SCERP Project Number: A-9
Principal Investigator: Peter Buseck
Arizona State University
OBJECTIVES OF THE PROJECT:
The primary objectives of the research were to characterize the particles in the ambient aerosol in U.S. towns along the Arizona-Sonora border and to distinguish background from anthropogenic sources. Particles were assigned to categories on the basis of their compositions. Information gained from the characterization of the particles was used to determine probable sources of the different particle types. Another goal was to determine the relative amounts of particles originating from various anthropogenic and natural sources in Arizona and Mexico.
ACTIVITIES/RESULTS/MILESTONES:
SAMPLE COLLECTION
Daily aerosol samples were collected simultaneously from June 24 to
June 29, 1991 in the Arizona border towns of Douglas and Nogales as well
as in Tombstone, slightly further from the border. Due to equipment problems,
samples were not collected in Douglas on June 24, 25 or 26. The particles
were collected on stacked filter assemblies. The coarse fraction, consisting
of particles mostly larger than 2 Êm was collected on 8.0-Êm
pore-size, Nuclepore, polycarbonate filters. The fine fraction, consisting
of mostly particles smaller than 2 Êm, was collected on 0.2-Êm
pore-size, Nuclepore, polycarbonate filters. Low volume pumps were operated
at flow rates between 10 and 14 L/min for periods from 8 p.m. to 8 a.m.
daily.
SAMPLE PREPARATION
A section approximately 1 cm2 was cut from the center of each
filter. Each section was then mounted onto a 1.3-cm pin-type carbon-foil
stub using colloidal carbon. The sample was then coated with approximately
20 nm of carbon to improve conductivity.
SAMPLE ANALYSIS
The samples were chemically and physically analyzed using a JEOL model
JXA8600 electron microprobe automated with TRACOR- Northern TN 5500 and
TN 5600 systems. The instrumental settings were: accelerating voltage 20
KeV, beam current 500 pa, magnification 2000x (fine fraction) and 800x
(coarse fraction), and x-ray counting time was 100 sec/particle. Approximately
1000 particles were analyzed on both the coarse and fine fractions from
each sample. A total of about 32,000 particles were analyzed in this study.
The data collected for each particle included size, shape parameters, and
an x-ray spectrum from 0 - 20 KeV. The following elements were searched
for in the spectrum: Ca,S,Na,Cl,Fe,Si,K,Al,Mg,P,Ti,Cr,Zn,Cd,Nm,Cu,Ga,V,Pb,As,Ba,
Ge, and Se. The minimum detectable particle size is around 0.1 Êm.
PARTICLE TYPES
Non-hierarchical cluster analysis was used to find distinct particle
types that represent the variability of the individual particle compositions.
The maximum distance between a particle and a particle-type centroid for
a particle to be assigned to a particle-type was 200 in 26- dimensional
vector space. The particle types listed in figure 1 give the average compositions
of all of the particles assigned to that particle type. Particle type X
contains all of the particles which do not have x-ray peaks for any of
the elements that we looked for. These are interpreted to consist mostly
of soot particles. Particle types 1-24 have compositions that appear to
reflect a crustal nature. Particle types 25-30 are interpreted to have
a marine origin. Particle types 31-55 ar assumed to come from varied anthropogenic
activities as well as from secondary reaction in the troposhere. Crustal
particles are the largest contributors to the aerosol, comprising 44% of
the fine fraction and 73% of the coarse fraction. Sea-salt particles make
up 10% of both the fine and coarse fractions. The fine fraction is almost
39% anthropogenic or secondary whereas only 10% of the coarse fraction
is of this type. Both fractions had 7% of their particles unassigned to
particle types due to their unique compositions.
PRINCIPAL COMPONENTS
Discriminant analysis was used to assign the particles to particle types
in order to obtain the occupation for each cluster. Correlation analysis
was used to construct the particle occupation versus particle occupation
correlation matrix. The correlation matrix was then subjected to principal
components analysis. Nine principal components were identified (Figure
2), which account for 87% of the variability in the data set. Particle
types with eignvectors > 0.5000 for a particular principal component were
assigned to it; those with smaller coefficients were eliminated. A combination
of positive and negative correlation coefficients within a single principal
component indicates that the particle types represented by different signs
occur with opposite trends. PC1 contains all of the alumino-silicate and
Fe-rich particle types. These are interpreted as originating from the weathering
of soils, travel on unpaved roads, construction, and possible fly ash.
PC2 contains of the Ca-rich silicates except particle type 21. The composition
of these particle types is similar to that of cement. Their probable origin
is the lime processing plant near Agua Prieta, Mexico, there is also the
possibility of these particles originating from any site with cement production
or large scale construction using cement. PC3 probably represents marine
aerosol transported from the Gulf of California. PC4 has been identified
as emissions from either the copper smelter at Cananea or Nacozari, Mexico.
Assigning sources to the particle types in the other principal components
is much less straightforward and would require further research.
The FY91 SCERP-supported phase of this project:: A-9
The FY94 SCERP-supported phase of this project: AQ94-OF-6
Last updated 7/1/99