United States
Environmental Protection
Agency
Atmospheric Sciences
Research Laboratory
Research Triangle Park NC 27711
Research and Development
EPA/600/S3-86/015 May 1986
SERA Project Summary
Elemental Tracers Applied to
Transport of Aerosol from
Midwest to Northeast
Kenneth A. Rahn, Kenneth R. Wunschel, and Douglas H. Lowenthal
The main objectives of this Cooperative
Agreement were to use a new regional
elemental tracer technique to search for
the signature(s) of midwestem aerosol, to
use the resulting signature(s) to determine
whether midwestern aerosol reaches the
Northeast during summer, and to compare
the transport of perfluorocarbon tracer
gases with that of regional aerosol during
the CAPTEX experiment. Secondary goals
included testing various statistical aspects
of the tracer system and evaluating stabili-
ty of signatures during transport.
From samples in Ohio and Pennsylvania,
two midwestern signatures were found
and later used to show that aerosol is
routinely transported from Midwest to
Northeast in distinct pulses of 2-8 days.
The tracer system was validated
qualitatively during CAPTEX '83, where
pulses of tracer gas from Ohio appearing
in New England were always linked with
pulses of midwestern aerosol.
Factor and cluster analysis were found
to be only moderately useful for determin-
ing regional signatures. In eastern north
America, As/Se, noncrustal V/Se, and
In/Se have greater tracer power than
Sb/Se, Zn/Se, and noncrustal Mn/Se.
Long-term apportionments showed that
most elements in the Northeast come
mostly from the Northeast, that As and In
have large components from Canadian
smelters, and that sulfate and Se have
large components from the Midwest.
Results for sulfate generally agreed within
20% with models and directional studies.
Regional signatures were found to be
stable: during episodic transport from
Midwest to Northeast, all four quantifiable
elemental ratios changed by less than
25% on the average.
Regional least-squares apportionments
were found to be insensitive to a variety
of factors such as weighting scheme,
scales of signatures, random elemental
perturbations, duration of samples, and
local sources of V.
This Project Summary was developed
by EPA's Atmospheric Sciences Research
Laboratory, Research Triangle Park, NC, to
announce key findings of the research pro-
ject that is fully documented in a separate
report of the same title (see Project Report
ordering information at back).
Introduction
Since 1980, the Arctic/Tracer Research
Group at the University of Rhode Island
has been developing an elemental tracer
system to determine the source areas of
pollution aerosol at sites far downwind.
This capability is important for under-
standing long-range transport of aerosols
and gases because at distances of 1000
km and more, traditional meteorological
approaches such as trajectory analysis
and synoptic analysis become very
uncertain.
Elemental signatures have been used for
many years to determine sources of urban
aerosol because emissions from individual
sources or types of sources have very dif-
ferent compositions. By contrast, aerosols
from different regions are much more
similar to one another because most
regions contain similar mixes of sources.
Nevertheless, significant elemental dif-
ferences between regions can be found
and exploited.
By late spring 1983, when this project
was being proposed to EPA, the current
seven-element tracer system (As, Se, Sb,
Zn, In, noncrustal Mn, noncrustal V) had
-------
been created and shown to be generally
promising. Regional signatures, however,
were still preliminary and needed to be
developed further. The principal goals of
this project were to seek to establish the
elemental signature(s) of midwestern
aerosol from samples taken in Penn-
sylvania and Ohio, use them to search for
midwestern aerosol reaching the northeast
during summer, and attempt to validate
the tracer system by comparing the
transport of perfluorocarbon tracer gases
with that of regional aerosol during the
Cross-Appalachian Tracer Experiment
(CAPTEX) of September-October 1983.
Secondary goals included testing various
statistical aspects of the tracer system
and evaluating the stability of elemental
signatures during transport.
All these goals were met. Two distinct
midwestern signatures were found, and
were used to show that aerosol is
transported routinely from the Midwest to
the Northeast in pulses of a few days in
duration. These pulses are related to par-
ticular configurations of large-scale
meteorology, are most frequent in sum-
mer, and often come simultaneously to
widely separated parts of the Northeast.
During CAPTEX, pulses of tracer gas in
Rhode Island and Vermont stemming from
releases in Dayton, OH were always
associated with pulses of midwestern
aerosol; i.e., the elemental tracer system
was validated qualitatively. High-volume
aerosol samples were found satisfactory
for our system of regional tracing in
eastern North America: six of seven
elemental ratios differed by less than 50%
between fine and total aerosol; during
episodic transport from Midwest to
Northeast, none of the four quantifiable
ratios changed by more than 25% on the
average.
The geographical boundaries of the
source regions are still only partially
delimited. The sharpness of their bound-
aries is not yet known but is suspected to
be fairly broad.
When necessary, signatures were de-
termined from as few as 4-5 samples
chosen carefully from much larger sets.
When available, as many as 48 samples
were used for signatures, however. Al-
though the smaller numbers of samples
appear to yield reliable signatures, the
greater numbers are clearly preferable.
Typical uncertainties of elemental ratios
in regional signatures are 40%.
The least-squares regional apportion-
ment procedure was shown to have little
or no sensitivity to weighting scheme,
scales of signatures, random perturbations
in samples or signatures, typical duration
of samples, strong local sources of V, and
number of samples used to apportion
sulfate. The apportionments did depend
moderately on the combination of
elements in the signatures and strongly on
having representative signatures from
each major region.
Long-term regional apportionments of
aerosol at Narragansett and Underhill
showed that most elements come mostly
from the Northeast, that As and In have
large components from the Canadian
smelters, and that sulfate and Se have
large components from the Midwest.
Absolute accuracy of regional coeffi-
cients cannot be fully verified at present,
because no independent standard exists.
Samples during CAPTEX confirmed the
midwestern coefficients semi-quantita-
tively: strong maxima of midwestern
aerosol in the Northeast were detected
each time a pulse of perfluorocarbon gas
from Dayton was measured. Where they
overlap, long-term regional apportion-
ments of sulfate in the Northeast agree
within 20% to results from transport
models and other field studies.
The Two Midwestern Regional
Signatures
The Lower Midwest (LMW) signature
was determined from a several-day pollu-
tion episode at Allegheny Mountain, PA
during August 1983. LMW represents
aerosol from the Ohio River Valley and
southward. The same signature was also
seen in the next southwestern episode at
Allegheny Mountain, and was later sensed
repeatedly at two sampling sites in Ohio.
Because LMW is strongly enriched in Se
relative to the other elements, it seems to
be derived primarily from coal combustion.
The Upper Midwest (UMW) signature
was determined from samples taken near
McArthur and Hartville, OH during winter
and spring 1984. It represents aerosol
from the Ohio River Valley and northward.
Compared to LMW, most elemental ratios
to Se are markedly higher in UMW. Thus,
UMW seems to be coal-based but with a
significant component from heavy in-
dustry. This agrees with the relatively
greater industrialization of the Upper
Midwest.
Transport of Aerosol from Midwest
to Northeast
The two midwestern signatures were
used to detect aerosol transported to one
or more sites in the Northeast according
to the following procedure:
(1) At each site, a series of daily aerosol
samples was taken and analyzed for
the seven tracer elements. From the
seven elemental concentrations,
each sample was apportioned into
its five regional contributions, two
from the Northeast, two from the
Midwest, and one from the Cana-
dian smelters. The result was five
"regional coefficients."
(2) Generalized northeastern and
midwestern regional coefficients
were then formed for each sample
by summing the two northeastern
coefficients and the two midwes-
tern coefficients. The summed coef-
ficients are proportional to the
amount of primary pollution aerosol
from the two major source regions.
(3) The generalized regional coefficients
and the Canadian smelter coef-
ficient were plotted for the series of
samples. The resulting figure
showed directly how the relative in-
fluences of the source areas varied
over the time series.
Such plots for New England invariably
have the same basic features:
(1) Northeastern aerosol is nearly al-
ways present and has relatively
steady concentrations (the north-
eastern "foreground").
(2) Midwestern aerosol is much more
variable. It usually appears as strong
pulses of 2-4 days in duration. Be-
tween pulses, midwestern coeffi-
cients are very low. Pulses come
roughly weekly.
(3) Canadian smelter aerosol appears
much less often than midwestern or
northeastern aerosol. It comes in
spikes of 1-3 days in duration which
are usually near-simultaneous in at
least Rhode Island and Vermont.
Most variations in regional coefficient
can be directly related to changes in large-
scale meteorology. When air comes to
New England from the north or east, or
stagnates, northeastern aerosol domi-
nates. When air flows in an organized
manner from the interior, a midwestern
component is superimposed on the north-
eastern aerosol. When air comes strongly
from the northwest, usually behind a cold
front, a Canadian smelter component is
often seen.
Deriving Regional Signatures by
Pattern-Recognition Techniques
Regional signatures are normally de-
termined by a combination of chemical
-------
and meteorological techniques. These
procedures are still evolving and have not
yet become standardized or fully objective.
Scientific judgment is still required to
select the most appropriate samples. To
remedy this, alternative multivariate ap-
proaches such as factor and cluster
analysis are being explored.
Cluster analysis of 100 samples from
two sites in Ohio gave a "UMW" signature
that did not differ from the original UMW
in any systematic way, but an "LMW"
with higher ratios than the original. Fac-
tor analysis of the same samples gave
another "UMW" and "LMW," again with
the "UMW" matching the original more
closely than the "LMW" did. In addition,
factor analysis of samples from Underbill,
VT provided estimates of a northeastern
and a transported midwestern signature.
The UMW and LMW signatures derived
from factor analysis of Ohio samples fit
the Vermont samples as well as the
original UMW and LMW did, but the UMW
and LMW derived from cluster analysis did
not work as well. The Vermont samples
were fit much more poorly by the
"northeastern" and "midwestern"
signatures derived from factor analysis of
the Vermont samples themselves.
Concerning the feasibility of using
pattern-recognition techniques to derive
regional signatures of pollution aerosol, it
thus seems that
(1) factor analysis within the source
area works better than cluster
analysis;
(2) factor analysis shows promise, at
least for determining the major
signature within a strong source
region. The indeterminancy and
arbitrary nature of the factor solu-
tion must be kept in mind, however.
Considering that there are at least some
circumstances under which factor analysis
gives reasonable signatures, its limits
should be explored in more detail.
Long-Term Regional
Apportionments in the Northeast
Extensive data from four seasonal
experiments during 1982-83 at Narra-
gansett, Rl and Underbill, VT have been
used to apportion sulfate and the seven
tracer elements among northeastern,
midwestern, and Canadian smelter
sources. The results showed that
(1) most of the noncrustal V, noncrustal
Mn, Zn, and Sb (50-98%) came
from the Northeast;
(2) As and In had significant compo-
nents (25-80%) from the Midwest
and/or the Canadian smelters, par-
ticularly during winter;
(3) the influence of the smelters was
greater at Underbill than at
Narragansett;
(4) annually, Se and sulfate at both sites
came roughly equally from the
Northeast and the Midwest;
(5) only 3-5% of the annual-average
sulfate came from the smelters;
(6) the apportionments at Narragansett
varied little with season, whereas
those at Underbill were much more
northeastern in winter than in
summer.
Seasonal average concentrations of the
tracer elements were generally accounted
for to within 10-15%, whereas individual
daily concentrations were accounted for
to 20-50%. Sulfate was accounted for to
the 10% level for seasonal averages but
to only 50-60% for individual daily
samples.
In summary, most trace elements in
northeastern aerosol actually came from
the Northeast. Only when an element was
strongly enriched in a distant source, such
as As and In in the Canadian smelters and
Se and S in the Midwest, did the distant
source contribute significantly relative to
the northeastern foreground.
Qualitative Validation of the
Elemental Tracer System During
CAPTEX '83
During CAPTEX '83, perfluorocarbon
tracer gas was released from Dayton, OH
and Sudbury, Ontario under meteoro-
logical conditions expected to bring it to
the Northeast. Because signatures of
aerosol from both Ohio and the Sudbury
Basin were being developed, it seemed
.natural to see whether aerosol from these
sources would be found to reach the
Northeast simultaneously with the tracer
gas. Tracer gas from Dayton should be ac-
companied by one or both of the mid-
western signatures; tracer gas from
Sudbury should be accompanied by the
Canadian smelter signature.
Daily samples from Rhode Island and
Vermont during CAPTEX (mid-September
through October 1983) were analyzed for
sulfate and trace elements and resolved
into regional components as discussed
above. The abundance of aerosol from the
region of release, as measured at each site
1-2 days after the release, were used to
predict whether the tracer gas should have
been sensed at the sites. Of the 12 cases
for which gas could be compared with
aerosol, 10 agreed clearly. In one of the
two other cases, midwestern aerosol was
found without tracer gas, but this type of
result was anticipated and did not con-
stitute a problem. In the other case,
smelter aerosol was absent, but moderate
tracer gas from Sudbury was found. On
this day, the strongest part of the tracer
plume passed rapidly and briefly north of
our sampling site This probably explained
why smelter aerosol was not observed.
Thus, qualitative predictions for tracer
gas during CAPTEX were accurate in 11
of 12 cases, and the other case can be ex-
plained reasonably.
Stability of Regional Signatures
The stability of regional signatures dur-
ing transport was examined in two ways.
First, the relative particle sizes of the tracer
elements were determined repeatedly at
Narragansett, to address the question of
how much elemental ratios would change
if all coarse aerosol and no fine aerosol
were removed during transport. Second,
actual changes in ratios were examined by
comparing signatures of aerosol in 34
midwestern episodes at Underbill, VT to
the two original midwestern signatures.
To evaluate the relative particle sizes of
the tracer elements and sulfate, 12 pairs
of simultaneous samples of total aerosol
and submicron aerosol were taken at
Narragansett during May-June 1984. The
X/Se ratios in total aerosol relative to the
fine component ((X/Se) total/(X/Se) sub-
micron), the factor by which the elemen-
tal ratio would decrease during transport
if all the coarse aerosol and none of the
fine aerosol were removed, were less than
1.5 for all ratios except noncrustal Mn/Se,
for which it was 2.4. By contrast, elemen-
tal ratios vary by factors of 6-170 over the
five North American signatures. Thus, the
maximum changes expected during trans-
port due to particle-size effects alone are
far below the ranges already found for
North American signatures.
During actual transport, elemental ratios
seem to be even more stable than this,
however. For 34 episodes of midwestern
aerosol observed 1000 km downwind in
Underbill, VT, average values of the As/Se,
Zn/Se, and noncrustal Mn/Se ratios were
within 25% of the mean of the mid-
western ratios, and the average value of
Sb/Se, after correcting for a 24% contribu-
tion from northeastern sources, fell within
10% of the midwestern mean. Thus, the
-------
best empirical evidence showed that all
four quantifiable elemental ratios changed
by less than 25% on the mean during
transport of 1000 km. Consequently, it
appears that regional signatures may be
applied with confidence to receptor
samples taken far from the source under
episodic conditions, even when the
aerosol samples are not size-segregated.
Statistical Tests of the
Regional Tracer System
A systematic series of tests of various
statistical aspects of the elemental tracer
system has confirmed its general stabili-
ty and the general validity of the results
derived from it:
(1) The number of samples used to
derive a signature from a specific
distribution has little effect on the
final value. From sets of 48 and 12
samples used to construct two of
the signatures, four random subsets
of five samples each gave X/Se
ratios that were statistically iden-
tical to the original signatures.
(2) According to a procedure for
evaluating colinearity, or "near-
dependency" of signatures, the five
signatures for North America are not
colinear.
(3) According to several tests, using dif-
ferent weighting schemes for the
least-squares apportionments does
not affect the resulting regional
coefficients systematically. In cases
where coefficients within the
Midwest and the Northeast change,
the total contribution from each
region remains stable.
(4) Scaling regional signatures upward
or downward by one to two orders
of magnitude does not affect un-
constrained apportionments (where
negative coefficients are allowed),
but may affect constrained appor-
tionments slightly.
(5) The number of samples in a series
used to apportion sulfate is not
critical, as long as it is above 20 or
so.
(6) By perturbing a Vermont aerosol
sample and associated signatures
randomly and repeatedly and appor-
tioning with the perturbed data, it
was shown that the perturbed coef-
ficients overlapped the unperturbed
coefficients; i.e., the tracer system
is stable with respect to long-term
effects of random variations.
(7) Within limits, length of sample does
not influence apportionments:
results from semi-weekly samples
are nearly identical to apportion-
ments of daily samples over the
same period.
(8) Strong local sources of individual
tracer elements affect apportion-
ments minimally: doubling or halv-
ing the concentrations of individual
elements changes the regional
coefficients by factors several times
smaller.
(9) Regional coefficients are robust
relative to the mix of elements in
the signatures: although removing
elements changes regional coeffi-
cients somewhat, it does not alter
the basic qualitative impressions
about the regional makeup of the
sample.
(10) Midwestern and northeastern
signatures are not interchangeable
in practice: mixtures need one sig-
nature from each area in order to fit
them well, and aerosols from a
single region must have a signature
from the region to be fit properly.
(11) Bias introduced by constraining
regional coefficients to non-negative
values is insignificant, i.e., com-
parable to or less than the overall
uncertainties of apportionment.
(12) Typical uncertainties (precisions) of
the 1 -2 largest regional coefficients
in a given sample are 25-40%;
uncertainties in minor coefficients
can reach or exceed 100%.
Recommendations for Future
Work
Although this study has helped to
demonstrate that regional elemental
tracers can now be used reliably to deter-
mine source areas of signature elements
and sulfate in aerosol of eastern North
America, the tracer system has not
reached its potential, nor have its limits
been probed. Much can still be done to ex-
pand and improve it. The following areas
of research seem most important and
promising:
(1) Add elements to the system, so that
more signatures can be apportioned
simultaneously and existing signa-
tures discriminated more reliably.
(2) Investigate the extent to which ex-
isting regional signatures would be
refined by more samples from more
sites. Several sampling sites per
major region, each with at least 100
samples, would be ideal.
(3) Determine additional signatures in
North America, especially in the
southern and western portions of
the Midwest. This will help resolve
the true effect of the more distant
regions of the Midwest on air quali-
ty in the Northeast.
(4) Apply the tracer technique to addi-
tional sites in New Hampshire,
Maine, and New York State. Only
then will the conclusions about
sources of pollution aerosol be
generally valid for "the Northeast."
(5) This program should be continued
until climatologically valid conclu-
sions about source areas can be
drawn for one or more sites in the
northeast. This kind of conclusion
requires data from at least five
years.
(6) The utility of regional elemental
signatures for sampling periods well
under one day should be investi-
gated. Situations under which the
sources of northeastern aerosol
may vary in significantly less than
24 hours include coastal storms
which pass rapidly over New
England and bring a cold sector, a
warm sector, and a cold sector in
less than a day, as well as frontal
conditions in general.
(7) Multivariate statistical techniques
for determining signatures should
be pursued until their prospects are
clarified.
(8) Synthetic data should be used to
help determine the limits of detec-
tion of regional signatures in in-
dividual samples.
(9) The relative amounts of signature
elements in submicron and total
aerosol should be determined in the
Midwest in the same fashion as
already done in the Northeast. This
will help evaluate potential changes
in signatures during transport from
Midwest to Northeast.
-------
(10) Actual changes in regional signa-
tures during transport should be fur-
ther studied, under both episodic
and nonepisodic conditions if
possible.
(11) The extent of the association be-
tween regional signatures and
large-scale meteorology should be
investigated in detail, for chemistry
and meteorology are inherently
complementary and will be used
together in future studies.
(12) The statistics of the regional tracer
system need to be examined in
greater detail than has been possi-
ble to data Two examples of impor-
tant topics are the extent of bias
introduced by constraining the re-
gional coefficients to non-negative
values and the mathematical basis,
strengths, limitations, and possible
bias of the regression procedure for
sulfate.
(13) In order to draw reliable conclu-
sions about the origin of trace con-
taminants in deposition, the tracer
technique must be extended to
precipitation. This will be one of
the most important applications of
the tracer technique.
•&U. S. GOVERNMENT PRINTING OFFICE:!986/646 116/20825
-------
Kenneth A. Rahn, Kenneth R. Wunschel. and Douglas H. Lowenthal are with the
University of Rhode Island, Narragansett, Rl 02882-1197.
Thomas G. Dzubay is the EPA Project Officer (see below).
The complete report, entitled "Elemental Tracers Applied to Transport of Aerosol
from Midwest to Northeast," (Order No. PB 86-168 812/AS; Cost: $16.95,
subject to change} will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Atmospheric Sciences Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for Private Use $300
EPA/600/S3-86/015
0000329 PS
U & ENVIR PROTECTION AGENCY
REGION 5 LIBRARY
230 S DEARBORN STREET
CHICAGO IL 60604
------- |