99.1501 W) at the typi-

cal times of day (12:00–18:00 local time) when the C-130

flew over the city. The average for the whole campaign was

used for T0 as there were only a few direct flyovers and both

instruments were not always operating during those times,

and also because the bulk aerosol composition was not highly

variable at T0. The relative concentrations of organic and to-

tal inorganic species are very similar. The fraction of organ-

ics from the MCMA-2003 campaign (Salcedo et al., 2006) is

larger than the fraction reported in this study, but this is due

to the MCMA-2003 study reporting a full day average. A

full day average would increase the mass fraction of organ-

ics because there are higher primary organics and lower am-

monium nitrate concentrations during evenings, nights, and

early mornings.

The spatial and vertical distribution of the species showed

significant differences. Figure 3c shows that away from the

city the total average concentration is

∼

1/4 of that over the

city, confirming the importance of the MCMA as a regional

source, and with an increased fraction of sulfate at the ex-

pense of nitrate. In general, sulfate was more of a regional

component to the aerosol with similar concentrations both

in the MCMA basin and in the regional airmass, while ni-

trate was localized to the city and in the near-outflow. This

is consistent with the conclusions of Salcedo et al. (2006),

who based them on the rapid variability and strong diurnal

cycles of nitrate in the city, versus the much more constant

and slowly varying levels of sulfate, as well as the fact that

OH+NO

could explain the rapid nitrate increases observed

Atmos. Chem. Phys., 8, 4027–

4048

, 2008

www.atmos-chem-phys.net/8/4027/2008/

P. F. DeCarlo et al.: Aerosol size and chemistry measurements during MILAGRO

4035

Fig. 4.

Spatial maps of

(a)

NR-PM

sulfate,

(b)

NR-PM

nitrate,

(c)

NR-PM

chloride, and

(d)

NR-PM

Organics. The dashed box

represents the area designated as “city” in Fig. 3. Several large SO

sources for the region are indicated in part (b). Volcanoes are designated

as red triangles, petrochemical refineries indicated by hollow circles. Both Popocatep

etl and Colima were active in 2007, while the last

known eruption of El Chich

on is 1982 (Smithsonian, 2007). Also note that the colorscale for part (d) is double that of parts (a) and (b), and

the scale of part (c) is 10% of parts (a) and (b).

Table 1.

Results of the Organic Aerosol to CO regression analysis for RFs 1, 2, 3, 9, 10, 11, 12. RFs 1, 2, 3, 12 had a city and regional

component; the rest of the flights are included for completeness, but do not represent the same range of conditions. CO background is given

by the X-axis intercept for the low CO (

200 ppbv) conditions. Slopes are the OA/CO ratio for each flight. The R

value is given to indicate

the quality of fit.

RF 1

RF 2

RF3

RF9

RF10

RF 11

RF12

Units (STP)

4 March

08 March

10 March

23 March

26 March

28 March

29 March

OA/CO (low CO)

g m

−

ppmv

−

77.9

81.1

84.8

62.5

32.5

47.0

80.2

Estimated CO background

ppbv

73.1

0.84

0.53

0.82

0.65

0.25

0.50

0.78

in the mornings, while sulfate formation from OH+SO

was

small compared to the concentrations observed. Kleinman et

al. (2008) analyzed the increase of the sulfate/CO for their

MILAGRO dataset from the G-1 aircraft in the MCMA out-

flow. However, given the lack of correlation of sulfate with

urban pollutants, this analysis may have led to an artificially

inflated rate of growth of sulfate/CO for this species in the

results of Kleinman et al., as CO decreased due to dilution

but sulfate did not due to its more regional character. Spatial

maps of NR-PM

sulfate, nitrate, and chloride are shown in

Fig. 4a-c. Both industrial sources and volcanic sources have

been identified as potentially contributing to the aerosol sul-

fate in the basin (de Foy et al., 2007; Johnson et al., 2006b;

Raga et al., 1999; Salcedo et al., 2006). The maps show the

more regional distribution of sulfate with significant struc-

ture indicating the influence of the large SO

sources from

industrial complexes (e.g. Tula and refineries near Veracruz)

and active volcanoes (e.g. Popocateptl). Emissions of SO

www.atmos-chem-phys.net/8/4027/2008/

Atmos. Chem. Phys., 8, 4027–

4048

, 2008

4036

P. F. DeCarlo et al.: Aerosol size and chemistry measurements during MILAGRO

Fig. 5.

Shows OA and NO

−

aerosol components divide by excess

CO to account for dilution. Part

(a)

shows a spatial map of the

OA/CO ratio. Part

(b)

shows a spatial map of the NR-PM

ni-

trate/CO ratio. Part

(c)

shows the average OA/CO and NO

/CO

ratio as a function of distance from Mexico City. The error bars

are the standard error of the mean for all of the measurements in

that distance bin. Points from the Yucatan portion of RF 9 were

excluded from this analysis due to the influence of local fires from

that region.

from Popocateptl were also monitored directly during MI-

LAGRO, when the volcano was a continous source of SO

and its emissions varied from a minimum of 0.56 Gg/day to

a maximum of 5.97 Gg/day (Grutter et al., 2008). Submi-

cron aerosol nitrate appears to be in the form of ammonium

nitrate, based on the ammonium balance and observed frag-

mentation pattern (see below), while the fractional contribu-

tion of organic nitrates to total nitrate appears to be small.

Nitrate shows a dominant source in the city basin, and cor-

relates strongly with CO in the city basin (

=0.79) but not

with HCN (

=0.10) indicating the dominance of the urban

non-fire source for this species. Mexico City is character-

ized by high concentrations of gas-phase NH

that favor its

co-condensation with nitric acid to form semivolatile am-

monium nitrate (San Martini et al., 2006). In contrast, fire

plumes around Mexico City have a large molar excess of

vs. NH

(Yokelson et al., 2007), which may explain the

limited ammonium nitrate formation. Nitrate shows a very

large fractional reduction in the outflow, most likely due to

evaporation upon dilution with regional air with low HNO

and NH

, or the loss of NH

to sulfuric acid or ammonium

bisulfate. Regional temperature gradients do not appear to

play a role, as the average temperature measured on the C-

130 is approximately 5

◦

C higher above the city box than the

regional air, which would favor evaporation in the city and

condensation away from the city. HNO

also reacts with

dust in the Mexico City area forming mineral nitrates in the

supermicron mode (Fountoukis et al., 2007; Moffet et al.,

2007; Querol et al., 2008) that would not be detected by the

AMS. However the dust spatial and temporal distribution is

highly variable and is unlikely to be the only cause of the

pronounced decrease in nitrate away from the city observed

in this study. Figures 5a–c show the ratio of organic and ni-

trate aerosol to excess CO to remove the effect of dilution in

the cleaner regional air. Excess CO is defined as the CO con-

centration above background, with the background value for

each flight given in Table 1. Additionally data with OA con-

centrations less than 2

g m

−

STP or NO

concentrations

less than 0.2

g m

−

STP were eliminated from the analysis

to reduce the impact of noise and of uncertainty in the CO

background (90% of the points used in the analysis had CO

excess

35 ppbv). Clearly NO

/CO shows a large reduc-

tion with distance from the Mexico City urban area, while

OA/CO does not. Although aerosol nitrate does not com-

pletely disappear, its ratio to CO decreases quickly and has

dropped by nearly a factor of 4 by the time the aircraft is

200 km from the city basin. The OA/CO ratio in the outflow

near the city is about 80

g m

−

STP ppm

−

. This is likely

due to a combination of rapid SOA formation from urban

emissions and mixing of biomass burning OA, and will be

analyzed in more detail in a subsequent publication. This ra-

tio is similar to the value found by Kleinman et al. (2008), for

their study of the near outflow on the DOE G-1 aircraft. It is

also much larger than values of 5–10

g m

−

STP ppm

−

for

urban POA (Aiken et al., 2007a; Zhang et al., 2005c), which

highlights the dominance of SOA in the pollution outflow

of the city, consistent with previous observations in Mexico

City (Kleinman et al., 2008; Volkamer et al., 2006, 2007)

and at several other locations (Zhang et al., 2007a). Both

the asymptotic value of OA/CO and the timescale of SOA

formation of approximately one day are similar to findings

reported for the outflow of the Northeastern US (de Gouw et

al., 2005, 2008; Kleinman et al., 2007; Peltier et al., 2007a),

and of the Po Valley in Italy (Crosier et al., 2007). The fact

that similar asymptotic values are observed despite lower

biogenic emissions being added to anthropogenic pollution

Atmos. Chem. Phys., 8, 4027–

4048

, 2008