EPA-600/4-77-034
June 1977
Environmental Monitoring Series
THE LOS ANGELES CATALYST STUDY SYMPOSIUM
Environmental Monitoring and Support Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL HEALTH EFFECTS RE-
SEARCH series. This series describes projects and studies relating to the toler-
ances of man for unhealthful substances or conditions. This work is generally
assessed from a medical viewpoint, including physiological or psychological
studies. In addition to toxicology and other medical specialities, study areas in-
clude biomedical instrumentation and health research techniques utilizing ani-
mals — but always with intended application to human health measures.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
SYMPOSIUM
FOR DISCUSSION OF
THE LOS ANGELES CATALYST STUDY
Proceedings
Hosted by
The U.S. Environmental Protection Agency
April 12 and 13, 1977
Environmental Monitoring and Support Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
DISCLAIMER
This report has been reviewed by the Environmental Monitoring and Support
Laboratoryf U.S. Environmental Protection Agency, and approved for publication.
Approval does not signify that the contents necessarily reflect the views and
policies of the U.S. Environmental Protection Agency, nor does mention of
trade names or commercial products constitute endorsement or recommendation
for use.
In general, the texts of the papers included in this report have been
reproduced in the form submitted by the authors.
11
-------
FOREWORD
Careful assessment of the relative risk of new man-made environmental
hazards is necessary for the establishment of sound regulatory policies. In
1974, the Environmental Protection Agency (EPA) established an interdiscipli-
nary research program to determine the environmental risk of emissions from
vehicles equipped with oxidation catalytic control devices. One major area of
emphasis in this program has been the operation of a long term roadside moni-
toring study to quantify ambient levels of catalyst vehicle emission products.
The Environmental Monitoring and Support Laboratory, Research Triangle
Park, North Carolina, has conducted the Los Angeles Catalyst Study since June,
1974, before the introduction of the 1975 model year catalyst equipped automo-
biles. Through a combination of careful site selection and study design,
best-technology sampling equipment, experienced personnel, and a carefully
created quality assurance program, an extensive and reliable data base has
been generated for a multitude of ambient pollutants relating to automotive
emissions. The Los Angeles Catalyst Study site has also proved to be an
excellent location for evaluating the performance parameters of the newer
methods applied to the measurement of catalyst emission products.
Information presented in this Symposium Proceedings document highlights
the various aspects of the Study operations. The critical area of trend
assessment is addressed along with a detailed discussion of quality assurance
measures used to establish the reliability of trend data. Overall, the Los
Angeles Catalyst Study is providing a sound data base upon which EPA can make
a responsible appraisal of the impact of the catalyst control technology oil
the ambient environment.
Thomas R. Mauser
Acting Director
Environmental Monitoring and Support
Laboratory
Research Triangle Park, North Carolina
111
-------
CONTENTS
FOREWORD iii
ACKNOWLEDGMENTS viii
INTRODUCTION 1
ORIGINS OF THE LOS ANGELES CATALYST STUDY 3
A.C. Trakowski
LOS ANGELES CATALYST STUDY DESIGN 7
Franz J. Burmann
PREVIOUS SULFATE STUDIES IN THE LOS ANGELES BASIN 21
Edward P. Parry*
Franz J. Burmann
DISCUSSION 41
Wm. Pierson
SELECTION RATIONALE FOR ANALYTICAL METHODS FOR THE LOS
ANGELES CATALYST STUDY 43
Richard J. Thompson
DETERMINATION OF LOW CONCENTRATIONS OF PARTICULATE NH*
WITH THE SPECIFIC ION ELECTRODE 51
Lloyd S. Shepard
George Colovos
Allen M. Miles
Charles E. Rodes
A COMPARISON OF THE THORIN AND MODIFIED METHYLTHYMOL BLUE
METHOD FOR THE DETERMINATION OF MICRO-AMOUNTS OF SULFATE 61
George Colovos
Edward P. Parry
Allen M. Miles
Martha Panesar
Charles E. Rodes
*In cases of multiple authors, the first listed person presented the paper at
the Symposium.
-------
X-RAY FLUORESCENCE ANALYSIS OF LACS AEROSOLS 83
jR.D. Giauque
R.B. Garrett
L.Y. Goda
George Colovos
Edward Parry
Charles E. Rodes
Franz J. Burmann
FUELS SURVEILLANCE IN SOUTHERN CALIFORNIA ANALYTICAL
RESULTS AND METHODOLOGY 97
Robert H. Jungers
MEGAVOLUME SAMPLER (MARK I) FOR COLLECTING RESPIRABLE
PARTICULATES AT THE LACS SITE Ill
Ralph I. Mitchell
W.M. Henry
COMPOUND IDENTIFICATION OF LOS ANGELES CATALYST STUDY SAMPLES 127
Edward P- Parry
Leo E. Topol
M.D. Lind
A.B. Barker
R.M. Housley
Franz J. Burmann
COMPARISON OF LEAD RECOVERY WITH AND WITHOUT THE USE OF A
LOW TEMPERATURE ASHER BEFORE EXTRACTION 141
Lloyd S. Shepard
April Price
Charles E. Rodes
DETERMINATION OF PERCENTAGE OF DIESEL TRUCKS AND CATALYST
EQUIPPED CARS 147
Edward P. Parry
Raymond A. Meyer
Charles E. Rodes
LABORATORY QUALITY CONTROL 157
George Colovos
Edward P. Parry
Lloyd S. Shepard
Charles E. Rodes
EXTERNAL LABORATORY QUALITY ASSURANCE FOR THE LOS ANGELES
CATALYST STUDY 175
John C. Puzak
Thomas Clark
vi
-------
THE QUALITY ASSURANCE PROGRAM EMPLOYED DURING SAMPLING 193
Charles E. Rodes
Bobby E. Edmonds
Malcolm C. Wilkins
PRECISION OF LACS SAMPLING AND ANALYTICAL METHODS 207
Gary F. Evans
COMPARISON OF THE DEGRADATION OF AMMONIUM ION ON HIGH
VOLUME GLASS FIBER FILTERS AND ON MEMBRANE FILTERS 265
George Colovos
Edward P. Parry
Charles E. Rodes
SULFATE CONCENTRATIONS AT TWO LOS ANGELES FREEWAYS 281
A.H. Bockian
George Tsou
Dennis Gibbons
Robert Reynolds
SUMMARY OF LACS CONTINUOUS DATA . 301
Gary F. Evans
Charles E. Rodes
NITRIC OXIDE, NITROGEN DIOXIDE, AND OZONE INTERRELATIONSHIPS
ACROSS THE FREEWAY 343
Robert K. Fankhauser
SUMMARY OF LACS INTEGRATED POLLUTANT DATA 359
Charles E. Rodes
Gary F. Evans
STATISTICAL ANALYSIS OF THE LOS ANGELES CATALYST STUDY DATA-
RATIONALE AND FINDINGS 415
George C. Tiao
Steven C. Hillmer
DISCUSSION 461
William Pierson
LOS ANGELES CATALYST STUDY-CONCLUSIONS AND RECOMMENDATIONS 463
Thomas R. Hauser
vii
-------
ACKNOWLEDGMENTS
The cooperation of the Rockwell Air Monitoring Center, under contract
to EPA, in the LACS operation and analysis is gratefully acknowledged. The
assistance of the California Department -of Transportation and the Wadsworth
Veterans Administration Hospital are also acknowledged for assistance in the
selection and preparation of the study site.
Conference arrangements were made by Northrop Services, Inc., under
Contract Number 68-02-2566.
viii
-------
INTRODUCTION
A symposium was held on April 12 and 13, 1977 at Raleigh, North Carolina
for discussion of the Los Angeles Catalyst Study. Welcoming remarks were
presented by Dr. Thomas R. Mauser, Acting Director of the Environmental
Monitoring and Support Laboratory, EPA, Research Triangle Park, North Carolina,
Dr. Mauser, who chaired the Symposium, introduced the first speaker, Albert C.
Trakowski, Deputy Assistant Administrator, Office of Monitoring and Technical
Support. Mr. Trakowski informally profiled the background details of the
Study. His comments and the formal papers which formed the basis of the
meetings are presented here in order of occurrence.
-------
ORIGINS OF THE LOS ANGELES CATALYST STUDY
A. C. Trakowski
Environmental Protection Agency
Washington, D. C.
It seems almost customary that meetings held where the knowledge is - are
opened with kind words by someone from that place where the knowledge "ain't".
On this occasion I'm very pleased to be that person, to have the opportunity
to meet you and at least in part participate in the review of this project
that had its origin just about four years ago. Other than tell you how nice
it is to be here to see what can be told from the data that has been in
collection for nearly three years, it might serve us all well to go over the
events that led into this project. It was, you know, set up for some clear
purposes. And those purposes, in turn, had their causative stimulations.
I'm not quite sure who or where it was that concern was first expressed
about sulfate emissions from automotive catalyst converters. It dates back
to mid or late 1972. But, in the early months of 1973 it was generally
recognized, at least in the technological community, that sulfuric acid emis-
sions from catalyst equipped cars could possibly cause a harmful effect on
both..the environment and on the health of people. In November 1973, the
Administrator of EPA, Russell Train, presented his views to the U.S. Senate
Public Works Committee concerning the introduction of catalytic converters on
1975 model year cars. Testimony was also presented at that Committee hearing
by various automobile manufacturers, the petroleum industry, environmentalists,
various consultants and other special interest groups. During his testimony
to the Committee, Mr. Train summarized the events leading up to his decision
to make no changes in the previously promulgated emissions standards for 1975
model year cars and to take no action that would inhibit the use of catalysts
on 1975 cars. His testimony included his recognition of possible risks from
the use of catalytic converters, but he expressed the belief that the use of
catalysts on cars held much promise for the reduction of atmospheric pollution.
He explained to the Senate Committee how a catalyst functions to reduce
the levels of hydrocarbon and carbon monoxide emissions. Mr. Train went on to
tell how the sulfur in the fuel is converted to sulfate by the catalyst and
that the amount of sulfate emitted from the cars to the atmosphere could
possibly be at a level that would present a new and different danger to the
health of exposed people. He discussed how EPA had projected the quantity of
possible sulfate emissions from catalyst equipped cars and its anticipated
effect on air quality. However, his bottom line was an acknowledgement of the
limitations of EPA's knowledge on the entire subject of the effect on the
environment of sulfate, platinum, and other emissions from catalyst equipped
cars.
-------
In making his decision to allow the go-ahead for use of catalysts, Mr.
Train was forced to consider several very difficult issues which demonstrated
the complexities that had to be faced in EPA's efforts to improve the quality
of the air environment and protect the nation's health. Some of these issues
were:
First, a comparison of the effects on public health with or without
catalysts on automobiles;
Second, methods to insure protection from sulfate emissions if further
studies indicated such actions were necessary;
Third, effects of the catalyst on the momentum of the nation's total
air pollution control effort;
Fourth, impact on energy conservation;- and
Last, effects on the complete range of technological options available to
control automotive emissions, including catalysts.
Addressing these issues before the Senate Committee in November 1973, Mr.
Train concluded that if EPA postponed the requirements to control automotive
emissions for a year, the nation's overall momentum in achieving health-
related air quality standards would suffer. A postponement would have required
that some transporation control plans would have had to have been made more
stringent.
Also, Mr. Train observed that EPA prohibition would very likely have
made the industry dependent on other technologies, some of which were promising,
but none of which were ready for mass production. The further we advance new
technology, the more we become aware of possible unintended side effects. It
was recognized that as other technologies came closer to massive application,
that they may likewise exhibit undesirable side effects.
Additionally, there was the point that the use of catalysts enabled the
use of more efficient automobile engines, which offered potential for reducing
gasoline consumption, thereby having a favorable effect on energy conservation.
Of course, his decision was based on the need for comprehensive protec-
tion of public health. Mr. Train stated that in the time then to come, EPA
would be able to learn much more about how to measure sulfate emissions from
cars, their impact on public health, and how to control these emissions.
So—it was at that time, in November 1973, that Mr. Train made the
decision to take no action that would prevent the use of catalysts on 1975
cars. But he simultaneously announced that EPA would carry out several tasks
to develop the information necessary to make future judgments concerning the
effects of his decision. These intended tasks were:
First, EPA would accelerate work on developing a test procedure of assured
reliability for measuring emissions of sulfates from cars;
-------
Second, EPA would consider all feasible measures for controlling sulfate
emissions from cars, including a regulatory approach for controlling sulfur
content in gasoline/ and other alternatives. In addition^ EPA would encourage
the petroleum industry to voluntarily make available unleaded gasoline of low
sulfur content for the 1975 model vehicles;
Third, EPA would improve its ability to estimate the impact of sulfate
and other emissions from cars on air quality and public health; and
Last, EPA would improve its understanding of the atmospheric chemistry
involved and establish an appropriate air monitoring program, to determine the
effects of catalyst equipped vehicles.
To carry out the monitoring program promised by this fourth task, a number
of options were analyzed on where to locate it, and a decision was made in
March 1974 to carry out the work in California. Subsequently, the Los Angeles
Catalyst Study was initiated in June 1974. The prime objectives of this study
were fourfold:
o To determine the impact of catalyst equipped cars on air quality for
automobile related pollutants.
o To establish the ambient levels of carbon monoxide, lead, and sulfur
dioxide in order to determine the surrogate sulfuric acid ambient
levels.
o To provide an active data base to compare the effects of pre- and
post-catalyst automobiles.
o To evaluate and improve field methods for detection and measurement
of air quality changes relatable to catalyst emissions.
Now we are here, nearly three years later, to see how we've done and
what we've discovered. Has the use of the catalytic converter affected the
ambient level of automobile pollutants - especially along-side heavily
travelled roads? As they said in the old days - "stay tuned to find out."
But before revealing the solution to our mystery, let's go into how the
sleuthing was planned and carried out. Our next speaker, Mr. Franz Burmann
will discuss the LACS study design, pre-study considerations, and essentially
take up where I leave off.
Thank you very much. You have my very best wishes for a most successful
symposium.
-------
LOS ANGELES CATALYST STUDY DESIGN
Franz J. Burmann
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
The Los Angeles Catalyst Study came into being because of public concern
over the possible adverse impact of emissions from catalyst-equipped auto-
mobiles on the ambient air. The study was to be a comprehensive long-term
investigation of the ambient levels of sulfuric acid aerosol, sulfates, and
other potential catalyst emission products in areas adjacent to a heavily
traveled freeway in Los Angeles. Based on this general objective, the
Environmental Monitoring and Support Laboratory proceeded to design a road-
side study to monitor pollutant levels on both sides of a freeway for a pro-
jected 3-year period. In order to obtain data prior to the introduction of
catalyst-equipped vehicles the study had to be designed and initiated in a
very short time. After delineating specific study objectives, the selections
of the pollutants and related measurements to be monitored were made. The
required quality control measures were determined and consideration given to
the need to maintain the desired level of data precision and interlaboratory
comparability. In that the ambient level measurement technology for sulfuric
acid was not available, a variety of sulfur compounds and related auto-
emission products were selected to provide the basic on which to estimate
sulfuric acid concentrations. In order to minimize interfering effects from
extraneous pollutants, high background levels, and unfavorable meteorology, an
intensive effort was made to select a site in Los Angeles that met a prescribed
list of siting requirements. The site selected on the San Diego Freeway in
West Los Angeles provided predominant cross-freeway wind flow, such that
identical upwind and downwind measurements could be made to assess the contri-
bution from the freeway. A short-term study was subsequently performed which
verified that during favorable meteorology periods the upwind "background"
site did provide measurements unbiased by the freeway emissions. Overall the
study design has been shown to be excellent for the assessment of the original
objectives.
STUDY DESIGN
To explain the Los Angeles Catalyst Study (LACS) design, it is desirable
to review briefly some of the estimates of increases (incremental,) sulfate
(SC%) exposure which were expected to result from the introduction of the
catalyst. The Environmental Protection Agency (EPA) estimated human exposures
from locally increased concentrations by considering ambient air guality and
-------
human activity patterns (Table I). The three independent techniques used to
estimate exposure were:
• Use of dispersion models developed for emissions of carbon monoxide
(CO) ,
• Use of lead (Pb) surrogate data/ i.e., the Pb/SO^ ratio of emission
rates to project ambient SOf levels,
• Use of the carboxyhemoglobin surrogate method. This method avoids
assumptions about traffic or human activity patterns but depends,
among other assumptions, on an assumed relationship between CO and
suspended particulate 30% emission rates.
Table I shows the estimates for short term incremental exposures to
suspected particulate sulfates and sulfuric acid emitted from vehicles
equipped with oxidation catalysts obtained by the CO dispersion model and
the Pb surrogate for various receptors under selected meteorological con-
ditions. Concentration estimates are shown for four different receptor loca-
tions. The concentration on automobile passenger or driver experience is
considered to be similar to that existing at a location in an outer lane.
The receptor labeled "Pedestrian Near Expressway" is at a location 3 meters
away from the outermost downwind lane. The third receptor, "Nearby Somes"
is assumed to be at a location 50 meters from the nearest downwind lane and
the receptor labeled "Homes about 5 blocks from Expressway: is assumed to be
500 meters downwind of the highway.
Estimates were made for two types of meteorological conditions. The
"Normal Meteorology" can be characterized by light to moderate wind speed of
4 meters per second and a slightly unstable atmosphere. The "Worse"
Meteorology" consists of a barely perceptable drift of the wind at 0.5 meters
per second and a slightly stable atmosphere. Two wind directions were con-
sidered for the estimates: (a) wind directly across the highway and (b) a
wind direction at an angle which resulted in the highest calculated concen-
tration for a given receptor. Based on the dispersion model used, the worst
angle varies from nearly parallel to the highway for near receptors, to
larger angles for more distant receptors.
In addition to those listed in the table, two further assumptions are
made in applying the CO dispersion model: (a) acid aerosols and fine
particulate sulfates disperse as does CO, and (b) a one-hour commuter trip
of 30 miles on a ten lane expressway with traffic flow of 20,000 vehicles
per hour is typical. The Pb surrogate estimates might be affected by the
settling of Pb particles near the roadside which would tend to underestimate
the predicted sulfate concentration. Conversely, if re-entrainment of settled
Pb particles occurs, overestimation of incremental sulfate is possible. How-
ever, as can be seen in TaJble I the projected sulfate concentrations by the
method are in the general range of values predicted by the CO dispersion
model. Data from the carboxyhemoglobin surrogate method (not shown) were also
in this general range.
-------
TABLE I. PREDICTED SHORT-TERM INCREMENTAL EXPOSURES TO SUSPENDED PARTICULATE SULFATES AND
SULFURIC ACID EMITTED FROM VEHICLES EQUIPPED WITH OXIDATION CATALYSTS*
vo
Incremental Exposure (microerrams per cubic meter)
Normal Meteorology Worst Meteorology
Human
Exposure
Automobile
Passengers
Pedestrians
Pedestrians
Nearby Homes
Homes About
5 Blocks
Away
Highway
Source
Expressway
Near Expressway
Urban Street**
Canyon
Expressway
Expressway
Method of
Estimation
Dispersion Model
Lead Surrogate***
Dispersion Model
Lead Surrogate***
Dispersion Model
Lead Surrogate***
Dispersion Model
Dispersion Model
Wind Directly
Across Highway
2
4 to 5
2
3 to 7
4 to 6
2 to 9
1
0.3
Wind at Wind Directly
Worst Angle Across Highway
7 22
5 20
_ __
2 19
0.3 5
Wind at
Worst Angle
124
88
__
33
6
*Emission assumptions: 25% of vehicle miles traveled on busiest multilane expressway (20,000 vehicles
per hour at 30 mph by vehicles) equipped with oxidizing catalysts emitting .05 grams of sulfuric acid
or sulfate per vehicle mile: Ohter 75% of vehicle miles involve no emissions of sulfuric acid and
suspended sulfate. Peak hourly values are tabled.
**Assumes 10,000 vehicles per hour instead of 20,000. Actual observed carbon monoxide values in medium
sized suburban shopping center indicate that peak hourly exposures to 4.5 to 5.3 yg/ra3 of carbon
monoxide and by inference 1 to 2 vg/suspended sulfates and acid aerosols would occur.
***Computed by multiplying observed lead levels by adjustments for differences in emissions per vehicle
mile and percentage of vehicle miles emitting the particulate of interest: (observed lead level) X
(.25) or .18 times observed.
-------
For pedestrians near expressways the incremental exposure estimates
ranged from 2 to 88 micrograms per cubic meter (iig/m^) of particulate sulfates
and sulfuric acid depending on the meteorology. The same calculation indicated
20 ygr/m3 for the worst meteorological conditions and the wind directly across
the highway. Even with normal meteorology the estimates ranged from 2 to 7
Ugr/m3. These predictions were, because of the feared health consequences, of
great concern to EPA. With these projections as a background, the Environ-
ment Monitoring Brach (EMB) of the Environmental Monitoring and Support
Laboratory (EMSL) initiated the design of the LACS.
OBJECTIVES
The primary objective of the study is to determine the impact of catalyst
equipped cars on ambient air quality. To achieve this goal it was decided to
attempt to obtain a data base for freeway contribution of emission products by
monitoring the ambient air adjacent to a heavily travelled highway before and
after the 1975 model year catalyst equipped automobiles were introduced. In
addition the study's objectives are:
• to determine the ambient levels of CO, Pb, and sulfur dioxide (802)
in order to determine the surrogate (predicted) sulfuric acid (H2SOi4)
ambient levels since technology for measuring ambient H2SO^ concen-
trations was not available, and
• to evaluate an improved field methodology for detection of air
quality changes attributable to catalyst emissions.
Since the 1974 model year catalyst-equipped automobiles appeared in California
in September or October 1974, it was important that the background sulfate
study be initiated as soon as possible to obtain the background data base.
Initially, it was estimated that after the appearance of these cars, at least
3 years of monitoring would be required in order to quantify the possible
increase in sulfate attributable to catalytic converters.
SITE SELECTION
The California area in and around Los Angeles was the primary area for
consideration for several reasons. Because of the more stringent California
standards a greater number of vehicles were to be equipped with catalytic
converters than in the rest of the nation. The average sulfur content in
California gasoline was approximately twice as high as compared to the
national average. Traffic density on some of the freeways in Los Angeles
approached 200,000 cars/day. The generally low background SC>2 levels in
Southern California reduced the possibility of formation of high sulfate
background levels.
With only 3 months to set up a full scale field study, the obstacles
encountered were formidable. No specific sulfur aerosol samplers were
available. Suitable sites were not known nor had been selected, and no
sulfate data related to the Los Angeles area had been compiled. A planning
meeting was held in late February 1974 with several pertinent groups in-
cluding EPA's Regional Office (RO) in San Francisco, the California Air
10
-------
.Resources Board (CARB) , the Los Angeles Air Polution Control District (LAAPCD),
the California Department of Transporation, and the California State Air Indus-
trial Hygiene Laboratory (AIHL). Early in March 1974 extensive discussions
were held at the National Environmental Research Center at Research Triangle
Park, North Carolina (NERC/RTP), to determine if the site requirements of the
Environmental Monitoring and Support Laboratory (EMSL), the Physics and
Chemistry Laboratory (PCL), and the Meteorology Laboratory (ML) could be com-
bined. However, it was determined that the 16 stations that were desired by
the ML could not be accommodated at the locations under consideration. These
particular locations were sought by the ML in order to validate the CO model
which had been used as a surrogate to predict sulfate/sulfuric acid levels. A
site near Camp Pendleton, located on Highway 5, which could accommodate all of
the sites was deemed unacceptable for determination of incremental sulfate
increases early enough because of the low vehicle count (fewer than 50,000
vehicles/day). The general consensus was that both needs —• early detection
of catalytic attrition products and validation of the CO diffusion model in
correlation with any sulfate increases — were not compatible.
At the end of March 1974 a site selection trip was conducted with the
assistance of Rockwell International, under contract to the Environmental
Protection Agency (EPA), and consultants of CARB. Criteria considered in re-
lation to an acceptable site were:
0 Predominant perpendicular winds
0 Probability of low oxides of sulfur (SO ) contribution from
stationary sources
0 Traffic count approaching 200,000 cars/day
0 Adequate area for site installation
0 A freeway elevation near grade
0 Adequate electrical power availability
0 Simple lease arrangement and accessibility
Initially the so-called 42 mile loop (Figure 1) of Los Angeles was inves-
tigate. This loop is formed by the San Diego Freeway, the Harbor Freeway,
and the Santa Monica Freeway. The loop was of interest since it has traffic
count sensors embedded in its freeway ever 3.2 kilometers. On-ramp and off-
ramp traffic counts are also recorded and the loop is equipped with traffic
control devices. This loop was travelled and carefully searched; however,
suitable sites were not found. It was felt that anything south of the vicin-
ity of Manchester Avenue would not be suitable because of the influence of the
airport and the many point sources such as refineries. A major refinery,
located just south of the airport, is probably one of the worst sulfur point
sources in the whole Southern California area.
On the San Diego Freeway south of the Harbor Freeway junction there
was a suitable site which was sufficiently open to accommodate the stations,
11
-------
Ventura Frw>
I I
KILOMETERS
Figure 1. Freeway system of Los Angeles area.
12
-------
Jbut it was rejected because of the high background levels from the many in-
dustrial point sources in the area. Also investigated was the area east of
the Santa Ana Freeway and the 605 Freeway (San Gabriel River Freeway), and
sites were observed which were sufficiently open. However, the wind field
was not very well defined in this area and because of the high smog levels in
the summer months, the sulfate background was expected to be too high to de-
termine increases due to the catalytic converters.
After aerial surveillance three potential site locations were indentified
The first two sites were located on the Ventura Freeway west of 405. Both of
these sites posed problems concerning accessibility and lack of available
power. The third potential site was located on the San Diego Freeway near the
Wilshire Boulevard off-ramp (Figure 2). On the west side of the San Diego
Freeway are the grounds of the Veterans Administration (VA) Hospital and on
the east side is the Veterans Cemetry. The sites on both sides of the free-
way are about 2 meters below freeway grade. The west side is an open field
with power readily available. On the east side, within the fence marking the
freeway right of way, a trailer location was available. About 40 meters from
the freeway edge due east is the VA cemetery. Wilshire Boulevard, which runs
east-west, is about one-half mile south of the location. After consideration
of various other possibilities, this last site was chosen for the study.
A graphical overhead view, Figure 3, shows the location of the four plat-
forms that were installed for placement of sampling equipment. Sites A and B
were designated "upwind" sites. The two "downwind"sites are labeled C and D.
These designations, of course, are only correct for the periods when favorable
meteorological conditions exist, i.e., when a Seabreeze is prevailing. How-
ever, the stations are indeed upwindand downwind very consistently during the
daytime hours in the summer months. Therefore, we were able to obtain data
on freeway contributions by subtracting background measurements from those
made at the downwind site. Although these characterizations do not apply for
24-hour integrated sampling, the designation has been kept to maintain con-
sistency. Data collected at Sites B and D provided an estimate of spatial
distribution and yielded some information of freeway contribution during per-
iods of prevailing landbreeze conditions. Figure 4 shows a graphic cross
sectional view of the approximately symmetrical placement of the four stations.
The distance from the edge of the freeway of Sites A and D is approximately
30 meters. The ambient air intake of the samplers is about 1 meter above
the freeway surface.
METHODOLOGY
Because CO, Pb, SO^r and sulfates were the primary pollutants of interest,
and methodology was readily available, it was decided to initially monitor
these pollutants. The samplers in Figure 4 shown above the dotted lines were
on site at the beginning of the study. The principal means of collection for
sulfates and Pb were the standard high-volume (hi-vol) sampler and the 102 mm
membrane sampler. For CO, nondispersive infrared analyzers were used. For
SC>2/ the collection procedure of the federal reference method was used. Hi-
vol and membrane samplers were programmed to operate initially from 6 to 10
&.m. in the morning, from 3 to 7 p.m., and/or a continuous 24-hour cycle.
The morning and evening 4-hour intervals were selected because of the assumed
13
-------
*>.
Figure 2. LACS site location in Los Angeles.
-------
PRIVATE PROPERTY
VA HOSPITAL PROPERTY
SCALE 1 IN.= 40FT
FENCE
FREEWAY *>'•'
RIGHT-OF-WAY
4
FREEWAY
RIGHT-OF-WAY
4
SAN DIEGO FREEWAY 8 LANES
"!',-
SEPULVEDA BLVD. 4 LANES
4 FT X 8 FT
PLATFORM
4 FT x 8 FT:
PLATFORM
8 FT X 14 FT SHELTER
"A" SITE
PREVAILING WIND
3 FT X 8 FT
PLATFORM
D" SITE
-------
p
TC
10
1
flET
JWER
r-V*T* *T« PREVAILING WIND »
m
Y WW^v/ \/\/ V/A/I
SAMPLER INLETS
1m ABOVE SAN DIEGO
X FREEWAY SURFACE \ FREEWAY SURFACE v
M_ 0 1 n ) ^ ff
IP3I >^" 2m ABOVE GRADE
|«— 8m-*. 1
• 30m b
\ SEPULVEDABLVD W
• — 8m — »|
* 3(lm »
SITE A
METEOROLOGICAL DATA
24-HR HI-VOL
4-HR HI VOL
4-HR MEMBRANE
CASCADEIMPACTOR
24-HR S02 BUBBLER
CO ANALYZER
SITEB
24-HR HI-VOL
4-HR HI-VOL
4-HR MEMBRANE
BATTELLEIMPACTOR
TOTAL SULFUR ANALYZER
NO/N02 ANALYZER
03 ANALYZER
DICHOTOMOUS SAMPLER
SITEC
24-HR HI-VOL
4-HR HI-VOL
4-HR MEMBRANE
CASCADE IMPACTOR
24-HR S02 BUBBLER
CO ANALYZER
BATTELLE IMPACTOR
TOTAL SULFUR ANALYZER
NO/N02 ANALYZER
03 ANALYZER
DICHOTOMOUS SAMPLER
TRAFFIC COUNT/SPEED
SITED
METEOROLOGICAL DATA
24-HR HI-VOL
4-HR HI-VOL
24-HR MEMBRANE
Figure 4. LACS study site composition and elevation.
-------
peak traffic density periods. For comparison purposes, membrane samplers
were operated simultaneously with the hi-vols. Andersen cascade impactors
were operated on a 24-hour basis to collect information on particle size.
In order to obtain long term SO^ trend data, SC>2 impingers were also operated
on a 24-hour cycle. The arrangement of the samplers was so designed to deter-
mine the contribution of sulfate/pollutants from the freeway. Particular
attention was given to obtaining valid meteorological data and two meteoro-
logical stations were installed. The upwind station was instrumented at the
10 meter level. To determine the validity of these measurements, a second
station at Site D was operated at approximately the 2 meter level. After it
was ascertained that the wind direction data from both stations agreed, the
recordings from the downwind site were discontinued.
The samplers listed below the dotted line in Figure 4 were added at Sites
A and C during the study period. The Battelle Impactor is operated on an ex-
perimental basis to obtain large samples of particulates (>1 gm/day) for pos-
sible compound identification work. It collects three size fractions at a
flowrate of approximately 25 m^/min. The two analyzers utilize well proven
detection principles: total sulfur is continuously monitored by flame photo-
metric detection, and chemiluminescence is employed by the WO/WO2 and 0$
instruments. The dichotomous samplers were early models with three channels—
fine, coarse, and total. However, only samples from the fine channel were
acceptable. The traffic count/speed system became operational in 1976. Both
the car count and the speed in 5 intervals is recorded for each of the 8 lanes.
To obtain an estimate of the number of catalyst equipped cars present in
the Los Angeles area, sales records and registration are surveyed on a quart-
erly basis. A satisfactory system tg directly identify catalyst equipped cars
passing the study site is highly desirable but has been impossible to imple-
ment. Since it was know that gasoline in California contains higher concen-
trations of sulfur (S) than the rest of the nation, fuel samples are being
collected and analyzed for S and Pb. The integrated samples being obtained
are analyzed by traditional gravimetric, wet chemistry and x-ray fluorescence
(XRF) procedures. The data routinely obtained by this surveillance program
are shown in Table II.
QUALITY ASSURANCE
To assure the high quality data required for the study, a rigid form-
alized quality assurance procedure was initiated at the beginning of the
study. Integrating samplers are subjected to monthly calibrations and ex-
ternal audit flow checks. Approximately 1% of the integrated samples are
analyzed by the contractor and our own laboratory. _Blind samples^ i.e.,
filter media spiked with known concentrations of SOi+, nitrate (NO$) , and Pb
are sent routinely for analysis to the contractor. The continuous analyzers
received daily zero checks, weekly span checks and monthly complete calibra-
tions. Auxiliary equipment such as mass flowmeters and calibrators are
checked monthly. As with all phases of the LACS, the quality assurance pro-
gram has undergone changes during the course of the study. The split and
blind sample phase was reduced substantially after the contractor laboratory
demonstrated the capability of achieving the desired results. However, since
site operation was transferred to contractor personnel in the summer of 1976
17
-------
TABLE II. LACS DATA BEING OBTAINEP - 1/76
Sampling
Method
Sampling
Interval(s)
Variables Measured
Hi-Vol
Membrane
Dichotomous
Cascade
3-7 pm &
24 hour
3-7 pm &
24 hour
24 hour
24 hour
TSP, Pb, NO , NH , SO
TSP, NO , SO , Al, Br, Ca,
Clr Fe, Pe, Si, Zn, S
TSP, NH , SO , Strong Acidity,
Al, Br, Ca, Cl, Fe, Pe, Si, Zn,
S (3 fraction/site-day)
TSP each stage (6 fractions/
site-date), SO , NO , Al, Br, Ca,
Cl, Fe, Pe, Si, Zn, S on backup
filter
Bubler
Gaseous
Analyzers
Meteorology
24 hour
Hourly
Hourly
SO.
CO, NO/NO , Total S, 0
Wind Direction/speed, ambient
Temperature/dewpoint
Traffic
5 min
Quarterly
Count by lane and speed interval
Mix by sales and registration data
18
-------
the external audit portion has been increased. In addition to ongoing moni-
toring, numerous studies in the field and in the laboratory have been conduct-
ed to determine critical parameters and to evaluate equipment, and where ne-
cessary to initiate changes.
REPORTS AVAILABLE
Figure 5 shows graphically the reports available and their relationship
to estimated precentages of cars equipped with catalytic converters as ob-
tained from registration information. These precentages do not represent
catalyst vehicle miles travelled (VMT). The estimated VMT is approximately
25% to 30%, which was one of the assumptions used in the original EPA estimate
for incremental sulfate exposure (Table I). All listed reports pertaining
to the period through the close of 1976 are available from the Office of the
Director, EMSL. It is hoped that the scientific community will avail them-
self of these reports and assist those involved in the LACS in making full
use of the data base, which as will be shown by the subsequent papers, is
extremely well suited for assessing effects on the ambient air quality by
mobile source emissions.
19
-------
Estimated
Percent
Cata!yst
Cars
15--
10--
Pre-catalyst
period
STUDY
OK-LINE
6/74
5--
introduction of
catalyst cars
Post-catalyst period
BACKGROUND
SUMMARY
REPORT
9/74
SUMMARY
REPORT
12/74
SUMMARY
REPORT
3/75
SUMMARY
REPORT
6/ 75
SUMMARY
REPORT
12/75
SUMMARY
REPORT
6/76
Figure 5. LACS program schedule related to precent catalyst cars.
-------
PREVIOUS SULFATE STUDIES IN THE LOS ANGELES BASIN
Edward P. Parry
Rockwell International
Newbury Park, California
and
Franz J. Burmann
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
During the initial phases of the Los Angeles Catalyst Study, a review
of sulfate studies and other pertinent data representing previous work in the
Los Angeles Basin was undertaken. This review was prepared at the Air and
Industrial Hygiene Laboratory under subcontract to the Rockwell Air Monitoring
Center. The results of this review are presented, together with other perti-
nent information concerning sulfate levels.
The distribution of 50£ sources in relationship to the Los Angeles mete-
orology and the Los Angeles Catalyst Study site is discussed. The climatology,
general wind patterns, and effect of seasons on general sulfate levels are
presented. Maximum two-hour average sulfate levels of 50 to 70 pg/m3 have been
reported .
A total of 14 air quality studies of the Los Angeles area is identified
in the review. Of these, six seemed pertinent to detailed discussion. These
studies are presented and discussed as background to the Los Angeles Catalyst
Study .
INTRODUCTION
During the early phases of the Los Angeles Catalyst Study (LACS), it
became apparent that information on the climatology of the Los Angeles Basin,
sulfate levels and their variability in space and time, reliability of pre-
vious sulfate measurements, and location of emission sources were most import-
ant to the future understanding and interpretation of the data. This in-
formation would be important in establishing background sulfate levels and in
generally evaluating the site chosen for the study. It would also reveal any
previous studies having objectives similar to those of the Los Angeles
Catalyst Study and make available the possible conclusions of these studies.
21
-------
Further, it would help in providing the necessary climatological character-
ization of the area.
To prepare this review, Rockwell Air Monitoring Center subcontracted
with the California Air and Industrial Hygiene Laboratory (AIHL) to search all
available sources, select pertinent studies, and evaluate the data and methods
employed. By the time this review was underway, initial site selection for
the LACS had been completed. The detailed evaluation was, therefore, limited
to those studies that included sulfate/SC>2/sulfur measurements performed in
areas of high traffic density and geographically related to the LACS site. In
addition, only reasonably recent studies were considered, i.e. after 1965,
since the methods of measurement of sulfate and sulfur dioxide (SO^) have
significantly changed and improved with time, and the general reliability of
the older data might be questioned. A total of 14 air quality studies was
identified in the Los Angeles Basin. Of these studies, six were considered
pertinent to the LACS and were discussed in some detail. The six studies are
listed in Table I. In addition to an analysis of these studies, some informa-
tion on the emissions from Los Angeles Basin sources and the climatology of
the area was also given. In the present presentation, a synopsis of this
information is described. The complete report is available from The Environ-
mental Protection Agency (EPA).
SULFUR DIOXIDE EMISSIONS
In 1971, an average of 250 tons of SC>2 per day was discharged into the
Los Angeles atmosphere (1). Table II provides a breakdown of these emissions
between the various stationary and mobile sources within Los Angeles County.
Of the 250 tons/day total emissions, 210 tons (84%) resulted from
stationary sources, principally chemical and metallurgical processing. Sulfur
dioxide emissions from power plants are usually of lesser significance but are
variable and depend on the availability of natural gas and low sulfur (<0.5%)
fuel oil. In Los Angeles Countyf during the April 15 to November 15 summer
months, all utilities are required to burn natural gas when available. Dur-
ing the winter (November 16 to April 14), these facilities are permitted to
switch to fuel oil when residential demands exceed gas supplies. During
periods of severe natural gas curtailment, daily emissions up to 885 tons of
SO? have been experienced (1). Any interruption in the availability of low
sulfur fuel and natural gas would, therefore, result in increases in SO2 and
particulate emissions.
The major stationary sources are located principally in the industri-
alized southwestern, south coastal, and the southern portions of the central
Los Angeles Basin and the San Fernando Valley areas.
In 1971, approximately four million vehicles moving throughout the Los
Angeles Basin contributed about 14% of the total SO2 emissions, while aircraft
contributed about 2% (2). Automotive emissions are generally constant through-
out the year with minor seasonal variations, but the diurnal and weekend traffic
densities vary markedly.
Sulfur dioxide is formed in the combustion of gasoline from various
22
-------
TABLE I. LIST OF DATA SOURCES
1. The California State Air Resources Board (ARB) sponsored
Aerosol Characterization Experiment (ACHEX) - Rockwell
International (G.M. Hidy, Principal Investigator), 1972-73.
2. The ARB sponsored Freeway Aerosol Study (FAS) - University
of California, Davis (T.A. Cahill and P.J. Feeney), 1972.
3. The Los Angeles County Air Pollution Control District
(LAAPCD), Los Angeles, California, 1965-73.
4. The Federal National Aerometric Surveillance Network (NASN)
- EPA, 1965-73..
5. The Federal Community Health and Environmental Surveillance
System (CHESS) - EPA, 1972-73.
6. The National Institute of Environmental Health Sciences and
EPA sponsored 5O2 to sul-fate Conversion Study (SDC) -
California Institute of Technology, Pasadena, California
(P.T. Roberts and S.K. Friedlander), 1973.
23
-------
TABLE II. MAJOR SOURCES OF SULFUR DIOXIDE IN
LOS ANGELES COUNTY (1971)
Source
Chemical /Metal 1 urgical
Petroleum
Power Plants
Total Stationary
Motor Vehicles
Aircraft
Total Transportation
Grand Total
Average
Daily Emission
Tons % of Total
120 48
55 22
35 14
210 84
35 14
5 _2
40 16
250 100
Source: Lemke, et al.. Reference (21
-------
orgranosulfur compounds not removed during refining. Motor fuel (gasoline) can
contain up to 0.14% sulfur by weight (3.5 gm/gal). Typical sulfur levels in
gasoline sold in California during 1973 are shown in Table III and range from
0.005 to 0.72%. With conventional (non-catalyst equipped) automobiles,
virtually all the sulfur in the fuel emerges as SO2; less than 2% is converted
to sulfate before emerging from the exhaust pipe (3,4).
CLIMATOLOGY OF THE LOS ANGELES BASIN
In the Los Angeles Basin during the summer months, subsidence inversions
reduce the mixing volume into which the pollutants are dispersed. The average
inversion height is about 2000 ft. (600m). In winter, surface inversions are
formed by radiative cooling. Although winter inversions occur about three
times as often as summer inversions, these are generally weaker and dissipate
more quickly during the day. Winter inversions result in higher levels of
primary pollutants, such as CO and 50%, probably because of the lower inversion
heights.
During periods of extremely low inversions (surface to 500 ft.), which
can occur up to 50% of the time during August through December (5) , the
pollutants from stationary source stacks can be injected into the atmosphere
above the inversion. In such cases, the levels of these pollutants may be
abnormally low in the vicinity of the sources and result in abnormally high
values some distance downwind when the pollutants reach a region with good
vertical mixing.
The daytime onshore wind flow pattern for the Los Angeles Basin under
low-level inversion conditions is shown in Figure 1*. At night and early
morning, the wind typically reverses direction to blow offshore due to
nocturnal radiative cooling (Figure 2). In Figures 1 and 2, the small arrows
with the numbers before and after indicate wind variability and average
speed. The number behind the arrow indicates the variability class. This is
approached in 22%° wind sectors in a 360° wind circle. If 75% of the wind
directions fall within a single 22%° wind sector, the wind variability is
described as zero. If an additional 22%° wind sector on each side of the
original 22%° wind sector (total now 67%°) is required to account for 75% of
the wind directions, the variability is described as 1. The addition of
another 22%° sector on each side to include 75% is described as 2 (112%° = 2;
157%° = 3; 202%° = 4; 247%° = 5; 292%° = 6, etc.). Double wind arrows imply
that the wind occurred nearly all the time from either of the two directions.
The average wind speed in miles per hour is also given for each location and
is the number in front of the arrow. The onshore wind flow pattern period
is longer in the summertime than in the winter. During the summer months
especially, the prevailing daytime onshore winds sweep the 50% from the major
stationary sources generally east inland through the basin.
In Figures 1 and 2, the location of the LACS site is given. As can be
seen, favorable winds blow perpendicular to the freeway for extended periods,
especially during the summer months. Because these winds blow in from the
ocean and no significant sources are in the immediate vicinity, it was
*Figures 1 and 2 obtained from Reference (6).
25
-------
TABLE III. SULFUR CONTENT OF GASOLINE IN CALIFORNIA
Gasoline
Sulfur Content - Wt %
Average Range
Premium
Southern California
Northern California
Regular
Southern California
Northern California
Unleaded
Southern California
Northern California
0.034
0.011
0.046
0.033
0.020
0.014
0.107-0.072
0.006-0.018
0.013-0.068
0.005-0.058
0.007-0.080
Source: Shelton E; Motor Gasoline: Summer 1973. Petroleum
Products Survey No. 83, U.S. Department of Interior,
Bureau of Mines, Bartlesville, Oklahoma (January 1974).
26
-------
POMONA -^5.5
X" LOS ANGELES
14.9
MOST FREQUENT WIND DIRECTION
AVERAGE
WS (MPH)
V IS A VARIABLE DIRECTION
Figure 1. Los Angeles Basin - Low level, inversion mean wind flow
regime for August at 1200 PST.
-------
MOST FREQUENT WIND DIRECTION
X •
WD
AVERAGE
WS (MPH)
VARIABILITY
V IS A VARIABLE DIRECTION
Figure 2. Los Angeles Basin - Low level, inversion mean wind flow
regime for September at 0600 PST.
-------
expected that the background sulfate levels would be low. In addition,
this portion of the freeway has a high traffic density (^200,000 cars/day).
These factors make the LACS site a favorable location for the study.
As indicated in the introduction, results of six Los Angeles studies
were deemed pertinent and were discussed in some detail in the AIHL review.
In the discussion below, brief summaries of these studies, including the study
sponsor, the objective, a brief description of the results, and an assessment
of the measurement methods are presented.
AEROSOL CHARACTERIZATION EXPERIMENT (ACHEX)
This study was sponsored by the California Air Resources Board and was
conducted in the summer and early fall of 1972. An additional study was
done at the same time of year in 1973. The purpose of the study was to
investigate the role of chemical transformation and its relationship to
source emissions in the evolution of atmospheric aerosol.
Simultaneous obervations were taken of aerosol properties, gaseous
pollutant concentration, and meteorological properties using a mobile labor-
atory and several fixed stations. Aerosols were collected to obtain as
short as two and as long as 24-hour samples for size distribution and chemical
composition analysis.
Electron spectroscopy for chemical analysis (ESCA) was used in 1972 as a
means of estimating changes in sulfate levels because a sufficiently sensitive
method to analyze sulfate in the two-hour filter sample aqueous extracts was
not available. ESCA is an analytical technique that can differentiate an
"oxidized" sulfur (sulfite and sulfate) from a "reduced" sulfur (sulfur and
sulfide). In 1973, a procedure was developed and sulfate data were obtained
on two-hour samples by aqueous extraction of the filter and subsequent colori-
metric analysis. The pollutant gases monitored included SO^, hydrogen sulfide
(H^S) , ozone (O$) , nitric oxide (NO) , oxides of nitrogen (NOX) , methane (CHi+) ,
total hydrocarbons and ethylene and acetylene.
The conclusions from analyzing the 1972 ACHEX data demonstrate that lead
(Pb) and bromide (Br) can be correlated well with automobile traffic, but
there is no correlation of sulfate levels as measured by "oxidized sulfur."
Reduced sulfur, however, correlates well with automobile traffic and is sug-
gested as a traffic indicator. Since two-hour aerosol samples were taken,
diurnal plots of gaseous and particulate species were obtained. Figure 3
shows the diurnal variation of several pollutants during sampling at the
Harbor Freeway very near downtown Los Angeles. The increased concentrations
of lead and bromide during times of high traffic density are obvious. The
maximum oxidized sulfur values of 3.5 ygr/m3 would correspond to a sulfate
level of 11 pgr/m3.
The conclusions reached from the 1973 ACHEX data are:
1. In most cases, sulfate and SO2 diurnal peaks occurred at the same
time for both SO2 source-enriched sites and pollutant receptor sites.
In no instance did peak SO2 and sulfate levels correspond to peak Pb
concentrations.
29
-------
1.00
2.50
2.00
1.50
1.00
0.50
0.00
1 I 1 " 1
— I | | j ± -imiAjr.L-.rl .. f ^ * J * * ' ' . I r"^T^^ i j
212223241 2345678 9101112131415161718192021
TIME OF DAY (PST)
Figure 3. Diurnal patterns for pollutants sampled at the Harbor
Freeway on September 19-20, 1972 (ACHEX).
30
-------
2. The ratio of small participate sulfate (<0.5 mm) to total particulate
sulfate for measurement Immediately downwind of major SO sources was
small and reasonably constant throughout the day. However, at sites
further downwind, the above ratio showed diurnal variations. During
the times when total sulfate was a maximum, the small particulate
sulfate was large and in one case was 60% of the total.
3. The apparent conversion of gas phase to particulate sulfur was seen
to generally increase with distance downwind from SO^ sources areas.
4. Finally, particulate sulfate levels from 50 to 70 yg/zn3 were observed
as maximum two-hour average values at downtown Los Angeles, West
Covina, and Dominguez Hills.
Sampling techniques were validated by comparing results obtained on
different samplers. For example, the diurnal total filter (Gelman GA-1
membrane filter) data for sulfate during Phase II composited over 24-hours
showed good agreement with sulfate values for 24-hour high-volume (hi-vol)
samplers (Whatman 41). During the ACHEX experiment, the 24-hour hi-vol filters
were analyzed for sulfate using the turbidimetric method (Intrasociety
Committee Method 42401-01-697). The two-hour samples were analyzed for sulfate
(in the 1973 experiment) by a method involving the measurement of the change
of aJbsorJbence of a barium-dinitrosulfanazo III complex in the presence of
sample sulfate. The sensitivity of this method is about 2 yg/ml. Some work
was done to validate this method by comparison with other methods. The data
showed that at sulfate concentrations of 6 yg/ml and above, the agreement
with the turbidimetric method was within t 20%. Data obtained showed that
samples analyzed both by ESCA for S+ ("oxidized sulfur", mostly sulfate) and
for sulfate by water extraction did not agree by better than a factor of 2.
However, comparison of diurnal patterns for S+ values and water soluble
sulfate values (both samples from the same filter) suggested that the ESCA
result can be used in a comparative manner to evaluate relative changes in
sulfur levels. X-ray fluorescence analysis was used for analysis of sulfur,
lead, bromine and zinc.
FREEWAY AEROSOL STUDY (FAS)
This study was also sponsored by the California Air Resources Board and
was conducted by the staff of the Crocker Nuclear Laboratory, University of
California (Davis). The purpose of this study was to assess the dispersal
of lead (and other emissions) from automobiles, and the relation of these
emissions to meteorological conditions and traffic density. This represents
the only Los Angeles Basin study known which had as its principal objective
the assessment of vehicular emission under ambient conditions. Lead emission
was of major interest but other elements were also determined. However,
assessment of particulate sulfur emissions from this study is significantly
limited, as discussed later. No pollutant gas data were obtained. Four sites
on heavily travelled Southern California freeways were used and sampling was
done using multistage impactors at various locations both upwind and downwind
across the freeway. The rotating stages of the Lundgren impactor permitted
two-hour sample resolution. The after filter was also changed on a two-hour
schedule. Sampling was conducted during nine days between February and August
31
-------
2972. In addition to the particulate collection and analyses, simultaneous
traffic information and meteorological data were obtained from nearby stations
of the California Department of Transportation (Caltrans).
The only analytical technique used in this study for analysis of the
particulates on the impaction media and filters was ion induced x-ray fluores-
cence analysis (IXRFA). Lead, bromine, and sulfur were determined by this
technique on the various stages of the Lundgren impactor. For the impactor
surface, Mylar covered with a very thin paraffin film was used. Whatman 41
cellulose filters were used as after filters. Interlab and inter-method
agreement for Pb and Br were within 10%. The comparative results were ob-
tained for Pb by atomic absorption and for Br by x-ray induced fluorescence.
On the impactor stages, the accuracy for sulfur was within 10% for standard
samples and within 35% for natural aerosols. The use of a Whatman 41 cellulose
filter as the after filter precluded the accurate determination of sulfur
because of absorption effects. Thus, sulfur levels were only obtained on the
impactor stages. It is estimated that this accounts for only about 50% of the
total particulate sulfur. Thus, little can be said about the fate and
distribution of automotive particulate sulfur emissions from this study.
A typical data presentation is shown in Figure 4, where the variations of
traffic density, wind speed, and lead concentrations collected by the various
samplers is given as a function of time during one study period. Inspection
of all the data suggests that under some conditions of wind speed, the upwind
sampler will show higher lead levels, and the lead levels do not vary uniformly
as a function of distance from the roadway. The results suggest that mechan-
ical and thermal turbulence created by the freeway may cause some unexpected
effect and that care should be used in designing experiments and interpreting
the data.
LOS ANGELES COUNTY AIR POLLUTION CONTROL DISTRICT DATA BASE
As part of its charter to control air quality, the Los Angeles County
Air Pollution Control District (LAAPCD) (now called Southern California Air
Pollution Control District) has monitored the air in the Los Angeles Basin
for a number of years. The LAAPCD data were evaluated to assess the general
levels, variability, and distribution of the pollutants of interest. The
LAAPCD operates a number of stations in which gaseous pollutants are measured
and 24-hour particulate samples are collected with hi-vol samplers. Glass fiber
filters are used. These filters are subsequently analyzed for sulfate, lead,
and other pollutants. A turbidimetric method was used for the sulfate
analysis, while atomic absorption analysis was used for lead and zinc. Carbon
monoxide was determined by non-dispersive infrared analysis and sulfur dioxide
was determined by continuous conductometric analyzers. In Figure 5, the
pattern for the annual averages for the 1966-1973 periods are given for
several pollutants. The Sp/Sg value is the ratio of particulate sulfur to gas
phase sulfur. The mean sulfate value for the period 1966-1972 for West Los
Angeles is 7.8 yg/m3. For downtown Los Angeles, the mean value is 13.1 yg/m3
and for Lennox 11.9 yg/m3. The slope of the plot is positive with recent
values in the 11-16 yg/m range. Monthly arithmetic mean plots of the LAAPCD
data indicate that the peak sulfate values occur more often in the summer
months, suggesting that the higher solar irradiation at that time of year gives
32
-------
900
1300 1700
8-14-72
2100 I 0100 0500 0900 1300 1700
| 8-15-72
TIME OF DAY
Figure 4. Diurnal variation in lead by impactor location at San
Diego Freeway on August 14, 1972 (Freeway Aerosol Study)
33
-------
D)
C/)
"a
CO
0.25
0.20
0.15
0.10
0.05
20
00
& 15
10
« 15<
l> 10
^S.O
1.5'
CO
i1-0
0.5
15
| 10
a
5.0
2.0
01
I1-0
* 0
110
80
65
WEST LOS ANGELES
SLOPE = 0.026
|=«U39
SLOPE = -0.476
SO2 = 15.2 ppb
SO| * 7.80 jug/m3
SLOPE = 1.11
Pb = 0.86
SLOPE = 0.03
H20 FILTER
INSTALLED
-SLOPE = -0.19
CO = 4.83 ppm
SLOPE - -0.34
Zn - 0.77
TSP=91.9M9/m3
66
67
68
69 70
YEAR
71
72
73
Figure 5. Pattern of annual averages for West Los Angeles from Los
Angeles County Air Pollution Control District Data.
34
-------
rise to a greater photochemical oxidation of SO2 in the presence of the other
atmospheric pollutants. Plots of other pollutants [carfcon monoxide (CO), Pb,
S02 and total suspended participates (TSP)"] also show cyclical behavior with
maxima generally in the winter months. This is probably caused by low level
winter inversions and the winter use of fuel oil in the basin.
NATIONAL AEROMETRIC SURVEILLANCE NETWORK DATA BASE
The National Aerometric Surveillance Network (NASN) is operated for
surveillance purposes by the U.S. Environmental Protection Agency. Comparison
of annual averages of TSPf sulfur dioxide, and sulfate with the previous
LAAPCD data are of interest. In Figure 6, plots of the annual average for
TSP and sulfate are given using both NASN and LAAPCD data. The two samplers
were immediately adjacent to each other, but operated on different schedules.
Significant differences exist between the two sets of data. For sulfate, the
NASN data show essentially a constant, concentration while the LAAPCD data
show an increasing concentration. The analysis of sulfate by the Environmental
Protection Agency for the NASN project has been done by the methylthymol blue
method.
Although the NASN sampling schedule was different from that of the LAAPCD,
the sample collection data coincided on 22 occasions for the 1965-1972 period.
Figure 7 shows a scatter diagram for those comparative results. Some large
discrepancies are evident. Analysis of the data shows that the values report-
ed by NASN are on the average 19% greater than the LAAPCD results for these
22 days of side-by-side sampling. Even though there were differences in
analytical method, sampling technique, high-volume flow calibration procedure,
type of filter, 24-hour sampling schedule, and sampler flow rates employed
during sampling, it is still difficult to propose a completely rational
explanation of these differences in the data.
COMMUNITY HEALTH AND ENVIRONMENTAL SURVEILLANCE SYSTEM DATA BASE
The Community Health and Environmental Surveillance System (CHESS) was
operated under contract by the U.S. Environmental Protection Agency. The
objective of the study was to relate pollutant levels to human health effects.
A number of stations were located throughout the Los Angeles Basin in which
various gaseous pollutants (including 802) and particulates were measured.
The 24-hour high-volume sampler with glass fiber filter was used to collect
particulate samples. Sulfate was analyzed using the turbidimetric method.
From CHESS data, based on analysis of 128 samples in triplicate, a precision
of i- 5.8% at the 95% confidence level, and an accuracy, based on standard
addition, of +_ 11.2% was reported. However, differences of up to 430% have
been found for analysis of two different strips of the-same high-volume
filter when analyzed by the CHESS contractor and the New York City Air Pollu-
tion Control District.
One of the CHESS stations was in Santa Monica, which is four miles east
of the West Los Angeles station of LAAPCD. In 1972, the average of 8 months
(from May through December) CHESS data for sulfate gave a value of 8.4 yg/m .
The West Los Angeles station of the LAAPCD gave an annual average for 1972
of about 12 vg/m*. However, in 1972 some rather high sulfate levels were
35
-------
18
17
16
15
« 14
4 "
8 12
» 11
10
9
8
7
1.8
n 1 a
^ 1.6
w 1-4
a.
1.2
170
160
n 150
S: uo
CO
l- 130
O)
51 120
110
100
LAAPCD
NASN
SULFATE
1 1
65 66 67 68 69 70
YEAR
71 72 73
Figure 6. Comparison of LAAPCD and NASN data from downtown Los Angeles
location: stations are immediately adjacent.
36
-------
CO
<
z
LAAPCD
Figure 7. Comparison of paired sulfate concentrations reported by
LAAPCD and NASN
37
-------
measured during the first of the year. Comparison of data covering the same
period of time show about a 20% difference between the two sets of data.
Fairly good agreement was also obtained between TSP values determined in the
CHESS program and by the LAAPCD.
SULFUR DIOXIDE CONVERSION STUDY
The Sulfur Dioxide Conversion Study (SDC) was conducted in 1973 by the
California Institute of Technology, Department of Environmental Engineering
Science. It was conducted to determine the conversion rate of SO2 to particle
phase sulfur (sulfate) in urban atmospheres. In this program, gas phase
sulfur and particulate phase sulfur concentrations were measured in the Los
Angeles Basin near major stationary sources and at a downwind receptor site.
Measurements of SO2 and sulfate were made on five days generally coinciding
with measurements made during the 1973 Aerosol Characterization Experiment
(ACHEX) (see earlier). The 5O2 measurements were made with a continuous
flame photometric analyzer. For the sulfate analysis a new technique was
developed at the California Institute of Technology wherein particulate was
collected on a glass fiber filter. The filter, after collection, was heated
to 1100°C and the emitted gaseous sulfur species was determined by a flame
photometric detector. Since data were obtained for the same time period
and in the same location for this study as for the ACHEX study, comparison
of this method with the ACHEX wet chemical sulfate method could be made.
Comparison shows that the new method gave low results consistently with differ-
ences as high as a factor of 2. However, the diurnal patterns obtained in
the two studies were similar.
From the collected data, Roberts and Friedlander (7) were able to
estimate the apparent rate of SO2 conversion in the Los Angeles Area. For
this, the LAAPCD data for S02 along wind trajectories between source-enriched
sites and Pasadena, the average ratio of particulate to total sulfur (gas
phase and particulate) at Dominguez Hills and Pasadena, and the estimates of
additional SO^ inputs along the trajectory were used. The values obtained
ranged from 1.2 to 13.0%/hour with a mean value of 7.2 +_ 3.8%/hour based on
data for three days in July 1973.
An additional aspect of the SDC study was an effort to determine the in-
fluence of auto exhaust catalytic converters on sulfate levels near roadways
and at the receptor site, Pasadena. Assuming all cars to be equipped with
catalytic converters, 50% conversion of sulfur in gasoline to sulfate by
each car, and gasoline containing an average of 0.075% sulfur*, an increase
of 76 yg.m3 SO^ above the levels due to other sources was calculated by these
investigators "near freeways at rush hour". About a 10% increase in sulfate
level was calculated for a low traffic density area (Pasadena) after two
model years of catalytic converter use.
CONCLUSIONS
The AIHL review has assembled information on the sulfur dioxide emissions
in the Los Angeles Basin, the climatology of the Los Angeles Basin, and the
*This is an upper limit to the sulfur content found in South Coast Basin
gasoline.
38
-------
results of several sulfate related studies performed in portions of the basin.
Data are given indicating that in 1971 84% of the S02 in the basin came from
stationary sources and 14% came from vehicles. About 2% came from aircraft.
The climatology of the area is characterized by subsidence inversion in
the summer months with an average inversion height of about 600 m. Surface
inversion occurs in the wintertime but generally dissipates quickly during
the early day. Winter inversions generally give rise to higher concentrations
of primary pollutants (CO, SO2)• The summer low subsidence inversions give
rise to reduced mixing volumes for extended periods of time and, with the
increased solar radiation, result in much higher concentrations of secondary
pollutants, including sulfate. During the day the prevailing wind is off the
ocean and at night it is generally off the land. The period of time in which
the flow is off the ocean is longer in the summer than in the winter.
The location of the LACS site is shown to be favorable, having extended
periods when the wind is perpendicular to the freeway (especially during the
summer) and having high freeway traffic density. The background sulfate is
also expected to be low on the basis of stationary source locations.
The results of the six previous studies which were evaluated indicate
that in some areas of the Los Angeles Basin, east of the sources, the maximum
two-hour average sulfate level can reach 50 to 70 ygr/zn3 in the summer. Jn
West Los Angeles near the LACS site, the average value is in the 7-14 ygr/m3
range, although significant discrepancy exists in the past data.
One study was done to relate ambient pollutant concentrations (mainly
lead) to traffic emissions. The results suggest that mechanical and thermal
turbulence can cause unexpected effects when monitoring near freeways.
REFERENCES
1. Lemke, E.E., Shaffer, N.R., Thomas, G. and Verssen, J.A., "Air Pollution
for Los Angeles County", Air Pollution Control District, Los Angeles
County, Los Angeles, California (January 1967).
2. Ibid, January 1971.
3. Pierson, W.R., Hammerle, R.H. and Kummer J.T., "Sulfuric Acid Aerosols
from Catalyst-Equipped Engines", presented at the Automotive Engineering
Congress, Detroit, Michigan (February 25-March 1, 1974).
4. Beltzer, M., Campion, R.J. and Petersen, W.L., "Measurement of Vehicle
Particulate Emissions", presented at the Automotive Engineering Congress,
Detroit, Michigan (February 25-March 1, 1974).
5. Hidy, G.M., et al., "Characterization of Aerosols in California, Interim
Report for Phase I" Rockwell International Science Center, Thousand Oaks,
California (April 30, 1973).
39
-------
6. Smith, T.B., Blumental, D.L., Stinson, F.R. and Mirabella, V.A.,
"Climatological Wind Survey for Aerosol Characterization Program", Final
Report, PO 262-1366, MRI, Inc., Altadena, California (1972).
7. Roberts, P.T. and Fried-Zander, S.K., "SC>2 to Sulfate Conversion Study
(SDC)", California Institute of Technology (1973).
40
-------
DISCUSSION
My name is Pierson. Ford Motor Company.
Mr. Trakowski has recounted the events leading to the project we are here
to discuss. He noted that about 4 years ago was the time of origin of concern
about sulfuric acid from catalyst-equipped cars. He said he was not sure who
it was who first expressed concern "
The information is in the record of the EPA resuspension hearings con-
ducted by Eric Stork in 1975. It was Ford Motor Company who first expressed
concern. The problem was discovered in 1972 in my lab at Ford on samples
obtained from Dow Chemical Company. As to how notice of the problem was
given: I informed Otto Manary at Dow in a letter Dated October 20, 1972.
Herbert L. Misch informed the EPA in a letter of February 1973.
The detailed story has been given in a number of publications. I would
like for this to appear in the record of this meeting.
41
-------
SELECTION RATIONALE FOR ANALYTICAL METHODS
FOR THE LOS ANGELES CATALYST STUDY
Richard J. Thompson
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
The first considerations in the selection of analytical methods were the
data needed and the sampling that could be implemented. Upon consideration
of the data output and the monitoring strategy, the methods were selected on
the basis of reliability [considering both reproducibility and repeatability],
precision [with standards and with real samples], training [technology trans-
fer] , and cost [time, equipment, and personnel]. The analytical methodologies
to be employed were to be fitted within the framework of a quality assurance
program and implemented by contract personnel in more than one contract
laboratory.
INTRODUCTION
My distinguished colleague has described to you the objectives of the Los
Angeles Catalyst Study (LACS) and the information needed to attain these
objectives. This presentation is given to outline the rationale for selecting
analytical methods, the two prime governing factors being accuracy obtainable
and practicalities imposed.
PROCEDURES
The first consideration in such a selection, of course, is the nature of
the data needed and the sampling that can be implemented. To obtain the data
output necessary from the samples obtained, factors used in selection for the
LACS were: reproducibility (within a laboratory)/ repeatability (between
laboratories); precision (obtained with standards and with environmental
samples); technology transfer (training of contractor personnel); and cost
(time, equipment and personnel). Implementation of a quality assurance
program was essential. These factors are summarized in Table I .
Ideallyf one has inter-laboratory standardization data upon which to base
judgment (and from which standardized methods can be secured) and one has
standard samples for an audit program. The only body of available knowledge
of the type necessary at the time of initiation of the LACS study consisted
of unpublished information gained by experience and based on published methods
not subjected to formal or structured comparisons. The methods are shown in
43
-------
TABLE I. GOVERNING FACTORS: ACCURACY AND PRACTICABILITIES
SELECTION CRITERIA CAS OF FEBRUARY 1974)
Parameter
Desired
(Other Than Journals)
Obtainable
Reproducibility
Precision
Repeatability
Methods
Standards
Technology Transfer
Quality Assurance
Intensive Ruggedness Testing
S R M ' s
Inter-Laboratory Comparisons
Standardized Methods
S R M ' s
Formal Courses
Standardized Programs
EMSL Experience
Lab Reports
Informal Sample Splits
In-House Write-ups
Commercial/Internal
EMSL
In-House Quality Control
and Sample Splits
44
-------
Table II. They had been employed internally for thousands of samples, and
information had been painfully accrued as to what readily identifiable mistakes
could be made routinely, how they could be avoided or minimized, and how
effective internal quality control could be effected.
In the main, the methodologies used internally, including quality control
procedures, were implemented by two prime mechanisms: documents and personal
contact. The documents included published and in-house texts; personal con-
tacts included training in Environmental Monitoring and Support Laboratory
(EMSL) labs and verbal interchange by telephone.
The methods used have been subsequently investigated extensively else-
where and found to be state-of-the-art methods, with some exceptions.
Quality assurance will be presented later, but, as summarized in Table
III the following steps were taken:
Special filters were supplied by EMSL which had been obtained to specifi-
cations.
Sample splits with EMSL were imposed at a high level (10%). Discrepancies
were noted and corrected in the case of sulfate and nitrate; the ammonium ion
had and has been proved to be elusive in the samples obtained. (As will be
mentioned later, blind audits were initiated when samples became available.)
The internal checks used by EMSL were adopted routinely by the contractor
lab, and included incorporation into the samples sets: open and blind dupli-
cate strips of randomly selected samples; standards for each nth sample; and
spiked duplicates. Automatic re-runs are made of samples containing over a
pre-set level of a component. Overall, over a quarter of the analytical burden
is for quality control.
As a specific example, the case of platinum was of interest. Platinum,
a possible environmental contaminant of hazard to human health was considered,
and the following was needed to assess the potential problem.
0 The environmental level had to be determined,
• A method had to be demonstrated on real samples.
• A rountine method had to be proven for environmental samples.
In an elemental survey of particulate matter, platinum was found at some
4 x 10~11 g/ra3 by SS/MS, which was very imprecise. The geochemical levels
were anticipated to be found in particulate from ambient air (some 2 x 10~12
g/m3) using the geochemical platinum content and an average ambient TSP
assumed to be of the same platinum content. As was stated in an in-house
report, "the positive analysis of ambient air for noble metals is anticipated
to be quite challenging, if indeed possible."
A massive sample was obtained from a public building near the LACS site.
The 1.5 kg sample contained a considerable amount of non-suspended particulate
matter, such as beetles, etc.
45
-------
TABLE II. METHODS CHOSEN (FEBRUARY 1974)
Parameter
Sample Preparation
Analytical Method
Lead
Nitrate
Sulfate
T S P
SO.
(LTA +) Acid Digestion
HO Reflux
HO Reflux
&
Conditional
72 ± 1°F & 45 ± 5% RH
Atomic Absorption
Colorimetric (Cu/Cdf
Griess-Ilsovay)
M T B
Mass
Bubbler/West-Gaeke
46
-------
TABLE III. INITIAL QUALITY ASSURANCE
Analyzed Filters Bought to Specifications Furnished
Duplicate Samples Were Run By EMSL and Contractor (10%)
EMSL Quality Control Procedure Furnished, Contractor QAB Plan Examined
— Duplicates (open and blind)
— Standard Every Nth Sample
— Spiked Duplicate
— Automatic Rerun of Values Over a Selected Level
47
-------
The sample was homogenized and halved. One half was spiked with known
stable isotopes of platinum, palladium, and ruthenium (a suitable isotope of
rhodium was not available) and oxidized in a muffle furnace with incremental
temperature increases programmed. The residue was dissolved in analyzed aqua
regia; a known gold carrier added; and the precious metals obtained by
reductions with zinc. The gold button bearing the precious metals was dis-
solved in analyzed aqua regia, diluted to volume, and aliquots distributed to
participating laboratories. The aliguots were analyzed by EMSL, Stewart Labs/
and Battelle Northwest.
RESULTS
The information obtained is given in Table IV. Fortunately, soluble
platinum species were not identifiable in catalyst equipped automobile
effluents, and adverse biological effects have not been noted. A routine
method for platinum at atmospheric levels does not appear existent nor
obtainable at the current state-of-the-art.
48
-------
TABLE IV. ENVIRONMENTAL PLATINUM
Sample Method Pt Level (g/m3)
Suspended Particulate Method
Geochemical 1
Massive Sample 2
SS/MS
Literature
IDMS
NAA
AAS
XRF
OES
4 x 10-11
2 x 10~12
2-5 x 10~12
<9 x 10"11
<3 x ID'11
NS
NS
1 An average Los Angeles particulate loading was assumed to contain platinum
at the average geochemical level given in the literature.
2 An air volume was assumed from the average Los Angeles particulate loading
and the mass of the sample.
49
-------
DETERMINATION OF LOW CONCENTRATIONS OF PARTICULATE
WITH THE SPECIFIC ION ELECTRODE
Lloyd S. Shepard, George Colovos, and Allen M. Miles
Rockwell International
Newbury Park, California
and
Charles E. Rodes
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
The development of a method employing an ion sensitive electrode and
the Gran-plot for the microdetermination of NH^ in particulates will be
discussed. For the successful application of this method without loss of
NH^ as NH% during the determination, addition of a small amount of NHi+
(about 0.1 ppm) to the recipient solution is necessary. Response of the
electrode and the slope of the Nernstian plot should be determined carefully
before use of the method. These experimentally determined electrode para-
meters should be used for the Gran-plot rather than theoretical values to
avoid erroneous data. The sensitivity, precision and accuracy of the developed
method are given and comparisons between this method and the automated indo-
phenol method are discussed. Results of analyses of both synthetic and
ambient samples are presented.
INTRODUCTION
Several papers presenting applications of the ammonia sensitive electrode
have been published. T. R. Gilbert and A. M. Clay (1) first and later
R. F. Thomas and R. L. Booth (2) studied its application to the determination
of ammonia in water and wastewater samples. J. M. Bremner and his co-workers
(3,4) have used electrodes for the determination of ammonia in soil extracts
and others (5,6) have used them for ammonia determination in the presence
of urea. In all the above studies, the electrode response is directly pro-
portional to the logarithm of the concentration. This approach is quite
acceptable for relatively high concentrations, but for low concentrations it
is somewhat difficult to apply direct potentiometry because the electrode
response becomes too slow to reach the equilibrium potential at a reasonable
time. This problem can be minimized by utilizing a standard addition
technique, thus decreasing the response time of the electrode.
51
-------
In addition to the faster response of the electrode, the standard addition
method offers the advantage of applicability to the Gran-plot technique (7)
for the estimation of the amount of ammonia present in the sample. The Gran-
plot technique is based on the linear extrapolation to zero of a plot of the
exponential form of the Nernst equation. This in the case of the ammonia
sensitive electrode is given by equation 1
= * • 10- - [NH3]Q 1
where k is a constant related to the normal potential (E ) and the slope
of the electrode response, E , is the measured potential, [NH$] and [##3]
are the original and the added concentrations of ammonia, and S is the slope
of the electrode response.
This combination of standard addition and Gran-plot techniques has
been used by several investigators in connection with various ion sensitive
electrodes, such as fluoride (8,9), cesium (10), lead (11) and others. Most
recently, W. Selig, J. W. Frazer and A. M. Kray (12) have used this approach
for the microdetermination of ammonia with an ammonia sensitive electrode
titration system interfaced with a minicomputer. With this system they were
able to determine accurately ammonia in the range of 28 ppb to 2.8 ppm of
ammonia nitrogen. The manual version of the above technique has been investi-
gated by us and, through these studies, we were able to develop a method for
the determination of submicrogram amounts of ammonia in solution. The
developed method was then applied to the determination of particulate ammonium
collected with the aid of a dichotomous sampler on 37 mm Teflon filters.
In this paper, the result of our studies are reported.
EXPERIMENTAL
INSTRUMENTS AND APPARATUS
An ammonia selective electrode (Orion M-95-10) connected to a Beckman
Expandomatic SS-2 pH meter was used throughout this study. The sensitivity
of the SS-2 pH meter was increased greatly through the use of a Heathkit
(EU-200-01) amplifier connected with a Heathkit (EU-200-02) potential off-
set module by which the signal could be appropriately offset and amplified
to be recorded on a strip chart recorder.
All the titrations were conducted in a thermos ta ted, specially con-
structed glass cell. The temperature of the titration cell was kept con-
stant by circulating water from a constant temperature water bath (Haake,
Inc. Model FJ) .
REAGENTS
All chemicals were Baker analyzed reagents unless otherwise specified.
Ammonium chloride solution (0.1N) was prepared by dissolving 5.3 g NH^Cl
in a liter of water. Stock solution of ammonium (1000 ppm NH$fN) was
52
-------
prepared by dissolving 3.82 g of NHkCl in a liter of water. The working
standard (100 ppm NH^-N) was prepared by diluting 10.0 ml of the solution
to a liter with ammonia free distilled water. The conditioning solution
was prepared by diluting 2.5 ml of the working standard solution to a liter
with ammonia free distilled water. Solutions of 0.1 or 2.5N NaOH and
l.ON HC1 were prepared by diluting appropriately concentrated solutions of
NaOH and HC1, respectively.
PROCEDURE
The ammonia electrode must be kept in solution at all times. Between
titrations, the electrode was kept in a solution of 0.1N NaOH. When it was
not in use for any extended period of time, the electrode was stored in a
solution of 0.1N NHiiCl. Each time the electrode changed solutions, it was
rinsed thoroughly with distilled water and wiped dry with a Kimwipe. Care
was taken not to damage the hydrophobic membrane. Membrane damage can
be detected from the electrode response time. The response of electrodes
with damaged membranes is usually very slow.
The procedure requires independent, accurate determination of the slope
of the Nernstian response of the electrode. This was accomplished by adding
5.0 ml of distilled water spiked with 10 yl of l.ON HC1 into the titration
vessel, covering with a sheet of parafilm and stirring the solution for
approximately five minutes. After this, the electrode was removed from the
0.1 NaOH (kept at room temperature, 22°C-25°C), washed well with distilled
water, dried with a clean Kimwipe, and placed in the titration vessel, making
sure there were no air bubbles trapped on the electrode membrane. The desired
levels of ammonia in the solution were obtained by adding small (]il) volumes
of working standard into the solution with the aid of a micropipet. Initially,
5.0 yl of working standard was added and the solution was stirred for one
minute. 0.5 ml of 2.5N NaOH was then pipetted into the titration vessel,
and the potential recorded until the change in potential for a one minute
time interval was less than 0.5 millivolts. This amount of change defines a
plateau in the voltage vs time plot. Another 5 yl of working standard was
added and the new potential was measured again at the plateau. The addition
of ammonia was continued in increments of 5 yl until a stable potential
reading for 180 nanomoles was obtained. From these data the slope of the
Nernstian response of the electrode was determined by plotting the measured
potentials against the logarithm of the corresponding ammonia concentration.
The concentration of ammonia in the conditioning solution (blank) was deter-
mined as follows:
Two ml of the conditioning solution (0.250 ppm NH^'N) and then 10 yl of
l.ON HC1 were added into the titration vessel. The solution was then stirred
for approximately two minutes, and then 3.0 ml of distilled water were added.
The concentration of ammonia in the final solution was about 0.1 ppm NH$'N
or a total of 36 nanomoles of NH^in 5 ml. This solution was covered and
stirred for three minutes and then 5 yl of 100 ppm NH$'N (36 nanomoles NH$)
and 0.5 ml of 2.5N NaOH were added. The potential was measured at the
plateau for five 5 yl incremental additions of 100 ppm NHfN as described
previously in the determination of slope. The original concentration of
ammonia in the conditioning solution was determined by the Gran-plot of the
53
-------
measured potential versus the concentration of ammonia added in it. The
analysis of actual samples was performed in a similar manner, as follows:
First, 2.0 ml of the conditioning solution followed by 10 yl of l.ON
HC1 were added in the titration vessel and the obtained solution was stirred
for three minutes. Second, 3.0 ml of the sample was added; the solution was
stirred for at least two minutes to reach thermal equilibrium (25 C) and
the electrode potential was measured after incremental addition of 5 yl of
100 ppm NH^'N. The same technique as described for slope determination was
used to obtain electrode potential. It is very important that the potential
reading for any one addition of ammonia be taken at the potential vs time
plateaus or when the change in potential for a one minute time interval is
less than ± 0.5 mV. The concentration of ammonia in the sample was determined
by subtracting the ammonia concentration of the conditioning solution (deter-
mined previously) from the total concentration of ammonia obtained from the
Gran-plot.
RESULTS AND DISCUSSION
Accurate determination of the electrode potential and use of the cor-
rect slope of the electrode response with respect to concentration are the
two most important factors of the Gran-plot technique. Accurate determination
of the electrode potential is accomplished by utilizing the combination of
electrometer (pH meter) and amplifier, described in the experimental part of
this paper. With this, the electrode potential can be determined with an
accuracy better than 0.2 mV which corresponds to a relative percent accuracy
of better than 0.5%. The slope of the electrode response, theoretically,
should be given by RT/nF, where R is the gas constant, T is the absolute
temperature, n the number of the electrons, and F is the Faraday constant.
In most cases the calculated (from the expression RT/nF) and the experimental
values of the slope are so close that either one can be used in electro-
chemical calculations such as the Gran-plot technique. However, in the case
of the ammonia electrode used in this study (Orion M-95-10) , the experi-
mentally determined slope was found to be considerably different (usually
lower) from the theoretically calculated value. The average slope (Em vs log
of twenty independent experiments conducted at 25 C over a period of
ten days was found to be equal to 55 mV with a relative standard deviation
of 2.3%. The range of the above slope value was 52.0 to 57.3 mV which is
indicative of the day-to-day variation.
Figure 1 shows the effect of the slope value on the Gran-plot of one
experimental run. The slope of the electrode response was experimentally
determined to be 55 mV which, when it is used in the Gran-plot, results in
an intercept equivalent to 32 nanomoles of ammonia. If, however, in the
same run, another slope were used for the application of the Gran-plot
technique, the resulting intercept would be substantially different. As
shown in Figure 1, the intercept of the Gran plot for an electrode slope
equal to 52 mV is equivalent to 21 nanomoles and the one corresponding to
the theoretical value of 59 mV is equivalent to 46 nanomoles. This clearly
demonstrates that an error of 100 to 200% may result from the use of an erron-
eous value of electrode slope. For this reason the slope of the electrode
response should be determined separately before and after the application of
54
-------
UJ
2
T
T
T
THE THREE PLOTS (a), (b), AND (c) WERE CONSTRUCTED
FROM THE EXPERIMENTAL VALUES OF ONE RUN BY
USING THE RT/nF VALUES OF 52. 55 AND 59 mV
RESPECTIVELY.
(0
(b)
(a)
160
200
NANOMOLES NH3
Figure 1. Effect of the RT/nF value tslope) on the intercept of the Gran-plot,
55
-------
the Gran-plot technique in the determination of ammonia by an ammonia sensitive
electrode. It should be noted here that the ratio Em/S is dimension!ess and,
therefore, the units in which the numerator and denominator of the ratio are
expressed, are unimportant as long as they are the same. If, however, the
theoretical value of the slope (mV) of the electrode response is used, the
measured potential (Em) should be expressed in mV rather than in arbitrary
units. This can be accomplished by calibrating the instrument against a
standard voltage source.
During the development of this method, it was observed that analysis
of synthetic solution containing ammonia in the range of 16-36 nanomoles
produced fairly precise results (3-4% relative standard deviation) but not
accurate (more than 12% error). Since the results were consistently lower
than the expected value, it was hypothesized that losses of NH$ (gas) from
the solution may be the cause of this problem. This hypothesis was corrobo-
rated by the fact that the difference between the expected and the obtained
value was greater at low concentrations. To test this hypothesis, the use of
conditioning solution was employed. This solution contained enough ammonia
to bring the concentration up to a level not affected by the small loss
which may take place during the performance of the analytical procedure. The
use of conditioning solution appeared to eliminate the problem and, therefore,
it was adopted in this procedure.
Table I gives the results of the application of this method in the
analysis of synthetic samples. From this it appears that the method pro-
duces accurate results for concentration above 16 nanomoles per 3 ml of the
sample. To further demonstrate the validity of the developed method, samples
of aqueous extracts of particulates collected at the sites of the Los Angeles
Catalyst Study were analyzed. Table II shows the results obtained by the
electrode method in relation to those obtained by the automated indophenol
procedure. The good agreement between the two methods shows them to be
equivalent and, therefore, application of this technique to the determination
of ammonia in aqueous extracts of particulates should be acceptable. Cur-
rently, the described procedure is being used for the determination of parti-
cipate ammonium collected by the dichotomous samplers operating at the Los
Angeles Catalyst Study sites.
REFERENCES
1. Gilbert, T.R. and Clay, A.M., "Determination of Ammonia in Aquaria
and in Seawater Using the Ammonia Electrode," Anal. Chem. 45, 1757 (1973).
2. Thomas, R.F. and Booth, R.L., "Selective Electrode Measurement of
Ammonia in Water and Wastes," Env. Sci. and Tech. 7_, 523 (1973).
3. Bremner, J.M. and Tabatabai, M.A., "Use of an Ammonia Electrode for
Determination of Ammonium in Kjeldahl Analysis of Soils: Commun.
Soil Sci. Plant Anal. 3_, 159 (1972).
4. Banwart, W.L. , Tabatabai, M.A. and Bremner, J.J., "Determination of
Ammonium in Soil Extracts and Water Samples by an Ammonia Electrode,"
ibid 3_, 449 (1972) .
56
-------
TABLE J. DETERMINATION OF AMMONIA IN SYNTHETIC
SOLUTIONS BY THE GRAN-PLOT TECHNIQUE
Nanomoles NH$
8.04
16.1
20.7
33.2
61.7
Average
6.7 (a)
16.0 (b)
20.7 (d)
33.2 (c)
58.7 (d)
Found
Standard
±
±
+
+
+
Deviation
2.4
0.6
2.1
1.1
1.0
(a) Average of eight determinations
(b) Average of six determinations
(c) Average of nine determinations
(d) Average of three determinations
57
-------
TABLE II. DETERMINATION OF AMMONIUM IN
AQUEOUS EXTRACTS OF PARTICULATES
ppm
Ammonia Automated Indophenol
Sample # Electrode Method % Diff.
1 8.37 8.86 - 5.7
2 5.09 4.99 + 2.0
3 7.65 7.55 + 1.3
4 5.09 5.25 - 3.1
5 5.23 5.05 + 3.5
6 5.96 6.50 - 8.7
7 15.96 12.56 +11.9
8 41.52 41.68 - 0.4
9 41.12 42.75 - 2.0
58
-------
5. Woodis, T.C. and Cummings, J.M., "Determination of Ammonia - Nitrogen
in the Presence of Urea with an Ammonia Electrode," J. Assoc. Off. Anal.
Chem. 56_, 373 (1973).
6. Rogers, D.S. and Pool, K.H., "Analysis of Urea Nitrogen in the Presence
of Ammonia Using an Ammonia Gas Sensitive Electrode," Anal. Letters 6_,
801 (1973).
7. Gran, G., "Determination of the Equivalence Point in Potentiometric
Titrations II," Analyst 77, 661 (1952).
8. Liberti, A. and Mascini, M., "Anion Determination with Ion Selective
Electrodes Using Gran's Plots. Application to Fluoride," Anal. Chem.
41, 676 (1969).
9. Selig, W., "Microdetermination of Fluoride Using Gran's Plot," Mikrochim.
Acta, 87 (1973).
10. Coetzee, C.J. and Basson, A.J., "A Potentiometric Determination of
Cesium Ion," Anal. Chem. 56, 321 (1971).
11. Brand, M.J.D. and Rechnitz, G.A., "Computer Approach to Ion-Selective
Electrode Potentiometry by Standard Addition Methods," Anal. Chem. 42,
1172 (1970).
12. Selig, W., Frazer, J.W. and Kray, A.M., "Microdetermination of Ammonia
Using an Ammonia Gas-Sensing Electrode and a Minicomputer," 2_, 675
(1975) .
59
-------
A COMPARISON OF THE THORIN AND MODIFIED METHYLTHYMOL BLUE
METHOD FOR THE DETERMINATION OF MICRO-AMOUNTS OF SULFATE
George Colovos, Edward P. Parry, Allen M. Miles, and Martha R. Panesar
Rockwell International
Newbury Park, California
and
Charles E. Rodes
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
The method developed by Brosset (Thorin method) for the microdetermination
of sulfate is discussed and compared with the modified Methylthymol Blue
method. Improvements made to the original Thorin method which increase the
sensitivity, are discussed and the disadvantages presented. The sensitivity
of the Thorin method has been increased without addition of sulfate (Brosset)
to the reagents by increasing the excess of the Thorin reagent. The decrease
of stability with time, the cost per sample and the reliability of measurement
are the disadvantages of the method. The sensitivity of the conventional
Methylthymol Blue method has also been increased by modifying the flow process
so that no dilution of the sample occurs. Comparison of the sensitivity,
precision and accuracy of the Thorin and modified Methylthymol Blue methods
is made for the analysis of both synthetic and ambient samples.
INTRODUCTION
The property of thorin Co-[(2-hydoxy-3,6-disulfo-l-naphthyl)azo] benzene-
arsenic acid) which forms weak complexes with barium ions has been utilized in
the determination of sulfate. J.S. Fritz and S.S. Yamamura (1) adapted first
the use of thorin as indicator in the titrimetric determination of sulfate.
They suggested use of an 80% ethanol-water mixture as the titration medium
and 5 x 10~3M Ba(C10i+)2 as titrant. The sensitivity of this analysis was
10 vg/ml SO^ in a total volume of 25 ml. The sensitivity has been further
studied and improved by other investigators. For instance, O. Menis, D.L.
Manning, and R.G. Ball (2) replaced the 80% ethanolic titration medium with
50% isoamyl alcohol-water mixture and they also introduced a spectrophotometrie
method for the detection of the end-point.
G. Persson (3) first developed a direct spectrophotometric determination
of sulfate utilizing the barium-thorin complex. The reaction of sulfate with
61
-------
this complex results in formation of insoluble barium sulfate and release of
free thorin. The extent of this reaction can be monitored spectrophotometrie-
ally at 520 nm, where the molar absorptivity of the barium-thorin complex is
much higher than that of the free thorin. The decrease of the absorbence at
520 nm is, therefore, a measure of the amount of sulfate present in the
sample. Persson automated this method and applied it to the determination of
sulfur dioxide in air after absorption in a dilute solution of hydrogen
peroxide.
Most recently, C. Brosset and M. Perm (4) modified Persson's method and
applied this modified version to the analysis of water soluble sulfate of
airborne particulates. This version of the thorin method employs use of
automatic repipets which allow rapid analysis of aqueous sulfate without the
use of an autoanalyzer. This latter method has been utilized by us for the
analysis of sulfate in suspended particulate samples collected by means of a
dichotomous sampler. In the course of adopting this method of analysis, we
investigated, to some extent, the chemistry involved and we also compared
this against the methyl thymol blue method (MTB) , which is used routinely in
our laboratory for the analysis of similar samples. In this paper the results
of these studies are reported and discussed.
EXPERIMENTAL
INSTRUMENTS AND APPARATUS
A Perkin-Elmer Model 202 Ultraviolet-Visible spectrophotometer was used
throughout these studies. This instrument has a double beam design with a
photometric accuracy of 1% in the range of 0.0 - 1.0 absorbence units. The
peak-to-peak noise of the instrument is not more than +_ 0.005A at 1.0 absorb-
ence unit. The photometric cells were Coleman spectrosil cylindrical cuvets
with a transmission match of 1.5% at 240 mm. The cells had an internal light
path of 20mm. For this particular analysis the spectrophotometer was adjusted
as follows:
Both spectrophotometric cells were filled with reference solution (see
below), then placed in the spectrophotometer and the instruments were adjusted
to read zero at 520 nm. It is essential that the spectrophotometer be zeroed
in this manner before any absorbence measurement for standards or samples are
made. All spectrophotometric measurements were made against this reference.
The dispensing of samples dictated the need for specialized equipment
that would be both semi-automatic and precise. Hence, the apparatus shown in
Figure 1 was constructed. This apparatus consists of two piston operating
pipets arranged in a way that precise mixing of an aliquot of the sample with
the barium reagent can be achieved. For this, the tip of the double pipet
is first submerged in the sample solution. Then the stopcock is opened and
2.0 ml of the sample is drawn into the sample loop using syringe A. The stop-
cock is closed and 5 ml of the barium-p-dioxane solution is drawn into syringe
B which is then delivered via the sample loop into a mixing vessel. Finally,
the stopcock is opened and the liquid of syringe A is also expelled into the
mixing vessel.
62
-------
SYRINGE A
5 cc
2cc
STOP COCK
SAMPLE LOOP
(2ml)
PROBE
GROUND GLASS
VALVES
SAMPLE
BARIUM- P-DIOXANE REAGENT
GROUND GLASS VALVES
10 cc—
5cc —
1
SYRINGE B
Figure 1. Double repipet used for mixing the sample with barium-p-dioxane
reagent,
63
-------
REAGENTS
All the chemicals used for the preparation of the necessary solutions
were Baker analyzed reagent grade chemicals unless otherwise specified.
Barium stock solution was prepared by dissolving 525 mg of barium perchlo-
rate dihydrate in 250 ml of 0.1N HC10k. Barium-p-dioxane solution was prepared
by diluting 5.0 ml of the stock barium perchlorate solution to 500 ml with p-
dioxane.
Thorin dye solution was prepared by dissolving 250 mg of thorin in a
beaker containing 10 ml 0.01N NCIO^. After dissolution of the dye was com-
pleted, the contents of the beaker were transferred quantitatively to a 100
ml volumetric flask and diluted to the mark with 0.01N
The ion-exchange resin (Dowex 50W-X8) used to remove the interfering
cation must be in a fully swollen condition. This condition was achieved by
soaking the resin in distilled water and drying by suction.
Sulfate standard solution was prepared by appropriate dilution of stan-
dardized sulfuric acid solution to a final concentration of 0.005N. This
solution is equivalent to a SO^ concentration of 240 ppm. It was used for the
preparation of the standards in the thorin procedure. Usually the concentration
of the standards used was in the range of 0.5 to 10 ppm.
PROCEDURE
A 10 ml aliquot of the sample solution to be analyzed was transferred
into a 15 ml centrifuge tube containing 1 gram of the Dowex 50W-8X ion-
exchange resin. The sample solution was then shaken with the resin for one
hour. The samples were centrifuged for five minutes at 3000 rpm and 2.0 ml of
the supernatant solution of each sample was pipetted into 25 ml volumetric
flasks. After this, 5.0 ml of barium-p-dioxane solution was added into the
volumetric flasks and the flasks were capped to prevent excessive evaporation
of the p-dioxane. The pipetting and the mixing of the sample and barium-p-
dioxane solutions were performed with the aid of the double pipet mentioned
previously. Finally, an aliquot of 0.100 ml of the thorin solution, precisely
measured with the aid of a micropipet, was added into the volumetric flasks.
The resulting solutions were shaken well and then, after allowing time (three
seconds) for bubbles to clear, the absorbence was measured at 520 nm. Absorb-
ence measurements were made within three minutes after the addition of the
thorin solution to sample. Blanks were prepared by using distilled water
instead of standard solutions. Standard solutions and blanks were carried
through all the steps of the analytical procedure simultaneously with actual
samples. The reference solution was prepared by pipetting 4.0 ml of the
distilled water supernatant used in the preparation of blanks, 10.0 ml of
blank-p-dioxane (no barium) solution, and 0.200 ml of the thorin solution into
a 25 ml volumetric flask. The resulting solution was shaken well. This
reference is stable for more than eight hours.
RESULTS AND DISCUSSION
Early studies (3,4) have shown that the spectrophotometric method of
64
-------
analysis of sulfate by thorin was insensitive and inaccurate in the low con-
centration range (0 - 2.0 yg/ml SC%). This was found to be the result of a
substantial negative deviation of the response of the method to low con-
centrations of sulfate. This lack of response at the low concentration range
was explained by Brosset, et al., by assuming incomplete precipitation of
barium sulfate at this concentration range. To compensate for this, they added
sufficient amounts of sulfate to the thorin reagent solution thus eliminating
this problem, our studies have shown, however, that the same result can be
brought about by changing the thorin-to-barium ratio. Figure 2 shows a
typical calibration curve obtained by using our approach. The linearity of
this curve is excellent (correlation coefficient 0.9999). To further show
the consistency of this approach, the lower part of the calibration curve was
studied separately. Figure 3 shows this part of the curve which was studied
in increments of 1 yg sulfate (equivalent to 0.5 yg/ml in the extract). Even
this part of the curve shows an extremely high linearity (correlation coeffi-
cient 0.998) and, therefore, the reason for the non-linear response obtained
by Brosset, et al. can be attributed to incomplete complexation of the barium
by thorin which results in relatively large concentrations of free barium ions
in the solution. It is obvious that if free barium ions are present the
sulfate will not react stoichiometrically with the barium-thorin complex, thus
causing deviation from linearity. Figure 4 demonstrates the formation of the
barium-thorin complex. In this, the absorbence of solutions containing
constant concentration of thorin and variable concentrations of barium ions
was measured at 520 nm against a reference solution containing identical
concentrations of thorin. Therefore, the increase in absorbence represents
the formation of the barium-thorin complex, and the plateau shows the com-
pletion of this reaction. The thorin-to-barium molar ratio at the intercept
of the rising and horizontal lines is 1.9:1 which indicates the formation of
1:2 barium-to-thorin complex. Since the absorbence reading is a measure of
the formation of the complex, the difference between the observed and the
anticipated absorbence (intercept of the two lines) at the equivalence point
should be a measure of the degree of dissocation of the complex. Therefore,
this difference can also be utilized for the estimation of the concentration
of free barium ions at the equivalence point. If A is the maximum
absorbence obtained with excess of barium and Aequ^v *he observed at the
equivalence point, the concentration of the freeqbarium ([Ba \free> at the
eguivalence point will be
A
FBa++l = fBa^l . = £2HiZ [Ba(th;2] (1)
[Ba *free L Jeguiv A iim*v
lUcLX
and since [Ba(th)2] = [Ba
FtlclX
A - A . ,.
++ max equiv fna 1
iBa I free ~ A equiv (2)
where [Ba(th)2] and [Ba++] . are the maximum concentration of the barium-
thorin complex ancl the total concentration of barium at the equivalence point,
respectively-
in the experiment of Figure 4, A^ and A^^ are 0.775 and 0.700,
respectively, and therefore,
65
-------
0.7
0.6
UJ
u
< 0.5
00
cc
5
0.4
0.3
0.2
0.1
OBTAINED WITH 2.0 ml STANDARDS,
5.0 ml BARIUM-P-DIOXANE REAGENT
AND 0.100 ml THORIN SOLUTION.
X = 520 nm
6
8
10
12
>4
14 16 18 20
Figure 2. Calibration curve for the range 0-20 ygr
66
-------
0.8
0.7
0.6
g 0-5
<
00
QC
ffi
< 0.4
OBTAINED WITH 2.0 ml STANDARDS,
5.0 ml BARIUM-P-DIOXANE REAGENT
AND 0.100 ml THORIN SOLUTION.
X = 520 nm
0.3
0.2
Figure 3. Calibration curve for the range 0-5
67
-------
0.8
0.6
UJ
u
<
CO
oc
§ 0.4
0.2
FORMATION OF THE BARIUM-THORIN
COMPLEX. TOTAL VOLUME 14 ml;
MOLES OF THORIN IN 14 ml 8.68 X 10';
= 6.35 X 10-3M. THORIN-TO-
BARIUM MOLAR RATIO AT INTERCEPT
1.9; X-520 nm
0.1
m£ OF 6.85 x 10-
Figure 4. Spectrophotometric demonstration.
0.2
68
-------
tBa ]^M " °-107 [fla ]
free Jequiv,
Therefore, almost 11% of the barium at the equivalence point remains uncom-
plexed and it would cause non-linear response for low concentrations of sulfate.
Obviously this can be avoided by adding sufficient excess of thorin which will
complex all the available barium. As was mentioned previously, this approach
did eliminate the non-linearity problem yielding calibration curves consis-
tently similar to those presented in Figures 2 and 3, provided that the absorb-
ence reading of each solution was taken within three minutes after the addition
of the thorin reagent. When the absorbence was measured differently, the
obtained results were not as precise as those shown in Figures 2 and 3. This
was found to be caused by a change in the absorbence of the sample solutions
with time.
Figure 5 presents absorption spectra of barium-thorin (1:2) and plain
thorin solutions taken at 0, 10, 20, 30 and 40 minutes after their preparation.
During this period of time, the spectrum of the plain thorin solution remained
unchanged but that of the solution containing the complex showed a significant
decrease in absorbence. The decrease in the absorbence of the solution
containing the complex can be attributed to either catalytic decomposition of
the free thorin existing in solution or precipitation (or decomposition) of the
barium-thorin complex. It is rather evident that the decrease in the concen-
tration of the free thorin and barium-thorin species in a solution can be
estimated as follows by measuring its absorbence at two different wavelenghts.
(th)i' [Ba(th)2]i (th)2 ' [Ba(th)2]2
of the free thorin and the barium-thorin complex at wavelengths Xj and \2,
respectively, then the absorbence of these two wavelengths should be given by
equations (4) and (5) below
AI = £(th)i °(th) + £[Ba(th)2]1 °[Ba(th)2] (4)
A2 = Z(th)2 °(th) + &[Ba(th)2]2 C[Ba(th)2] (5)
where C and CT -, are the concentrations of the free thorin and of
(th) [Ba(th)2\
the barium-thorin complex, respectively.
The concentrations of the two species can then be estimated from the
absorbence values obtained at the two wavelengths, provided that the molar
z»h&s-\v*r*4- i*T»V^i^oc* P * £ r ...T <312u £r« y _*_»_ i T 3JT6 K.TIOWTI • -T/3S
absorptivities £(th)1' *(th)2 [Ba(th)2]i [Ba(th)2]2
values of the molar absorptivities of these two species can be estimated
independently by measuring the absorbences of solutions containing known
concentrations of plain thorin and of barium-thorin complex at the two wave-
lengths \i and X2.
Figure 6 presents graphically this type of estimation for an experimental
set similar to the one described previously in Figure 5. From this, it can
be calculated that the decrease of free thorin concentration during the 60
minute period, accounts only for 7.7% of the total decrease in concentration,
whereas the change in the concentration of the complex accounts for the rest.
69
-------
1.0
0.9
0.8
u, 0.7
u
1 0.6
DC
| 0-5
<
0.4
0.3
0.2
0.1
400
450
500
550
600
Figure 5. Spectral change of the barium-thorin complex with time.
70
-------
10
20
30 40
TIME (min)
50
60
Figure 6. Rate of decrease of concentration of the thorin and barium-thorin
species from solution.
71
-------
TMs indicates strongly that this reaction proceeds via a mechanism of pre-
cipitation of the barium-thorin complex. This indication is corroborated by
the fact that a substance of similar color with that of the complex pre-
cipitates out when solutions containing the complex are left overnight. For-
mation of the same reddish precipitate was also observed in the automated
thorin procedure developed by Persson (3). Because of this kinetic phenomenon,
the measurement of the absorbence should be made as soon as possible after the
addition of the thorin reagent and it should never be delayed for more than
three minutes. It should also be mentioned here that during these studies a
substantial increase in stability occurs when about three-fold excess of
thorin is used. Thus, we adopted the procedure in which the barium-to-thorin
ratio is 1:3. This assures complexation of all the available barium and it
also increases the stability of the samples. The results of the application
of this procedure to the analysis of synthetic samples, as well as comparison
against the modified methylthymol blue method (5) will be presented below.
Table I presents precision data obtained by analyzing multiple samples
at two concentration levels. The relative standard deviation for the low and
high concentration levels was found to be 5.2% and 1.6%, respectively. The
decreased precision at the low level can be attributed to a small change of
absorbence between the blank and the sample. Also the rate of precipitation
of the barium complex may have a role in this since the concentration of
that complex is considerably higher at the low concentration than at the
higher.
Table II presents results for the analysis of high-volume sample extracts
by the thorin and the automated methylthymol blue methods. Because of the
relatively high concentration of sulfate in these extracts, the samples had
to be diluted for the thorin procedure. The relative difference of the two
values was found to be in the range of 0.5 to 15%.
Table III show*data obtained from the analysis of aqueous extracts of
four-hour membrane filter samples. Although the correlation data shown in
this table indicate a bias of about 10%, the overall agreement appears to be
good especially for the higher concentrations.
In Table IV, x-ray fluorescence sulfur data obtained from the analysis
of dichotomous samples are compared with those obtained by the thorin method.
The 1000 series filter numbers correspond to the sample of fine particles
and the 3000 series to the total particles. As it is shown in Table IV,
better agreement was obtained in the analysis of the fine particles than in
that of the totals. This may be the result of particle size effect in the
x-ray fluorescence analysis of the total samples. Despite these differences,
the agreement appears to be fairly good.
All these analytical data have shown clearly that the thorin method
generates fairly accurate results when it is used properly. However, a close
comparison of this method with the modified methylthymol blue procedure (5)
revealed the following advantages of the latter over the thorin procedure:
(1) the cost per analysis is significantly lower for the automated methylthymol
blue method; (2) the manpower required is lower by a factor of five; and (3)
the expense for chemicals is minimal if it is compared with the cost of
72
-------
TABLE I. REPEATED ANALYSES OF SYNTHETIC SULFATE
SOLUTIONS BY THE THORIN METHOD
Present
4.8
4.8
4.8
4.8
4.8
9.60
9.60
9.60
9.60
9.60
9.60
9.60
9.60
9.60
9.60
(]ig SOl^)
Found
4.99
5.03
4.75
4.43
4.64
9.62
9.66
9.52
9.48
9.66
9.60
9.30
9.80
9.70
9.79
% Diff.
3.96
4.79
-1.04
-5.63
-3.33
0.2
0.6
-0.8
-1.3
0.6
0.0
-3.1
2.1
1.0
1.9
73
-------
TABLE II. RESULTS OF SULFATE ANALYSIS OF HIGH-VOLUME SAMPLES BY BOTH
THE THORIN AND THE AUTOMATED METHYLTHYMOL BLUE METHOD
Sample No.
63191
63192
63193
63194
63195
63196
63195
63292
63293
63294
23393
23394
23395
12478
12481
12484
Thorin
32.13
29.37
18.50
19.63
12.88
16.69
14.26
27.51
19.75
18.79
16.10
17.80
10.90
35.23
35.78
18.13
yg/ml
Technicon
31.3
30.9
16.1
17.0
14.0
16.8
13.7
29.7
19.8
17.6
18.8
19.8
11.8
34.1
31.8
17.9
% Diff. *
+ 2.6
- 5.0
+13.8
+14.4
- 8.4
- 0.6
+ 4.0
- 9.6
- 0.3
+ 6.6
-15.4
-10.6
- 7.8
+ 3.2
+11.8
+ 1.3
aver.
74
-------
TABLE III. COMPARISON OF SULFATE VALUES OBTAINED BY THE MTB (RANGE 1-10)
& THORIN (BROSSET) METHODS FOR 4-HOUR MEMBRANE FILTER SAMPLES
(LACS SITE A 15-19)
Sampling Date
125
130
133
136
139
142
145
148
154
157
160
SO% ug/nr
MTB
9.56
7.17
10.07
11.03
5.50
8.54
17.04
32.66
13.26
10.49
20.81
I
Thorin
11.93
6.41
12.40
10.89
7.30
10.31
17.28
34.51
14.41
11.29
20.59
LINEAR LEAST SQUARES DATA:
Intercept -1.13
Slope -0.898
Standard Error -1.03
Coefficient of Correlation -0.99
75
-------
TABLE IV. RESULTS OF SULFATE ANALYIS OF DICHOTOMOUS SAMPLES BY
X-RAY FLUORESCENCE AND THORIN PROCEDURES
Filter No.
1136
3136
1129
3129
1124
3124
1130
3130
XRF *
1.36
1.34
3.01
1.57
4.17
3.20
3.34
2.00
pg/ffl3 S
Thorin
1.46
1.66
2.89
2.29
4.61
4.01
3.64
2.19
% Diff.
7.1
21.3
4.1
37.3
10.0
22.5
8.6
9.1
Average 15.0%
* Samples analyzed independently by Lawrence Berkeley Laboratory
76
-------
chemicals used in the thorin method which in 1976 was approximately 70? per
sample. In addition, but most important, the p-dioxane which is used in the
thorin method is a health hazard and prolonged exposure can cause severe
damage to humans. For these reasons, an additional effort was made to deter-
mine better the equivalency of the two methods.
To check the correlation between the two methods, actual particulate
samples collected on 27 mm Teflon filters were analyzed. The filters were
extracted according to established procedure and aliguots of the extract
analyzed by both methods. The results given in Table V show the agreement
between duplicate analyses on the same extract by the thorin method. As
indicated in the column labeled "% Diff.", duplicate results may deviate from
the average value by a maximum of £ 6.4%. Table VI shows the comparison
between the average values from Table V and the values obtained by the MTB
method. The correlation between the two methods is better than 95% (correlation
coefficient 0.99) , although a bias of about 9% between the two methods does
exist. Finally, the two methods were compared in a determination of the back-
ground levels of sulfate (blank) in three different filtering media. For this,
individual filters were extracted with water and each extract was analyzed by
both methods. From Table VII it becomes apparent that the standard deviation
of the thorin results is much higher than those of the methylthymol blue
method. This can be attributed to the lower precision of the thorin procedure
because the sensitivity of both methods is about the same.
From all these experiments it was concluded that either method can be
used for the analysis of particulates collected by the dichotomous sampler.
However, because of the above mentioned advantages, the automated methylthymol
blue method was selected for the analysis of the Los Angeles Catalyst Study
dichotomous samples which are also analyzed for total sulfur by x-ray
fluorescence at the Lawrence Berkeley Laboratory at the University of California
(6). Statistical analysis of the results obtained by wet chemistry (modified
MTB) and x-ray fluorescence revealed good correlation between the two methods.
REFERENCES
L. Fritz, J.S. and Yamamura, S.S., "Rapid Microtitration of Sulfate" Anal.
Chem. 27_, 1461-1464 (1955).
2. Menis, O., Manning, D.L. and Ball, R.G., "Automatic Spectrophotometric
Titration of Fluoride, Sulfate, Uranium and Thorium", Anal. Chem. 30_,
1772-1776 (1958).
3 Persson G.A., "Automatic Colorimetric Determination of Low Concentrations
of Sulphate for Measuring Sulphur Dioxide in the Ambient Air", Air and
Water Pollut. Int. J. 10, 845-852 (1966).
4 Brosset, C. and Ferm, M., "An Improved Spectrophotometric Method for the
Determination of Low Sulphate Concentrations in Aqueous Solutions
Swedish Water and Air Pollut. Research Laboratory, 402, 24 Gothenburg,
Sweden (1974).
77
-------
TABLE V. DUPLICATE ANALYSIS OF SULFATE BY THE THORIN METHOD IN AQUEOUS
EXTRACTS OF PARTICULATE COLLECTED ON 37 mm TEFLON FILTERS
Sample No.
1204
3204
1205
3205
1206
3206
1207
3207
1208
3208
1209
3209
usr so-
1st Run
189.72
176.95
134.81
152.57
207.99
236.84
227.01
243.36
107.05
140.5
223.75
229.53
2nd Run
182.65
187.79
145.61
144.86
200.46
241.18
257.87
244.40
102.59
-
213.79
243.33
ug SOT;
Average
186.18
182.37
140.21
148.72
204.22
239.01
242.44
243.88
104.82
140.5
218.77
236.43
%Diff.
+ 3.8
- 5.9
- 7.7
+ 5.2
+ 3.7
- 1.8
-12.7
- 0.4
+ 4.3
-
+ 4.6
- 5.8
78
-------
TABLE VI. COMPARISON OF SULFATE DATA OBTAINED BY THE THORIN
METHOD AND THE AUTOMATED METHYLTHYMOL BLUE METHODS
Sample No.
1204
3204
1205
3205
1206
3206
3207
1208
3208
1209
3209
Thorin
186.18
182.37
140.21
148.72
204.22
238.01
243.88
204. #2
140.5
218.77
236.43
yg/SOij
MTB
178.53
166.67
135.37
149.82
205.98
227.54
228.80
101.60
127.86
197.55
219.30
% Diff.
+ 4.2
+ 9.0
+ 3.5
- 0.7
- 0.8
+ 4.9
+ 6.4
+ 3.1
+ 9.4
+10.2
+ 7.5
79
-------
TABLE VII. DETERMINATION OF THE BACKGROUND SULFATE LEVELS (BLANK) IN THREE
DIFFERENT FILTERING MEDIA BY THE THORIN AND THE SENSITIVE MTB METHODS
Total ygr SO *
Thorin MTB
Standard
Type of Filter Average Deviation Average Deviation
Gelman Acropore (120 nun) 45.76 9.38 43.42 4.15
Millipore Cellulose
Ester (102 mm) 24.52 4.85 28.25 0.22
Fluoropore (37 mm) 0.476 0.15 0.388 0.05
* Average of independent analysis of five filters.
80
-------
5. Colovos, G. Panesar, M.R. and Parry, E.P., "Linearizing the Calibration
Curve in Determination of Sulfate by the Methyl thymol Blue Method",
Anal. Chem. 48_ (12), 1963 (1976).
6. Giauque, R., private cojurounication, Lawrence Berkeley Laboratory,
University of California C1976).
81
-------
X-RAY FLUORESCENCE ANALYSIS OF LACS AEROSOLS
R.D. Giaugue, R.B. Garrett, and L.Y. Goda
Lawrence Berkeley Laboratory
Berkeley, California
G. Colovos and E. Parry
Rockwell International Air Monitoring Center
Newbury Park, California
C. Rodes and F. Burmann
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
The concentrations of eight elements: sulfur, lead, bromine, chlorine,
zinc, silicon, calcium, and iron in aerosol specimens collected during the
Los Angeles Catalyst Study program, have routinely been determined by x-ray
induced x-ray fluorescence analysis. Correlations among pairs of elements
have been observed. Also, at the sites used for the study, maximum levels of
particulate sulfur due to emissions from vehicles equipped with catalytic
converters are estimated. Good agreements typically obtained between total
sulfur from the method and wet chemical methylthymol blue sulfate data suggest
that over 90% of the particulate sulfur is present as sulfate. The capa-
bilities and limitations of the x-ray fluorescence technique, as it applies to
analysis of aerosol specimens collected during this program, is discussed.
Results from specific experiments are included.
INTRODUCTION
The analytical technique of x-ray fluorescence (XRF) lends itself to
simultaneous nondestructive determination of a broad range of elements in
aerosol specimens collected on filter media. The technique is quantitative
and high sensitivities are obtainable (^30 ng/cm2 using two minute counting
periods).
During the Los Angeles Catalyst Study (LACS) program, the concentrations
of eight elements were routinely ascertained for three types of specimens: •
total aerosols (4- and 24-hr); size segregated aerosols acquired with dichot-
omous samplers (24-hr); and Anderson cascade after filter specimens (24-hr)
for which particles <0.7\i are collected.
83
-------
DISCUSSION OF METHOD
Figure 1 is a schematic diagram of the XRF system used for analysis of
LACS aerosol specimens. Radiation provided by an x-ray tube impinges upon
the filter media containing the collected particulate matter. A fraction of
the photons, if of sufficient energy, produce vacancies in the inner shells
of atoms within the specimens, and in turn characteristic x-rays are emitted,
These x-rays are detected by a semiconductor detector, sorted by their
energies, and the elemental concentrations are determined from the intensities
of the x-rays.
In addition to the above interactions, a portion of the excitation
radiation is scattered. For analyses of aerosols collected on filter media,
the intensity of the scattered x-rays is often quite large in contrast to
those of the characteristic x-rays. In order to obtain high intensities,
near monochromatic excitation radiation with energies slightly greater than
the K or L absorption edges of the elements to be analyzed is used.
EQUIPMENT
The x^ray system in the direct excitation mode is shown in Figure 1. It
consists of a. low power silver anode x~ray tube and transmission filters. The
cryostat design allows the excitation radiation to pass through the vacuum
cryostat, permitting minimum x-ray attenuation, as an air path of only 1.5-mm
exists between the specimen and the detector window (0.002-cm beryllium). In
the determination of low atomic number elements: silicon (Si), sulfur (S),
chlorine (Cl), and calcium (Ca), the x-ray tube is operated at 9 KV and 200
Vamps, and a 0.00025-cm silver filter is used. The x-ray tube is operated
at 37 KV and 600 \iamps, and a 0.007-cm silver filter preceded by a 0.025-cm
vanadium filter are utilized for the determination of heavier atomic number
elements: iron (Fe), zinc (Zn), bromine (Br), and lead (Pb).
Table I list the theoretical detection limits (presuming no overlapping
x-rays) of the x-ray system for aerosol specimens collected on cellulose
ester membrane filters. These detection limits are based on two 2-minute
analysis periods. An air flow of 2.5 m3/cm2 (typical sampling rate for 24-hr
membrane filters) is presumed.
CALIBRATION METHOD
STANDARDIZATION
To calibrate a spectrometer for analysis of air particulate specimens
collected on filter media, it is necessary to determine the sensitivity of
the system for each of the elements to be analyzed. The sensitivity, S, for
an element i_ may be expressed:
S. = Ii/m. (1)
where S. = the sensitivity in counts/sec per ygr/cm2/
J. = The intensity, counts/sec, of the x-ray line from an elemental
1 thin film (matrix effects presumed to be negligible); and
84
-------
00
Scale
H H
icm
Collimators
Specimen
X- ray tube
Ag anode
Collimator
Figure 1. Schematic diagram of x-ray fluorescence system.
-------
TABLE I. THEORETICAL DETECTION LIMITS
Element Spectral Line
Si Ka 0.03 pgr/m3
S Ka 0.02
Cl Ka 0.03
Ca Ka 0.06
Fe Ka 0.09
Zn Ka 0.03
Br Ka 0.01
Pb Ka 0.03
86
-------
UK = the mass of the elemental thin-film, yg/cm2.
Relative excitation-detection efficiencies of the system for x-ray lines
from Ca, Fe, Zn, Br , and Pb were determined utilizing multielement solution
deposits on thin substrata (1) . One of the elements, copper (Cu) , in each
multielement standard solution served as an internal standard. Actual
standardization of the system was achieved using a single element thin-film
standard which was prepared by vacuum vapor deposition of pure Cu onto a thin
substrate.
Standardization for Cl was made with a basic physical approach (calcu-
lations using x-ray cross section and fluorescent yield data from the
literature) .
Thin film deposits of Si and S, for which matrix effects of low energy
x-rays from the elements would be negligible, are difficult to prepare.
Consequently, thick pure element discs of Si and S were used to standardize
for analysis (1) . The mass, m (\ig/cm2l of a thick element disc is
expressed: tnick
m thick = S-W/fV"*! + VSCV (2>
where y and y = the total mass absorption coefficient of the element
for the excitation and fluorescent radiations, respectively; and
fj and ^2 = the angles formed by the excitation and fluorescent radiations
with the surface of the disc.
Since m . represents the mass of which only 25% of the radiation
(excitation x Fluorescent) is not attenuated, the mass of the disc for
standardization purposes is 0.25 x m . . ,.
* thick
Cross checks on standardization procedures are reported in detail else-
where (1) . Standardization for Ca, Fe, Zn, Br, and Pb are estimated to be
accurate to within 1% or better, for Si and S to within 3% or better, and for
Cl to within 5%.
DETERMINATION OF OVERLAPPING SPECTRAL PEAK CORRECTIONS
The analysis program uses a fixed number of channels to measure charac-
teristic x-ray line intensities of each element to be determined. Peak over-
laps were established from x-ray spectra generated from thin deposits of
each element. S K and Pb M x-rays are the most critical overlap, as they are
quite close in energy.
Small particles containing Pb from automotive emissions can collect as
conglomerations on the filter. This has been observed with scanning electron
microscope (9EM) photographs by Dr. Thomas Hayes of Lawrence Berkeley Laboratory,
Additionally, large particles containing Pb (>lv) are collected when sampling
near the freeway. In each of the above two cases, particle size effects exist
for Pb M x-rays. It was determined that using a direct relationship between
high energy Pb L and low energy Pb M x-rays to predict the Pb M overlap with
S K x-rays produced S results which were low due to an overcorrection;
consequently, S K and Pb M x-rays are simultaneously unfolded for the S
87
-------
analysis. Figure 2 illustrates this overlap for a 24-hour aerosol specimen.
Listed in the figure are the element concentrations, \ig/m^, for this specimen
(2.5 m3 of air was sampled (cm2).
All other spectral peak overlaps were usually very minor, and, in most
cases, the corrections amounted to a few percent or less.
Approximately 2.5 m3 of air was sampled per 1 cm2 of filter area for
aerosol specimens collected over 24-hr periods. The amount of particulate
Pb collected usually varied from 2 to 30 yg/cm2. Consequently, a small
fraction of the low energy x-rays from Si, S, and Cl are attenuated by the
automotive particulate loading. Corrections were made in the analysis
program for absorption effects due to these loadings. From the Pb concen-
tration determined, using the higher energy Pb L a x-ray line (10.5 keV) ,
absorption corrections are calculated on the assumption that these loadings
are distributed uniformly across the filter and the composition of the Pb
compounds in the particulate matter is PbBrCl. For S K x-rays the absorption
correction was 3% for 5 yg, and 10% for 20 pg of Pb/cm2. Absorption
corrections applied for aerosol specimens collected over 4-hr periods, in
nearly all cases, were 2% or less.
ABSORPTION CORRECTION FOR SULFUR DUE TO PENETRATION OF SULFUR PARTICLES INTO
FILTER MEDIA
Giaugue, et al., have reported (2) that attenuation of S K x-rays, due
to penetration of the S containing particles into the filter media, is much
less than 5%. This was based on experimental determinations of S using
different energies of excitation radiation which acted as variable depth
probes.
More recently, Loo et al., (3), have substantiated our conclusions by
comparing XRF data for monodisperse CuSO^ and K2SOi+ aerosols collected on
both cellulose ester membrane filters and on very small pore Nucleopore
polycarbonate (surface collection media) filters.
In the XRF program used for the LACS program a 2% correction is made for
the attenuation of S K x-rays due to particle penetration.
ABSORPTION CORRECTION DUE TO PARTICLE SIZE EFFECTS
The source of Si is usually wind-blown soil dust. Particles containing
Si are usually greater than several microns. A sizeable fraction of the Si
K x-rays (1.74 keV) are attenuated in sedimentary particles containing
silicon. Initially, two dozen 24-hr membrane aerosol samples were analyzed
using different energies of excitation radiation (Zr, L, Ag L, and Ni K
x-rays), which served as variable depth probes to ascertain average particle
size corrections to be applied for the Si analysis. The actual procedure
used is reported elsewhere (2). These corrections for aerosol samples
collected on membrane filters were typically around a factor of 2 (i.e., over
half of the Si K x-rays are absorbed in the sedimentary particles).
For the S analysis, no corrections were made for particle size effects.
88
-------
2000-
o
O
00
VO
1000-
Figure 2.
2 3
X-ray energy (keV)
X-ray spectrum obtained on a 24-hr membrane filter specimen using
Ag L and Bremsstrahlung excitation radiation for the determination
of the elements Si, S, and Ca. Also illustrated are the overlapping
S K and Pb M x-ray spectral peaks.
-------
Typically, over 90% of the S containing particles are in the size range of
2 microns or less. This fact is substantiated in the size range of microns
or less. This fact is substantiated by S data acquired on samples collected
with the dichotomous air samplers and also by S data ascertained during the
ACHBX programs (4), which were carried out in California during 1972 and
1973.
SPECTRAL BACKGROUND
The background under each of the x-ray lines used for analysis is
referred to as the spectral background. It is related to the intensity of
the scattered excitation radiation. The background arises from two sources:
(1) the filter itself and (2) the beryllium window between the filter and the
x-ray detector. The contribution of each of these sources is easily
established by acquiring spectra from the beryllium window only, and from the
window plus a blank filter. The actual particulate mass loadings collected on
the filters are quite small in contrast to the mass of the blank filters.
RESULTS
Spectral data acquired for analysis are recorded on magnetic tape.
Computations removing spectral background and unfolding overlapping spectral
lines are made using a Control Data 6600 computer. The analysis program
requires less than 40 K of core space. Element concentrations, ygr/m , were
calculated using equation 3
C.. m
(\ig/m^)i= -^-x Ab±x^-x^ (3)
s i
where C. and C are the characteristic x-ray count rates from element i and
a thin-film standard, respectively; Ab. is the absorption correction (for
automotive particulate loadings); m is the mass of the standard, ygr/cn?2/
K. is the ratio of the x-ray system sensitivity factor for element i to that
oT the standard element; V is the air volume sampled, m3/ and A is the area,
cm2, over which the particulate deposit is collected on the filter.
Typically, analyses were carried out using two 2-minute counting periods.
Analytical results, including errors due to counting statistics, were put
on punch cards. A control sample was analyzed each day as check of the x-ray
system. Analytical results determined for six elements: Si, S, Ca, Fe, Br,
and Pb, for which counting statistics were not the limiting factor, were
reproducible to within ± 4%.
Additionally, quality assurance analyses have been made on 13% of the
aerosol specimens collected on membrane and cascade after filters. This was
achieved by determining the concentrations of five elements: Ca, Fe, Zn, Br,
and Pb, using a different x-ray spectrometer. Agreements to within 5% were
typically obtained when counting statistics were not the limiting factor.
Quality assurance analysis for elements with low energy x-rays: Si, S, and
Cl, was not possible with the second x-ray system.
Also, due to the small size limitations of dichotomous filter specimens
(15-mm OD punched out discs), quality assurance analyses could not be carried
90
-------
out using the other x-ray spectrometer.
Correlations among four pairs of elements have been observed. Tables II
and III list correlations established for air particulate specimens collected
on membrane filters between January 1976 through March 1976. For this time
period, 96 specimens were analyzed. Ratios and correlations established
between Ca/Si and Fe/Si are in very close agreement with data reported by
Flocchini et al. (5).
A comparison was made betwwen XRF S data and wet chemical methylthymol
blue sulfate data for 4-hr and 24-hr aerosol specimens collected on membrane
filters during the period 1/13/76 to 6/29/76. The data for the 4-hr samples
are shown in Figure 3. A linear regression of the 219 data points yields the
equation S - 0.32 x S0k = + 0.08 with a coefficient of determination of 0.95.
For the 24-hr samples, 56 data points were used and are shown in Figure
4. (Data for 33 additional specimens were not included as they had varying
degrees of brown spots on the reverse side of the filter. This can be
attributed to condensation getting onto the filter. In such cases, sulfate
is absorbed into the filter and low XRF sulfur data is realized.) A linear
regression yielded the equation S = 0.29 x SO^ = + 0.27 with a coefficient
of determination of 0.93. For these specimens the XRF sulfur data could be
slightly low due to heavier particulate loadings which increase x-ray
absorption effects.
At the sites used for the study, maximum levels of particulate S due to
emissions from vehicles equipped with catalytic converters is estimated.
Data obtained for the 24-hr air particulate specimens collected on Anderson
cascade after filters (particles •? 0.7) were used for the estimation.
Paired sets of samples collected at site A (approximately 30-m back and 1-m
above the west side of the San Diego freeway) and at site C (approximately
8-m back and 1-m above the east side of the freeway) between 7/15/75 and
4/6/76 were analyzed by XRF. For 38 paired sets of specimens the ratio of
the difference between sulfur, AS, and lead, APJb, were plotted ((S - S )/
(Pb - Pb ) = AS/AP2),) . A linear regression was performed on 45 of the 48
data points. The equation of the curve obtained was AS - 0.20 x APJb - 0.12
with a coefficient of determination of 0.75. Using a value of 3.6 yg/m3 for
Pb (this was the mean value 3.6 ± 1.6 for all cascade after filter Pb data),
we obtain a S value of 0.7 yg/m3 or 2 yg/m3 as SO^ ~. If an estimate of 20%
is used for the miles driven by vehicles equipped with catalytic converters,
an approximate average daily SO" impact of 9 yg/m3 is calculated for 100%
conversion to catalytic equipped vehicles. This estimation is most likely
high, as some of the particulate sulfur is produced by diesel vehicles. Also
particulate sulfur can be produced by other processes (e.g., gas-particle
interactions.)
For 26 paired sets of cascade after-filter specimens collected during
the time period 4/9/76 to 7/14/76, a linear regression was performed on the
AS/APJb ratios. The eguation of the curve obtained was AS = 0.12 x &Pb - 0.10
with a coefficient of determination of 0.46. This curve yields a lower SO$
impact than the curve for the earlier time period.
91
-------
TABLE II. CORRELATIONS FOR PARTICULATE SPECIMENS
Element
Si
Ca
Fe
Correlation with Si
(1
0
0
.00)
.93
.96
Ratio to
(1.00)
0.21
0.28
Si
TABLE III. CORRELATIONS FOR PARTICULATE SPECIMENS
Element
Pb
Br
Zn
Correlation with Pb
(1.00)
0.95
0.74
Ratio to Pb
(1.00)
0.40
0.03
92
-------
- 4 Hour LACS Membrane Data
S = 0.32 x SC>4 + 0.08
r2=0.95 219 paired values
5 10 15
3.0 x S /zg/m3
Figure 3. Comparison of XRF sulfur data with wet chemical sulfate data for
4-hr aerosol specimens collected on membrane filters.
93
-------
25-
eo
'if
CO
15
10
0
24 Hour LACS Membrane Data
S = 0.29 x SOJ + 0.27
r2 = 0.93 56 paired values
0
10 15
3.0 x S JJL g/m3
20
25
Figure 4. Comparison of XRF sulfur data with wet chemical sulfate data for
24-hr aerosol specimens collected on membrane filters.
94
-------
During October and November, 1976, 15 pairs of cascade after-filter
aerosol specimens were collected over 4-hr periods during peak traffic
hours with the hope of amplifying any relationship which might exist between
AS and APi>, but such was not evident.
Air particulate specimens collected with the dichotomous samplers were
collected on 37-mm teflon filters. For XRF, 15-mm discs were cut from the
edge of the deposit area. For large particle elements, si, Ca, and Fe, the
sum of the results for the two collected size fractions were usually low by a
factor of 2 in contrast to total filter results. This can be attributed to
non-uniform distribution of the deposit on the large particle stage. The
larger particles tend to be concentrated toward the center of the filter.
Similar comparisons of small particle elements: S, Br, and Pb, often had
discrepancies of 10 to 20% or more.
During October 1975, air particulate specimens were collected in tandem
on membrane filters. The back-up filters showed substantial losses of Cl
(0.57 vg/cm2), which is an impurity in the filter. The amount of Cl lost was
the same for both 4-hr and 24-hr collection periods. Additionally, there
was a small amount of S (0.05 \ig/cm2) collected on the backup filters, which
could be a result of SO2 gas interacting with the filter. Both the S and Cl
x-rays are less than 3 keV. Since the XRF method only measures those x-rays
which arise from the surface of the filter, the total change, for the entire
thickness or mass of the filter, would be much larger.
Four blank 2.54-OD membrane filters were each washed with 50-60 ml of
water, air dried, and analyzed. The Cl values decreased by 0.59 vg/cm2.
This value is in close agreement with the Cl lost in the tandem sampling
experiment.
ACKNOWLEDGEMENTS
The authors are indebted to Don Malone for designing and constructing
the spectrometer utilized for this study. We are grateful to Fred Goulding
and members of the Electronic Research and Development Group for upgrading
and maintaining the x-ray system. We thank David Gok for writing the basic
computer program used.
DISCLAIMER
This report was prepared for the U.S. Energy Research and Development
Administration under contract W-7405-ENG-48 with funding provided by the
Environmental Protection Agency. Any conclusions or opinions expressed in
this report represent solely those of the author(s) and not necessarily
those of the Lawrence Berkeley Laboratory nor of the U.S. Energy Research
and Development Administration.
REFERENCES
1. R.D. Giaugue, R.B. Garrett, and L.Y. Goda, "Calibration of Energy
Dispersive X-ray Spectrometers for Analysis of Thin Environmental Samplers"
in X-ray Fluorescence Analysis of Environmental Samples, T.G. Dzubay, Ed.,
Ann Arbor Science, p. 153-164 (1977).
95
-------
2. R.D. Giaugue, R.B. Garrett, L.Y. Goda, J.M. Jaklevic, and D.F. Ma lone,
"Application of a Low Energy X-ray Spectrometer to Analysis of Suspended
Air Particulate Matter", in Advances in X-ray Analysis, R.W. Gould et al.,
Ed., Kendall/Hunt Pub. Co., Vol. 19, p. 305-320 (1976).
3. B.Y. Loo, R.C. Gatti, B.Y.H. Liu, C.S. Kim, and T.G. Dzubay, "Absorption
Corrections for Submicron Sulfur Collected in Filters", ref. 1, p. 187-202.
4. R.D. Giauque, L.Y. Goda, and R.B. Garrett, "X-ray Fluorescence Analysis
of ACHEX Aerosols", Lawrence Berkeley Laboratory, Berkeley, Calif.,
Report LBL-4414 (1975).
5. R.G. FloCctdni, D.J. Shadoan, T.A. Cahill, R.A. Eldred, P.J. Feeney, and
G. Wolfe, "Energy, Aerosols and Ion-Excited X-ray Emission", in Advances
in X-ray Analysis, W.L. Pickles et al., Ed., Plenum Press, Vol. 18,
p. 579-597 (1975).
-------
FUELS SURVEILLANCE IN SOUTHERN CALIFORNIA
ANALYTICAL RESULTS AND METHODOLOGY
Robert H. Jungers
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
Several constituents in fuels are of primary interest, especially those
occurring naturally or those that are directly added during manufacture.
These constituents include lead, sulfur, manganese, phosphorus and hydrocarbon
content. Methodology for these analyses fall into one of three categories
(1) available and usable, (2) available but archaic, and (3) non-existent with
development necessary. Analytical data from various studies between southern
California and other regions in the United States as well as current analytical
methodology is discussed.
INTRODUCTION
The advent of the catalyst equipped vehicle which requires the use of
unleaded gasoline initiated the surveillance of motor vehicle gasoline
specific to the Los Angeles area. Initial interest focused on sulfur in
gasoline as a possible contributor to atmospheric pollution from engine
emissions. Next, attention was directed to lead and phosphorus content of
unleaded gasoline, a known poison from vehicle catalysts. Unleaded gasoline
by definition has limited lead content, with an allowed maximum of 0.05 grams
lead per U.S. gallon (g Pb/gal). Phosphorus must be limited to a maximum of
0.005 grams per U.S. gallon (g P/gal). Presently interest has expanded
beyond these areas to include consideration of aromatic content, manganese
content and other factors. Development of new and more efficient analytical
methodology was facilitated by widespread interest in having more available
data coupled with the constraint of a non expandable workforce.
DISCUSSION
ANALYTICAL
Table I shows the result of 136 unleaded gasoline samples collected
nationwide in the summer of 1974. The average sulfur content is 0.028 weight
percent (wt. %Sj with Region IX (San Francisco) having an average value of
0.018 wt.*S. Table II represents sulfur data from 140 samples taken in New
York State (NYS) from NYS police vehicles during 1976. These vehicles are
fueled from regular service stations along the NYS throughways. The average
97
-------
TABLE I. SULFUR IN UNLEADED GASOLINE SUMMER S FALL 1974
TOTAL SAMPLES 136
CONCENTRATION 0.028 weight percent sulfur
REGION NO. OF SAMPLES SULFUR WT. %
I 24 0.029
II 14 0.023
V 22 $).031
VI 23 0.035
VII 22 0.037
VIII 23 0.023
IX 8 0.018
98
-------
TABLE II. NEW YORK STATE FUELS SURVEY SULFUR IN UNLEADED GASOLINE
DECEMBER 1975 to NOVEMBER 1976
Date
Dec.
Jan.
Feb.
Mar.
Mar/Apr
Apr/May
May
May/Jun
Jun.
Jul/Aug
Aug/Sept
Sept/Oct
Oct/Nov
1975
1976
1976
1976
1276
2976
1976
1976
1276
1976
1276
1976
1276
No . Samples
10
10
10
10
10
10
10
10
10
10
10
10
10
Sample Number
Average Value
Range - O.QQ6
Avg. Sulfur (wt percent)
0.0161
0.0152
0.0136
0.0176
0.0172
0.0208
0.0199
0.0152
0.0178
0.0186
0.0208
0.0162
0.0180
- 140
- 0,017% Sulfur
to 0.043% Sulfur
99
-------
sulfur content is shown as 0.017 wt.%S. This value should be compared with the
average values of sulfur as shown in Table III for California. This Table
compares sulfur content of gasoline sampled in Los Angeles and San Francisco
at the same time in 1975. The average sulfur content of unleaded gasoline
ranges from 0.027 wt.%S in San Francisco to 0.034 wt.%S in Los Angeles. This
is nearly double the previously presented New York average of 0.017 wt.%S.
The analyses of both the regular leaded and premium leaded gasoline grades
collected in Los Angeles and San Francisco indicate higher sulfur content in
Los Angeles. Table IV presents data from unleaded gasoline sampled in southern
California between January and July 1976. The average sulfur value is 0.026
wt.%S with a monthly average range of 0.019 to 0.036 wt.%S. The average lead
value is 0.005 g Pb/gal with a monthly average range of 0.003 to 0.008 g Pb/gal.
The average aromatic value is 42.7% with a monthly average range of 33.6% to
46.3%. Aromatic content of gasoline is of great importance to both the refiner
and the environmentalist. Higher aromatic content can increase the octane
number of the gasoline. As organic metallic fuel additives are reduced the
addition of aromatics is one way to keep the octane number up and avoid the
customer complaint of "spark knock." The environmentalist sees the increase
in aromatics as a potential health hazard. Table V and Table VI present an
analytical comparison of gasolines collected in Los Angeles and San Francisco
during the winter of 1975-1976 and the summer of 1976. It should be noted
that on the the whole the sulfur content is higher in the Los Angeles samples
and that the aromatic content is higher in the summer than winter.
METHODOLOGY
A summary of analytical methodology as used in our laboratory is shown
in Table VII. Three separate groupings are listed: one, methods available
and used; two, methods available but extremely slow and tedious; and three,
methods developed in house which are designated as American Society of Testing
and Materials (ASTM) standard methods, potential ASTM standard methods and
survey methods used to expedite analyses for regulatory support.
The sulfur method by gas chromatograph flame photometric detector (GC/FPD),
has increased our capability by at least a factor of five, 12 to 60 samples
per day. The phosphorus method using the flameless atomic absorption spectro-
meter with a graphite furnace (AAS/HGA) is capable of surveying 80 samples
per day while 10 samples per day by the molybdate-hydraine method is average.
Table VIII indicates a comparison between the AAS/HGA and the molybdate-
hydrazine method in iso-octane standards. Figure I, a bar graph, presents a
comparison of 45 unleaded gasoline samples by both methods. The molybdate-
hydrazine method indicated all samples to be 0.001 g P/gal. while the AAS/HGA
method showed the following phosphorus content distribution.
27-0.001
7-0.002
5-0.003
1-0.004
5-0.005
Both methods indicate that the 45 samples are at or less than 0.005 g P/gal.
This is the maximum amount allowed in unleaded gasoline. Therefore, this is
100
-------
TABLE III. COMPARATIVE STUDY OF GASOLINE IN REGION IX FOR SULFUR CONTENT
SUMMER 1975
GRADE
Premium
Regular
No Lead
Premium
Regular
No Lead
SIGH
0.024
0.094
0.068
0.062
0.090
0.085
SAN FRANCISCO
LOW
0.001
Q.003
0.003
LOS ANGELES
0.002
0.033
0.010
BASIN
AVG.
0.009
0.035
0.027
BASIN
0.024
0.051
0.034
NO. OF SAMPLES
19
15
18
19
12
16
All results reported in weight percent sulfur.
101
-------
TABLE IV. SOUTHERN CALIFORNIA UNLEADED GASOLINE FUELS SURVEY
JANUARY - JULY 1976
Month
JAN
FEB
MAR
APR
MAY
JUN
JUL
No. of
Samples
2
3
27
5
16
14
3
% S
0.027
0.033
0.020
0.036
0.031
0.028
0.019
Avg. Values
g Pb/gal
0.007
0.007
0.005
0.008
0.005
0.006
0.003
% Aromatics
33.6
42.3
42.2
46.3
42.7
43.3
45.7
102
-------
TABLE V. MVMA FUEL SURVEY
Winter 1975 - 1976
LOS ANGELES
Prem Pb
Hi
Lo
Avg.
Rep Pb
Hi
Lo
Avg,
No Lead
Hi
Lo
Avg.
SAN FRANCISCO
Prem Pb
Hi
Lo
Avg.
Reg. Pb
Hi
Lo
Avg.
No Lead
Hi
Lo
Avg.
% S
.107
.018
.043
.145
.006
.052
.058
.006
.025
.021
.002
.014
.081
.00.7
.035
.038
.002
.017
g Pb/gal
3.830
.790
2.237
2.050
.370
1.259
.007
.002
.004
3.670
1.180
2.277
1.970
.420
1.331
.015
.003
.007
RON
99.2
93.8
98.0
96.0
93.1
94.0
96.1
90.8
92.3
100.2
96.4
98.6
95.9
93.2
94.4
96.6
91.1
92.9
% Aromatics
34.8
23.4
28.8
32.7
23.3
27.4
37.3
27.5
32.7
37.4
18.6
28.4
32.8
23.9
27.1
38.6
19.6
30.6
103
-------
TABLE VI. MVMA FUEL SURVEY
Summer 1976
Los Angeles
% S g Pb/gal RON % Aromatics
Reg. Pb
Hi
Lo
Avg.
No Pb
Hi
Lo
Avg.
.143
.029
.066
.048
.012
.026
2.810
0.520
1.487
.007
.001
.004
94.1
92.0
93.3
95.8
91.0
92.6
35.4
23.7
31.0
44.1
26.9
34.1
SAN FRANCISCO
Reg. Pb
Hi .077 3.360 95.3 34.7
Lo .016 .600 93.7 25.7
Avg. .042 2.056 94.4 28.9
No Pb
Hi .046 .018 96.2 42.5
Lo .012 .001 91.5 28.5
Avg. .030 .006 92.9 33.8
104
-------
TABLE VII. ANALYTICAL METHODS
I Available and Used
Fluorescent Indicator Absorption ASTM D - 1319
DEPENTANIZATION ASTM D - 2001
API Gravity § 60 F ASTM D - 287
Reid Vapor Pressure ASTM D - 323
DISTILLATION ASTM D - 86
Sulfur, Lamp ASTM D - 1266
II Available - Slow and Tedious
Sulfur, Lamp ASTM D - 1266
Phosphorus, Colormetric ASTM D - 3231
Lead, Gravimetric ASTM D - 526
III Developed In-House
Lead, AAS ASTM D - 3237
Lead, FTK &S™ *> - 3348
Sulfur, GC/FPD
Phosphorus, AAS - HGA
Manganese, AAS-
105
-------
TABLE VIII. COMPARISON OF AAS/HGA
MOLYBDATE-HYDRAZINE METHODS
G/GAL P G/GAL P
MOLYBDATE-HYDRAZINE EGA
METHOD SCREENING METHOD
.002 .004
.002 .004
.002 .002
.002 .002
.001 .002
.001 .001
.001 .002
MOLYBDATE-HYDRAZINE VS. SCREENING METHOD USING ISO OCTANE
STANDARDS
106
-------
30
Figure 1. Phosphorus in gasoline HGA screening method.
^r ^f
o n
n" ^~ *•'
°-
s
^f
tr\
\ii
LL
0
010
z
0
x
x
x
^
x
/
f
/
/
/
/
/
x
x
x
x
x
X
x
X
x
x
x
^
F
x
X
X
I
I
I
I
gP/gal xio"3
-------
considered to be an acceptable screening method. Manganese as an antiknock
additive has become increasingly attractive to refiners as lead content is
reduced by law. The manganese additive, Ethyl MMT or methyl cyclopentadienyl
manganese tricarbonyl, is in ever increasing use. The present recommended
use by Ethyl is 0.0625 g Mn/gal or 1/16 g Mn/gal. Table IX presents com-
parative data obtained by additions to unleaded gasoline at the low end of
concentration. Comparative data between our method and the method used by
Dupont showed comparability at the average recommended concentration. This
method and variations of this method are being collaboratively tested by ASTM
as a future standard method.
CONCLUSION
Our laboratory is confident that as the need arises for more data on fuels
and fuel chemistry analytical methodology we can meet this challenge.
108
-------
TABLE IX. RECOVERY OF SOLUBILIZED AQUEOUS MANGANESE
STANDARDS FROM GASOLINE
BRAND
ARCO
CITGO
PHILLIPS 66
TEXACO
MARATHON
EXXON
STANDARD
UNION 76
LOCATION
OREGON
TEXAS
N. CAROLINA
TEXAS
INDIANA
NEW YORK
GEORGIA
CALIFORNIA
GMS/GAL
AMOUNT
ADDED
0.004
0.001
0.004
0.004
0.001
0.001
0.001
0.001
GMS/GAL
AMOUNT
RECOVERED
0.004
0.001
0.003
0.003
0.001
0.001
0.001
0.001
COMPARISON OF DUPONT & MOLECULAR
CHEMISTRY LABORATORY METHODS
GRAMS PER GALLON
MANGANESE
DUPONT METHOD
GRAMS PER GALLON
MANGANESE
MOLECULAR CHEMISTRY
METHOD
0.078
0.026
0.072
0.032
109
-------
MEGAVOLUME SAMPLER (MARK I) FOR COLLECTING
RESPIRABLE PARTICULATES AT THE LACS SITE
Ralph I. Mitchell and W.M. Henry
Battelle, Columbus Laboratories
Columbus, Ohio
ABSTRACT
A large volume sampler for the collection of gram quantities of respi-
rable particulates has been designed and calibrated. The sampler, which is
being used at the Los Angeles Catalyst Study site, has a daily sampling rate
of approximately 1,000,000 cubic feet (28, 143 m3) and has two impactor stages
with cutoff sizes of 3.6 and 2.1 }im. Particulates smaller than 2.1 ym are
collected on high purity aluminum precipitator plates which are removable
for sample recovery. The main advantages of the sampler are that samples
are collected in three size fractions, there is minimal contamination from
the collection substrate, and samples are partitioned into respirable
fractions of sufficient size for detailed chemical analysis and bioassay
screening.
INTRODUCTION
The hazard associated with the inhalation of airborne particulates
depends upon their physical, chemical and/or biological properties. These
properties determine their deposition site within the respiratory tract and
their interaction after they have been deposited. One of the major mechan-
isms for particle removal within the lung is by ciliary action. Non-soluble
particles which penetrate beyond the ciliated portion of the lung ("respirable
particulates") will remain in contact with the lung tissue for extended time
periods and consequently may be capable of producing very deleterious effects.
In order to correlate potential health hazards of air pollutants with
epidemiological data, it is necessary to not only have a sampler which has
size-selecting characteristics of the human respiratory tract but one which
will collect a sample large enough for detailed chemical analysis with
minimal contamination from the collection substrate. The quantity of
particulates collected in most 'respirable' dust samplers is not sufficient
for bioassay and, for many of the trace materials, more of the specific
component can be found in the collection substrate than in the collected
sample. The sampler described herein has a sampling rate of at least 15 times
greater than commercially available samplers. (1,2) and the total sample is
collected either on a Teflon or high purity aluminum substrate which minimizes
any potential contamination.
Ill
-------
Another major advantage to this sampler is that the collection charac-
teristics of the first stage duplicates the ACGIH (3) respirable curve. The
other samplers which use this distribution as their design criteria have
limited sampling rates (4, 5), and the sample fraction of interest is collected
on a filter substrate.
Because of the urgency of the catalyst problem the megavolume samplers
were taken to the Los Angeles Catalyst Study (LACS) site before they were
thoroughly evaluated and calibrated in the laboratory. However, over one year
of operation has been obtained with the original prototype sampler and the
experience obtained has shown several design changes which would improve the
ease of operation of the sampler. The current version of the sampler (Mark
II) has incorporated these design changes. This paper presents some of the
calibration and collection data of the LACS sampler as well as some of the
collection data with the Mark II version to show some of the overall merits of
such a sampling system.
DESCRIPTION OF SAMPLER
The sampler body is constructed of stainless steel and is approximately
4-1/2 feet tall and 2 feet square. Figure 1 is a photograph of one of the
LACS samplers located beside the San Diego Freeway. Figure 2 is a schematic
diagram which shows the principle of operation. The sampler is essentially a
large cascade impactor which utilizes an electrostatic precipitator to collect
the fines which are not impacted on the last collection stage. Air is pulled
through the sampler by means of a 1000 cfm high pressure blower; however, the
flow characteristics of the blower are such that the total flow is reduced to
690 cfm with the operating pressure drop. The blower is driven with a 1 hp
motor which pulls up to 12 amperes at 115 volts.
Air enters the sampler at the top perimeter which serves as the scalping
stage and has a cutoff size of approximately 20 ym. The air then enters the
impaction plate assembly containing four 16-gage stainless steel plates with
slots which serve as impactor jets and collection targets. The first two
plates have a cutoff size of 3.6 ym. The jet stage has 110 slots which are
1/4 x 1 inch with rounded corners. The distance between the collection stage
and the jet stage is 1/2 inch and the target width is 1-1/8 inch. The next
two plates (second stage) have a cutoff size of 2.1 \im and they are 1/8 of an
inch apart. The jet stage consists of 374 rectangular slots 1/16 x 1 inch.
The impaction surface for each slot is 7/16 of an inch wide. The impaction
plate assembly is constructed of anodized aluminum and can be disassembled
by the removal of two screws holding an end plate. Both the impaction jets
and collection plates are Teflon coated to minimize substrate contamination
of the sample. This also facilitates in the removal of the sample.
After passing through the impaction plates, the smaller particles are
collected by means of a specially designed precipitator plate assembly which
accommodates for ready removal of the plates for sample recovery. These are
55 high purity aluminum collector plates approximately 11 x 17 inches and in
the latest version these are coated with a conducting Teflon film.
There is a precipitator ionizing section which has a positive potential
112
-------
sampler located at 9 LACS site.
113
-------
3.5 /zm stage
-1.7/xm stage
Scalping stage
cut off - 20/xm
Electrostatic
precipitator
Figure 2. Schematic of high-volume air sampler.
114
-------
of approximately 8000 volts. The precipitator plates also carry the same
voltage and each plate alternates from positive to negative. The clean air
passes through a transition section which connects directly to a high
pressure blower.
CALIBRATION OF IMPACTOR STAGES
The main criterion of this sampler was to have the collection charac-
teristics of the first stage duplicate or closely approximate the ACGIH
respirable curve. In order to obtain the high volumes required, it was
necessary to use long slots. The impaction characteristics of these slots
were determined by means of a miniature impactor which would accommodate
either one or two slots of the same size of those to be used in the full-scale
sampler. This impactor was designed so that the jet-to-slot spacing could be
varied over a wide range. Particulates which did not impact on the impaction
stage were collected on a 142-mm glass fiber backup filter.
The impaction efficiency curves for the impactor jets were determined
using monodispersed aerosols of dibutyl phthalate containing a highly
fluorescent dye. The aerosols were generated using a Bergman-Liu generator
and the particles were passed through a heated duct to insure complete
evaporation of the diluting solvent (ethanol). The particle size of the
aerosol was continuously monitored by sampling with a cascade impactor. The
impactor sampling showed that the aerosols were monodispersed and in the
approximate size range. The impactor slides were coated with "Aerosol OT"
and the exact size of the droplets before impact could be determined with a
microscope knowing the flattening coefficient of the droplets.
Figure 3 show the impaction characteristics of the 1/4 inch by 1 inch
slots used in the first stage of the megavolume sampler. This figure shows
the effect of changing the target width from 1/2 to 1 1/8 inch^ With the
1 1/8 inch wide impaction target the impaction parameter, V , was found
to be 0.35 which yields a cutoff size of 3.6 \im. The impaction parameter,
1/2
V , for the second stage which has a target surface which is 7 times wider
than the jet width is 0.45. It produces a cutoff size of 2.1 ym.
FIELD COLLECTION DATA AND COMPOSITION ANALYSIS
The following presents the results obtained from the first four days of
use of the sampler at the LACS site. One sampler was placed upwind and the
other sampler placed on the downwind side of the freeway. During this sampling
period a 24 hour sample was taken at each site with a special Battelle cascade
impactor which was designed for the particle size determinations of automobile
exhaust.
Table I shows the weights obtained for the collection stages for the two
sampling sites. The data show that the particulate concentration of the air
at the down stream site is about 20 percent greater than at the upstream site.
Figure 4 shows the particle size distribution of the pollutants at the
upstream site obtained with the special cascade impactor. Figure 5 shows
similar data obtained at the down stream site. On both of these figures the
115
-------
IUU
80
"c
0)
o
l_
(U
Q.
>; 60
o
c.
0)
u
s—
•*-
LU
.2 40
"o
-------
TABLE I. PARTICULATE COLLECTION AT L.A. SITE
Site Location
Size Fraction
Time Period
Mass, g
Percent
Site A
(Upwind) 3.6-20 pro (first stage) 'v
2.1-3.6 pm (second stage) ^
<2.1 pm (electrostatic plates) *v
TOTAL
4 days
4 days
4 days
(100
(100
(100
hr)
hr)
hr)
1
0
3
5
.440
.845
.370
.655
25
14
59
100
.5
.9
.6
.0
Site C
(Downwind)
3.6-20 pm (first stage)
2.1-3.6 pm (second stage)
2.1 pm (electrostatic plates)
TOTAL
days (100 hr)
days (100 hr)
days (100 hr)
1.245
0.945
4.495
6.685
18.6
14.2
67.2
100.0
-------
a>
CO
*c.
o>
o
0>
CL
0)
E
3
yy
98
95
90
80
70
60
50
40
30
20
10
5
2
1
0.5
0.2
O.I
0.05
O.OI
^rr
^- •
c
— - -
)
— —
*^
^x
X"
^
^-n
>^
A H
o s
^
*a
gh-vo
peciol
exhou<
^
^"
lume
auto
>t irr
^
so
mo
tpai
X
mp
bill
cto
X
le
>
r
xS
r
^
3
*^^
0.2 0.3 0.4 0.6 0.8 I 2346
Equivalent Particle Diameter, micrometers
8 10
Figure 4. Particle size distribution of upwind pollutants at San Diego
Freeway site.
118
-------
c
a>
o
Q?
O)
3
O
99
98
95
90
80
70
60
50
40
30
20
10
2
1
0.5
0.2
O.I
0.05
0.01
0
•WHMHMBBBI^^IIM
^-cr
••^v^^^^^^^^—
••••••••••••
. '
•^^••••i
-X
•^^^
'r
••••
X**
^^M
••M—
^M
x
•M
—^ —
c
./•*
^^
r
-
^•^^^^^^B
> ^
^
^•^^^^
^^••MBI
>x
^M^
X
^••n
X"
•MM
^
A High-volume sampler
O Special automobile
exhaust impac
tor
•••
{
^Ml
^
.2 0.3 0.4 0.6 0.8 1 2 34 68 1C
Equivalent Particle Diameter, micrometers
Figure 5. Particle size distribution of downwind pollutants at San Diego
Freeway site.
119
-------
data obtained with the megavolume sampler is shown. In both cases the particle
size data obtained with the megavolume sampler agrees with that of the special
cascade impactor.
Table II shows some preliminary chemical analysis obtained with these
samples. Since these samples were taken, several pounds of sample have been
collected. The tentative conclusions which can be drawn from these data are:
(1) Several elements show concentration change with size. Those of
crustal origin such as Iron (Fe) are higher in the larger size
ranges. Those of man's activities such as Sulfur (S), Lead (Pb)
and organics are higher in the small sizes.
(2) A traffic effect is shown by the Pb contents of downwind versus
upwind collections.
(3) Nitrite (NO^) accounts for ^70 percent (weight basis) of the total
Nitrogen (N) in the large particle size while ammonium (NH^~) accounts
for ^>75 percent of the total N of the very fine (>2.1 um sizes.)
(4) Sulfate (S0i+) accounts for most of the S in the two larger particle
sizes,_while the total S, before and after benzene extraction, and
the SOi+ to total S ratio in the fine fraction indicates possibly
1 to 2 percent organo sulfur compounds in the < 2.1 pm size.
COLLECTION CHARACTERISTICS OF THE MARK II SAMPLER
The second generation of the sampler has incorporated most of the design
limitations of the prototype sampler which evolved during its field use.
The major problem area has been the reliability of the electrostatic power
supply. This part of the sampler has been completely redesigned. Another
change has been in the cutoff sizes of the impactor stages. The current
cutoff sizes are 3.5 and 1.7 \im.
REPRODUCIBILITY
REPRODUCIBILITY TESTS USING AMBIENT AEROSOLS
In order to check repeatability and reproducibility of the two samplers
in the field with ambient aerosols, the samplers were positioned about 20 feet
apart outside the laboratory. Because of building obstruction and possible
variance in meterological condition, the sampler positions were interchanged
daily in order to minimize sample bias. The samplers were operated for 14
days. Table III shows the weights of particulates removed from each collec-
tion surface. The alcohol scrub was used to recover ami particulate residue
left on Stage 3 after the normal scrape recovery. These results show that
the two samplers have similar collection characteristics.
REPRODUCIBILITY TESTS WITH FLUORESCENT AEROSOLS
After determining that essentially identical data could be obtained
with the full-scale sampler as compared with single slots, the sampler was
120
-------
TABLE II. ANALYSES OF COLLECTED ^ARTICULATES FROM L.A. BACKGROUND
CATALYST STUDY SITE - RESULTS IN PERCENT
Total
Site A Pb S Fe
Upwind: First Stage 0.86 2.0 3.6
Second Stage 1.13 5.6 2.9
Electrostatic 0.93 8.7 ,,.1.0
[8.9]Cb}
Site C
Downwind: First Stage 2.43 2.0 3.8
Second Stage 3.16 4.7 3.1
Electrostatic 3.74 5.5 1.9
Benzene-
Soluble . +
Organics S SO= C H N NO~i NHi±
8.0 2.4 6.45 7.3 1.5 2.2 7.75 0.87
9.0 6.5 16.8 7.9 2.4 2.5 2.85 2.55
16.6 9.4 21.3 6.5 4.3 3.1 6.5 3.55
4.3 1.9 6.3 8.6 1.7 1.9 7.4 0.69
8.8 5.2 14.6 10.6 2.3 2.3 5.5 2.17
16.0 6.1 17.4 9.4 3.4 2.5 3.0 2.45
(a) After benzene extract.
(b) Recheck determination.
-------
TABLE III. REPRODUCIBILITY OF THE MARK IT SAMPLERS
IN THE COLLECTION OF AMBIENT AEROSOLS
Sampler A Sampler B
1st Impactor Stage C3.5 to 20 ]im} 1.177 g 1.074 g
2nd Impactor Stage (1.7 to 3.5 \m} 0.996 g 0.995 g
3rd ESP Stage (_<1.7 ]im) 9.110 g 9.073 g
Alcohol Scrub of Precipitator Plates 0.800 g 0.860 g
Total 12.083 g 12.002 g
122
-------
challenged with a fluorescent aerosol. The aerosol was generated by means
of a conventional aerosol can containing the dye dissolved in Freon .
The aerosols were sprayed into the wind tunnel containing the massive
volume samplers and a special cascade impactor. A high-volume filter sampled
the effluent from the megavolume samplers in order to obtain the overall
collection efficiency.
Figure 6 is a particle size distribution plot of the fluorescent aerosol
generated by Can "E" as obtained by the Battelle cascade impactor. This plot
shows that the mass median diameter of the aerosol is 2.8 pro. The plot also
shows the two data points obtained with the massive volume samplers. Twenty-
eight percent of the aerosol mass was collected on Stage 1 (3.5 ]im) and the
cascade impactor distribution shows about 32 percent is larger than that size.
This is very good agreement considering some of the material is removed by
the scalping stage. The second data point for 1.7 ym fell exactly on the
distribution curve (78 percent of the particles were larger than that size).
COLLECTION EFFICIENCY OF PRECIPITATOR UNIT
Another change in the megavolume sampler was to switch from a positive
to a negative ionizing corona and to coat the precipitator plates with a
conducting type Teflon to minimize the difficulty in removing the collected
'particles from the precipitator plates. The collection efficiency of the
megavolume sampler containing Teflon-coated percipitator plates was
determined with the Minnesota aerosol analyzing system. Ambient air was
sampled and the particle concentration was measured at the sampler inlet and
outlet.
Figure 7 is a plot of the collection efficiency obtained over the
particle size range of 0.013 to 0.75 yra. This plot shows that the average
collection efficiency for the submicron particles is over 90 percent. For
this test the voltage on the precipitator plates was 5500 and the ionizing
wire carried a negative potential of 10,000 volts.
CONCLUSIONS
The massive volume sampler provides a means by which gram quantities of
particulates can be collected from low concentration ambient atmospheres.
The collected particulates are classified according to the ACGIH criteria for
respirable aerosols and there is a further classification to separate the
primary and secondary aerosols. The sampler has a high collection efficiency
and minimizes the possibility of contamination from the collection substrate.
Samples obtained using this collector are ideally suited for an iterative
biological- chemical testing scheme for the identification of possible
causative agents for airborne insult to human health.
REFERENCES
1. Andersen 2000 Inc., Atlanta, Ga., "Sales Literature", March (1972).
2. MISCO, (Macrochemical Specialities Company) , Berkeley, California,
Sales Literature.
123
-------
g>
i
4>
o>
3
E
3
o
99
98
95
90
80
70
60
50
40
30
20
10
5
2
I
0.5
0.2
O.I
0.05
0.01
Stage 2
0.2 0.3 0.4 0.6 Q8 I 2 34
Droplet Diameter, micrometers
8 10
Figure 6. Particle size distribution data of fluorescent aerosol package
Can "E".
124
-------
c
O)
o
J_
O)
D.
100
80
O
O
Ul
u
n 60
UJ
.2
o
2 40
O
O
20
0.01
0.05 0.10
Particle Size, micrometers
0.50
1.0
Figure 7. Collection efficiency of Mark II massive-volume sampler obtained
with the Minnesota aerosol analyzing system on ambient particles.
-------
3. Aerosol Technology Committee, AIHA, "Guide for Respirable Mass Sampling,
AIHA Journal, March-April (1970).
4. Lippmann, M. and W. B. Harris, "Size Selective Samplers for Estimating
'Respirable Dust Concentrations", Health Physics, Vol. 8, pp 155-163
(1962) .
5. Partride, J. and H. J. Ettinger, "Calibration of a Spinning Disc Aerosol
Generator and Two Stage Samplers, Report LA 4066, Los Alamos Scientific
Laboratory, Los Alamos, N. M. (1969).
126
-------
COMPOUND IDENTIFICATION OF LOS ANGELES CATALYST STUDY SAMPLES
Edward P. Parry and Leo E. Topol
Rockwell International
Newbury Park, California
M.D. Lind, A.B. Harker, and R. M. Housley
Rockwell International
Thousand Oaks, California
and
Franz J. Burmann
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
With the use of a Battelle massive sampler, discrete samples have been
collected at two sites, upwind and downwind of the San Diego Freeway. These
samples have been analyzed by X-ray diffraction, X-ray emission spectroscope
using a scanning electron microscope, photoelectron spectroscopy, and wet
chemistry. The Battelle sampler separates the samples into three size frac-
tions, the first two separations being by impaction and the final separation
by an electrostatic precipitator. Material from all three stages has been
analyzed. The electrostatic precipitator samples have also been compared
with Teflon filter samples prepared simultaneously. In addition, laboratory-
generated aerosols were prepared to evaluate their behavior as a function of
temperature in the photoelectron spectrometer and their detectability by X-ray
diffraction. Several sulfates have been identified in the Battelle samples.
In addition, artifact formation in the Battelle sampler has been noted for
long-term samples. The results are presented and discussed.
INTRODUCTION
The Los Angeles Catalyst Study (LACS) is a program designed to deter-
mine possible increases in sulfate levels as a result of automobiles equipped
with catalytic converters. Since the effect of sulfate on health varies
with the individual sulfate compound, the identification of the specific
sulfate species in participate samples collected in areas of heavy automobile
traffic is desired.
Some preliminary work has been done on samples collected with a Battelle
massive sampler using various analytical techniques. While lead sulfate was
127
-------
present, no specific ammonium compounds could be identified in the collected
particulate. Results also strongly suggested that artifact formation was
occurring in the Battelle sampler* Some experiments have been done in an
attempt to define this.
In this paper, the nature and limitations of the techniques used will be
discussed and a potentially useful technique to infer compound information
using photoelectron spectroscopy will be described. Finally, some of the
preliminary results will be discussed.
TECHNIQUES - NATURE AND LIMITATIONS
The techniques which have been employed in this study and some brief
comments about them are given below and are summarized in Table I.
X-RAY DIFFRACTION (XRD)
X-ray diffraction, employing Debye-Scherrer powder patterns, was used
to identify the crystalline components of aerosols. Amorphous substances do
not yield XRD lines. The aerosol sample can be examined directly on the
filter substrate or can be separated from the collector and packed into a
glass capillary. The use of the filter has the obvious advantage of minimizing
handling and artifact formation but it has two problems. First, the Teflon
filters generally used yield XRD lines, which complicate the analysis. Second
the amount of sample exposed to the x-ray beam may be too low to produce
detectable lines. We have found that about 10 pg/cm2 of (NHt+) 2^04. will give
strong, well-defined diffraction lines. Increased detectability of crystalline
aerosols collected on a filter can be obtained either by folding the filter
several times so that the x-ray beam penetrates more layers of sample, or by
separating the deposit from the filter which concentrates the sample to a
smaller volume. This separation can be accomplished by dissolving away the
filter, but one must be careful that the sample is not altered in the process.
In addition to the problem of decreasing intensity of the line with the quantity
of compound present, the lines broaden as the crystallites become small. The
crystallite size, where line broadening commences, will vary with each com-
pound and must be determined but, generally, is around 0.03-0.1 pm. Thus, for
nucleation mode particles, which are about 0.04 pm and are expected to be
present in high concentration near sources such as a freeway, XRD may not be
amenable. However, this must be determined. If nucleation mode particles of
ammonium nitrate and sulfate do not yield identifiable XRD patterns, then the
technique is limited to the accumulation mode size particles or recrystallized
material.
Although XRD is one of the few tools available to us that give direct
evidence as to the identity of a component, the analysis is complicated since
aerosols are composed of numerous constituents and the XRD patterns are quite
complex. To simplify this task, elemental analysis is utilized. Another
technique, fractionation of the mixture with solvents specific to certain
sulfates can be used to simplify the diffraction patterns. The negligible
formation of artifacts in this fractionation process must be established.
128
-------
TABLE I. TECHNIQUES AND PERTINENT INFORMATION
Technique
Sample Required
Limitations
Information Obtained
X-Ray Diffraction
(XRD)
X-Ray Photo-
electron
Spectroscopy (XPS)
Scanning Electron
Microscope with
EDAX
Wet Chemistry
(Inorganic)
Wet Chemistry
(Organic)
10-50 vgm/cm2
On filter or
without substrate
5 vigzn/cm2
2-5 ygm/cni2
10 m gins/cm2
For crystals of "^ 0.1
Jim size. Mixture of
many compounds yeild
complex pattern.
For surface analysis
(10-40A), clean substrate
stable over wide tempera-
ture range (non-backed
Teflon).
For analysis (to 0.5
\im) , elements heavier
than Na.
Identification of compound
in bulk sample.
Identification of Element
(except H) and oxidation
state; from temperature
behavior can get indication
of volatile sulfates,
nitrates, etc.
Elemental analysis of speci-
fic crystals.
Qualitative amounts of bulk
species.
Acid, base and neutrals;
aliphatic, aromatic and
oxygenated fractions.
* Extracted into 10 ml
-------
X-RAY PHOTOELECTRON SPECTROSCOPY (XPS or ESCA)
XPS measures the kinetic energies of photoelectrons expelled from a
sample irradiated with monoenergetic x-rays. The spectrometer measures the
intensity of the electrons emitted as a function of their kinetic energy. The
electron binding energies are characteristic for each element and so enable
the method to be used for elemental analysis. The binding energies for an
element are modified by the valence electron distribution, i.e. oxidation
state. In general, atoms of higher oxidation states have higher binding
energies with respect to the neutral atom (positive chemical shift), and for
reduced states they will show a negative shift. The XPS technique can, there-
fore, Jbe used for both elemental analysis (for elements heavier than He) and
for determination of oxidation state. The relative amounts of the elements
can be obtained from the spectral amplitudes. However, to obtain quantitative
results, an element in the sample must be analyzed by another technique. Since
XPS analyzes only a 10-40A depth of the exposed particles, the surface
concentrations are obtained and these may not be similar to bulk concentrations.
Since the technique is a surface tool, particulates imbedded in the
filter matrix will not be analyzed. Teflon filters are, therefore, the best
substrates. Fiber glass cannot be used because of the uneven nature of the
surface and the impurities in the filter. Sample concentrations of about
5 vg/cm2 are sufficient for analysis, and as with XRD the response signal can
be better defined with longer exposure times.
Although XPS does not identify compounds directly, such information can
be gathered by observing the temperature of volatilization of the species
present. For this, a means of varying the temperature of the sample from
-150° to +250° is used. Sulfuric acid will volatilize rapidly in the vacuum
of the XPS spectrometer (10~* Torr) at about 0°C and can be determined by the
disappearance of the S(VI) peak. NH^NO^ will also start to volatilize at
about 0°C and both the +5 and -3 nitrogen spectra will decrease. The ammonium
sulfates are more stable and will persist to temperatures well over 100°C.
For these thermal measurements, unbacked Teflon filters can be used up to 250°C.
SCANNING ELECTRON MICROSCOPE (SEM)
The SEM utilizes a well-focused electron beam to scan over the target
sample. Secondary electrons emitted from the sample yield its image and
topography on a screen. Sample areas of 100A (.01 \im) can be resolved and
samples of a few lagm/cm2 are reguired. The surface of the sample must be
conducting which can be readily accomplished, if necessary, by covering the
sample with a conducting material such as graphite, aluminum or gold. The
SEM is used to obtain morphological information which helps determine if
artifact formation has occurred, e.g. recrystallization to large crystals,
crusting, etc. Samples can be examined on a filter substrate or in special
containers. Since the SEM is also run under vacuum, a cold stage is available
for volatile samples.
The SEM, used in the present studies, also has an "energy dispersive
analysis by x-ray fluorescence" (EDAX) capability. In addition to the
secondary electron beam, x-rays, characteristic for the various elements are
130
-------
also emitted from the sample. This method permits analysis of elements
heavier than sodium. This analysis is qualitative for large areas, but is
more definitive for single particles especially if standards for the elements
of interest are available. The depth of penetration is about 0.5 Vm since
self-absorption of the x-rays occurs.
WET CHEMISTRY
The methods used on standard LACS samples have been employed in this
study for analysis of NHk, SO^ and NO^.
ORGANIC ANALYSES
For organic analyses, the method of Hoffman and Wynder (1) is being tested.
About 100 mg of sample is required and the organic particulates are separated
from the rest of the sample by extraction in ether. The ether is mixed with
2N H^SOt^ which removes the basic fraction. The ether layer is then extracted
with 2N NaOH which removes the acid fraction. The neutral organics remain
in the ether. The neutral portion can be fractionated into aliphatic, aromatic
and oxygenated fractions by column chromatography with n-hexane, n-hexane +
benzene, and ether, respectively.
USE OF XPS AND XRD TO INFER NATURE OF COMPOUNDS
In the preceding section a technique to infer compound structure from
XPS data at different temperatures was briefly described. By obtaining XPS
spectra of pure compounds as aerosols on filters at different temperatures
and determining the relative decrease of the individual species with tempera-
ture, useful information can be obtained about volatilization ratios of various
ions. From this information, inference can be made about the compounds in the
sample. For example, Figure 1 shows the temperature behavior of a synthetic
sample prepared by collecting on a filter sulfuric acid aerosol generated by
using a Baird Atomic generator* The size distribution was measured with a
Thermo-Systems Electrical Aerosol Analyzer.
The increase in signal between about -40° and 0° is probably caused by
evaporation of other volatile material from the surface, thus exposing more
sulfate. The decrease between 0° and about 30° is the result of the loss of
sulfuric acid while the gradual decrease beyond this point is probably the
decrease caused by some ammonium bisulfate contamination in the sample caused
by air exposure.
The volatilization temperature ranges for the various synthetic aerosol
compounds commonly expected in ambient particulate are shown in Table II.
In addition to the XPS work, filters of synthetic aerosol samples were
examined by x-ray diffraction to determine how concentration of submicron size
(0 075 - 0 1 vm) aerosols affect the x-ray diffraction pattern. The lowest
concentration of aerosols that yield an identifiable XRD pattern with 12-hour
exposure, and the corresponding ambient concentrations required to yield such
samples with 2-hour sampling time at 100 1/min on 47 mm Fluoropore filters
are shown in Table III. For the ambient aerosols, the assumption was made
131
-------
10
8
AMPLITUDE
(NORMALIZED
TOPb)
I
I
I
N"
so;
C/36
NOg
I
0 -40 0 40 -80
TEMPERATURE (°C)
Figure 1. XPS of ambient aerosol (electrostatic precipitator)
132
-------
TABLE II. VOLATILIZATION TEMPERATURE RANGE OF
AEROSOL COMPOUNDS IN XPS SPECTROMETER
f!0~8 torr. vacuum)
Temp. Range Studied
Aerosol (T°C)
H2SO^ -90 to +40
(NHii)2SOi+ -130 to +200
NH^HSOi) -60 to +100
NH^NOs -130 to +110
NHyCl -90 to 100
Volatilization
Initial (T°C)
0
30
0
0
50
Temperature Range
Final (T°C)
(20) *
200
C200) *
100
(120) *
* Estimated
133
-------
TABLE III. DETECTION LIMIT OF SUBMICRON AEROSOL BY XKD
Aerosol
Quantity Seen
(yg/cm2)
Ambient Concentration *
11
10
> 50
> 40
NHi+HSO-
50
40
> 50
> 40
NHi+Cl
> 40
Does not form readily
Does not form readily
* For two-hour sampling at 100 1/min with 47 mm filter.
134
-------
TABLE IV. ANALYSIS OF LACS SAMPLES *
Impactor Dust
Both Sites
EDAX(4)
Mg
S
Al
Si
K
Ca
Fe
ESCA<2>
Fe
N
S _
(87% SO")
C
XRDU)
Mg(NH4)2(S04)2-
6H20 (3)
Si02
no PbS04
Electrostatic Precipitator Dust
Both Sites
EDAX(4)
Al
Si
S
Ca
Cl
Fe
K
Pb
Br
ESCA
N (amine- and/or
amide- types at
-25 to 25°C),
(Ammonium and
nitrate at -25°C
only)
C
S (86% SO!)
(14% S or S")
Pb
XRD
Many well -defined
lines obtained.
Patterns similar
for both sites.
PbS04 both sites
* Samples collected with a Battelle Sampler
(1) Samples from both stages and both Sites A and C gave similar results
(2) Analysis of the small particles of second stage of Site C
(3) Analysis of a single large (100 ym) crystal particle from second impactor stage from Site C
(4) Energy dispersive analysis by X-ray fluorescence (with scanning electron microscope)
-------
that presence of other ions does not decrease the sensitivity of XRD. The
detection limits given in Table III were determined with unfolded filters.
However, the sensitivity of the XRD technique has been improved considerably
by folding the filter sample several times, thus resulting in exposure of
multilayers of Sample-to the x-ray beam. With this technique, 10.8 ygr/cm • of
(NHi+)2SOi+ were recently analyzed in a 4-hour exposure versus over 12 hours
for a single filter layer. The synthetic aerosols used in these experiments
consisted primarily of 0.075 ym particles but scanning electron micrographs
showed particles of up to 1 pm size and a non-uniform distribution on the
filter.
ARTIFACT FORMATION IN SAMPLE COLLECTION
In the original studies (2), composite samples, collected with the
Battelle massive sampler, were used. Results suggested that possible changes
occurred in the sample during and/or after collection as determined by crystal
growth and bulk caking of the sample. These changes could arise by (1) water
condensation on the sample with subsequent dissolution, reaction and recrys-
tallization; (2) reaction of participates with the sampler collector surface;
(3) reaction of collected particulates with reactive gases; and (4) reaction
between individual particles.
Some preliminary experiments have been done in which various sample
collection conditions have been used. Results of scanning electron micro-
graphs and chemical analysis obtained from samples collected on the electro-
static precipitator plates have been compared with samples collected on Teflon
filters. Differences have been found but to date the exact causes for the
differences have not been identified.
In order to obtain the filter sample as described just above, an opening
in the side of the Battelle sampler, below the impactor stages, was made. A
sample was drawn from this opening, through a Teflon filter at about 100 1/min.
Thus, both the filter and the electrostatic precipitator plates were sampling
the same material. Some comparison of the precipitator samples has also been
done with samples collected using a dichotomous sampler. The small amount of
particulate on 4-hour dichotomous filters makes the use of the XRD and XPS
techniques rather difficult.
The use of the Battelle massive sampler requires that the particulate
fraction collected on the precipitator plates undergo significant handling
before a sample is ready for analysis. For most analytical techniques
employed, particulate samples collected on Teflon or membrane filters can be
used directly. For this reason, direct collection of samples on filters has
an advantage. However, a longer collection time is required to obtain
sufficient sample if a dichotomous sampler is used.
AMBIENT SAMPLE ANALYSES
During the early part of the program/ Battelle samplers were operated
continuously at sites A and C and the material collected was scraped off and
composited. Preliminary work &as done with approximately a six month
composited sample. Results are summarized in Table IV.
136
-------
All the x-ray diffraction patterns contained numerous well-defined lines
indicating the presence of crystalline compounds. Samples from all the
impactor stages resulted in similar patterns, but these patterns differed
substantially from those of the electrostatic precipitator. Large single
crystals (* 100 vm) of Mg CNH^)2(SOk)2-6H2O have been found in all the impactor
samples, indicating crystal growth after particle size separation. In
addition, the sample material was caked and crusted, indicating that compound
identification with this sample and possibly with this collection device may
not be representative of the compounds in the air.
Ammonium sulfate has not been identified in any of the samples, but lead
sulfate was found in both precipitator samples. Further, both precipitator
samples contained large amounts of sulfur and a significantly larger quantity
of lead was present at site C than A. About 86% of the sulfur is present as
sulfate with the remainder in the elemental or reduced form. About one-third
of the sulfate observed by XPS is volatile at temperatures slightly above
ambient in the vacuum of the XPS instrument. In addition to the sulfur and
lead, the chief elements present in the precipitator samples are Al, Si, Cl,
K, Ca and Fe. None of the common sulfates of these cations has been identified
by XRD. XPS measurements showed the presence of NH^, NO"§ and amine or amide
species at -25°C, but at room temperature the NHi+ and NO$ had volatilized
completely in the vacuum of the XPS spectrometer. Significant concentrations
®f organic matter were also present in the precipitator samples.
The Battelle sampler has more recently been operated at the peak period
(3:00 to 7:00 pm) for two successive days at site A. About 250 mgs of sample
was obtained from the precipitator plates, and this was approximately 88% of
the total material recovered from the sampler.
Photoelectron spectroscopy was used to obtain spectra at various temper-
atures to suggest possible compounds present. Figure 2 shows the behavior of
the XPS signal as a function of temperature for the various species of interest.
The intensities are ratioed to lead. Of most interest, perhaps, is the fact
that only about 29% of the surface sulfate is volatile at 100°C in the vacuum
of the spectrometer (10~8 Torr). This might suggest that the majority of
sulfate is in various forms of non-volatile compounds (e.g. Na2SOii, etc.).
However, the temperature was not taken high enough to rule out the presence of
a small amount of ammonium sulfate. Because the nitrate and ammonium peaks
decrease together, it suggests that the majority of the ammonium ion is in the
form of ammonium nitrate.
The x-ray diffraction pattern did not show lines for lead sulfate or for
ammonium sulfate. However, it must be remembered that the technique can be
limited by the size of the participate and the amount of sample present.
High-volume filter samples have been extracted with ether and about 50%
of the sample was dissolved. Using the scheme of Hoffman and Wynder (1) for
one sample, the percentage of total organic and acidic, basic and neutral
fractions is given in Table V.
137
-------
8
yj
O
ui -3
Hh-
00
t 4
<
NH4HS04(?)
0
-100
1
I
-50
0 50
TEMPERATURE (°C)
Figure 2. XPS for H2SOn aerosol,
100
150
-------
TABLE V. ORGANIC PARTICIPATE ANALYSES
Fraction Weight % of Total Sample
Total Organics 57.3
Basic Negligible
Acidic 28
Neutral 18.6
Insoluble 10.7
139
-------
CONCLUSION
In summary/ samples collected from the precipitator plates (and from
side stream filters) of the Battelle sampler after 8-hours of operation appear
caked and contain large particles. The use of a dichotomous sampler does not
produce a large enough loading in a 4-hour period for XRD or XPS. A technique
has been described in which the nature of compounds present in a sample can be
inferred by x-ray photoelectron spectroscopy. Synthetic aerosols have been
generated and used to verify and calibrate the technique. Preliminary work
on ambient samples suggests a significant amount of sulfate is non-volatile.
In some samples, lead sulfate has been identified by x-ray diffraction.
Preliminary work suggests that a significant amount of the freeway aerosol
sample may be organic.
REFERENCES
1- Hoffman, D. and Wynder, E.L., "Orangic Particulate Pollutants", in
Air Pollution, A.C. Stern, erf., Vol. 2, p. 191, Academic Press, N.Y. 1968.
2. "Preliminary Report - Compound Identification of LACS Samples (Los Angeles
Catalyst Study)", prepared by Rockwell International Air Monitoring Center
for Environmental Protection Agency, August 1976.
140
-------
COMPARISON OF LEAD RECOVERY WITH AND WITHOUT THE USE
OF A WW TEMPERATURE ASHER BEFORE EXTRACTION
Lloyd S. Shepard and April Price
Rockwell International
Newbury Park, California
and
Charles E. Rodes
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
Results of Pb analysis with and without the use of a low temperature
asher are presented. Comparable results can be obtained by these two sample
preparation methods although indication of slight losses during the low temper-
ature ashing step occur.
INTRODUCTION
The conventional method of trace metal analysis in particulates involves
nitric acid extraction followed by atomic absorption (AA) determination of
the metal(s) of interest. This method is effective provided that the particles
containing metallic compounds are not covered with carbon containing substances
such as soot, oils and/or other organic compounds which may decrease the
efficiency of the acid extraction. This problem can be overcome easily by
prior ashing of these samples with a low temperature asher, which is proven
to eliminate possible losses of volatile metals (i.E. Eg, Zn, Cd, Pb) which
do take place with conventional muffle furnace ashing. At the early stages
of the Los Angeles Catalyst Study (LACS) , an effort was made to determine if
application of low temperature ashing is necessary in the analysis of par-
ticulate lead (Pb) collected close to the San Diego Freeway by means of high
volume samplers.
EXPERIMENTAL PROCEDURE
Duplicate strips of high-volume samples collected at the LACS sites were
analyzed for Pb with and without the use of low temperature asher with the
following procedure. The first strip was ashed at low RF energy and with
oxygen passing over the sample at a low rate (4 cc/min) for approximately 10
minutes or until ashing was complete. The strips were than transferred to
acid-washed 50-ml beakers and the ashing boat rinsed with a few ml fflV03
141
-------
(1:10 diluted) to remove any dislodged particles. The samples were then
boiled for 20 minutes, filtered and diluted to 24 ml for four hour samples or
to 100 ml for 24 hour samples.
The treatment of the second strip was identical as above with the omission
of the ashing step.
The solutions thus obtained were analyzed for Pb by atomic absorption by
employing a 303 atomic absorption spectrophotometer.
RESULTS AND DISCUSSION
Twenty-six duplicate strips of exposed filters from the LACS project were
analyzed with and without prior ashing by the low temperature asher according
to the procedures described above. Ten of the duplicate strips were 24 hour
integrated samples and the remaining 16 were four hour integrated samples.
Table I lists the results of the analysis of the duplicate samples. The
average range of the duplicate analysis with and without ashing is 0.20 yg/ml
with an upper control limit of 0.65 pg/ml. These compare very well with the
previously established operational control limits for analysis of duplicate
lead samples, 0.28 yg/ml and 0.91 \ig/ml, respectively. Only one value (sample
No. 8) is out of control. The average percent difference with the exception
of four samples, very low in Pb (sample Nos. 14, 16, 17 and 20), is 3.1 with
a standard deviation +_ 2.8 which is not very uncommon with this type of sample.
The comparison of the results with and without ashing is also presented
in Figure 1. The correlation coefficient was found to be 0.998 and the slope
and intercept of the least squares fit were 0.972 and 0.058, respectively.
The slight deviation of the slope from unity may suggest a loss of Pb during
the ashing step. Statistically, however, the results obtained with or without
the use of the low temperature asher should be considered equivalent. These
results clearly show that low temperature ashing is not necessary for the
atomic absorption determination of Pb in the LACS samples.
142
-------
TABLE I. DUPLICATE LEAD ANALYSIS WITH OR WITHOUT LTA
SAMPLE
NO.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
SAMPLING TIME
(hr)
24
24
24
24
24
24
24
24
24
24
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
NOT ASHED
/ig/m1
5.18
5.18
7.34
9.34
4.68
7.02
8.50
11.77
5.28
5.92
0.69
12.33
12.59
0.32
12.68
0.79
0.50
1.64
4.05
0.34
8.60
12.85
9.48
9.52
4.21
5.76
ASHED
/xg/ml
4.93
5.12
7.34
9.14
4.93
6.92
8.00
10.89
4.96
5.79
0.69
12.58
12.08
0.20
12.11
0.97
0.42
1.49
4.00
0.40
8.54
12.77
9.34
9.52
4.46
5.79
%DIFF.*
-4.94
•1.16
0.00
-2.16
5.20
-1.43
-6.06
-7.77
-6.25
-2.22
0.00
2.01
-4.13
-46.15
-4.60
20.45
-17.39
-9.58
-1.24
16.22
-0.70
-0.62
-1.53
0.00
5.77
0.52
*%DIFF.=
ASHED - NOT ASHED
AVERAGE
143
-------
15
2
i
E
I
o»
CORRELATION COEFFICIENT = 0.998;
SLOPE = 0.972; INTERCEPT - 0.058;
STANDARD ERROR - 0.233
fig
10
- UNASHED
15
Figure 1. Comparison of lead recovery with and without the use of a low
temperature asher before extraction.
144
-------
TABLE I. DUPLICATE LEAD ANALYSIS WITH OR WITHOUT LTA
Sample
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
*—~~—^~~^~^*f*~f"
^^^WM>«MM*^-V
* o. r> ,' Cf
Sampling Time
('hours)
24
24
24
24
24
24
24
24
24
24
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
— .
H^ HHM.^ •*-"— IBiMBlW^M^^ "• '• .—^ •
Ashed -Not Ashed
Not Ashed
]ig/ml
5.18
5.18
7.34
9.34
4.68
7.02
8.50
11.77
5.28
5.92
0.69
12.33
12.59
0.32
12.68
0.79
0.50
1.64
4.05
0.34
8.60
12.85
9.48
9.52
4.21
5.76
— —
_______———
Ashed
pg/ml
4.93
5.12
7.34
9.14
4.93
6.92
8.00
10.89
4.96
5.79
0.69
12.58
12.08
0.20
12.11
0.97
0.42
1.49
4.00
0.40
8.54
12.77
9.34
9.52
4.46
5.79
-_
- . I" "
% Diff. *
- 4.94
- 1.16
0.00
- 2.16
5.20
- 1.43
- 6.06
- 7.77
- 6.25
- 2.22
0.00
2.01
- 4.13
-46.15
- 4.60
20.45
-17.39
- 9.58
- 1.24
16.22
-0.70
- 0.62
- 1.53
0.00
5.77
0.52
-----
_ - - - —
Average
145
-------
DETERMINATION OF PERCENTAGE OF DIESEL TRUCKS AND
CATALYST EQUIPPED CARS
Edward P. Parry and Raymond A. Meyer
Rockwell International
Newbury Park, California
and
Charles E. Rodes
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
The percentage of catalyst equipped cars passing the study site is a
critical parameter. Several techniques to obtain this information, such as
infrared surveillance, passing vehicle identification, etc., were considered.
The infrared surveillance required considerable research, and there were no
available personnel capable of vehicle identification. Los Angeles County
new car registrations have been studied since September 1974 to determine the
percent of catalyst equipped automobiles within the county. The monthly
increase rate, 0.5%, has been quite constant over the period. These data
alone do not establish the catalyst equipped car ratio at the freeway study
site and must be modified by a use factor. This factor is presently under
investigation. Diesel truck exhaust contains a high sulfate level which
could perturb the data. Therefore, a diesel truck count was made for 15
consecutive days in September 1976. Truck movement was determined by the hour
and direction, bata are reported for 13 days, during which trucks were
counted from 0800 - 1200 and 1500 - 1900, and two days, which were counted
for 24 hours. Both weekdays and weekends were included in the study.
INTRODUCTION
The number of cars which are equipped with the catalytic converter is an
obvious parameter which must be known in order to project the impact of
sulfate emission from these cars on ambient sulfate levels. Beginning with
1975 models, it was mandatory in California to have the catalytic converter
on automobiles from most manufacturers in order to meet California emission
standards. Thus, 1975 and later models of most makes of automobiles were
equipped with the catalytic converter, while 1974 and earlier models were not.
In the initial stages of the Los Angeles Catalyst Study (LACS), an investi-
gation was undertaken to evaluate various techniques which could be used to
147
-------
determine the number of vehicles passing the LACS site which were equipped
with the catalytic converter.
Another source of freeway particulate sulfate is diesel trucks. The
exhaust of diesel trucks is reported to contain a rather large concentration
of sulfate/ and the sulfate emission per mile is very large. In order to
determine the potential contribution of these trucks to freeway sulfate levels,
a rather recent determination of the number of diesel trucks passing the LACS
site has been made.
In this paper, the approaches which were investigated, those finally
selected and the results of the determination of both diesel trucks and cars
equipped with catalytic converters are discussed. In addition, improvements
in the method to determine the percentage of catalyst equipped cars are
discussed.
DIESEL TRUCK STUDY
With a little experience, identification of diesel trucks by visual
observation is straightforward. In this study, college students were
employed to count the diesel trucks as they passed the site. Small hand
held counters were used, and the personnel were stationed at the side of the
freeway at the LACS site. The study was divided into two phases. In the
first phase, trucks were counted during the two peak periods, 0800-1200 and
1500-1900, from Monday, 13 September 1976 through Sunday, 26 September 1976.
In the second phase, trucks were counted for two 24-hour periods, one on
Sunday, 3 October 1976 and the other on Thursday, 7 October 1976. These
periods were chosen to study all types of traffic flow. In Figure 1 a bar
graph is presented showing the distribution of northbound and southbound
traffic for the peak periods over the two week time span of this study. In
general, there is significantly more traffic in the 0800-1200 time period.
The total trucks for the two peak periods is, in nearly every instance, some-
what less than 1000. In Figure 2, the diurnal traffic pattern is shown for
the two 24-hour periods. From about 0600 in the morning until 1800 at night,
on weekdays, a sawtooth pattern is found with traffic density above 50 trucks
per hour. Sunday truck traffic is small as would be expected. The total
24-hour truck count for both northbound and southbound for the Thursday period
was 2,252. The total number of vehicles for the same period (both north.
and southbound) was approximately 175,000 cars. The truck traffic thus amounts
to about 1 to 2% of the total traffic.
APPROACH TO DETERMINE THE NUMBER OF CATALYST EQUIPPED CARS
Several different approaches were initially proposed. These included
the use of time-lapse photography, thermal imagery, methods to mark 1975
cars uniquely at the time of registration for later identification, car
registration information and others. Some of these techniques were shown not
to be feasible, as preliminary evaluation and investigation of the techniques
uncovered very difficult problems (e.g. unique tagging of 1975 vehicles).
Some of the other approaches were considered in a little more detail. These
are described below:
148
-------
700
I I
600
a soo
o
Q.
CC 400
UJ
Q.
M
§
CNJ
-6
o
o
in
cc
i-
300
200
100
SOUTHBOUND
NORTHBOUND
01 in * • •• • • • •• • • •" * • * "* •
-»^f»~l^«-^-l-»-n^e/J^
i nil i
CO
3
33
v>
Figure 1. Diesel truck peak period study.
149
-------
100
80
cc
O 60
DC
111
a.
CO
*
40
20
THURSDAY,
7 OCTOBER 1976
SUNDAY,
3 OCTOBER 1976
V ^**
1 1 1 1 1 1 1 1 1 1 1 I 1 1
•*
1 1 1 1 1 1 1 1 1
D
6 8 10 12 14 16 18 20
HOUR OF DAY
22 24
Figure 2. Twenty-four hour truck count.
150
-------
THERMAL IMAGERY
_ Catalytic converters increase the temperature of the exhaust gas by a
significant amount. The use of thermal imagery to scan the tail pipe temper-
ature of passing automobiles remotely was considered. Systems for this pur-
pose are available commercially. A disadvantage of the system is cost. A
question also exists with regard to the resolution of the system since there
is some variation in operating temperature of catalytic converters, and in
muffler and tail pipe designs which will cause variations of the temperature
of the exiting gas from various car makes. Another question concerned the
possibility of measuring tail pipe temperatures for cars in all traffic lanes,
especially in heavy traffic, without interference from surrounding cars. A
third question which had to be answered before the technique could be used
involved the ability to differentiate exhaust pipe heat sources from others
including reflected infrared energy, engine heat sources and others.
This technique has the advantage that it identifies catalytic converter
equipped cars directly and without assumptions. The disadvantage is that
the perfection of the technique would take considerable research and develop-
ment and the cost of equipment is rather high. Further work was not carried
out with this technique.
VISUAL RECOGNITION OF MODEL YEAR
Some years ago it was possible to recognize the model year of practically
every car on the road by the distinguishing characteristics introduced by
the manufacturer with each new model year. This could form the basis for
differentiating 1975 model cars from older ones and, thus, indirectly getting
the number of cars equipped with the catalytic converter. Two approaches
could be considered in this proposed method. One would be real-time identi-
fication by having people stand on the side of the freeway, or on an overpass
near the LACS site, and physically count the cars of interest. Another
approach which has obvious advantages is to use time-lapse photography to
photograph a statistical sample of automobiles as they pass the site and per-
form the identification at a later date. A valid statistical sample is
obviously a necessity. If the photographs were good, the ability to study
cars as required in order to identify small differences between models is
an advantage. In addition, the photographs form a permanent record which
could be used for later verification and checking as necessary. The time-
lapse photographic approach has some obvious advantages in obtaining good
photographs of cars in all traffic lanes (particularly at night) and in ob-
taining assurance that a valid statistical sample is being used.
The determination of make and model year of present automobiles is not
trivial. Although it might have been a casual observation, it appeared to
us as if model year distinction of various makes of automobiles has become
more difficult in recent years because only very minor changes are being
made to the exterior for succeeding model years. To determine how diffcult
this distinction might be, inquiries were made of various private and public
agencies. The list included: California State Highway Patrol, Los Angeles
City Police Department (both Traffic and Auto Theft Divisions), the Automobile
Club of Southern California (AAA), editors of several automobile publications,
151
-------
such as Road and Track, several automobile retailers and used car outlets,
zone offices of major automobile manufacturers, etc. All of the people
contacted stressed the fact that during the past few years external changes
between model years were very subtle and in -some cases non-existent. These
contacts have indicated that an effort directed toward visual identification
of traffic passing the site would not be successful. Because of the diffi-
culties uncovered in the assessment of this approach, no further work was
done.
LEADED-UNLEADED GASOLINE SALES
All catalyst equipped cars are required to use unleaded gasoline.
Because unleaded gasoline is more expensive, it might not be too unreasonable
to assume that drivers of non-catalyst equipped cars will, in general, not use
unleaded gasoline. If, in addition, it is assumed that the overall gasoline
mileage of the mix of 1975 and later models is about the same as the pre-1975
models, the use of gasoline sales might be an approach to determine the per-
centage of catalyst equipped cars. However, the difficulty of validating the
assumptions required and the fact that this approach does not give percent-
age of catalyst cars passing the LACS site makes it questionable. No further
work was, therefore, done.
REGISTRATION DATA
During the early stages of the LACS program, an approach was proposed
in which the license plates of automobiles passing the site would be photo-
graphed and the model year subsequently identified by checking the plate
against state registration data. This approach needed a.research and develop-
ment effort in order to develop techniques to obtain legible photographs of
license plates of cars in all lanes of the freeway, especially at night
during heavy traffic. In addition the State Department of Motor Vehicles
(DMV) charges $1.75 to identify make and model year from each complete
license plate number. The cost of this program thus seemed prohibitive
and was, therefore, not carried further.
New car registration information was next considered. For a very
nominal charge, the State Department of Motor Vehicles will provide a
monthly summary of car registrations. In addition, several private companies
buy the registration information from the state, reduce the data, and sell
reports of interest to automobile sales agencies and others. One such
company, the Donnelly Corporation, produces the Motor Registration News (MEN).
This is published weekly and lists all new car registrations. In addition,
a portion of the vehicle identification number (VID) is given. The complete
VID contains model year identification information. However, as given in
the MEN, the year identification portion for all cars except Fords, Lincolns,
and Mercurys is truncated, and is not available.
During the initial phases of the program (late 1974), model year identi-
fication was mandatory since both 1974 and 1975 cars were being sold as new
cars (registered for the firs^t time) . Because of necessity, the 1974-1975
split for the Ford-Lincoln-Mercury group was assumed to be the same as for
all other cars. This approach thus gave the 1975 cars registered in Los
152
-------
Angeles County from the time of their introduction, with a constant auto-
mobile population of 3.75 million in Los Angeles County and an estimate of
1975 cars not equipped with the catalytic converter, the percent of cars
equipped with the catalytic converter could be estimated. Figure 3 shows a
plot of catalyst converter equipped cars (as obtained from cumulative
registration information) as a function of time.
In about April 1975, it was learned that the Donnelly Corporation had
the complete vehicle identification number (including make and model year
information) for all cars in their computer files. It thus became possible
to obtain from them the cumulative total of post-1974 vehicle registrations
in Los Angeles County as well as the cumulative total of post-1974 vehicles
which do not have the catalytic converter. This group includes BW, Colt,
Honda, Mazda, Porsche, and Subaru. Data from this source were first ob-
tained in May of 1975. These data are also plotted in Figure 3, and are
identified as "Donnelly Data." The small discontinuity in the plot is caused
by the two different methods used. Considering the overall reliability of
the method, this discontinuity is trivial.
The data show that as of January 1977, 15.5% of the cars registered in
Los Angeles County were equipped with the catalytic converter. The increase
averages about 0.6% per month.
A serious flaw has been recognized in this approach since it began to be
used. The information obtained is the percentage of cars, which are equipped
with the catalytic converter, registered in Los Angeles County and not the
percentage of converter equipped cars which pass the LACS site. A TRW report
prepared for the California Air Resources Board ("A Mobile Source Emission
Inventory System for Light Duty Vehicles in the South Coast Air Basin" -
draft copy) contains data which indicate that the newer cars are driven
much more than older cars. Therefore, registration information may not be
adequate in estimating the number of new cars which are traveling on free-
ways. In the report, it indicates that for 1975 cars, the miles traveled
per vehicle on a normalized basis is 2.04. To obtain this number, the total
normalized miles driven by 1975 cars is divided by the total normalized
number of 1975 vehicles in the study. For 1974, this normalized value is
1.90, and the value gradually decreases to 0.98 for 1969 model cars and 0.40
for 1964 cars. Because of the greater usage of the newer cars, it is likely
that the number of cars equipped with the catalytic converter that travel
the freeway (and, therefore, statistically pass the LACS site) is greater
than, for example, the January 1977 value of 15.5%. On the basis of this
usage, the January 1977 actual percent of converter equipped cars passing the
site is more likely between 20 and 30%.
In the beginning of this section, the possibility was discussed of ob-
taining make and model year identification from the State Department of Motor
Vehicles after photographing the license plates of automobiles passing the
LACS site. Because of the charge which the state imposes for identifying
the vehicle from the license plate number, this approach, if all license
plates had to be identified, appeared to be expensive. It has since been
found that the State Department of Motor Vehicles has issued license plates
serially since the introduction of the "blue background" plates in 1968.
153
-------
to
cc
LU
16
14
£ 12
o
o
o
t-
U
X
to
5
o
h-
z
01
O
DC
LU
O.
10
I I I I
i I I I r
DONNELLY DATA
I I I I I I I I I I I I
I I I I I I I I I I I I
CO
m
"O
D
m
o
(O
•si
•si
Ul
(O
sj
tn
CO
m
TJ
(O
sj
cn
O
m
O
(O
•si
at
(O
Z
CD
sj
cn
CO
m
•o
3
«
O
m
O
oo
_>
CP
Figure 3. Percent catalytic equipped autos registered in Los Angeles County.
154
-------
The number of plates, therefore, requiring identification may be significantly
decreased making this approach a more viable one.
The "blue background" plates (used in 1968 and later) have three numbers
and three letters (e.g. 248 JGN) . The DMV has indicated that the letter
portion of the plate is serialized and that in September 1974, all license
plates sent from the manufacturer to the distributors were marked MXX or
with other first letters which follow M in the alphabet. Allowing one letter
for the local stock at hand in the DMV offices, it is, therefore, possible to
say that all license numbers starting with letter A through K are pre-1975
model, and license plates starting with L to Z may be 1975 or later.
The assignment of L to Z class to post-1974 cars is not definitive, since
consideration must be given to the possibility that pre-1975 cars may have
been re-registered for some reason (e.g., out of state cars) and, therefore,
have a license plate number in the L to Z class. In addition, one must con-
sider the fact that not all post-1974 cars have the catalytic converter.
It appears as if techniques can be used to definitize the number of
catalytic equipped cars in the L to Z license plate class, making this an
approach which can determine the actual number of cars equipped with the
catalytic converter which pass the LACS site. Application of this technique
is currently Jbeincr implemented.
155
-------
LABORATORY QUALITY CONTROL
George Colovos, Edward P. Parry, and Lloyd S. Shepard
Rockwell International
Newbury Park, California
and
Charles E. Rodes
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
The internal quality control applied to laboratory analyses of the Los
Angeles Catalyst Study manual samples is presented. Included in this discus-
sion are the use of quality control reference samples, quality control on
calibration parameters, statistical treatment of the data, establishment of
operational quality control limits, development of quality control reports,
and establishment of necessary rules for acceptance or rejection of sample
data. An early warning system presently in use and future plans for on-line,
internal quality control is also presented.
INTRODUCTION
The importance of applying quality assurance control practices to
laboratory procedures was recognized very early by chemists; several texts of
analytical chemistry devote chapters to this subject. Essentially, the pur-
pose of quality assurance is to answer the question of whether data generated
by an analytical procedure can be regarded as typical samples from a single
population of data. If such data can be so regarded, statistical control can
be assumed. The most commonly used method of determining accurate represen-
tation consists of control charts. Control charts are sequential plots of
various quality characteristics. For example, qualities shown might be a
day-to-day plot of the average content of copper (Cu) in an ore, the normality
of a standard solution, the calibration parameters of an instrument, etc.
Control charts give a continuous record of the quality characteristic and
trends in data. Also, sudden lack of precision can be made evident and causes
may be sought by use of the charts. The necessity of comprehensive quality
assurance techniques in air quality data generated either in the field or in
the laboratory are very well known and have been recognized widely (1) . No
study can be considered complete without the application of some type of
quality assurance procedure. In the case of the Los Angeles Catalyst Study
157
-------
(LACS) project, the chemical analysis of integrated samples is vital for the
interpretation of air quality data associated with use of catalytic converters
on automobiles. For this reason, a comprehensive plan for quality assurance
of chemical analysis was conceived and implemented (2).
In this plan, measures for controlling all the possible parameters
affecting the quality of data were proposed and implemented. Considered
parameters included: chemicals, sample identity, transportation, storage,
analytical procedures, personnel, and calibration. In this paper an attempt
is made to present the most important features of this program and also to
analyze some of the results and discuss some future plans for computerized
on-line laboratory quality control.
DISCUSSION
As mentioned previously, the quality assurance plan for chemical analysis
of LACS samples included all the parameters associated with the quality of the
laboratory results. Although some of the listed parameters are very important
for the overall quality of the data, they will not be discussed here because
the types of controls recommended for them were more procedural that statis-
tical. Therefore, they could not be used as quality control indicators.
Basically, the control limits for these parameters were set forth by estab-
lishing procedural protocols which assured the integrity of the samples and/or
the application of sound and proven techniques.
The adherence to these protocols were checked by employing quality control
of the calibration parameters and by utilizing quality control of the calibra-
tion parameters and by utilizing quality control samples. Performance during
analyses was checked in the same manner.
The calibration parameters of any analytical procedure can be considered
as a measure of accuracy. Therefore, quality control procedures pertinent
to each parameter ascertain the degree of uniformity of laboratory data. If
the arrived at degree of uniformity is combined with use of primary calibration
standards, accuracy obtained for laboratory data can be checked by employing
quality control on calibration parameters. Table I lists standards used for
calibration of procedures utilized in the analyses of the LACS samples An
out-of-control situation of any given batch of analyses can be traced back
to these stndards and a determination can be made as to whether the cause of
the out-to-control condition was due either to wrong standards or to some
other factor. In order to insure complete traceability, all steps involved
in the preparation of standards, such as weighings,dilutions, date of pre-
parations, etc, are recorded in a special log book.
The use of quality control samples reveals information about the pre-
cision of the analytical procedure. Table II gives the types of control
samples employed in analysis of LACS samples along with the frequency of their
application. With the exception of the spiked samples, control samples are
measures of analytical precision. Control charts made from these data can
show trends of precision and out-of-control conditions for any given analysis.
The types of control charts utilized in this study are shown in Table III.
158
-------
TABLE I. STANDARDS FOR ANALYTICAL PROCEDURES
Method
Standard
1. Mass Determination
2. SC>2 Bubbler
3. Sulfate
4. Nitrate
5. Ammoni ton
6. Metals (Pb )
NBS traceable weights.
NBS certified S02 permeation tube.
Ha^S^O*, reagent grade chemical
standardized with iodine-thiosulfate.
Standard Na2SOit solution prepared
gravimetrically.
Standard NaNO$ solution prepared
gravimetrically.
Standard NH^Cl solution prepared
gravimetrically-
Standard metal salt solutions.
159
-------
TABLE II. INTERNAL QUALITY CONTROL OF THE INTEGRATED SAMPLES
OF THE LOS ANGELES CATALYST STUDY
Type of Analysis
Instrumentation
Quality Control Checks
5.
Mass Determin-
ation
A.
B.
Unexposed
Exposed
, NO~l S
Analytical Balance
Analytical Balance
Analytical Balance
Techicon Auto-
Analyzer II
Metals (Pb )
Atomic Absorption
Calibration checked daily
against a standard weight.
10% are reweighed
10% are reweighed
Calibrated daily against stand-
ard solutions.
Control checks per tray of 40
samples.
1. Extract from previous tray.*
2. Blank extract.**
3. Standard solution.
4. Duplicate exposed strips.
Calibration check daily against
standard solutions.
Control checks per run
1. Two repeat
extract.
2. Two blank extracts (one
spiked).
3. Two standard solutions.
4. Two duplicate exposed strips.
Calibrated daily against a 50.0
ml Type A volumetric flask.
Calibrated daily against stand-
ard solutions.
Control checks per tray of 40
samples.
1. Duplicate solution from pre-
vious trays.*
2. Old samples.
3. Standard solution.
4. Another standard solution
*In the first tray, one of the first samples is moved to the end of the
tray.
**Once a week one spiked blank.
Preparation of
Bubblers
Analysis of SO2
Bubblers
fiepipette
Technicon Auto-
Analyzer II
160
-------
TABLE III. QUALITY CONTROL GRAPHS
Type of Analysis
Level & Type of Control
Type of Graphs
Mass Determination
NO 3, NH*, Pb++
S02
10% Reweighed
Slope, intercept,
and error of the
calibration curve
Extracts from previous
tray
Blanks
Duplicate exposed
strips
Spiked samples
Slope, intercept,
and error of the
calibration curve
Duplicate solutions
from previous trays
Old samples
Standard solutions
Spiked samples
R-charts for each
one.
R-chart for each
parameter.
R-charts
R- and S-charts
R-charts
% recovery
R-charts for each
parameter
R-charts
R-charts
R-charts
% recovery
161
-------
Most of the charts are range charts (R-charts)(which represent either the dif-
ference between two successive values, i.e. repeated analysis of the same
extract, duplicate analysis of samples, etc.) or the difference of the experi-
mental value from the expected, (standard) solutions. Figures 1 and 2 show
the range charts for the slope of the sulfate standard curve and the duplicate
analysis of sulfate in the same extract, respectively. All the chart plotting
and the statistical calculations associated with them are done by the PDP-11/
35 computer established in the chemistry laboratory greatly facilitating
handling of the large number of data generated by internal quality assurance.
It should be noted that all the individual data used for the construction of
control charts are stored in the computer. Therefore, any other statistical
treatment of these data, if necessary, can be done easily.
It is evident that internal quality control through the use of control
charts can show if the analytical procedures are under statistical control.
However, their use does not completely insure the validity of generated data.
Additional information on the validity of data can be obtained by an external
quality control system utilizing unknowns and split samples. The external
quality control program outlined in Table TV has been implemented for the LACS
program and has been discussed in detail by J. Puzak (3) and G. Evans (4).
However, the proper use of an internal quality control system can minimize the
chances of error and can also reduce the cost of the necessary corrective act-
ions if it is found that the analysis has gotten out-of-control. Establish-
ment of operational control limits which allow the analysis to determine
out-control conditions is required in this program; the control limits est-
ablished from previous data of internal quality control were used as opera-
tional control limits. When range charts are utilized for quality control,
the lower control limit (LCL) is automatically defined as zero and the upper
control limit (UCL) is defined as 3.27 "R", where K" is the average range in the
chart. Operational_control limits used in these studies were calculated from
the overall range (R) obtained by incorporating all the control data since
the initiation of the program and is given by 3.27R.
Tables V - X show the operational control limits for various control
samples and calibration parameters used in the internal quality control of
analytical procedures. In these tables, two values are shown for each control
parameter. The two values correspond to two reporting periods (Reports VIII
and IX) . The first (Report VIII) shows the values which were established at
the beginning of the reporting period. The second (Report IX) illustrates the
values established by incorporating the data generated in this period with
those established in the previous reporting period. Out-of-control conditions
during one reporting period were determined by comparing the quality control
values against the operation control limits established at the end of the pre-
vious reporting period. Usually the reporting period was equivalent to one-
quarter of the year. Thus, the operational control limits were revised on a
quarterly basis by using all the data gathered from the beginning of the pro-
gram until the end of the last quarter. The broad base of data used for the
estimation of the operational control limits eventually resulted in values
which did not fluctuate significantly from quarter-to-quarter, although some
of the individual data may vary significantly. The variation indicates that
the present operational control limits for all practical purposes should be
considered as constants and that no quarterly revision of them is necessary
162
-------
RANGE. CHART OK SO4 REPEATED SAMPLES— GROUP 2
-------
RANGE CHflFO OF THE SLOPE OF THE SOI STANDARD CURVE—GROUP 2
UCL
R MEfiN
ORDF.R R
OF *100
RESULTS
1 1 . 07
2 0. 61
3 0. 44
4 O. 59
t- 1. 30
•'-• 1. 43
7 0. IS
8 0. 49
7' 0. 72
10 O. 36
RANGE
YG/ML/Rt
6. 00 —
S. SO —
!j. 60 —
5. 4O —
fj. 2O —
5. 00 —
4. 80 —
4. 60 —
4. 40 —
4. 20 —
4. OO —
3. BO —
3. 60 —
3. 40 —
3. 20 —
3 OO —
?. SO —
2. 60 -1
2. 40 —
2. 20 —
2. 00 —
1. 80 —
1. 60 —
1. 40 —
1. 20 —
1. 00 — . ft.
0. 30 —
0. 60 — *
0 40 — *
'<>. 20 . ~
0. 00 —
DATE ORDER R DATE ORDER R DATE ORDER R DATE
Of- »10O OF *1OO OF *100
RESULTS RESULTS RESULTS
26 11 1.19 30 21 1.37 40 31 1.39 42
26 12 0. 61 30 22 ' 1. 27 40 32 0. 07 43
27 _ . 13 . 0.20 -30 . 23 1..42 4O 33 0. 12 43
27 14 1. 86 3O 24 1. 95 41 34 O. 47 43
27 IS 0.68 30 25 0. 52 41 35 0 57 43
29 16 1. 6S 35 26 1. 22 41 36 1. 06 44
29 17 1. 86 35 27 0. 24 41 37 O. 14 44
29 18 O. 95 35 23 0. 4o 42 38 1.11 44
29 19 0. 63 35 29 1. 46 42 39 1. 32 44
29 2O O. 26 40 30 1. 11 42 4O 1. 31 44
HtftN H - 0. 90S YG/ML/Rt UCL = 2. 967 YG/ML/RE SI D. DEV. - O.
#
* *
#
* * » « * »
* # ft # *
4* y*
«
* * * * * *
* » * *
* * ' .... » ..*... ._*....
*
ORDER R DATE
OF »1OO
RESULTS
41 0.45 44
42 1. 54 47
43 O. 36 47
44 0. 61 47
45 0. 01 47
46 1. 68 48
47 2. 11 48
48 0. 38 48
49 1. 15 48
50 1. 42 48
566 YG/ML/RE
...
#
* *
* »
* *
*
* « *
»
»
RANGE
YG/ML/RE
— 6. 00
— 5. 80
— 5. 60
— 5. 40
— 5. 2O
— 5. 00
— 4. 80
— 4. 60
— 4. 40
— 4. 20
— 4. OO
— 3. 80
— 3. 60
— 3. 40
— 3. 20
— 3. 00
— -2. 80
— 2, 60
— 2. 40
— 2. 20
— 2. 00
— 1. 80
— 1. 60
»TT- 1. 40
— 1. 20
— 1. 00
— 0. 80
— 0. 60
— 0. 40
.— 0.20
— 0. 00
10
20 25
ORDER OF RESULTS
30
35
40 '45
50
Figure 2. Range chart of the sulfate standard calibration curve.
-------
TABLE IV. EXTERNAL QUALITY CHECKS
Analysis
Exchange Program
1. Unexposed
Filters
2. Exposed
Filters
3. Bubblers
Unexposed filters preweighed by the 2 per day
QAB/EMSL of EPA, weighed daily and
results reported to EPA on a weekly
basis.
8" x 3/4" strips from randomly se- Once every
lected filters corresponding to 3 weeks
approximately 5% to the total num-
ber of exposed filters sent to EPA
for analysis of SO^, NO^, NHi^ and
Pb++. Rockwell reports analytical
results simultaneously with the
submission of samples.
In_addition, 10 blind spiked with SO^, ^2 per day
NC>1 and OTty and 7 spiked with Pb
filter strips sent from QAB/EMSL
analyzed by Rockwell along with regular
samples. Results reported to EPA on a
weekly basis. Checks on the mass deter-
mination done occasionally by sending
exposed filters to EPA.
Blind solutions containing SQ
-------
TABLE V. CONTROL LIMITS FOR THE
GRAVIMETRIC ANALYSIS OF HI-VOL FILTERS
UCL
Unexposed Exposed Unexposed Exposed
Q.C. Report VIII 0.31 0.52 1.01 1.69
Q.C. Report IX 0.31 0.52 1.02 1.70
166
-------
TABLE VI, CONTROL LIMITS OF THE NO"l ANALYSIS
UCL
Slope
Intercept
Standard
Error
Reruns
Standard
Solutions
Blanks
Recovery
Duplicates
Q.C. Report
VII
8.24xlO~3
0.23
8.8xlO~2
0.42
0.72
0.20
0.95
; 0.40
Q.C. Report
IX
7.89xlO~3
0.22
8.9xlO~2
0.41
0.72
0.20
0.96
0.39
Q.C. Report
VIII
2.69xlO~2
0.75
0.29
1.39
2.36
0.66
UCL =1.01
LCL =0.89
1.29
Q.C. Report
IX
i -2
2.58x10
0.73
0.29
1.35
2.34
0.66
UCL = 1.01
LCL = 0.90
1.26
167
-------
TABLE VII. CONTROL LIMITS OF THE SOj, ANALYSIS
UCL
Slope
Intercept
Standard
Error
Reruns
Standard
Solutions
Blanks
Recovery
Q.C. Report
VIII
1.00xlO~2
0.60
0.20
0.92
0.86
0.80
0.96
Q.C. Report
IX
0 . 99xlO~2
0.59
0.20
0.90
0.82
0.78
0.96
Q.C. Report
VIII
3.26xlO~2
1.96
0.65
3.01
2.80
2.61
UCL - 1.
LCL = 0.
Q.C. Report
IX
3.22xlO~2
1.92
0.64
2.93
2.69
2.55
01 1.01
90 0.91
Duplicates 0.68
0.67
2.22
2.18
168
-------
TABLE VIII. CONTROL LIMITS OF THE NH^ ANALYSIS
UCL
Slope
Intercept
Standard
Error
Reruns
Standard
Solutions
Blanks
Recovery
Duplicates
Q.C. Report
VIII
J.49xlO~3
7.3xlO~2
2.5JC10"2
0.10
0.42
9.1xlO~2
0.95
5 0.10
Q.C. Report
IX
1.42xlO~3
6.9xlO~2
2.4xlO~2
0.10
0.37
9.0xlO~2
0.95
0.10
Q.C. Report
VIII
4.88xlO~3
0.24
8.1xlO~2
0.34
1.36
0.30
UCL - 1.04
UCL = 0.86
0.34
Q.C. Report
IX
4.65xlO~3
0.22
7.9xlO~2
0.33
1.22
0.29
1.03
0.87
0.34
169
-------
TABLE IX. CONTROL LIMITS OF THE LEAD ANALYSIS
UCL
Q.
Slope
Intercept
Standard
Error
Reruns
Standard
Solutions
Blanks
Recovery
Duplicates
C. Report
VIII
0.59
0.11
8.6xlO~2
0.23
0.38
0.00
0.93
0.24
Q.C. Report
IX
0.55
0.11
8.0xlO~2
0.24
0.33
0.00
0.93
0.21
Q.C. Report
VIII
2.92
0.37
0.28
0.76
1.22
0.00
UCL = 0.99
LCL = 0.86
0.77
Q.C. Report
IX
2.81
0.34
0.26
0.77
2.08
0.00
0.99
0.87
0.70
170
-------
TABLE X. CONTROL LIMITS OF THE ANALYSIS OF SO2 BUBBLERS
Q
Slope
Intercept
Standard
Error
Reruns
Samples
Premeation
Tube
Standard
Solutions
,C. Report
VIII
1.78xlO~4
1.01xlO~2
2.66xlO~3
1.4 xlO~2
1.8 xlO~2
0.9 xlO~2
Q.C. Report
IX
1.76xlO~4
1.01xlO~2
2.64xlO~3
1.4 xlO~2
2.8 xlO~2
n n -2
0.9 xlO
Q.C. Report
VIII
5.82xlO~4
3.29xlO~2
8.68xlO~3
4.6 x 10~2
9.2 xlO~2
0.11
3.0 xlO~2
Q.C. Report
IX
5.77xlO~4
3.31xlO~2
8.62xlO~3
4.5 xlO~2
9.0 xlO~2
0.12
3.0 xlO~2
171
-------
unless the analytical procedures are changed drastically.
Based on the comparison of quality control data obtained during one an-
alytical run with the operational control limits, a determination can be made
about the statistical control of generated data. If the calibration para-
meters were found to be out-osf-control, the entire run was considered to be
out-of-control and the reason for this was investigated. Essentially, the
calibration parameters can get out-of-control either by employing incorrect
standards or because of instrument malfunction. In the first case, corrective
action consisted of normalizing the results for the difference in the stand-
ards. In the second case, the samples were re-analyzed. If calibration par-
ameters were found to be within control but some of the control samples were
above the operational control limits, the run was considered to be question-
able and the reason for the deviation was investigaged. Again, this problem
can be attributed to either the use of an incorrect sample preparation pro-
cedure or to an instrument malfunction. In the first case, the run was con-
sidered to be within control; no corrective action was taken. In the second
case, only the part of the run in which the malfunction occurred (not the
entire run) was considered to be out-of-control and requiring re-analyses of
the involved samples.
It is evident that any quality assurance procedure, by definition, is an
after-the-fact method of evaluating a process. Thus, the corrective actions
dictated by the procedure may have serious economic and/or technical impact
on the total program. For this reason, quality assurance steps should be
taken expeditiously and the corrective actions required applied before irre-
versible damage, such as invalidation of data, takes place.
In the analysis of LACS samples, a "real-time" internal quality control
method was established by which the analyst could evaluate if a particular
analysis were in statistical control. The method was accomplished by trans-
lating the operational control limits of each control sample or calibration
parameter to expected instrument readings which allowed the analyst to eval-
uate the control data during the performance of the analysis. For example,
in the automated analyses, the operational control limits were converted to
percent chart reading expected to be obtained for each control sample and,
therefore, "real-time"quality control was achieved. The disadvantages of this
approach are that the process has to be watched constantly and also that it
requires some subjective judgment to be made by the operator. To avoid
these limitations a computerized quality control is planned as soon as the
analytical instrumentation is interfaced with the laboratory computer. Most
of the software for this step has already been developed. The hardware re-
quired for the interface has been purchased and installed in the computer.
This system will be operational in the very near future. Internal quality
control in the chemistry laboratory will therefore, be performed objectively
by the computer on real-time.
172
-------
REFERENCES
(1) N. Berg and J. Walling, "Quality Control Practices in Processing Air
Pollution Samples", Research Triangle Park, North Carolina (1973).
(2) "Quality assurance Plan for Chemical Analysis for Los Angeles Catalyst
Study", prepared for EPA by Rockwell International, AMC 15 December 1975.
(3) J. C. Puzak, T. A. Clark, and E. Gardner "External Quality Control for
the Los Angeles Catalyst Study Contractor", presented at the Los Angeles
Catalyst Study Symposium, 12 April 1977.
(4) G. F. Evans and C. E. Rodes, "Precision of Sampling and Analytical
Methods", presented at the Los Angeles Catalyst Study Symposium, 12 April
1977.
(5) G. F. Evans and C. E. Rodes, "Summary of Continuous Data", presented at
the Los Angeles Catalyst Study Symposium, 13 April 1977.
173
-------
EXTERNAL LABORATORY QUALITY ASSURANCE FOR
THE LOS ANGELES CATALYST STUDY
John C. Puzak and Thomas A. Clark
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
In order to assess the precision and accuracy of the results reported by
the analytical contractor, Rockwell Air Monitoring Center, two separate
external quality assurance programs are conducted by the Environmental Monitor-
ing and Support Laboratory. The Quality Assurance Branch supplies the con-
tractor with blind quality control samples, which simulate actual field samples
containing sulfate, nitrate, lead, and sulfur dioxide. The Analytical Chem-
istry Branch of Environmental Monitoring and Support Laboratory performs
duplicate [split] sample analyses on approximately 1% of the hi-vol and mem-
brane samples previously analyzed by Rockwell. The blind sample program
provides information on the precision and accuracy of the analytical method-
ology of the contractor. The split sample program provides information on the
comparability of the Rockwell laboratory with the Environmental Monitoring
and Support Laboratory reference laboratory, Analytical Chemistry Branch.
INTRODUCTION
Before awarding the contract for the Los Angeles Catalyst Study (LACS),
the Environmental Monitoring and Support Laboratory (EMSL) of the Environmental
Protection Agency (EPA) , was concerned about the quality of the data that
would be produced during the study. In order to verify and document the
quality of the data, two forms of quality assurance were contained in the
original contract. An internal quality assurance program was to be conducted
by the contractor and an external quality assurance program would be conducted
by EMSL. The external quality assurance for the analytical laboratory was to
consist of two parts, split sample analysis and blind audits.
A split sample analysis quality assurance program was initiated in July
of 1974 when collection of air samples began in LACS. Aliquots of the air
samples are analyzed by the contractor, Rockwell Air Monitoring Center (RAMC),
and the Analytical Chemistry Branch (ACE), EMSL. The split sample program was
begun to determine the comparability between analyses performed by RAMC and
EMSL and to determine analytical variability on real air samples.
In November of 1974 a blind audit quality assurance program was initiated
for RAMC by the Quality Assurance Branch (QAB), EMSL. The blind Audit program
175
-------
was initiated to investigate and document the precision and accuracy achieved
by RAMC during the analysis of ambient air samples collected in LACS.
Both the blind audit and split sample program have changed since their
beginnings in 1974 to accommodate fluctuations in the number of samples
collected in LACS, variability of the analytical results, and EMSL's growing
experience in conducting external audit programs. Table I gives the pollutants,
types of audit sample and frequencies of both the blind audit and split sample
programs used in the last quarter of 1976 and the first quarter 1977 for RAMC
analyses.
Sulfate and nitrate are spiked on the same filter for the blind audit
samples: lead (Pb) is spiked on a separate filter. RAMC must then extract
and analyze these samples as they would an ambient air high volume (Hi-Vol)
sample. Sulfur dioxide (SO2) blind samples are special freeze-dried tetra-
chlormercurate-sulfite solutions. RAMC must thoroughly rinse the audit sample
from a sealed vial with absorbing reagent before analyzing as a normal 50 ml
bubbler sample.
The samples used for blind audits have all been developed by QAB as part
of its program for assisting analytical laboratories to improve their in-house
analysis capabilities. Each lot of samples is analyzed by QAB and a corrob-
orative laboratory before it is accepted for use in the blind audit program.
The QAB analysis, the corroborative analysis, and the attempted spike must
agree to within 5% and the relative standard deviation, for a spike level must
be less than 2.5% or the samples are rejected.
The split samples consist of duplicate 3/4" x 8" filter strips cut from
exposed Hi-Vol filters, collected at the LACS stations. These filters are
sent to ACB and RAMC for analysis. Sulfur dioxide air samples are not split
because collected samples decompose with time.
PURPOSE OF SPLIT AND BLIND AUDIT PROCEDURES
The main use of the blind audit results is to document chronologically
the precision and accuracy achieved by RAMC. EMSL has decided that a meaning-
ful way to display the chronological precision and accuracy is through mean
and range control charts. The control chart format conveniently displays pre-
cision and accuracy versus time and also provides criteria to judge whether
the analytical methods used by RAMC exhibits acceptable variability or are
subject to greater than normal fluctuations. The common usage of control
charts to detect and immediately correct any analytical irregularities cannot
be applied because of the one to two week delays involved in transmitting
the results to EMSL. Corrective action on analytical problems is based on
RAMC's internal quality control data.
A second major use of the blind audit data is to determine if any portions
of the analytical methods employed by RAMC need improvement. This use will be
illustrated in the Results section which discusses low-level sulfate variability
and the corrective measures that were needed.
The third use of the blind audit results is to point out a large bias or
176
-------
TABLE I. BLIND AUDIT AND SPLIT SAMPLE FREQUENCIES
Pollutant
Sulfate
Nitrate
Lead
Sulfur Dioxide
Sample Type
Hi-Vol Filter
Hi-Vol Filter
Hi-Vol Filter
Bubbler
Audit Frequency
Blind Audit - Split Sample
10 /week
10 /week
7/week
10/bi-weekly
6/month
6/month
6 /month
None
177
-------
variability that might occur across all levels for a pollutant during an
analytical period. If this situation occurs, both the EPA project officer and
RAMC are immediately notified. This situation has never occurred during the
blind audit program.
Since the blind audit samples are not real air samples, but synthetic
samples, the split sample program serves to show variabilities encountered on
analysis of real samples. Secondly the split sample results serve as a mea-
sure of the reproducibility on real samples between two laboratories which are
using the same analytical methods. Initially the split sample program was
conducted at a much higher analysis rate than is shown in Table I, but the
larger time delay between analyses, the information obtained in the blind
audit results, and the continuing comparability between ACS and RAMC, allowed
EMSL to cut back to the current split sample rate.
Summary reports for both blind and split sample analysis are periodically
made to the EPA project officer and to RAMC. Split sample results are re-
ported at the end of each quarter, while blind audit results are reported at
twenty-five week intervals.
SPLIT AND BLIND AUDIT PROCEDURES
SPLIT SAMPLE PROGRAM
Samples to be used for split analysis are randomly selected by RAMC from
the Hi-Vol samples collected in LACS. Extra 3/4" x 8" strips are cut from the
Hi-Vol filters and sent to ACB where they are analyzed once per quarter.
Results from both analyses are compared by the EPA project officer. Limits
of +10% or 1 ug/m3, whichever is larger, are the acceptable limits for the
differences between the two laboratories for all pollutants. If differences
are outside these limits, further action such as reanalysis by one or both
laboratories, is usually taken. If significant differences do occur in the
split sample program the blind audit results for the same period are also
evaluated and corrective action is based on a combination of the results from
both audit programs.
BLIND AUDIT PROGRAM
Periodically (see Table I for frequency) a set of samples for sulfate/
nitrate, Pb, and SO^ are sent to RAMC for blind audit purposes. Table II
shows the quantity per level for each blind audit set. RAMC distributes the
audit samples as evenly as possible among the analysis days during the audit
period and intersperses the audit samples throughout their normal analysis
of LACS air samples.
On the last day of the audit period the results for all audit samples
analyzed are reported to QAB. When the results are received by QAB the mean
and range are determined for each pollutant level and plotted on their
respective control charts. The means and ranges reported for all sulfate,
nitrate, Pb, and SO^ blind audits conducted during a twenty-five week period
are used to determine a grand mean (x) and a mean range (R) for each pollutant
spike level. Outliers and out of control results are not included in the
178
-------
TABLE II. DISTRIBUTION OF BLIND AUDIT SAMPLES ACROSS ANALYTICAL RANGE
High Level
Mid Level
Low Level
Blanks
Total :
SO
-------
calculation of x and R. The grand mean and mean range can then be used to
calculate control chart limits which will be used for the next twenty-five
weeks of blind audits. Thus, the control limits are revised if necessary,
every twenty-five weeks.
The construction of control limits will be illustrated using sulfate
blind audit results. The control chart limits can be determined from the
equations given in Table III; x values for the control charts are the spike
value times the average recovery shown in the previous twenty-five weeks of
analysis. R is taken from a plot of mean range versus mean results for each
spike level for a particular pollutant. Figure 1 gives a plot of mean range
versus grand mean for RAMC blind audit sulfate results in the last twenty-five
weeks of 1976. The R to be used for control limit calculations are read from
the curve at the x-axis value corresponding to the spike levels to be used in
the next series of blind audits. Figure 2 gives an example of a sulfate mean
Q
and range control chart for a sample corresponding to about 20 yg SO^/m3. A
similar pair of control charts would be made up for each different sulfate
spike level.
Besides the control chart display the blind audit reports determine the
average recovery for each analytical method audited. The average recovery
for a method is defined to be the slope of the least squares linear regression
of the mean of the reported values versus the spike value.
Since there is so much data collected over a twenty-five week audit
period, plots of reported analysis versus spike value will be used to summa-
rize the blind audit results in the remainder of this paper.
RESULTS OF BLIND AND SPLIT SAMPLE AUDITS
SULFATE
A plot of RAMC mean sulfate analysis versus spike value for audits
conducted during the last twenty-five weeks of 1976 is shown in Figure 3.
Each triangle represents the mean of 15 to 18 RAMC analyses. The slope
(0.974) and small scatter about the regression line indicate very good
average recovery for the sulfate analytical method over its entire range.
The numbers in parentheses on the y-axis indicate the coefficient of
variation for RAMC analyses at the indicated sulfate level. These values,
below 6% over the entire range, show good precision for the RAMC sulfate
analyses. The values in parentheses on the x-axis are the ambient air
concentrations in yg SO^/m which correspond to the yg of sulfate found on
a 3/4" x 8" strip from a twenty-four hour Hi-Vol sample.
The agreement of results below'500 yg (10\ig/ml of filter extract) was
achieved only by reanalyzing the extracts by a sulfate procedure that is
10 times more sensitive than the routine sulfate procedure used by RAMC to
analyze LACS air samples. Table IV compares the analytical results for blind
audit samples containing less than 500\ig of sulfate/strip obtained by both
the routine and sensitive procedures. A marked improvement in both precision
and accuracy is achieved by using the sensitive procedure to analyze these
sulfate samples. Because of these findings RAMC now reanalyzes all filter
180
-------
TABLE III. CONTROL LIMITS FOR BLIND AUDIT MEAN AND RANGE CONTROL CHARTS
Range Control Charts
Upper Control Limit (UCL) = D^ R
Lower Control Limit (LCL) = D$ R
Mean Control Charts
UCL = X + A2 R
LCL = X - A2 R
R = Mean of Weekly Ranges
X = Mean of Weekly Means
Z>3, D^, and A^ are control chart factors from Duncan, Acheson J.
Quality Control and Industrial Statistics, 968-969, Fourth Edition
(1974), Richard D. Irwin, Inc.
181
-------
200
1 I 1 1 I I I I T
I I I I I I
I I I I
3500
(20 Mg/m3)
GRAND MEAN (x) RAMC
7000
(40 (ug/m3)
Figure 1. Range versus mean for RAMC blind sulfate analysis (1976)
182
-------
UCL (x) = 3675
O O 0
i i I I I Y i—i i i i I—i—M—I I I I I I i I—h-1—I—h
0 0
= 3623
x CONTROL CHART
LCLffl-3570
I-J
00
U)
_,—I O i T I O O
UCL(R) = 133.3
I I I I I t I I I I
R = 51.8
R CONTROL CHART
LCL(R)=0.(i
Figure 2. Sulfate Control Charts.
-------
7000
(1.5%)
LU
_j
Q.
<
-------
TABLE IV. COMPARISON OF RAMC ANALYSES OF LOW LEVEL SULFATE BLIND
AUDIT SAMPLES BY SENSITIVE AND ROUTINE SULFATE METHODS
Spike Total yg
ROUTINE METHOD
Mean Std.
Total Deviation
SENSITIVE METHOD
Mean Std.
Total Deviation
0 (Blank)
200 (1.0 yg/m3J
250
405
498 (2.5 yg/m3;
164
317
319
464
512
33
33
40
38
25
20
200
261
408
484
18
12
16
18
24
185
-------
extracts which fall below 10 yg SO^/ml by their sensitive sulfate procedure.
The split sample results from the first quarter of 1977 show agreement
between RAMC and ACB to within 5%.
NITRATE
A plot of the RAMC blind nitrate results versus the spike value for the
last twenty-five weeks of 1976 is shown in Figure 4. Again, each triangle
represents the mean of 15 to 18 RAMC analyses? the coefficient of variation
of the RAMC analysis is indicated in parentheses on the y-axisjand the am-
bient air concentration corresponding to a twenty-four hour Hi-Vol sample is
in parentheses on the x-axis. The slope (0.972) and scatter about the re-
gression line indicate good recovery down to about 300 ygr #03 per sample
(6 yg NO? /ml of filter extract) . Results below 300 ygr #03 are more variable
and are less accurate. A coefficient of variation of less than 4% over the
analytical range indicates that precision for nitrate analyses is excellent.
Since the actual nitrate levels found in the LACS air samples are
usually greater than 5 yg/m3 or 1000 yg/strip, no grreat effort has been made
to improve the method at low nitrate levels. Audit results for nitrate have
been similar to these since the blind audits began in November 1974.
The split sample results for samples collected in July and August 1974,
however, showed a difference between RAMC and ACB that was greater than the
acceptable limits. Table V gives examples of the split sample differences
that occurred in 1974. Investigation found that RAMC nitrate standard
solutions used in July and August had been improperly prepared. RAMC records
were complete enough so that recalculating the results was the only corrective
action that was necessary.
First quarter 1977 split sample results for nitrate indicate agreement
between RAMC and ACB to within 5%.
LEAD
Figure 5 gives two plots for RAMC blind Pb results versus the spike
values. The triangles represent audit data from 1976 and the squares rep-
resent audit data from 1975. Each triangle is the mean of 10 RAMC analyses
and each square is the mean of 18 to 30 RAMC analyses. The coefficient of
variation of RAMC analyses is shown in parentheses on the y-axis and the
ambient air concentration for a twenty-four hour Hi-Vol sample is shown in
parentheses on the x-axis. The slope for the 1976 results is 0.871 and the
slope for the 1975 results is 0.900. The RAMC blind Pb results have consis-
tently been 10% to 12% low since audits were started in 1975. RAMC is
investigating their Pb extraction and analysis procedure during the first
quarter of 1977 to see if recoveries can be improved. The coefficient of
variation is less than 3%, which indicates that precision of blind Pb analyses
is excellent.
The Pb split sample analyses performed in late 1974 and early 1975
showed the variable percent differences given in Table VI. A detailed
186
-------
£UUU
(1.5%)
I/SAMPLE
%**
H
uj 1000
^ (2.0%)
>
0
oc
400
(4.0%)
0
1 1 I I IT-;
— RAMC = 0.972 SPIKE f±_
&*
^^
"" / ~~
^"•^ f^^\
/'
—
h
~ xAA//
/i i i i i i i i i
0 200 1000
2000
SPIKED/SAMPLE
Figure 4. Comparison of RAMC blind nitrate results with spike values.
187
-------
TABLE V. NITRATE SPLIT SAMPLE DIFFERENCES - AUGUST 1974
RAMC fyg/m3; EMSL (\ig/m3) % Difference
10.97 9.90 + 10.3
8.74 7.20 + 19.3
21.72 18.30 + 17.1
19.45 17.50 + 10.6
7.03 5.90 + 17.5
15.40 13.10 + 16.1
11.29 8.80 + 24.8
188
-------
2000
(2.0%)
<
D)
I-
_l
3
UJ
OC
LJJ
>
u
1000
(2.0%)
200
(3.0%)
RAMC = (0.871) SPIKE (1976)
D RAMC = (0.900) SPIKE (1975)
0 200
1000
SPIKE MS/SAMPLE
2000
(10;ug/m3)
Figure 5. Comparison of RAMC blind lead results with spike values.
189
-------
TABLE VI. LEAD SPLIT SAMPLE DIFFERENCES (AUGUST - DECEMBER 1974)
RAMC EMSL % Difference
4.60 4.65 -1.1
4.30 13.55 -103.6
4.30 7.83 -50.2
9.50 12.40 -26.5
1.70 2.93 -53.1
2.70 2.88 -6.5
10.50 12.63 -18.4
7.00 6.89 +1.6
3.70 6.34 -52.6
190
-------
investigation of both ACB and RAMC Pb analytical procedures using real and
synthetic Hi-Vol samples showed an equipment malfunction and contamination
in the ACB extraction procedure. The RAMC results for LACS analyses were
accepted and the ACB extraction apparatus was modified and moved to improved
facilities.
First quarter 1977 Pb split analysis shows agreement between RAMC and
ACB to within 10%.
SULFUR DIOXIDE
Figure 6 gives a plot of RAMC SO2 blind audit results versus the spike
value. Each triangle represents the mean of 15 to 30 RAMC SO2 analyses. The
coefficient of variation of the RAMC analyses is given in parentheses along
the y-axis and the corresponding ambient air concentration is given in paren-
theses on the x-axis. The slope of the line (1.043) indicates good agree-
ment with the spike value. Results similar to these have been obtained since
the beginning of the blind audit program.
Because of the temperature instability of the SO2 bubbler samples, no
split sample analyses have been attempted.
CONCLUSIONS
The external quality assurance program consisting of split and blind
sample analyses has been quite useful in improving, verifying and quantifying
the quality of data obtained in LACS. Both the Pb and sulfate analytical
procedures have been improved because of the results obtained in the external
audit program. Detection and correction of analytical problems has been
timely enough to avoid invalidating large blocks of analytical data. Besides
upgrading analytical methods, a chronological history of accuracy and
precision is available from the start of this project. The results show that
an external quality assurance program can be a valuable part of an air
monitoring project.
191
-------
50
(2.5%)
a.
<
t/i
~S>
= 25
< (4-0%)
cc
UJ
10
(10.0%)
i i i r
RAMC = 1.043 SPIKE
i i i i i i i i i i
10
(35 jug/™3)
25
(85 jug/ro
SPIKE jug/SAMPLE
50
Figure 6. Comparison of RAMC blind sulfur dioxide result with spike values,.
192
-------
THE QUALITY ASSURANCE PROGRAM EMPLOYED DURING SAMPLING
Charles E. Rodes, Bobby E. Edmonds, and Malcolm C. Wilkins
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
This paper describes an extensive quality assurance program involving
sampling equipment used in the Los Angeles Catalyst Study. This quality
assurance activity is completely separate from quality assurance measures
taken during analysis of integrated samples. The program generally consists
of dynamic pollutant level checks of the continuous instruments and flow
calibration checks of the integrating samplers. The discussion covers the
types of internal and external checks used and typical data from various
categories of samplers. It also presents results from an external test of
field equipment performed by a person not involved in the field program.
External audits generate audit information on the accuracy of the sampling
equipment. These audits also provide comparability between routine on-site
checks and reference materials kept at the Environmental Protection Agency,
Research Triangle Park, North Carolina.
INTRODUCTION
The Environmental Monitoring and Support Laboratory (EMSL) has conducted
an extensive Quality Assurance (QA) program on all monitoring activities of
the Los Angeles Catalyst Study (LACS) . The purposes of the QA programs
utilized during LACS ambient air monitoring operations are to: (1) provide
the precision required using a given method; and (2) provide comparability
information to relate, through primary and secondary standards, the data
collected by contractor and Environmental Protection Agency (EPA) laboratories.
This study is comprised of essentially two separate operations — one carried
out on-site, using continuous and integrated sampling in the field, and one
which carries out analysis of integrated samples in the laboratory. Separate
QA programs are therefore followed.
The sampling methodologies can be divided into two categories: (1) those
that monitor continuously and produce in situ hourly averages, and (2) those
that collect samples integrated over a selected time interval. The continuous
samplers are calibrated by introducing known quantities of the pollutant of
interest or comparing their performance with reference samplers. The flow
rates of integrated samplers are calibrated against reference flow devices.
Additional Quality Assurance measures in support of the monitoring effort
include:
193
-------
• keeping daily record logs of maintenance and subsystem sampler
performance,
• providing written operation and calibration procedures for all
equipment,
• conducting unscheduled performance tests on instrument operations by
an individual other than the site operator, and
• data validation.
CONCLUSIONS
The quality of data being generated by the Los Angeles Catalyst Study
(LACS) is monitored very closely. The quality assurance elements employed in
the LACS sampling operations are the same as those employed in all other
Environmental Monitoring Branch (EMB) studies.
As demonstrated by examination of the routinely collected internal audit
and calibration information and the results of external audits, the data
quality of all measurements at the LACS is very good compared to most field
studies. In any large study such as LACS where there are multiple operators,
sites, and samplers, some problems usually occur. The quality assurance
programs serve the function of not only minimizing errors and quantifying
precision, but also pointing out problem areas for subsequent correction.
Without a quality assurance program, the data collected are virtually inde-
fensible — for nearly any purpose. The QA information is indispensible in
that it defines specifically which data sets are within study tolerance limits
for validation purposes, and provides a means in some cases of correcting
invalid data.
The only element remaining to be formulated in the LACS sampling QA
program is a procedure to compile the collected check and calibration data to
provide readily available summary QA statistics along side the data summaries
being presented.
DEFINITIONS
For the purposes of this study, the following nomenclature for several
types of calibration materials and audits which are referred to in this paper
was established.
External audit: An independent calibration test performed by an individual
other than the routine site operator(s). This test is typically per-
formed with separate working standards and calibration apparatus supplied
by the individual performing the audit.
Internal audit: A calibration test performed by the site operator on the
sampling equipment other than a scheduled multi-point calibration.
Routine zero and span checks on continuous samplers and periodic one or
two point flow checks are internal audit procedures.
194
-------
Transfer standard.- Any standard which is calibrated against primary standard
materials. This includes the National Bureau of Standards (NBS) trace-
able volumetric flowmeter in the laboratory, the gravimetrically cali-
brated sulfur dioxide (SO2) permeation tubes, and the cylinders of com-
pressed gas calibrated in the laboratory against NBS traceable materials.
Working standards: A standard used routinely and often prepared in quantity,
which is calibrated against the "transfer standard." This includes the
routine span cylinders used to check and calibrate the continuous instru-
ments, and the flow orifices used to calibrate the integrating samplers.
INTERNAL SAMPLING QUALITY ASSURANCE PROGRAM
For the LACS program, all quality control measures that do not fall under
the definition of "external audit" are considered "internal." These "internal"
audits will be discussed for continuous and integrating sampler methodologies.
CONTINUOUS POLLUTANT SAMPLERS
The carbon monoxide (CO), nitrogen oxide (NO), nitrogen dioxide (NO2) and
ozone (O^) analyzers are calibrated against compressed gas working standards.
The CO instruments which operate at 0-50 ppm full scale are calibrated using
cylinders of 40 ppm CO in air, which are certified against a "transfer stand-
ard" cylinder. The NO channel (0-1.0 ppm full scale) of the NO/NO2 instrument
is calibrated by diluting a "working standard" mixture of 100 ppm NO in nitrogen.
The NO2 channel (0-0.5 ppm full scale) is calibrated by gas phase titration
(GPT) of the "working standard" NO mixture with 03 to produce specific concen-
trations of NO2. The O$ samplers are also calibrated using the NO "working
standard," but utilizing reverse GPT between the NO mixture and a generated
source of 03. Since there is one "working standard" cylinder of each type gas
(CO and NO) at each instrumented site, a "transfer standard" cylinder of each
type is used to provide a common reference between sites.
The Total Sulfur (S) analyzers (0-0.10 ppm full scale) are calibrated
using SO2 permeation tubes, which qualify as both "transfer standards" and as
"working standards." These tubes are calibrated in the laboratory gravi-
metrically on a recording microbalance and sent to the field as needed. They
typically have an SO2 output of 0.2 yl/min to provide calibration points in
the 0-0.10 ppm range.
All continuous pollutant instruments are calibrated with a "zero" and
four upscale points. These calibrations are performed monthly with checks of
the "zero" and a single 80% span (upscale) point made on a more frequent
basis. Table I lists the continuous instruments and the frequency of these
checks.
The frequency of 1-point spans is determined from past history of the
type of instrument being used, balanced against the total time required to
perform the check. The latter item not only requires manpower, but also
results in data lost during the span operation. The Total Sulfur analyzers at
LACS require over 1 hour for a zero and single point span operation.
195
-------
TABLE I. FREQUENCY OF CONTINUOUS SAMPLER CHECKS
POLLUTANT PRINCIPLE 5-POINT CALIBRATION "ZERO" 1-POINT SPAN
CO NDIR* MONTHLY DAILY DAILY
NO CHEMILUMINESCENCE MONTHLY DAILY WEEKLY
NO2 CHEMILUMINESCENCE MONTHLY DAILY WEEKLY
O3 CHEMILUMINESCENCE MONTHLY DAILY WEEKLY
TOTAL S FPD** MONTHLY DAILY WEEKLY
* NON-DISPERSIVE INFRARED
** FLAME PHOTOMETRIC DETECTION
-------
Especially for the CO analyzers the monthly calibration becomes little
more than a linearity check, since the span adjustments are reset as needed.
These tolerances were established subjectively based on study requirements and
past performances of the same types of analyzers.
The % chart values in Table II indicate for the operators how precise the
unadjusted readings of a "zero" or upscale point should be before corrective
action is taken. Again these limits are based on past records and cost-
effectiveness. If a zero or span gas input is made and the sampler response
falls within the respective "Tolerance" listed, no adjustments are made. If
the response falls outside the tolerance band, but within the "Recalibration
Limit," the sampler "zero" or "span" is readjusted. If the response falls
outside the "Recalibration Limit," the sampler is considered as working impro-
perly, and a service check performed, followed by a complete recalibration.
The plot in Figure 1 of the CO "zero" drift data for a typical month
illustrates the type of information available from daily checks. Note that
only 2 unadjusted "zero" checks (5th and 6th) were outside the acceptable
±1.0% band. Apparently on the 5th day an improper "zero" calibration was
performed which provided an erroneously high reading. Readjustment by the
operator followed by a correct "zero" on the 6th resulted in a negative "zero"
that was outside the readjustment band. The problem did not recur during the
rest of the month.
A procedure for adjustment of data possibly affected by out-of-tolerance
response is also delineated. If the instrument "zero" and "span" are within
the "Recalibration Limit," the data collected are adjusted for monotonic zero
change only, back to the previous acceptable "zero" check. No attempt is made
to correct for span errors, since the instruments have linear response charac-
teristics. If the instrument checks are outside the "Recalibration Limit" all
data is voided back to the last acceptable "zero" and "span."
As a check on span gas stability and to provide comparability between
"working standard" gas cylinders, "transfer standard" cylinders are calibrated
periodically in the laboratory and sent to the field site. These "transfer"
cylinders are then used to check the calibration of the "working standards"
and calibrate new ones as old ones become exhausted. The "working standard"
calibrations are also checked just before being depleted against the transfer
standard to verify stability.
The flowmeters on the continuous samplers are not calibrated individually
(except in special cases), as their calibration is inherent in the dynamic
sampler calibration. The mass flowmeters built into the calibration systems
are calibrated on a quarterly basis.
INTEGRATING POLLUTANT SAMPLERS
The high-volume samplers (hi-vols), 102 mm membrane samplers, and dicho-
tomous samplers are calibrated for flowrate measurement. In all cases the
"transfer standards" for flow measurement are calibrated against NBS traceable
197
-------
TABLE II. LACS INSTRUMENT ADJUSTMENT TOLERANCES
POLLUTANT
CO
NO
N02
°3
TOTAL S
ZERO TOLERANCE* SPAN TOLERANCE* RECALIBRATION LIMIT**
% CHART
±1.0
±1.0
±1.0
±1.0
±2.0
% CHART
±3.0
±3.0
±3.0
±3.0
±5.0
% CHART
±5.0
±5.0
±5.0
±5.0
±5.0
* BEFORE ADJUSTMENT IS REQUIRED.
** SAMPLER MUST BE RECALIBRATED IF UNADJUSTED SPAN OUTSIDE THIS LIMIT.
198
-------
UNACCEPTABLE PERFORMANCE
8 10 12 13 16 18 20 22 24 26 28 30
Figure 1. Typical zero drift for CO sampler.
-------
positive displacement devices. The hi-vol and membrane samplers are cali-
brated with audit orifices. These orifices are calibrated using special
Rootsmeters which have metric (m3) scales readable to either 2 or 3 decimal
places, depending on the flow range required. The hi-vols operate typically
at 1.4 m^/min, while the membrane samplers operate at 0.14 m^/min. The dicho-
tomous samplers because of their lower flow rate (14 SL/min fine channel and
0.4 Z/min coarse channel) must be calibrated against a dry-test meter or
bubble flow tube.
Flow calibrations were scheduled on a monthly basis at the beginning of
the study, but were dropped back to quarterly periods based on collected data.
The data from these flow calibrations are processed by a Hewlett-Packard 9821
programmable calculator which has a plotter and printer. The audit devices
such as the calibration orifices are calibrated in the laboratory, and plots
with interpolated tables generated by the calculator. A typical membrane
audit orifice calibration plot is shown in Figure 2. The programmable calcu-
lator has several advantages in a quality assurance program over hand process-
ing of calibration data. This type of calculator: (1) standardizes formats
and procedures, (2) eliminates errors in calculation, (3) utilizes statis-
tically valid best fit procedures that are tedious by hand, (4) is faster, (5)
generates an interpolated table to eliminate chart reading errors, and (6)
allows technicians unfamiliar with fluid mechanics or statistics to provide
the best possible calibration data.
One of the statistics generated by the calculator, the correlation
coefficient R, is used to determine the degree of scatter of the data points
about the best fit line. It was determined that an acceptance level could be
established for each type of calibration, using a "typical" operator known to
perform careful calibrations. Setting these limits provides a guideline for
the technician to determine if his calibration is comparable to those gene-
rated in the past. If not, a recalibration is required. At present only the
calibration of the hi-vol audit orifice using the Rootsmeter and the calibra-
tion of a hi-vol in the field with the audit orifice have correlation coeffi-
cient limits established. An R value of 0.9995 must be obtained for the 5-
point orifice calibration and 0.9990 for the 5-point sampler calibration for
these types of calibrations to be acceptable for the LACS. Coefficient accept-
ance limits for the other methods are currently being established.
LACS EXTERNAL SAMPLING QUALITY ASSURANCE PROGRAM
The need for audit checks performed by an individual other than the
routine site operator and apparatus not associated with the study were recog-
nized from the beginning of the LACS. Because of initial resource limitations
and the confidence in the site operations by EPA operators, it was felt that
a full scale external audit program could be omitted and only certain key
aspects checked externally. After March, 1976, the site operations were
turned over to the analysis contractor (Rockwell Air Monitoring Center), which
began collection and analysis of LACS samples. An external audit program was
then felt to be necessary to: (1) demonstrate comparability, (2) pinpoint
problems, and (3) check the contractor's performance.
200
-------
MEMBRRNE RUDIT
ORIFICE CRLIBRRTIDN
S/N I H
DRTE 3/IE/77
20. B C
752 . 0 MM HE
DFERHTDR J^C.
3 7 B B 1011 I2I31HI
Delta P, in. H2°
Figure 2. Sample calibration plot.
7
-------
In December of 1976, an employee of EMB performed external audit checks
(usually 2 span points) on all LACS samplers using test equipment and stand-
ards brought from Research Triangle Park, North Carolina. The continuous
samplers were checked with span gas concentrations; the integrating samplers
were checked for flowrate accuracy; and the flowmeters in the calibration
systems also checked for flowrate accuracy. A comprehensive report was pre-
pared listing the % differences and the number of samplers not operating
within "acceptable limits." Table III is a summary of the results obtained
from this external audit. Overall the results indicate generally acceptable
operation of all LACS samplers as exhibited by the CO audit data in Figure 3.
The two problem areas pinpointed were the somewhat poor performance of the
Total Sulfur analyzer and the number of hi-vol and membrane sampler calibra-
tions not falling within the "acceptance" limits.
The out-of-limits calibrations of the hi-vols and membrane samplers were
readily traced to an inadvertent skipping of the previous quarterly hi-vol and
membrane sampler calibration. Figure 4 is a sample plot of the hi-vol audit
data. Although the average % differences were fortunately not substantial,
the omission of the routine calibration is unacceptable, and a revised opera-
tions check sheet has been instituted along with provisions for periodic
review by EPA personnel.
Except for the Total Sulfur analyzers, the continuous samplers performed
extremely well as demonstrated by the typical CO audit plot shown in Figure 5.
The Total Sulfur analyzers were expected to have a greater degree of vari-
ability during audit checks because of the very sensitive operational range
(0.10 ppm full scale). However, examination of the comparison of the Total
Sulfur monthly averages with those of the SO^ bubblers operated at the same
sites (see Figure 6) indicated that there were at times biases in one of the
measurements. The line drawn through the data points represents the regres-
sion obtained in 1975, and all the points are within approximately ±8 ygr/m^
of this line except for those grouped above and below the line. These "out-
lier" points are all Site C values from 1976. Closer examination of the Total
Sulfur sampler and associated calibrator at Site C resulted in the discovery
of an intermittent leak in the calibrator assembly. This leak had apparently
been causing problems all during the year, and was only discovered by exchang-
ing the calibration system at Site C with that from Site A. This leak has
since been corrected, but demonstrates that even with the best QA program some
problems can still occur.
ACKNOWLEDGEMENT
The external audit data in this report were collected and compiled by Mr.
Thomas A. Lumpkin of the Environmental Monitoring Branch, Environmental
Monitoring and Support Laboratory, United States Environmental Protection
Agency.
202
-------
TABLE III. SUMMARY OF EXTERNAL AUDIT RESULTS
K)
Instrument
Audited
Total
Sulfur
Nitric
Oxide
Ni trogen
Dioxide
Ozone
Carbon
Monoxide
Mass
Flowmeters
Number
Audi ted
2
2
2
2
2
10
Number
Operational
2
2
2
2
2
10
Audit
Level
0 . 06 ppm
0 . 40 ppm
0.20 ppm
0.20 ppm
20.0 ppm
2500 SCCM
Acceptable Limits
For Difference
- 0.009
* 0.06
- 0.03
- 0.03
- 3.0
* 125
ppm
ppm
ppm
ppm
ppm
SCCM
Number Out-
side Limits
1
0
0
0
0
1
Average
Difference
- 0.009 ppm
- 0.012 ppm
- 0.005 ppm
- 0.007 ppm
0.28 ppm
- 48 SCCM
High Volume
Sampler Flow
Rate
10
10
45 SCFM
10% of Audit
Value
-1.79 SCFM
Membrane/
Cascade
Sampler Flow
Rate
4 SCFM ± 10% of Audit
Value
0.068 SCFM
-------
N)
O
*>.
O AUDIT
• CALIBRATION
FIELD AUDIT
PROJECT NAME LACS
STATION
SITE C
INSTRUMENT
MSA CARBON —
MONOXIDE
DATE
12/6/76
RANGE
50.0 PPM
AUDITOR
T. LUMPKIN
30 40 50 60
AUDIT CO CONCENTRATION, PPM
70
30
90
Figure 3. Comparison of CO audit with last calibration.
-------
O
Ui
I I
O AUDIT
• CALIBRATION
HI VOL FIELD AUDIT
PROJECT NAME LACS
SITE B
HI-VOL S/N_
CAL DATE
6618
6/8/76
AUDIT DATE_
AUDITOR
12/2/75
T. LUMPKIN
20 30 40 50 60 70
DICKSON READING
80
90
100
110
Figure 4. Comparison of hi vol audit with last calibration.
-------
SO2 bubbler,
Figure 5. Comparison of monthly averages by site.
-------
PRECISION OF LACS SAMPLING AND ANALYTICAL METHODS
Gary F. Evans
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
Two special studies have been conducted at the Los Angeles Catalyst Study
sites in an effort to identify and quantify the major sources of error inherent
in the sampling and analytical methods employed.
First, co-located high volume samplers were operated concurrently for
twelve 24-hour periods at a location 700 feet upwind from the freeway. Follow-
ing equilibration and weighing, each exposed filter was analyzed in duplicate
(using center and end strips) for concentrations of lead, ammonium ion, nitrate
ion and sulfate ion by each of two laboratories. The results were used to
construct estimates of the amount of variability (i.e., "variance component")
between samplers, between laboratories and within laboratory, in addition, it
was possible to investigate effects of strip position (center versus end) and
the time lag between sample collection and chemical analysis on reported
pollutant concentration.
The second study consisted of operating five samplers (two high volume
samplers with glass fiber filters, two membrane samplers with cellulose ester
filters and a single membrane sampler with teflon filters) at each of two
sites (upwind and downwind) on a 4-hour interval (3-7 p.m.) for twelve days.
In addition to final weight, each filter was analyzed in duplicate for sulfate
ion concentration by each of two laboratories. Estimates were obtained for
the between sampler, between laboratory and within laboratory components of
variance. Finally, the effects of site and method on reported pollutant
concentration were investigated.
INTRODUCTION
Two external quality assurance programs are being conducted in the Los
Angeles Catalyst Study (LACS) to help measure and ensure the quality of data
from the high volume (hi-vol) samplers operated at the sites. In the "split-
sample" program, the analysis contractor, Rockwell Air Monitoring Center
(RAMC), provides the referee laboratory, Environmental Monitoring and Support
Laboratory (EMSL) , with a strip from a representative sample of one to five
percent of the total number of filters processed. The analyses obtained by
the two laboratories for lead (Pb) , ammonium ion (NHt^+), nitrate ion (NO^)
and sulfate ion (SO^=) are compared to provide some" measure of the interlabora-
tory reproducibility for a given filter. In the "blindsample" program, each
207
-------
laboratory is periodically provided with audit solutions of known concentra-
tions for each of the pollutants of interest so that some measure of the
accuracy of the analytical instruments used by each laboratory may be con-
structed. While these programs have produced much useful data, they are not
sufficient in scope to establish the total variability of the monitoring
system (including sampling) nor are they sufficiently comprehensive to allow
the identification of the specific components of the monitoring system to
which results are most sensitive (e.g., collection of samples, pre-conditioning
of filters, cutting of strips, preparation of solutions, operation of analyt-
ical instruments, etc.).
For these reasons, two short-term studies were designed and conducted at
the LACS sites. The first involved hi-vol samplers operated on a 24-hour
integrating period and was conducted in the summer of 1975. The second in-
volved both hi-vol and membrane samplers operated on a 4-hour integrating
period and was conducted in the summer of 1976.
TWENTY-FOUR HOUR HIGH VOLUME SAMPLER STUDY
PURPOSE
In the summer of 1975, a study involving 24-hour hi-vol samplers was
designed and conducted at the site of the LACS freeway monitoring project to
accomplish the following objectives:
9 To determine the variability inherent in the collection of samples
by comparing the results of duplicate (i.e., co-located, simultane-
oulsy operated) hi-vol samplers.
• To determine the between-laboratory variability in the analytical
process by comparison of results from RAMC and EMSL (same filter).
9 To determine the within-in laboratory variability in the analytical
process by comparison of duplicate strips and instrument readings
(same filter, same lab).
• To determine whether any general gradients in pollutant concentrations
exist across the area of an exposed filter.
• To isolate those components of the monitoring system which give rise
to the greatest amount of variability and which, should receive the
greatest emphasis in future quality assurance efforts.
• To assess the quality of LACS data generated from hi-vol samplers
[total suspended particulates (TSP), Pb, NH^+, NO$ and SO^=] by
estimating the precision of the total monitoring system for each
pollutant.
EXPERIMENTAL DESIGN
A temporary site, designated A' and located 700 feet from the freeway on
the upwind (ocean) side, was selected for this experiment (Figure 1). Site A'
208
-------
Figure 1. Aerial view of sampling sites.
-------
was selected because it would theoretically be removed from the area in which
concentration gradients exist for freeway-generated pollutants so that both
samplers may be presumed to be exposed to the same (background) pollutant
concentrations.
Two calibrated hi-vol samplers (designated #195 and #197, respectively)
were placed side-by-side at the site and operated concurrently on a 24-hour
(midnight-to-midnight) basis for a total of twelve days. The sampling days
were not consecutive as that would have required the operation of samplers in
tandem. The filters exposed during the first six sampling days were identified
by date and sampler and delivered to RAMC for initial processing.
After equilibration and weighing, RAMC cut a pair of standard 3/4 inch
strips from both sides of each filter and two pair of strips from the center
(see Figure 2). Strip position was specified for two reasons. First, it
was felt that maximum variability due to strips could be introduced into the
experiment by requiring that duplicate strips be as far removed from each
other as possible and secondly, by keeping track of the position from which a
strip was cat, soit would later be possible to test whether the center of exposed
filters tends to differ systematically in pollutant concentration from the
ends of the filters. These four pairs of strips were assigned identification
codes for the sampler (#195 or #197), analysis laboratory (RAMC or EMSL),
strip position (center or end), and sampling date. The twenty-four pairs of
strips designated EMSL were sent to EMSL for analysis.
Each laboratory prepared one member of each pair of strips for Pb analysis
(by atomic absorption) and the other member for analysis of ammonium ion (by
sodium phenolate), nitrate ion (by reduction-diazo coupling) and sulfate ion
(by methylthymol biiae (MTB) colorimetric). To the extent possible, each lab-
oratory was requested to introduce these solutions into the stream of their
routine analytical work, rather than to process them together as a batch.
The exposed filters from the final six days of sampling were sent to EMSL
for initial processing (equilibration, weighing, cutting and identification of
strips). These strips were then divided between the two laboratories and the
chemical analyses were completed.
The experiment was designed to isolate_each of_three major components of
the total monitoring system for Pb, NH^, NO^ and SOi+ in order to determine
whether and to what extent each component contributes to the variability in
reported pollutant concentrations. The potential sources of variation consid-
ered are defined below:
(1) Variation in Sample Collection — Assuming that the two high volume
samplers employed in the experiment were randomly selected from the- •
infinite population of such samplers, the concentrations reported
(averaged over laboratories and strips) may then be compared to test
whether the samplers yielded results which were significantly dif-
ferent and, if so, to provide an estimate of the amount of variabil-
ity (i.e., variance component) contributed by the factor "sampler."
(2) Variation Between Laboratories — Similarly, assuming that the two
210
-------
HI VOL I
HI VOL I I
1* OBTAIN FINAL WEIGHTS
Si CUT PAIRS OF
STRIPS
3. PERFORM PB
ANALYSES
4* PERFORM SO./ND-
ANALYSES
1 5 II 12 ISIS 13
13 13 13 IS 15 IS IS IS
IS IS IS
Figure 2. Twenty-four hour study sample analysis.
-------
laboratories employed in the experiment are a representative sample of the in-
finite population of such laboratories, the concentrations reported by lab-
oratory (averaged over sampler and strips) may then be compared to test whether
laboratories produced significantly different results and, if so, to provide
an estimate of the variance component for the factor "laboratory."
(3) Variation Within Laboratory — Since each laboratory processed two
strips from each filter, it -is possible to obtain an estimate of the
variance component for the factor "strips" (within filter, within
laboratory). This factor represents the variability contributed by
all the analytical procedures conducted within a given laboratory
(cutting strips, preparation of extraction solutions, and operation
of analytical instrument). One laboratory (EMSL) also reported
duplicate instrument readings for each of the extraction solutions
prepared in the experiment. Using this data it is possible to ob-
tain an estimate of the variance inherent in the analytical instru-
ment itself. Within laboratory variance, then, may be subdivided
into components for "strip preparation" (cutting and extraction)
and "instrument reading."
The experiment is balanced in the sense that the dependent variable (mea-
sured pollutant concentration for a given 24-hour period) is uniformly exposed
to two "crossed" independent variables (sampler and laboratory). Strip posi-
tion may be considered a "nested" variable (within laboratory).
Because of the manner in which the experiment was designed and conducted,
it was possible to evaluate the effects of two other potential sources of
variation in the monitoring system. Since each laboratory processed a strip
cut from the center and from the end of each filter (to insure maximum varia-
bility due to strip), a comparison was made of the results by strip position
(averaged over sampler and laboratory) to determine whether center strips
tended to differ from end strips in some systematic manner. Also, because one
laboratory (RAMC) processed all the end strips from one set of filters approx-
imately two months after they had processed the corresponding center strips,
it was possible to evaluate whether the time lag between sampling and analysis
date (within filter, within laboratory) had any impact on recorded concentra-
tions.
DATA SUMMARY
Total suspended particulate data appear in Table I. As explained in the
previous section, equilibration and final weighing were performed by RAMC for
the first six days of the experiment and by EMSL for the final six days. It
was not practical to have each laboratory obtain TSP values on the same filter,
so that a direct determination of the effect of laboratories on TSP results
is not available.
It is possible, however, to evaluate the effect of samplers. The percent
difference between duplicate samplers is tabulated for each day and summarized
by laboratory. The sampler differences are generally small (0.8% overall for
RAMC and 3.6% overall for EMSL) and neither sampler produces results which are
consistently higher than the other. A paired-t test was performed for each
212
-------
TABLE I. DATA SUMMARY TSP (]ig/m3)
RAMC EMSL
Sampler
Date #195
6/04/75 76.0
6/06/75 72.6
6/08/75 62.3
6/10/75 89.3
6/12/75 76.6
6/16/75 74.8
Overall 75.3
Paired-t
Corr. Coeff.
Sampler % Dif.1 Sampler
#197 Date #195
76.8 -1.1 6/23/75 79.3
69.5 +4.4 7/03/75 105.0
63.1 -1.3 7/05/75 75.4
87.1 +2.5 7/07/75 76.8
77.2 -0.8 7/11/75 106.1
74.5 +0.4 7/14/75 76.9
74.7 +0.8 86.6
+ 0.82 MS
0.98
Sampler % Dif.1
#197
82.4 - 3.8
107.4 - 2.3
71.2 + 5.7
75.3 + 2.0
98.3 + 7.6
66.4 +14.7
83.5 +3.6
+ 1.38 NS
0.94
„ , n;r _ 2<#195 - #197) v inn
#195 + #197
NS = Difference is not significant at P = 0.90
213
-------
group (by laboratory) to determine whether or not the sampler differences were
statistically significant once the variability due to date has been discounted.
Since neither computed paired-t value was significant, we may conclude that
no evidence appears in this data to suggest that sample collection introduces
a bias in the determination of TSP concentration at the LACS sites.
Tables II through V contain the results obtained for Pb, NHi+, NO$, and
SO~^, respectively. For these pollutants, duplicate (end and center) results
were reported by each laboratory for each of the duplicate samplers. Overall
averages by sampler, laboratory, and strip position appear at the bottom of
each table. Because the experimental design is balanced, these averages may
be considered indicators of the magnitude of the effects produced by the in-
dependent variables in the experiment (i.e., sampler, laboratory and strip).
ANALYSIS OF VARIANCE
The statistical technique known as analysis of variance was applied to
determine whether either samplers or laboratories were significant sources
of variation in the data for Pb, NH^, NO$ and SO^. The data for each pollutant
were treated as a 2x2 factorial experiment (samplers "crossed" with laborator-
ies) -conducted in twelve complete blocks (sampling days). For the purpose of
this analysis, center and end strips results were considered as simply being
within cell replicates.
The results of the four analyses of variance appear in Table VI. The
significance of the F-values (right hand column) is denoted by asterisks. As
would be expected, sampling date is a large source of variation and is sign-
ificant in every case. Of much more interest, however, is the fact that sampler
was not a significant source of variation for any of the four pollutants tested.
Hence, as was the case with TSP, we may conclude that two properly calibrated
hi-vol samplers operating side-by-side and concurrently will produce results
which are statistically indistinguishable._ Laboratory was a significant
factor for Pb, NHi^ and NO-^, but not for
VARIANCE COMPONENT
Using the mean square values obtained in the analysis of variance and
considering the sampler and laboratory effects to be random rather than fixed,
RAMC and EMSL are thought of as a random sample of an infinite poulation for
possible analysis laboratories and samplers. Number 195 and number 197 are a
random sample of an infinite population of possible high volume samplers. It
is easy to obtain estimates of the variance in reported pollutant concentrations
which is due to the sampler factor and the laboratory factor, respectively.
The estimates of these "variance components" are obtained as follows.
/\
a _ , = Mean square due to samplers - Mean square error
Sampler * -f- a
ff „. . = Mean square due to laboratories - Mean square error
•* — *
48
214
-------
TABLE II. DATA SUMMARY Pb (]ig/m3 )
Date
6/04/75
6/06/75
6/08/75
6/10/75
6/12/75
6/16/75
6/23/75
7/03/75
7/05/75
7/07/75
7/11/75
7/14/75
Average
——————— —^ — — — _^
Sampler #195
RAMC F.MKT.
End
0.70
0.30
0.50
0.70
0.70
0.70
0.40
3.40
2.70
2.10
2.70
0.60
1.29
Center
0.90
0.60
0.60
0.60
0.60
0.60
0.50
3.70
3.00
2.70
2.70
0.20
1.39
Center
0.94
0.55
0.57
0.73
0.70
0.68
0.69
3.80
3.38
2.47
3.08
0.62
1.52
End
0.95
0.52
0.55
0.76
0.56
0.67
0.66
3.74
3.40
2.34
2.86
0.58
li'47
- -
Sampler #197
RAMC EMSL
End
0.80
0.40
0.50
0.70
1.00
0.70
0.50
3.60
3.00
2.50
2.50
0.60
1.40
Center
0.80
0.40
0.60
0.50
0.70
0.60
0.60
2.90
2.70
2.40
2.30
0.40
1.24
Center
0.86
0.51
0.56
0.52
0.70
0.79
0.65
3.55
1.51
4.50
2.67
0.46
1.44
End
0.93
0.45
0.58
0.64 E
0.71
0.72
0.69
3.33
3.09
4.75
2.44
0.50
1.57
E = Estimated (Sample Lost)
OVERALL SAMPLER AVERAGES: #195 = 1.42, #197 =1.41
OVERALL LAB AVERAGES : RAMC = 1.33, EMSL =1.50
OVERALL STRIP AVERAGES : End = 1.43, Center =1.40
215
-------
TABLE III. DATA SUMMARY
(\ig/m3)
Sampler
RAMC
Date
6/04/75
6/06/75
6/08/75
6/10/75
6/12/75
6/16/75
6/23/75
7/03/75
7/05/75
7/07/75
7/11/75
7/14/75
Average
End
0.63
1.04
1.33
1.76
1.80
2.07
1.97
0.76
0.41
0.27
1.03
0.78
1.15
Center
1.63
2.19
2.32
2.64
2.59
3.10
1.97
0.78
0.42
0.29
1.06
0.80
1.65
#195
Center
0.30
0.70
0.90
1.30
1.40
1.80
1.60
0.50
0.50
0.00
0.50
0.50
0.84
EMSL
End
0.10
0.70
0.80
1.30
1.60
1.70
1.50
0.30
0.00
0.00
0.50
0.50
0.76
Sampler
RAMC
End
0.61
1.08
1.31
2.04
1.93
2.31
1.88
0.72
0.40
0.24
0.84
0.79
1.18
Center
1.45
2.22
2.32
2.63
2.71
3.13
1.91
0.89
0.42
0.29
0.95
0.78
1.64
#197
EMSL
Center
0.30
0.80
1.00
1.50
1.40
1.70
1.10
0.30
0.20
0.30
0.50
0 40
0.79
End
0.20
0.60
0.80
1.50
1.50
1.80
1.40
0.50
0.20
0.30
0.90
0.40
0.84
OVERALL SAMPLER AVERAGES: #195 = 1.10, #197 = 1.11
OVERALL LAB AVERAGES : RAMC = 1.41, EMSL =0.81
OVERALL STRIP AVERAGES : End = 0.98, Center =1.23
216
-------
TABLE IV. DATA SUMMARY NO$
Sampler #195
RAMC EMSL
Date
6/04/75
6/06/75
6/08/75
6/10/75
6/12/75
6/16/75
6/23/75
7/03/75
7/05/75
7/07/75
7/11/75
7/14/75
Average
End
6.97
4.14
1.95
1.85
1.13
0.41
1.44
15.93
9.63
10.02
9.55
6.14
5.76
Center
7.90
3.39
1.51
1.48
1.00
0.24
1.46
15.96
10.31
10.31
9.70
6.12
5.78
Center
8.40
3.80
1.90
1.80
1.30
0.50
1.70
16.10
10.60
11.30
10.60
6.20
6.18
End
8.30
3.50
1.60
1.70
1.30
0.50
1.50
14.40
10.10
10.70
9.90
5.40
5.74
Sampler #197
RAMC EMSL
End
7.52
3.62
1.77
1.68
1.42
0.43
1.50
14.83
10.20
10.22
8.93
5.76
5.66
Center
7.79
4.39
1.34
1.46
1.12
0.23
1.41
15.99
9.93
10.48
9.49
6.26
5.82
Center
7.10
3.90
1.70
1.70
1.40
0.50
1.60
14.90
10.10
17.80
9.60
6.00
6.36
End
6.60
3.70
1.60
1.90
1.50
0.50
1.60
14.60
9.70
19.00
9.40
5.60
6.31
OVERALL SAMPLER AVERAGES: #195 = 5.87, #197 = 6.04
OVERALL LAB AVERAGES : RAMC = 5.76, EMSL =6.15
OVERALL STRIP AVERAGES : End = 5.87, Center =6.04
217
-------
TABLE V. DATA SUMMARY SOk= (v.g/m3)
Sampler #195
RAMC EMSL
Date
6/04/75
6/06/75
6/08/75
6/10/75
6/12/75
6/16/75
6/23/75
7/03/75
7/05/75
7/07/75
7/11/75
7/14/75
Average
End
12.5
14.2
18.2
22.5
24.1
24.3
23.6
13.9
10.0
8.9
17.2
17.4
17.2
Center
14.5
16.2
20.4
24.8
25.4
25.7
24.3
14.0
10.3
9.1
16.9
18.1
18.3
Center
13.3
15.4
19.6
22.7
23.8
27.5
23.9
14.8
10.4
9.3
16.3
17.5
17.9
End
12.2
14.9
17.6
22.5
24.7
26.8
23.3
13.3
10.2
8.4
16.3
16.6
17.2
Sampler #197
RAMC EMSL
End
14.0
14.4
18.6
24.4
23.5
26.2
23.9
14.7
9.9
9.8
14.1
14.8
17.4
Center
14.5
16.1
20.4
24.4
24.9
26.4
24.6
14.7
10.0
9.8
14.7
15.4
18.0
Center
13.6
15.3
19.7
23.5
24.6
24.1
23.2
14.0
10.0
17.1
14.8
14.2
17.8
End
12.6
14.5
18.2
24.9
25.3
26.0
22.9
13.8
9.8
18.1
14.3
14.2
17.9
OVERALL SAMPLER AVERAGES: #195 = 17.7, #197 = 17.8
OVERALL LAB AVERAGES : RAMC = 17.7, EMSL =17.7
OVERALL STRIP AVERAGES : End = 17.4, Center 18.0
218
-------
TABLE VI. ANALYSIS OF VARIANCE
Pollutant Source
Date
Sampler
Pb Lab
Sampler x Lab
Error
Corrected Total
Date
Sampler
NH^+ Lab
Sampler x Lab
Error
Corrected Total
Date
Sampler
WO 3- Lab
Sampler x Lab
Error
Corrected Total
Date
Sampler
SOti= Lab
Sampler x Lab
Error
Corrected Total
D.F.
11
1
1
1
80
94
11
1
1
1
81
95
11
1
1
1
81
95
11
1
1
1
1
95
Sum of Squares
123.19591
0.00036
0.67683
0.00873
10.96030
134.84213
39.14085
0.00540
8.64000
0.00094
8.13131
55.81850
2217.33030
0.68851
3.67775
0.97405
101.01143
2323.68204
2653.80083
0.26042
0.00375
0-96000
176.94833
2831.97333
Mean Square
11.19963
0.00036
0.67683
0.00873
0.13700
3.55826
0.00540
8.64000
0.00094
0.10039
201.57548
0.68851
3.67775
0.97405
1.24705
241.25462
0.26042
0.00375
0.96000
2.18455
F -Value
81.74688**
0.00265
4.94024*
0.06371
35.44556**
0.05379
86.06729**
0.00934
161.64126**
0.55211
2.94915*
0.78108
110.43689**
0.11921
0.00172
0.43945
* Source is significant at p = 0.90
**Source is significant at p = 0.99
219
-------
Jt is also possible to obtain an estimate of the variance component for
filters strips by means of the following equation:
where n = number of pairs of center and end strips
c. = reported concentration center strip of pair i
e, = reported concentration end strip of pair i
It should be noted that this estimate is apt to be on the conservative (high)
side in this experiment since the position of the filter from which duplicate
strips were cut was controlled (center and end) and since the estimate contains
the variance due to the final step in the measurement process (analytical
instrument reading).
As previously mentioned, one laboratory (EMSL) reported duplicate instru-
ment readings for each of the extraction solutions prepared in the experiment
These duplicate data appear in Table VII. Only the first value in each case
was used in the data analyses which have been discussed. Using this duplicate
data, however, it is possible to obtain an estimate of the variance inherent
in the analytical instrument itself. This estimate is computed as follows:
a2 = - E - (Ril ~ Ri2)2
Within Instrument n . _7 2
where n = number of pairs of instrument readings
R. = 1st instrument reading for pair i
P
12 = 2nd instrument reading for pair i
The estimates of the variance components for these four factors (sampler,
laboratory, strips and instrument readings) appear in Table VIII. Since only
EMSL data were available for duplicate instrument readings and because RAMC
data for duplicate strips are confounded with analysis data (see next Section),
only EMSL data were used to estimate the variance components for strips and
instrument readings. The estimate for strips was "corrected" by subtracting
the corresponding estimate for readings. All of the data were used to estimate
the variance_components for samplers (zero for all pollutants) and laboratories
(zero for SO^ only).
.... , . . . , _., 100 x standard deviation...
The coefficient of variation (CV = —— ; is shown in
mean concentration
220
-------
TABLE VII. DUPLICATE INSTRUMENT READINGS (EMSL)
Site
195 Qc
195 Qe
197 Qe
197 Qc
195 Qe
195 Qc
197 Qe
197 Qc
195 Qe
195 Qc
197 Qe
197 Qc
195 Qe
195 Qc
197 Qe
197 Qc
195 Qe
195 Qc
197 Qe
197 Qc
195 Qe
195 Qc
197 Qe
197 Qc
195 Qc
195 Qe
195 Qc
195 Qe
195 Qc
195 Qe
195 Qc
195 Qe
195 Qc
195 Qe
195 Qc
195 Qe
197 Qe
197 Qc
197 Qe
197 Qc
197 Qe
197 Qc
197 Qe
197 Qc
197 Qe
197 Qc
197 Qe
197 Qc
6/04/75
6/04/75
6/04/75
6/04/75
6/06/75
6/06/75
6/06/75
6/06/75
6/08/75
6/08/75
6/08/75
6/08/75
6/10/75
6/10/75
6/10/75
6/10/75
6/12/75
6/12/75
6/12/75
6/12/75
6/16/75
6/16/75
6/16/75
6/16/75
6/23/75
6/23/75
7/03/75
7/03/75
7/05/75
7/05/75
7/07/75
7/07/75
7/11/75
7/11/75
7/14/75
7/14/75
6/23/75
6/23/75
7/03/75
7/03/75
7/05/75
7/05/75
7/07/75
7/07/75
7/11/75
7/11/75
7/14/75
7/14/75
(155)
(155)
(155)
(155)
(157)
(157)
(157)
(157)
(159)
(159)
(159)
(159)
(161)
(161)
(161)
(161)
(163)
(163)
(163)
(163)
(167)
(167)
(167)
(167)
(174)
(174)
(184)
(184)
(186)
(186)
(188)
(188)
(192)
(192)
(195)
(195)
(174)
(174)
(184)
(184)
(186)
(186)
(188)
(188)
(192)
(192)
(195)
(195)
7404257
7404257
7404258
7404258
7404273
7404273
7404272
7404272
7404290
7404290
7404291
7404291
7404164
7404164
7404163
7404163
7404193
7404193
7404192
7404192
7404309
7404309
7404308
7404308
7404466
7404466
7404850
7404850
7404868
7404868
7404425
7404425
7404889
7404889
7404549
7404549
7404465
7404465
7404849
7404849
7404869
7404869
7404426
7404426
7404888
7404888
7404548
7404548
yg/m3
NHh+
0.3-0.5
0.1-0.1
0.2-0.2
0.3-0.3
0.7-0.6
0.7-0.7
0.6-0.6
0.8-0.7
0.8-0.8
0.9-0.9
0.8-0.8
1.0-1.0
1 . 3-1 . 3
1.3-1.3
1.5-1.4
1.5-1.5
1.6-1.6
1.4-1.3
1.5-1.5
1.4-1.4
1.7-1.7
1.8-1.8
1.8-1.8
1.7-1.7
1.6-1.6
1.5-1.4
0.5-0.6
0.3-0.3
0.5-0.5
BMD-BMD
BMD-BMD
BMD-BMD
0.5-0.5
0.5-0.5
0.5-0.4
0.5-0.5
1.4-1.4
1.1-1.1
0.5-0.5
0.3-0.3
0.2-0.2
0.2-0.2
0.3-0.3
0.3-0.3
0.9-0.9
0.5-0.5
0.4-0.4
0.4-0.4
8
8
6
7
3
3
3
3
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
1
1
16
14
10
10
11
10
10
9
6
5
1
1
14
14
9
10
19
17
9
9
5
6
NOi~
.4-8.3
.3-8.0
.6-6.4
.1-7.1
.5-3.4
.8-3.6
.7-3.6
.9-3.9
.6-1.6
.9-1.8
.6-1.5
.7-1.7
.7-1.6
.8-1.7
.9-1.8
.7-1.7
. 3-1 . 3
.3-1.2
.5-1.5
.4-1.3
.5-0.5
.5-0.5
.5-0.5
.5-0.5
.7-1.6
.5-1.5
.1-15.9
.4-14.4
.6-10.6
.1-10.1
.3-11.0
.7-10.4
.6-10.2
.9-9.7
.2-5.9
.4-5.2
.6-1.6
.6-1.6
.6-14.6
.9-14.9
.7-9.8
.1-10.1
.0-19.0
.8-17.8
.4-9.4
.6-9.2
.6-5.6
.0-6.0
13
12
12
13
14
15
14
15
17
19
13
19
22
22
24
23
24
23
25
24
26
27
26
24
23
23
14
13
10
10
9
8
16
16
17
16
22
23
13
14
9
10
18
17
14
14
14
14
SOU~
.3-13.
.2-12.
.6-12.
.6-13.
.9-14.
.5-15.
.5-14.
.3-15.
.6-17.
.6-19.
.2-18.
.7-19.
.5-22.
. 7-21 .
.9-24.
.5-24.
.7-24.
.8-24.
.3-25.
.6-24.
.8-30.
.5-26.
.0-25.
.1-25.
. 9-24 .
.3-23.
.8-14.
.3-13.
.4-10.
.2- 9.
.3-9.
.4- 8.
.3-16.
.3-16.
.5-17.
.6-16.
.9-22.
.2-23.
.8-13.
.0-14.
.8-9.
.0-10.
.1-18.
.1-17.
.3-13.
.8-14.
.2-14.
.2-14.
5
3
7
7
9
1
5
5
7
6
2
7
5
7
6
5
7
1
3
6
2
7
1
1
1
5
9
6
4
8
3
7
4
4
5
5
9
0
8
0
8
0
1
1
9
8
2
8
Pb
0.94-0.92
0.95-0.95
0.93-0.92
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
86-0.84
52-0.55
55-0.55
45-0.46
51-0.53
55-0.56
57-0.57
58-0.58
56-0.56
76-0.76
73-0.73
lost
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
3.
3.
3.
3.
2.
2.
3.
2.
0.
0.
0.
0.
3.
3.
3.
1.
4.
4.
2.
2.
0.
0.
52-0.51
56-0.54
70-0.69
71-0.72
70-0.71
67-0.67
68-0.71
72-0.71
79-0.77
69-0.67
66-0.64
80-3.72
74-3.74
38-3.29
40-3.29
47-2.40
34-2.29
08-3.00
86-2.79
62-0.62
58-0.60
69-0.68
65-0.64
33-3.28
55-3.50
09-3.03
51-1.49
75-4.62
50-4.34
44-2.68
67-2.60
50-0.49
46-0.44
221
-------
the table for each variance component and for the total variance (sum of the
components) for each pollutant. This measure expresses the variance relative
to concentration so that comparisons may be made. It is interesting to note,
that, except in the case of NH^ (where the variance due to laboratories is
blown up by the confounding with analysis date), the largest source of varia-
tion in the total monitoring system for each pollutant is due to strips.
Some of the variance between strips is undoubtedly due to real differences in
concentration, but it seems probable that much of this strip-to-strip varia-
bility is introduced in the filter cutting and extraction processes. Perhaps
strip preparation is a component of the high volume sampler monitoring system
that should be given high priority in future method improvement and quality
assurance efforts.
At the bottom of Table VIII, estimates of the variance in the measurement
of TSP are given for each laboratory. These estimates were computed according
to the following equation:
-2 _ 2 n 2 2
0 TSP ~ n 2 (Sil ~ Si2)
where n = number of pairs of duplicate sampler results
S, = reported concentration for sampler #195
S. = reported concentration for sampler #197
Since the paired-t tests for samplers were not significant, this variance
is interpreted to be error inherent in the equilibration and weighing processes
rather than due to any real difference between duplicate samplers. As already
discussed, a direct evaluation of the effect of laboratories on TSP measure-
ment is not available, but a simple test may be applied to determine whether
the variance in TSP measurement is consistent from laboratory to laboratory.
This test consists of dividing the larger variance estimate by the smaller
and comparing the quotient with the tabulated F-value at the appropriate
degrees of freedom and confidence level. The test, shown at the bottom of
Table VIII is positive in this case implying that measurement variance for TSP
is indeed lab- dependent. It should be mentioned that the set of filters which
was pre-processed by EMSL was mailed to North Carolina from Los Angeles and
this may have introduced some variability. Since the lab factor doesn't enter
into the computed variance estimates, they may be too small to represent the
total monitoring system error. To partially compensate for this, the larger
of the two estimates (CV=5.2%) should be considered the more realistic figure.
The coefficient of variation in the total monitoring system for four of
the five pollutants are shown in bar graph format in Figure 3. The ammonium
ion results are not included since, as explained in the next section, they
were severly distorted by an unexpected effect due to the time lag between
sample collection and chemical analysis.
222
-------
TABLE VIII. ESTIMATES OF VARIANCE COMPONENTS (vg/m3)2
Variance Component
Between Instrument Readings
(within strips f filter, lab)
Between Strips
(within filter, lab)
Between Labs
(within filter)
Between Samplers
NO (duplicate hi-vols)
Overall
TSP
Pollutant
Pb NHu+ NO-,' SOU =
a CV* a CV* a CV* a CV*
.00116 2.4% .00125 3.2% .01042 1.7% .16823 2.3%
.05908 17.2% .01542 11.2% .13771 6.2% .29927 3.1%
.01128 7.5% .17793 38.2% .05071 3.8%
— — — —
.07152 18.9% .19460 39.9% .19884 7.5% .46750 3.9%
RISC EMSL
CT CV* a CV*
1.35 1.5% 17.20 5.2%
100 a
X
a (EMSL) T a CRAMC) = 12.74 > Ffr r. „, = 8.47
(0,0).01
-------
12.0
CV 8.0
4.0
20.0
16.0
CV 12.0
8.0
4.0
16.0
12.0
CV
8.0
4.0
16.0
12.0
CV
8.0
4.0
^•^•^••MMI^^^H
Figure 3.
TSP
Pb C
17.2%
7.5%
•
:
^— —
• ,
0% ==^—
SAMPLER LAB ST
N03
6.:
3.8% IT
i i u
°% i 1 1
SAMPLER LAB ST
SOJ
3.
o% o% mi
SAMPLER LAB ST
Twenty-four hour hi-vol. varia
2.4%
RIP ANALYSIS O
2%
1.7%
&£%
RIP ANALYSIS O>
l% •> w
2.3%
• •• l I
RIP ANALYSIS C
nee components (CV= -3, .
5.2%
* • * • *
IVERALL
18.9%
'**•••
• • • • •
VERALL
7.5%
[•••••^
• * •
• • •
• • •
• • •
• • •
*/ERALL
3.9%
• • • • •
• * • • •
* » • * •
* • * • •
^^••Mi
)VERA
100%;
m^^B^B
LL
*
224
-------
STRIP POSITION AND ANALYSIS DATE EFFECTS
Jn Tables IX though XIX, the data for Pb, NH^+, NO^ and SOk=, respectively,
are rearranged to facilitate an evaluation of the effect of strip position
(center vs. end) on reported pollutant concentrations. The data in the tables
are classified according to laboratory but not sampler since only the former
appears to exert an influence on results.
The data are further classified into the first six and the final six
sampling days of the study (designated sets I and II, respectively) . This was
done because "analysis date" (the date on which the chemical analysis of the
sample was performed), though not originally intended to be a variable in the
experiment, is apparently confounded (i.e., statistically inseparable) with
strip position in the case of the RAMC analyses of the set I samples. The
center strips from set I were all processed about two months earlier than the
end strips making it impossible to know whether any differences between center
and end strips were due to strip position effects, analysis date effects or
both. TMs problem did not arise with the RAMC analyses for set II nor with
any of the EMSL analyses. Hence, by observing the apparent strip position
effect for the RAMC set I data in the context of the strip position behavior
for the remaining data, it is possible to draw inferences about possible
analysis date effects as well as strip position effects.
Once again, the paired-t statistic was used to test whether the differ-
ences in concentration between center and end strips are statistically
significant. The test was performed for each of the four data subsets and
also for the pooled data in which analysis date was not a confounding variable
(RAMC set II and EMSL sets I and II).
The pattern of center vs. end differences for Pb appears to be random
with mean zero for all of the data and none of the computed paired-t values
are significant. One may conclude, therefore, that neither strip position
nor analysis date have a significant bearing on Pb results.
The results for NHi^+, however, are quite revealing. The RAMC data for
set I center strips are considerably higher than the corresponding end strips
(approximately 50 percent overall difference and a paired-t value which is
significant with more than 99% probability). The remaining data (where
possible analysis date effects are removed) exhibit a slight tendency for
center strips to exceed end strips in concentration (about 4 percent overall
difference with a paired-t value falling short of significance at the 95%
probability level). Thus, it appears that while the center of a filter may be
slightly higher in concentration than the ends, the time between sample
collection and analysis is a much more critical factor in determining NHi+
concentrations. Apparently, NHi++ is lost during filter storage. This finding
is consistent with the fact that the RAMC analysis for NH^+ has consistently
been much higher than the EMSL analysis for the same filter in the split-sample
program (RAMC performs its analysis several weeks before EMSL). It is inter-
esting to note, however, the even the set II data in which both RAMC and EMSL
performed their analyses on approximately the same date, the RAMC results are
still a good deal higher than the corresponding EMSL results. Perhaps some
difference exists between the filter storage environments in the two laboratories
225
-------
TABLE IX. STRIP POSITION AND ANALYSIS DATE COMPARISONS
Pb
RAMC
Analysis Date
Set I
Sample Date
6/4/75
6/6/75
6/8/75
NO
* 6/10/75
6/12/75
6/16/75
Sampler
195
197
195
197
195
197
195
197
195
197
195
197
Overall
8/21/75
End
0.70
0.80
0.30
0.40
0.50
0.50
0.70
0.70
0.70
1.00
0.70
0.70
0.64
Paired-t
Corr.
Coeff.
6/19/75
Center
0.90
0.80
0.60
0.40
0.60
0.60
0.60
0.50
0.60
0.70
0.60
0.60
0.63
-0.34
0.48
2 (c-e)
c+e
%Dif.
+ 25.0
0.0
+ 66.7
0.0
+ 18.2
+ 18.2
- 15.4
- 33.3
- 15.4
- 35.3
- 15.4
- 15.4
- 1.6
EMSL
Analysis Date
8/28/75
Center
0.94
0.86
0.55
0.51
0.57
0.56
0.73
0.52
0.70
0.70
0.68
0.79
0.69
8/28/75
End
0.95
0.93
0.52
0.45
0.55
0.58
0.76
Lost
0.56
0.71
0.67
0.72
0.67
+1.00
0.94
Jc~e' v- inn
c+e
%Dif.
- 1.1
- 7.8
+ 5.6
+ 12.5
+ 3.6
- 3.5
- 4.0
+ 22.2
- 1.4
+ 1.5
+ 9.3
+ 2.9
continued
-------
TABLE IX. CONTINUED
RAMC
EMSL
\>
X)
Analysis Date
8/27/75
Set II End
Sample Date Sampler
6/23/75 %*
7/3/75 ***
7/5/75 ^
7/7/75 ***
7/11/75 ^
7/14/77 ***
Overall
Paired-t
Corr. Coeff.
Pooled Data
0.40
0.50
3.40
3.60
2.70
3.00
2.10
2.50
2.70
2.50
0.60
0.60
2.05
(excluding
8/27/75
Center
0.50
0.60
3.70
2.90
3.00
2.70
2.70
2.40
2.70
2.30
0.20
0.40
2.01
-0.41
0.96
RAMC-Set I)
2(c~e) T 10Q
c+e
%Dif,
+ 22.2
+ 18.2
+ 8.5
-21.5
+ 10.5
- 10.5
+ 25.0
- 4.1
0.0
- 8.3
-100.0
- 40.0
- 2.0
Center Avg.
1.69
Analysis Date
8/28/75
Center
0.69
0.65
3.80
3.55
3.38
1.51
2.47
4.50
3.08
2.67
0.62
0.46
2.28
End Avg.
1.73
8/28/7
End
0.66
0.69
3.74
3.33
3.40
3.09
2.34
4.75
2.86
2.44
0.58
0.50
2.37
-0.59
0.94
%Dif.
-2.3
•5 2(°-e> X 100
_> A 0. U W
c+e
%Dif.
+ 4.4
- 6.0
+ 1.6
+ 6.4
- 0.6
- 68.7
+ 5.4
- 5.4
+ 7.4
+ 9.0
+ 6.7
- 8.3
- 3.9
Paired-t
-0.64
* Difference is significant at P = 0.95
**Difference is significant at P = 0.99
-------
TABLE X. CONTINUED
RAMC
EMSL
00
Analysis Date
Set II
Sample Date Sampler
6/23/75 ^
7/3/75 ^
7/5/75 %*
195
197
7/11/75 %*
7/14/75
Overall
Paired-t
Corr. Coef .
Pooled Data
8/25/75
End
1.97
1.88
0.76
0.72
0.41
0.40
0.27
0.24
1.03
0.84
0.78
0.79
0.84
(excluding
8/25/75
Center
1.97
1.91
0.78
0.89
0.42
0.42
0.29
0.29
1.06
0.95
0.80
0.78
0.88
+2.66*
0.99
RAMC-Set I)
2(c-e)
.. ^ J.UU
c+e
%Dif.
0.0
+ 1.6
+ 2.6
+ 21.1
+ 2.4
+ 4.9
+ 7.1
+ 18.9
+ 2.9
+ 12.3
+ 2.5
- 1.3
+ 4.7
Center Avg.
0.84
Analysis Date
8/22/75
Center
1.60
1.10
0.50
0.30
0.50
0.20
0.00
0.30
0.50
0.50
0.50
0.40
0.53
End Avg.
0.81
8/22/75
End
1.50
1.40
0.30
0.50
0.00
0.20
0.00
0.30
0.50
0.90
0.50
0.40
0.54
-0.12
0.88
%Dif.
+3.6
2(c-e) x 10Q
c+e
%Dif.
+ 6.5
- 24.0
+ 50.0
- 50.0
+200.0
0.0
0.0
0.0
0.0
- 57.1
0.0
0.0
- 1.9
Paired-t
+1.56
* Difference is significant at P = 0.95
**Difference is significant at P = 0.99
-------
TABLE X. STRIP POSITION AND ANALYSIS DATE COMPARISONS
RAMC
Analysis Date
Set I
*mi^mimiliiiii**illi*iii**iii*mmmm^^^^*ma^^^^^^*^^^
Sample Date
6/4/75
6/6/75
6/8/75
6/10/75
6/12/75
6/16/75
^^^•^^M^^BM^p^^^AMAvoq^^^HB
Sampler
195
197
195
197
195
197
195
197
195
197
195
197
Overall
8/15/75
End
V^^^MhlM^-^^«m*IMW^11111111^^M
0.63
0.61
1.04
1.08
1.33
1.31
1.76
2.04
1.80
1.93
2.07
2.31
1.49
Paired-t
Corr.
Coeff.
0.96
6/16/75
Center
^^^^^^^^^^^H^^^M^^^H^M«VW*^«i^^
1.63
1.45
2.19
2.22
2.32
2.32
2.64
2.63
2.59
2.71
3.10
3.13
2.41
19.40**
2 (c-e) , nn
c+e X
%Dif.
^^.^^^•••^••^•^^^••^^^^^•^^l^p.^M
+ 88.5
+ 81.6
+ 71.2
+ 69.1
+ 54.3
+ 55.7
+ 40.0
+ 25.3
+ 40.0
+ 33.6
+ 39.9
+ 30.2
+ 47.2
EMSL
Analysis Date
8/21/75
Center
^^^^^^^^•••MVMMB^^^^^M*^-^^^^—
0.30
0.30
0.70
O.SO
0.90
1.00
1.30
1.50
1.40
1.40
l.SO
1.70
1.09
8/21/75
End
•••^••••••^•••••^••^•W— H««-_^v^»^H
0.10
0.20
0.70
0.60
0.80
0.80
1.30
1.50
1.60
1.50
1.70
1.80
1.05
+1.10
0.98
2(c~e) x 100
c+e
%Dif.
HMimiiHiiiiiiiiii^^^^^Bmiiiiiiiiiiiiiii-miiiiii^^-miiiiiip^^^^MM^^^^^
+100.0
+ 40.0
0.0
+ 28.6
+ 11.8
+ 22.2
0.0
0.0
- 13.3
- 6.9
+ 5.7
- 5.7
+ 3.7
continued
-------
TABLE XI. STRIP POSITION AND ANALYSIS DATE COMPARISONS
#03 (vg/m3)
RAMC
Analysis Date
Set I
8/15/75
End
6/16/75
Center
2(c-e)
c+e
%Dif.
EMSL
Analysis Date
8/21/75
Center
8/21/75
End
2(c-e)
c+e
%Dif.
Sample Date Sampler
6/4/75
6/6/75
6/8/75
6/10/75
6/12/75
6/16/75
195
197
195
197
195
197
195
197
195
197
195
197
Overall
Paired-t
Corr. Coef.
6.97
7.52
4.14
3.62
1.95
1.77
1.S5
1.68
1.13
1.42
0.41
0.43
2.74
7.90
7.79
3.39
4.39
1.51
1.34
1.48
1.46
1.00
1.12
0.24
0.23
2.65
-0.60
0.99
+ 12.5
+ 3.5
- 19.9
+ 19.2
- 25.4
-27.7
- 22.2
- 14.0
- 12.2
- 23.6
- 52.3
- 60.6
- 3.3
8.40
7.10
3.80
3.90
1.90
1.70
1.80
1.70
1.30
1.40
0.50
0.50
2.83
8.30
6.60
3.50
3.70
1.60
1.60
1.70
1.90
1.30
1.50
0.50
0.50
2.73
+1.95
0.99
+ 1.2
+ 7.3
+ 8.2
+ 5.3
+ 17.1
+ 6.1
+ 5.7
- 11.1
0.0
- 6.9
0.0
0.0
+ 3.6
continued
-------
TABLE XI. CONTINUED
RAMC
EMSL
N>
U)
Analysis Date
Set II
Sample Date
6/23/75
7/3/75
7/5/75
7/7/75
7/11/75
7/14/75
Sampler
195
197
195
197
^97
195
197
\157
195
197
Overall
8/25/57
End
1.44
1.50
15.93
14.83
9.63
10.20
10.02
10.22
9.55
8.93
6.14
5.76
8.68
Paired-t
Corr. Coef.
Pooled Data
(excluding
8/25/75
Center
1.46
1.41
15.96
15.99
10.31
9.93
10.31
10.48
9.70
9.49
6.12
6.26
8.95
+2.38*
0.99
RAMC-Set I)
2(c-e) ,00
c+e
%Dif.
+ 1.4
- 6.2
+ 0.2
+ 7.5
+ 6.8
- 2.7
+ 2.9
+ 2.5
+ 1.6
+ 6.1
- 0.3
+ 8.3
+ 3.1
Center Avg.
7.16
Analysis Date
8/22/75
Center
1.70
1.60
16.10
14.90
10.60
10.10
11.30
17.80
10.60
9.60
6.20
6.00
9.71
End Avg.
6.91
8/22/75
End
1.50
1.60
14.40
14.60
10.10
9.70
10.70
19.00
9.90
9.40
5.40
5.60
9.33
+2.01
0.99
% Dif.
+3.6
2(c-e) ,00
c+e
%Dif.
+ 12.5
0.0
+ 11.2
+ 2.0
+ 4.8
+ 4.0
+ 5.5
- 6.5
+ 6.8
+ 2.1
+ 13.8
+ 6.9
+ 4.0
Paired-t
+3.32**
* Difference is significant at P = 0.95
**Difference is significant at P = 0.99
-------
TABLE XII. STRIP POSITION AND ANALYSIS DATE COMPARISONS
so!;
U)
NJ
RAMC
Analysis Date
Set I
Sample Date
6/4/75
6/6/75
6/8/75
6/10/75
6/12/75
6/16/75
Sampler
195
197
195
197
195
197
195
197
\157
197
Overall
8/15/75
End
12.5
14.0
14.2
14.4
18.2
18.6
22.5
24.4
24.1
23.5
24.3
26.2
19.7
Paired-t
Corr.
Coef.
6/16/75
Center
14.5
14.5
16.2
16.1
20.4
20.4
24.8
24.4
25.4
24.9
25.7
26.4
21.0
+ 6.24**
0.97
2 (c-e)
c+e
%Dif.
+ 14.8
+ 3.5
+ 13.2
+ 11.2
+ 11.4
+ 9.3
+ 9.7
0.0
+ 5.3
+ 5.8
+ 5.6
+ 0.8
+ 6.4
EMSL
Analysis Date
8/21/75
Center
13.3
13.6
15.4
15.3
19.6
19.7
22.7
23.5
23.8
24.6
27.5
24.1
20.3
8/21/75
End
12.2
12.6
14.9
14.5
17.6
18.2
22.5
24.9
24.7
25.3
26.8
26.0
20.0
+ 0.69
0.98
2 (c-e)
c+e
%Dif.
+ 8.6
+ 7.6
+ 3.3
+ 5.4
+ 10.8
+ 7.2
+ 0.9
- 5.8
- 3.7
- 2.8
+ 2.6
- 7.6
+ 1.5
continued
-------
TABLE XII. CONTINUED
RAMC
EMSL
NO
Co
Analysis Date
Set II
Sample Date
6/23/75
7/3/75
7/5/75
7/7/75
7/11/75
7/14/75
Sampler
195
197
195
197
195
197
197
197
197
Overall
8/25/75
End
23.6
23.9
13.9
14.7
10.0
9.9
8.9
9.8
17.2
14.1
17.4
14.8
14.9
Paired-t
Corr. Coef.
Pooled Data
(excluding
8/25/75
Center
24.3
24.6
14.0
14.7
10.3
10.0
9.1
9.8
16.9
14.7
18.1
15.4
15.2
+ 3.12**
0.99
RAMC-Set I)
2(c~e) x 100
c+e
+ 2.9
+ 2.9
+ 0.7
0.0
+ 3.0
+ 1.0
+ 2.2
0.0
- 1.8
+ 4.2
+ 3.9
+ 4.0
+ 2.0
Center Avg.
17.0
Analysis Date
8/22/75
Center
23.9
23.2
14.8
14.0
10.4
10.0
9.3
17.1
16.3
14.8
17.5
14.2
15.5
End Avg.
16.7
8/22/75
End
23.3
22.9
13.3
13.8
10.2
9.8
8.4
18.1
16.3
14.3
16.6
14.2
15.1
+ 2.02
0.99
% Dif.
+1.8
2(c-e) Y 10C
c+e
%Dif.
+ 2.5
+ 1.3
+ 10.7
+ 1.4
+ 1.9
+ 2.0
+ 10.2
- 5.7
0.0
+ 3.4
+ 5.3
0.0
+ 2.6
Paired-t
+2.31*
* Difference is significant at P = 0.95
**Difference is significant at P - 0.99
-------
which affects NH^+ loss or perhaps there are some differences in the way the
analysis method is applied in the two laboratories which accounts for this
difference. At any rate, it is suggested that further experimentation be
conducted to substantiate NH^+ loss on storage and inter laboratory differences
in the sodium phenolate method for NH^+ analysis.
For the N0$ data in which analysis date is not a factor (RAMC set II,
EMSL set I and II) , center strips showed a tendency to exceed end strips
(about 4 percent overall difference with a paired- t value which is significant
at the 99% confidence level) . Although the RAMC set I data appears to run
counter to this pattern (later analyzed strips exceeding earlier analyzed
strips) , the paired-t value was not significant so that no conclusion about
analysis date effect may be drawn.
The pooled S0^= data (excluding RAMC set I) showed a strip position
effect similar to that for NH^+ and NOJ (about 2 percent overall difference,
significant at the 95% probability level) , The RAMC set I data for S0^~
suggests that lag time between sampling and analysis has increased the center
vs. end differences (about 6 percent overall difference with more than 99%
probability of significance) . It is possible that some S0^= loss occurs
during filter storage, but it is apparently a much less serious problem than
with NHi++.
FOUR-HOUR HIGH VOLUME AND MEMBRANE SAMPLER STUDY
PURPOSE
High volume and membrane samplers are operated for a 4-hour integrating
period (3-7 p.m.) at the LACS sites because this period is characterized by
maximum diurnal traffic density and wind directions (Seabreeze) which are
favorable to the detection of the freeway contribution to ambient pollutant
concentrations.
In the summer of 1976, a study was undertaken to provide estimates of the
variances associated with the components of 4-hour particulate sampling and
subsequent analysis for SO^= using both high volume and membrane samplers.
By combining the precision tests for the two sampling methods into a single
concurrent experiment, a secondary objective of obtaining data useful for
establishing inter-method comparability was accomplished at little additional
expense. Also, conducting the study simultaneously at site A (100' upwind)
and site C (25' downwind) provided a test of the cross-freeway difference in
pollutant concentration and method precision.
The objectives of the study, then, were to provide information concerning
the following:
(1) 4-Hour Hi-Vol Precision. To estimate the between sampler, between
laboratory and within laboratory components of variance for partic-
ulate sampling and subsequent SOi+= analysis using hi-vol samplers
from 3-7 p.m. at the LACS sites.
234
-------
(2) 4-Hour Membrane Precision. To estimate the between sampler, between
laboratory and within laboratory components of variance for partic-
ulate sampling and subsequent SO^= analysis using membrane samplers
from 3-7 p.m. at the LACS sites.
(3) Method Comparability. To determine the extent of agreement in
concentration level and measurement precision between samples col-
lected by hi-vol and membrane samplers operated on a 4-hour basis
at the LACS sites and analyzed for TSP and
(4) Site Comparability. To determine what differences exist in concen-
tration level and measurement precision between LACS sites A and C
for TSP and S0i±= for the 3-7 p.m. sampling interval.
EXPERIMENTAL DESIGN
The placement of samplers for the study is shown in Figure 4. A pair
of hi-vol samplers using glass fiber filters, a pair of membrane filters
using cellulose ester filters, and a single membrane sampler using teflon
filters were placed at each of the two sites (A and C) . Though not routinely
used in the LACS, the teflon filters were included in the study as a control
since both glass fiber and cellulose ester filters are suspected of being
subject to an artifact SO^= formation (via SO^ conversion on the filter media) .
The ten samplers were operated from 3-7 p.m. for twelve days. As indicated in
Figure 4, the prevailing wind direction was perpendicular to the freeway (from
site A toward site C) from 3-7 p.m. during each day of the study.
The cutting and assignment of filter partitions is illustrated in Figure
5. Once again, one center and one end strip from each glass fiber filter
were assigned to each of the participating laboratories (RAMC and EMSL) .
Since chemical analysis did not appear to be a large source of variation in the
24-hour hi-vol study, the laboratories reported a single analysis per strip
in this study- Membrane filters (102 mm in diameter) are much smaller in
area than glass fiber filters (8 in. x 10 in.) and cannot be partitioned
further than quarters for chemical extraction and analysis. Thus, SOl+= was
the only pollutant determined other than TSP. Diagonal quarters from each_
membrane filter were assigned to each of the laboratories for a single SOi+=
determination .
DATA SUMMARY
The study results are presented by method, site and pollutant in Tables
XIII through XXIV. Again, paired- 1 tests were computed to determine whether any
statistically significant biases in TSP concentration exist between a pair of
samplers of the same type and operated at the same location. No such biases
were found. Overall averages by sampler, laboratory and strip position appear
at the bottom of each table of SOi+= results.
ANALYSIS OF VARIANCE
For purposes of analysis, the S0^= data was first considered to consist
of four subsets, classified by sampling method and site, as shown in Table 25.
235
-------
MET
TOWER f
Ui
PREVAILING WIND-
30 FT
/
SAMPLER INLETS
3-4 FT. ABOVE
FREEWAY SURFACE
**
FREEWAY SURFACE
•/>•
.*•*"
'• :. -'•: v.V--.^v--V/.':-.V.:iv''- •'-' >:
.-•••: :.••;.:•/:;•.:•: •;.'.::.: j^25 n^
4_ 100 FT
"A" SITE
HI VOL SAMPLERS -- 2
(GLASS FIBER FILTERS)
MEMBRANE SAMPLERS - 2
(CELLULOSE ESTER FILTERS)
MEMBRANE SAMPLER - 1
(TEFLON FILTERS)
SEPULVEOA BLVD.
110 FT-
"B" SITE
"C" SITE
HI VOL SAMPLERS - 2
(GLASS FIBER FILTERS)
MEMBRANE SAMPLERS - 2
(CELLULOSE ESTER FILTERS)
MEMBRANE SAMPLER - 1
(TEFLON FILTERS)
"D" SITE
Figure 4. Sample collection.
-------
U)
HIGH VOLUME SAMPLERS
(GLASS FIBER FILTERS)
A: RAMC
-B: RAMC
-c: EMSL
-D: EMSL
MEMBRANE SAMPLERS
(CELLULOSE ESTER
FILTERS)
MEMBRANE SAMPLER
(TEFLON FILTERS)
A E
C
•
D ^
R
r
n
" A: RAMC
B: RAMC
-c: EMSL
LD: EMSL
B
£
-A: RAMC
B: RAMC
-c: EMSL
A: RAMC
-B: RAMC
-c: EMSL
LD: EMSLLD: EMSL
-A: RAMC
B: RAMC
- c: EMSL
-D: EMSL
Figure 5. Sample analysis.
-------
TABLE XIII. HI-VOL AT SITE A FOR TSP (\ig/m3)
Date
8/26/76
8/27/76
8/28/76
8/29/76
8/30/76
8/31/76
9/ 1/76
9/ 4/76
9/ 5/76
9/ 6/76
9/ 7/76
9/ 8/76
Average
Paired-t
Corr. Coeff.
Sampler 1
80.0
125.9
74.6
74.6
105.3
114.7
100.1
66.2
46.3
29.2
85.7
117.0
85.0
-1.14
0.95
Sampler 2
90.6
104.4
85.4
82.7
111.5
119.1
99.7
76.5
46.4
31.2
84.9
121.9
87.9
238
-------
TABLE XIV. HI-VOL AT SITE C FOR TSP (\ig/m3)
Date
8/26/76
8/27/76
8/28/76
8/29/76
8/30/76
8/31/76
9/ 1/76
9/ 4/76
9/ 5/76
9/ 6/76
9/ 7/76
9/ 8/76
Average
Paired-t
Coir, Coeff.
Sampler 1
127.0
160.0
153.8
116.7
140.1
141.6
128.9
124.4
92.9
93.6
135.8
165.5
131.7
+0.53
0.93
Sampler 2
126.9
156.7
130.3
121.7
141.5
144.3
131.9
116.3
104.2
90.9
138.8
161.0
130.4
239
-------
TABLE XV. MEMBRANE (CELLULOSE ESTER) AT SITE A FOR TSP (\ig/m3)
Date
8/26/76
8/27/76
8/28/76
8/29/76
8/30/76
8/31/76
9/ 1/76
9/ 4/76
9/ 5/76
9/ 6/76
9/ 7/76
9/ 8/76
Average
Paired-t
Corr. Coeff.
Sampler 1
75.8
137.5
85.6
104.8
133.9
100.2
75.6
95.9
111.0
85.5
131.2
208.1
112.1
+1.58
0.93
Sampler 2
53.9
129.5
70.8
93.6
102.0
101.1
91.3
73.9
101.0
64.9
156.5
214.2
104.4
240
-------
TABLE XVI. MEMBRANE (CELIJJLQSE ESTER) AT SITE C FOR TSP (\ig/m3)
Date
8/26/76
8/27/76
8/28/76
8/29/76
8/30/76
8/31/76
a/ 1/76
9/ 4/76
9/ 5/76
9/ 6/76
9/ 7/76
9/ S/76
Average
Paired -t
Con. Coeff.
Sampler 1
179.8
194.3
138.1
134.3
174.5
159.6
159.1
103.1
114.3
179.1
216.6
270.5
168.6
-0.54
0.71
Sampler 2
111.7
179.2
229.8
134.6
155.5
150.2
141.7
128.5
140.2
185.4
255.0
284.3
174.7
241
-------
TABLE XVII. MEMBRANE (.TEFLON) AT SITE A FOR TSP
Date
8/26/76
8/27/76
8/28/76
8/29/76
8/30/76
8/31/76
9/ 1/76
9/ 4/76
9/ 5/76
9/ 6/76
9/ 7/76
9/ fi/76
Average
Result
19.8
43.7
43.8
47.2
354.8*
81.9
76.0
39.1
33.8
20.3
60.9
86.3
50.3
* Omitted from average
242
-------
TABLE XVIII. MEMBRANE (TEFLON) AT SITE C FOR TSP (\ig/m3)
Date
8/26/76
8/27/76
8/28/76
8/29/76
8/30/76
8/31/76
9/ 1/76
9/ 4/76
9/ 5/76
9/ 6/76
9/ 7/76
9/ 8/76
Average
Result
84.2
143.7
74.0
57.0
98.9
102.2
101.6
46.6
66.4
88.3
66.2
129.5
88.2
243
-------
TABLE XIX. HI-VOL AT SITE A FOR SO^. (vg/m3)
Sampler 1
RAMC
Date
8/26/76
8/27/76
8/28/76
8/29/76
8/30/76
8/31/76
9/ 1/76
9/ 4/76
9/ 5/76
9/ 6/76
9/ 7/76
9/ 8/76
Average
Str. a
16.1
39.3
23.4
14.1
13.5
15.4
23.2
11.5
10.3
6.8
8.0
11.6
16.1
Str. b
16.5
42.7
23.0
14.2
14.0
15.5
23.9
11.2
10.7
7.5
8.3
12.3
16.7
EMSL
Str. c
15.7
41.9
24.1
14.7
14.1
14.6
24.5
11.2
10.2
7.5
8.1
12.0
16.6
Str. d
16.1
41.9
22.8
14.4
13.2
16.1
22.7
10.6
10.4
7.7
8.3
12.5
16.4
Sampler 2
RAMC
Str. a
16.2
40.9
23.9
14.8
14.0
15.1
25.3
12.3
9.5
7.8
8.5
12.7
16.8
Str. b
17.0
43.0
22.8
14.9
13.6
15.9
25.0
11.8
10.7
7.9
8.1
12.3
16.9
EMSL
Str. c
16.9
45.8
24.2
15.6
14.2
16.1
25.2
12.4
10.6
7.9
8.7
12.5
17.5
Str.d
16.8
44.5
24.8
15.5
13.7
17.0
25.4
12.1
10.9
7.6
8.0
12.8
17.4
OVERALL SAMPLER AVERAGES: #1 = 16.4; #2 = 17.2
OVERALL LAB AVERAGES : RAMC = 16.6; EMSL =17.0
OVERALL STRIP AVERAGES : End = 16.7; Center =16.9
244
-------
TABLE XX. HI-VOL AT SITE C FOR SO4
Sampler 1
RAMC
Date
8/26/76
8/27/76
8/28/76
8/29/76
8/30/76
8/31/76
9/ 1/76
9/ 4/76
9/ 5/76
9/ 6/76
9/ 7/76
9/ 8/76
Average
Str. a
16.5
37.4
19.8
13.9
14.3
14.5
22.1
12.4
13.5
11.8
14.0
14.1
17.0
Str. b
16.0
38.3
20.0
13.9
14.0
14.4
22.9
11.8
13.4
11.3
13.3
14.4
17.0
EMSL
Str. c
16.1
40.4
21.5
13.9
13.0
14.0
22.9
13.4
13.4
11.7
13.5
13.6
17.3
Str. d
16.7
39.1
21.6
13.9
13.5
13.6
21.0
12.4
16.3
11.8
13.5
13.9
17.3
Sampler 2
RAMC
Str. a
15.2
36.8
19.7
13.5
14.4
14.7
21.7
11.7
12.6
10.6
12.3
13.8
16.4
Str. b
15.0
36.6
19.1
13.3
13.7
14.4
21.4
11.7
12.4
11.0
14.3
13.9
16.4
EMGT.
Str. c
15-7
40.5
15.9
15. 8
15.9
15.2
22.4
11.4
12.6
10.7
12.8
12.3
16.8
Str. d
15.4
38.1
22.0
15.8
16.2
14.6
24.1
11.0
12.0
10.7
12.6
12.6
17.1
OVERALL SAMPLER AVERAGES: #1 = 17.2; #2 = 16.7
OVERALL LAB AVERAGES : RAMC = 16.7; EMSL =17.2
OVERALL STRIP AVERAGES : End = 17.0; Center =16.9
245
-------
TABLE XXI. MEMBRANE (CELLULOSE ESTER) AT SITE A FOR SOk
Sampler 1
RAMC
Date
8/26/76
8/27/76
8/28/76
8/29/76
8/30/76
8/31/76
9/ 1/76
9/ 4/76
9/ 5/76
9/ 6/76
9/ 7/76
9/ 8/76
Qtr. a
13.0
38.5
20.0
11.9
12.5
11.6
18.8
10.1
7.6
6.7
6.8
10.5
Qtr. b
12.9
38.9
20.1
12.1
12.3
11.5
19.3
9.8
7.5
7.3
6.7
11.2
EMSL
Qtr. c
11.8
34.8
16.2
10.1
10.7
8.8
16.5
7.7
6.7
5.2
5.7
8.8
Qtr. d
11.8
38.2
17.1
10.7
10.7
9-9
16.5
8.0
7.0
5.5
4.6
8.8
Sampler 2
RAMC
Qtr. a
13.1
35.5
20.3
13.6
12.7
11.6
20.3
9.3
8.8
8.0
7.0
10.4
Qtr. b
13.5
45.9
19.7
12.9
13.2
11.3
20.0
9.0
8.3
7.2
7.4
11.6
EMSL
Qtr. c
10.7
39.5
22.2
11.3
9.7
9.0
16.7
8.0
6.3
5.1
4.8
8.8
Qtr. d
11.4
38.1
17.1
14.8
10.5
10.8
18.2
7.0
6.3
5.1
5.5
8.8
OVERALL SAMPLER AVERAGES: #1 = 13.1; #2 = 13.7
OVERALL LAB AVERAGES : RAMC = 14.3; EMSL =12.5
246
-------
TABLE XXII. MEMBRANE (CELLULOSE ESTER) AT SITE C FOR SO^
Sampler 1
RAMC
Date
8/26/76
8/27/76
8/28/76
8/29/76
8/30/76
8/31/76
9/ 1/76
9/ 4/76
9/ 5/76
9/ 6/76
9/ 7/76
9/ 8/76
Qtr. a
14.7
40.1
21.5
14.2
13.6
11.7
21.4
9.4
8.2
7.5
7.9
11.5
Qtr. b
14.0
40.1
20.8
12.7
12.0
12.7
20.3
10.2
8.1
6.7
7.9
12.9
EMSL
Qtr. c
9.4
33.4
15.9
10.4
9.4
8.6
11.5
5.1
5.3
4.6
3.8
8.4
Qtr. d
10.1
34.1
14.8
8.5
9.8
8.6
16.6
5.5
6.3
4.2
3.8
7.4
Sampler 2
RAMC
Qtr. a
14.4
42.8
22.2
12.3
12.3
13.1
21.3
9.4
8.5
7.0
6.9
11.1
Qtr. b
13.0
37.8
20.0
13.3
13.0
13.1
20.4
10.1
8.7
6.8
6.9
10.1
EMSL
Qtr. c
10.8
34.5
16.5
10.1
9.1
9.9
17.7
7.0
7.1
6.1
4.9
9.2
Qtr. d
12 . 343
35.4
17.2
12.3
10.9
10.3
18.5
7.7
6.4
4.8
4.9
7.9
a: estimated
OVERALL SAMPLER AVERAGES: #1 = 12.fi/ #2 = 13.5
OVERALL LAB AVERAGES : RAMC = 14.9; EMSL =11.4
247
-------
TABLE XXIII. MEMBRANE CTEFLON) AT SITE.A FOR S0t+ (vg/m3)
RAMC
Date
8/26/76
8/27/76
8/28/76
8/29/76
8/30/76
8/31/76
9/ 1/76
9/ 4/76
9/ 5/76
9/ 6/76
9/ 7/76
9/ 8/76
Qtr. a
10.3
34.0
15.8
9.7
9.9
9.1
16.7
6.6
6.5
0.4
3.8
9.4
Qtr. b
10.1
33.2
18.1
10.0
9.1
8.3
18.0
7.6
2.9
2.1
4.4
8.4
EMSL
Qtr. c
10.0
20.2
12.8
7.2
8.1
6.5
12.3
5.4
2.1
3.3
3.3
6.2
Qtr. d
5.4
23.8
12.8
6.0
8.3
7.8
14.4
5.8
4.6
2.3
3.3
7.1
OVERALL LAB AVERAGES: RAMC = 11.0; EMSL =8.3
248
-------
TABLE XXIV. MEMBRANE (TEFLON) AT SITE C FOR SO4 (\ig/m3)
RAMC
Date
8/26/76
8/27/76
8/28/76
8/29/76
8/30/76
8/31/76
9/ 1/76
9/ 4/76
9/ 5/76
9/ 6/76
9/ 7/76
9/ 8/76
Qtr. a
10.9
34.0
14.4
7.7
14.4
8.6
17.2
5.9
3.8
4.7
5.2
9.6
Qtr. b
10.5
31.2
17.1
7.0
14.8
9.4
17.2
5.3
4.0
4.5
4.8
10.5
EMSL
Qtr. c
6.7
22.1
13.9
5.6
16.1
8.3
14.6
5.0
2.7
2.6
1.2
5.8
Qtr. d
4.9
5.9
10.3
6.3
11.9
7.1
13.7
1.9
2.1
1.7
1.9
6.5
OVERALL LAB AVERAGES: RAMC = 11.4; EMSL =7.5
249
-------
TABLE XXV. SOi> RANDOM MODELS
Data Sets: Site
A C
Method HV T IJ
M(c) III IV
(DSL)ijk
where D: Day Effect ^ N (o,a2) i = 1,2,....,12
S: Sampler Effect ^ N (o,a2) j = 1,2
L: Lab Effect -v N (o,a2) k = 1,2
L
W (DSL): Experimental Error
-------
Since membrane teflon filters (M(t)) are not routinely employed in the LACS
and were not duplicated at each site, only the hi-vol (HV) and membrane cellu-
lose ester (M(c)) results are considered for the estimation of variance com-
ponents. The random effects model used to fit each data subset is shown in
Table XXV, along with the resultant table of expected mean squares required
for the estimation of variance components.
The results of analysis of variance are shown in Table XXVI for each of
the four data subsets. The two- and three-factor interactions are tested in
each case using the within mean square error (W(DSL)). Although the sampler
by laboratory interaction is significant at the 95% level for the hi-vol data
at site A, no interactions were significant at site C and the "pooled"
(including interactions) mean square error at site C exceeds that at site A.
Hence, it is probably reasonable to ignore interactions in further analysis.
Similarly, the membrane data at site C indicate some interactions, but since
none were evidenced at site A which has the larger pooled mean square error,
no further account of interactions was taken.
Using the pooled mean square error, tests were made for the significance
of the factors, sampler and laboratory in each data subset as shown by the
bracketted F-statistics in Table XXVI. in each case, sampler and laboratory
appear to exert a real influence on reported SO^= concentrations.
VARIANCE COMPONENTS
Estimates of the variance components associated with SOf measurements
for each of the four method-site classifications appear in Table XXVII. By
summing the estimates for the between sampler, between laboratory and within
laboratory components in each case, an estimate of the overall variance
associated with a reported SO^= concentration is obtained as shown in the
right hand column of the table.
Comparing estimates of variance components of the same type between sites
reveals differences as large as a factor of three. It should be noted, how-
ever, that the factors, sampler and laboratory are estimated with a single
degree of freedom since only two of each were used to generate each data sub-
set. Hence, if the variance ratios were tested by the F-statistic, the var-
iance components for sampler and laboratory between sites (within method)
would not be statistically different. These results, therefore, are pooled
for each method and the result is shown as a coefficient of variation for
comparison purposes. The within laboratory variance components, however, are
estimated with 82 degrees of freedom each and the estimate for the hi-vol data
at site C is statistically different than the corresponding estimate for site
A with the variance at site C exceeding that at site A by a factor of two.
Estimates of the overall variance in TSP concentration for each data
subset.are shown in Table XXVIII. Again, the separate estimates by site were
pooled and expressed as a coefficient of variation although the membrane
results at site C showed considerably more variability than those at site A.
In Figure 6, the 4-hour variance components appear in bar graph format,
expressed as coefficients of variation for comparison purposes.
251
-------
TABLE XXVI. SO^ ANOVA (RANDOM MODELS)
Source
Day
Sampler
Lab
DS
DL
SL
DSL
W(DSL)
(Pooled Error)
Source
Day
Sampler
Lab
DS
DL
SL
DSL
W(DSL)
(Pooled Error)
I. HV-Site A
df
11
1
1
11
11
1
11
48
95
(82)
II. HV-Site C
df
11
1
1
11
11
1
11
48
95
(.82)
(overall mean
SS
8197.40
12.69
3.19
8.33
7.58
1.73
3.01
18.07
8251 . 99
(38.71)
(overall mean
SS
4924.15
5.32
3.84
20.23
14.23
0.35
14.94
35.52
5018.58
(85.27)
= 16.79 vg/m3)
MS F
745.22
12.69 (27.00)**
3.19 ( 6.78)*
0.76 2.77
0.69 2.52
1.73 6.35*
0.27 0.73
0.38
86.86
(0.47)
= 16.90 ygr/m3;
MS F
447.65
5.32 ( 5.12)*
3.84 ( 3.69)*
1.84 1.35
1.29 0.95
0.35 0.26
1.36 1.84
0.74
52.83
(1.04)
NOTE: * indicates F is significant at P = 0.95
**indicates F is significant at P = 0.99
continued
252
-------
TABLE XXVI. CONTINUED
Source
III. M(c) - Site A
df
(overall mean
SS
13.39 ]ig/m3)
MS
Day
Sampler
Lab
DS
DL
SL
DSL
W(DSL)
(Pooled Error)
IV.
Source
Day
Sampler
Lab
DS
DL
SL
DSL
W(DSL)
(Pooled Error)
11
1
1
11
11
1
11
48
95
(82)
M(c) - Site C
df
11
1
1
11
11
1
11
48
95
(82)
7092.97
7.54
85.69
16.88
3.99
0.01
8.93
88.71
7304.72
(118.52)
(overall mean
SS
6612.94
9.63
292.60
8.58
29.30
18.03
5.94
49.00
7026.01
(110.84)
644.82
7.54
85.69
1.53
0.36
0.01
0.81
1.85
76.89
(1.45)
= 13.14 vg/m3)
MS
601.18
9.63
292.60
0.78
2.66
18.03
0.54
1.02
73.96
(1.35)
( 5.22)*
( 59.28)**
1.89
0.45
0.01
0.44
F
( 7.12)**
(216.46)**
1.44
4.93 **
33.38 **
0.53
NOTE: *indicates F is significant at P = 0.95
**indicates F is significant at P = 0.99
253
-------
TABLE XXVII. SOi) VARIANCE COMPONENTS ESTIMATES (cv= = •100%)
I. HV-Site A
II. HV-Site C
AVERAGE
III. M(c)-Site A
£n iv. M(c)-Site C
AVERAGE
% MSS - MSEp
S 48
0.25
0.09
0.17 (2.4 %)
0.13
0.17
0.15 (2.9 %)
"2 MSL - MSEp
ai,
0.
0.
0.
1.
6.
3.
48
06
06
06 (1.5 %)
76
07
91 (14.9 %)
O2
W
0.
1.
0.
1.
1.
1.
= MSEp
47
04
76 (5.2 %)
45
35
40 (8.9 %)
4'
0
1
0
3
7
5
0 0 2
= CF T o T 0
S L W
.78
.19
.99 ( 5.9 %)
.33
.59
.46 (17.6 %)
-------
Table XXVIII. TSP VARIANCE COMPONENT ESTIMATES (cv= — .100%)
12
L _ c )2
S}
12 1=1 2 li 21
I. HV - Site A 39.57
II. HV - Site C 34.84
AVERAGE 37.21 (5.6
III. M(c) - Site A 159.94
IV. M(c) - Site C 710.31
AVERAGE 435.13 (14.9
255
-------
cv
cv
12.0
8.0
4.0
16.0
12.0
8.0
4.0
HI-VOL-TSP
MEMBRANE (c) - TSP
5.6%
OVERALL
14.9%
CV
16.0
12.0
8.0
4.0
HI-VOL-SOf
5.2%
2.4%
1.5%
OVERALL
SAMPLER LAB
WITHIN
CV
20.0
16.0
12.0
8.0
4.0
MEMBRANE (c) - SOf
14.9%
2.£
8.9%
OVERALL
17.6
> • * * •
i • • • *
* • * • *
'SAMPLER LAB
WITHIN
OVERALL
Figure 6. Four-hour variance components (cv- £ 100%)
x ' °
256
-------
SAMPLING SITE AND METHOD EFFECTS
The TSP data by sampling site, method and date (averaged over duplicate
samplers) appear in Table XXIX. The fixed effects model shown in Table XXX
was used to fit the data for the purpose of testing site and method effects.
As shown in the analysis of variance, both site and method effects are appar-
ently real for TSP results. Partitioning the two degrees of freedom for method
reveals that while the membrane samplers produce significantly different
results depending on filter material used (cellulose ester or teflon) , the
average membrane results did not differ significantly from the hi-vol results.
Table XXXI shows the S0^= data by sampling site, method and date (aver-
aged over duplicate filter partitions, laboratory and sampler). A similar
fixed effects model was used to fit these data and the results appear in
Table XXXII. In this case, the site effect was not significant, implying
that there is no evidence in the data from this twelve day study to suspect
a significant contribution of SO^= from the freeway. The differences among
methods were significant, however, whether viewed as differences between
sampler types or between filter materials within membrane samplers.
The average TSP data are plotted by site and day of study in Figure 7.
While the three methods seem to generally fluctuate together, a great deal
of scatter remains. In general, the membrane teflon filters show the lowest
TSP loadings at either site while the membrane cellulose ester filters show
the highest. The cross- freeway differences are plotted at the top of the
figure. Although there is still a great deal of between method variability,
all differences are positive with an average freeway contribution of about
50 yg/m3 for the study.
Figure 8 is a similar set of plots for the average SOi+= data. The
between method correlation for SOi+= results is obviously very high (actually,
r>.99 for any pair of methods at either site). Also, it is interesting to
note that the order of methods is very consistent with the glass fiber filters
showing the largest SO^= concentration, followed by the cellulose ester filters
and, finally, the teflon filters. The cross-freeway differences for SO^= are
generally small (note that the scale is expanded) and, as previously discussed,
are not significantly different from zero for the twelve day study period.
CONCLUSIONS
The major conclusions which may be drawn from the 24-hour high volume
sampler study are as follows:
(1) Sample collection is not a significant source of error.
(2) Laboratory is a significant source of error for all pollutants
except
(3) Variability among filter strips is the largest source of error for
all pollutants.
(4) Variability among chemical analyses is relatively minor for all
pollutants.
257
-------
TABLE XXIX. AVERAGE TSP
Date
8/26/76
8/27/76
8/28/76
8/29/76
8/30/76
8/31/76
9/ 1/76
9/ 4/76
9/ 5/76
9/ 6/76
9/ 7/76
9/ 8/76
Overall
Average
HV
85.3
115.2
80.0
78.7
108.4
116.9
99.9
71.4
46.4
30.2
85.3
119.5
86.4
SITE A
M(c)
64.9
133.5
78.2
99.2
118.0
100.7
83.5
84.9
106.0
75.2
143.9
211.2
108.3
M(t)
19.8
43.7
43.8
47.2
354.8
81.9-
76.0
39.1
33.8
20.3
60.9
86.3
75.6
HV
127.0
158.4
142.1
119.2
140.8
143.0
130.4
120.4
98.6
92.3
137.3
163.3
131.1
SITE C
M(c)
145.8
186.8
184.0
134.5
165.0
154.9
150.4
115.8
127.3
182.3
235.8
277.4
171.7
M(t)
84.2
143.7
74.0
57.0
98.9
102.2
101.6
46.6
66.4
88.3
66.2
129.5
88.2
258
-------
TABLE XXX. TSP FIXED MODEL
Data Set: Average TSP c
Model: Y « y + D +
where D: Day Effect
S: Site Effect
M: Method Effect
ANOVA :
Source df
Day 11
Site 1
Method 2
HV vs. avg. M
M(c) vs. M(T)
DS 11
ri
DM 22
SM 2
DSM 22
71
Concentration by Date, Site and Method
S + M, + (DS) . . + (DM) .. + (SM) . + (DSM) .
j k ij IK jk ijk
(fixed) i = 1,2, ,12
(fixed) j = 1,2
(fixed) k = 1,2,3
SS MS F
61,090.55 5,553.69 4.29**
29,090.73 29,090.73 22.50**
40,469.38 20,234.69 15.65**
1 78.18 0.06
1 40,391.20 31.24**
20,464.84 1,860.44 1.44
44,666.47 2,030.29 1.57
7,922.78 3,961.39 3.06
28,447.95 1,293.09
232,152.70 3,269.76
NOTE: * indicates F is significant at P = 0.95
** indicates F is significant at P = 0.99
259
-------
TABLE XXXI. AVERAGE SO^
Date
8/26/76
8/27/76
8/28/76
8/29/76
8/30/76
8/31/76
9/ 1/76
9/ 4/76
9/ 5/76
9/ 6/76
9/ 7/76
9/ 8/76
Overall
Average
HV
16.41
42.50
23.63
14.78
13.79
15.71
24.40
11.64
10.14
7.59
8.25
12.34
16.79
SITE A
M(c)
12.28
38.68
19.09
12.18
11.54
10.56
18.29
8.61
7.31
6.26
6.06
9.86
13.39
M(t)
8.95
27.80
14.88
8.23
8.85
7.93
15.35
6.35
4.03
2.03
3.70
7.78
9.66
HV
15.83
38.40
19.95
14.25
14.38
14.43
22.31
11.98
13.28
11.20
13.29
13.58
16.91
SITE C
M(C)
12.34
37.28
18.61
11.73
11.26
11.00
18.46
8.05
7.33
5.96
5.88
9.81
13.14
M(t)
8.25
23.30
13.93
6.65
14.30
8.35
15.68
4.53
3.15
3.38
3.28
8.10
9.41
260
-------
TABLE XXXII. SQi+ FIXED MODEL
Data Set: Average SO^. concentration by Date, Site and Method
Model: Yj^ = y + D. + S , + M + (DS) . . + (DM) + (SM) ., + (DSM) . ..
•1 J K 1J IK JK JLJ/C
where D: Day Effect (fixed)
S: Site Effect (fixed)
M: Method Effect (fixed)
i = 1,2, ,12
j = 1,2
k = 1,2,3
MOV A:
Source df
Day 11
Site 1
Method 2
m vs. Avg. M 1
M(c) vs. M(t) 1
DS 11
DM 22
SM 2
DSM 22
71
SS
4108.79
0.29
642.23
474.65
169.58
36.66
131.29
0.54
37.01
4956.82
MS F
375.53 222.34**
0.29 0.17
321.11 191.14**
282.53**
99.75**
3.33 1.98
5.97 3.55**
0.27 0.16
1.68
69.81
NOTE: *indicates F is significant at P = 0.95
**indicates F is significant at P = 0.99
261
-------
Figure 7. TSP
concentrations.
262
-------
Figure 8. Sulfate concentrations.
263
-------
(5) Strips cut from the center of filters tend to be slightly (<5%)
higher in concentration than those cut from the ends,
(6) The time lag between sample collection and chemical analysis is a
serious source of error in the determination of
The major conclusions which may be drawn from the 4-hour high volume and
membrane sampler study are as follows:
(1) The overall TSP precision for 4-hour hi-vol samples is only slightly
greater than that for 24-hour samples (5.6% versus 5.2%).
(2) Though sample collection and laboratory are significant sources of
error for SOi+= determination from 4-hour hi-vol samples, the over-
all SOit= precision is only slightly greater than that for 24-hour
samples (5.9% versus 3.9%).
(3) The overall TSP precision for 4-hour membrane samples (cellulose
ester filters) is approximately three times that for 4-hour hi-vol
samples (14.9% versus 5.6%).
(4) The overall SO^= precision for 4-hour membrane samples (cellulose
ester filters) is approximately three times that for 4-hour hi-vol
samples (17.6% versus 5.i
(5) TSP concentrations appear to be dependent upon filter material
employed (i.e., cellulose ester or teflon), though not necessarily
upon sampler type (i.e., hi-vol or membrane).
(6) SOi)= concentrations are very dependent upon filter material employed
with the following ranking in the order of descending concentration:
hi-vol glass fiber, membrane cellulose ester, membrane teflon.
264
-------
"•' COMPARISON OF THE DEGRADATION OF AMMONIUM ION ON HIGH VOLUME
GLASS FIBER FILTERS AND ON MEMBRANE FILTERS
George Colovos
Edward P- Parry
Rockwell International
Newbury Park, California
and
Charles E. Rodes
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
The results of a study on the degradation of particulate ammonium ion
collected on various filtering media^are presented. The mechanism for the
loss of ammonia from glass fiber filters ij given. This mechanism is based
on two assumptions: (1) that the residual alkalinity of the glass fibers is
responsible for the loss of ammonium and; (2) that the rate of this loss is
determined by the diffusion of this residual alkalinity from within the fiber
to the surface. The effect of storage on the analysis of ammonium ion par-
ticulate collected on cellulose ester filters and Teflon filters is also dis-
cussed . c,
,>
INTRODUCTION
Undesirable artifacts associated with the analysis of particulates col-
lected on glass fiber filters have been reported for species such as sulfate
ion (SO^) and nitrate ion (NO$). Several studies are currently underway to
examime the jphenomena connected with the transformation of sulfur dioxide
(S0~z) to SOi+ and nitrogen dioxide/nitrogen oxide (NO2/NO) to NO$ on the filter.
The degradation of particulate ammonium ion (NHi+) collected on such
filter media came to attention through the external quality control (split
sample program) applied to the analysis of the high-volume samples of the Los
Angeles Catalyst Study (LACS). The wide disagreement between the two labora-
tories (Rockwell-Air Monitoring Center and Environmental Monitoring and Sup-
port Laboratory/Environmental Protection Agency) participating in the analysis
of ammonium split samples suggested the possibility of loss of ammonium from
the filters during storage. This hypothesis was also supported by the known
265
-------
high alkalinity of the glass fiber material, which can cause such a loss of
ammonia from the samples. To test this hypothesis, old LACS particulate
samples collected on glass fiber filters previously analyzed for ammonium
were analyzed again. The second determination yielded consistently lower re-
sults, verifying the loss of particulate ammonium. At the same time, however,
it was realized that these results were insufficient for explaining the
mechanism of degradation of the ammonium species on glass fiber filter samples.
This paper presents the results of the study subsequently conducted to
examine the loss of ammonium from particulates collected on glass fiber fil-
ters.
Additional experimental work for verifying the proposed mechanism of de-
gradation of particulate ammonium from glass fiber filters is suggested in
this paper. Some preliminary observations on the stability of particulate
ammonium collected on cellulose ester and Fluoropore filters are also pre-
sented.
EXPERIMENTAL
High-volume samplers operating at the LACS sites were utilized to col-
lect the particulate samples used in this study. The glass fiber filters
that were used were specially made for the LACS study by Gelman, and were
very similar to the Gelman type A/E filters. Three groups of three filters
samples containing variable amounts of particulates were collected and an-
alyzed in this study- Immediately after the collection, each group of sam-
ples was carried to the laboratory for processing. Upon arrival several 3/4"
x 8" strips were cut from the exposed portion of each filter. These strips
were used for the study of the kinetics of the degradation of the ammonium.
The first strip from each filter was extracted immediately and the extract
stored in the refrigerator for analysis. The other strips were folded in
half with the exposed side in, placed in glassine envelopes and kept in the
constant humidity and temperature weighing room (RH 40 +_ 5% and 72 F) for
subsequent analysis. Additional extractions of sample strips were performed
on the 2nd, 3rd, 4th, 7th (or 8th), 21st, (or 22nd), 28th, and 35th (only for
the third group) day after completion of sampling and stored again in the re-
frigerator for analysis. The extracts were analyzed with the next batch of
regular LACS samples. The automated indophenol method Vas used for determin-
ing the ammonium concentration in the extracts.
The same procedures were used for the analysis of the membrane sampJes.
In this case, each filter was divided into four quarters. The filter quar-
ters were stored and analyzed the same way as the glass fiber samples.
RESULTS AND DISCUSSION
\
Tables I, II, and III give the results of the determination of particu-
late ammonium on high-volume glass fiber filters as a function of storage
time. The analytical data are expressed in yg of NH* per ml of the extract
(50 ml total). No effort was made to convert the data to other concentration
units because, as will be shown, the degradation is a first order reaction
(the rate constant of which is given in t~^) and therefore arbitrarily chosen
266
-------
TABLE I. DEGRADATION OF AMBIENT PARTICULATE AMMONIUM
ON GLASS FIBER FILTERS
SAMPLING PERIOD 265-266 JULIAN DATA
(GROUP I)
Extraction
Date
266
267
268
269
270
273
273 DUP
280
288
294
Storage Time
(Days)
0
2
2
3
4
7
7
24
22
28
Estimated OH
Concentration
NH^ Concentration in yg/ml >
Sample #1 Sample #2 Sample #3
25.9
25.3
24.2
24.4
22.9
22.7
22.7
22.3
20.6
20.0
6.0
24.4
24.2
23.8
23.9
23.5
22.2
22.4
8.9
9.2
8.4
6.2
22.5
22.0
22.0
22.5
22.3
20.8
20.0
9.6
9.0
6.8
4.4
(*) 2/22 of the sample extracted in 50 ml of aqueous acetate buffer.
267
-------
TABLE II. DEGRADATION OF AMBIENT PARTICULATE AMMONIUM
ON GLASS FIBER FILTERS
SAMPLING PERIOD 179-180 JULIAN DATE
(GROUP II)
Extraction
Date
180(**)
180(***)
181
182
183
184
188
194
201
208
Storage Time
(Days)
0
0
1
2
3
4
8
14
21
28
Estimated Off"
Concentration
+ (*)
NHij. Concentration in \ig/ml
Sample #1 Sample #2 Sample #3
7.7
7.7
7.0
6.4
6.6
6.0
5.5
5.2
4.8
4.2
3.7
10.1
9.5
8.4
8.8
8.0
8.2
6.8
6.6
5.3
6.4
4.5
8.8
9.8
8.8
8.6
8.3
7.8
7.4
6.2
6.6
5.3
4.6
(*) 1/12 of the sample extracted in 50 ml of aqueous acetate buffer
(**) Strip from the end of the filter
(***) strip from the middle of the filter
268
-------
TABLE III. DEGRADATION OF AMBIENT PARTICULATE AMMONIUM
ON GLASS FIBER FILTERS
SAMPLING PERIOD 172-173 JULIAN DATE
(GROUP III)
Extraction
Date
173(**)
173(***)
174
175
176
177
180
188
194
201
208
Storage Time
(Days)
0
0
1
2
3
4
7
15
21
28
35
Estimated OH
Concentration
+ (*)
NHii Concentrations in yg/ml
Sample #1 Sample #2 Sample #3
5.6
6.0
4.9
4.3
4.6
4.2
3.9
3.0
2.4
1.7
1.1
5.0
5.3
5.2
5.0
3.7
4.4
4.4
4.0
2.9
2.4
2.3
1.1
3.6
4.8
4.5
4.5
4.4
4.0
4.1
3.8
2.6
2.1
1.5
0.9
3.9
(*) 1/12 of the sample extracted in 50 ml of aqueous acetate buffer
(**) Strip from the end of the filter
(***) Strip from the middle of the filter
269
-------
concentration units can be used. The concentration of hydroxyl ion (OH ) was
estimated from the empirical rate curve. The following assumptions were made
in this estimation: (1) the empirical rate curve should level off after re-
action of all the available OH~; and (2) the difference between the high (zero
storage time) and the low points of the curve should be a measure of the a-
mount of OH~ present in each filter. The low point of the curve was estimated
by visual extrapolation of the empirically best fit curve. Due to the similar
formal weights of the NH% and OH~ (18 and 17, respectively) and the relative
imprecision of the curve extrapolation, no effort was made to convert the dif-
ference between high and low ammonium concentration to the equivalent OH con-
centration.
Figures 1-3 graphically present the data given in Table I - III. All
of the curves in these figures are not the empirically best curves, but
those based on the postulated mechanism of the ammonium degradation.' Jt is
evident that the agreement between the theoretical curve and the experimental
result is rather close. These figures also show that the percentage of the
total loss of ammonium ranges from about 35% for the most concentrated samples
to about 82% for the less concentrated samples. However, as is shown in Table
IV, the total loss of ammonium does not appear to depend on the initial am-
monium concentration and, therefore, the loss should be attributed to reaction
between NH* and some species present in the sample. If it is assumed that the
reactive species is the OH~ of the glass fibers, then the following mechanism
can be postulated:
OH. "^ ) OH~ ,
(ff) (s)
NH~u<' >NH* + H (2)
and
(3)
or by combining (2) and (3)
+
OH + H+
i- uti HWo t + H9O . (4)
I C* 1 v f- I ^ /
\ * /
According- to this mechanism, the reaction shown in Equation (1) is the
rate determining step and represents the diffusion of the OH~ from inside the
glass fiber f>H(g)) to the fiber-particulate interface (OH~(s)). Therefore, if
this is the reaction mechanism, the rate expression should be given bu Eaua-
tion (5): y y
where [off" ] is the concentration of OH~ inside the fiber at time t. Also,
270
-------
NJ
CLJRVF
NO. K (DAV1)
1 0.02
2 0.03
3 0.04
4 0.05
15 20
TIME (DAYS)
25
30
35
Figure 1. Changes of the OTjf concentration in high-volume glass fiber filter
samples with time.
-------
NJ
XI
15
20
TIME (DAYS)
25
30
35
Figure 2. Changes of the NH* concentration in high-volume glass fiber filter
samples with time.
-------
NO
16 -
10
15 20
TIME (DAYS)
25
35
Figure 3. Changes of the NH% concentration in high-volume glass fiber filter
samples with time.
-------
TABLE IV. TOTAL LOSS OF AMMONIUM FROM THE SAMPLES
Sample #
1
2
3
4
5
6
7
8
9
NHi+ ]ig/ml
Initi al Cone.
,15.9
14.4
12.5
7.7
10.1
8.8
5.6
5.3
4.8
Total loss
6.0
6.2
4.4
3.7
4.5
4.6
5.0
3.6
3.9
Average 4.66 ± 0.94 St:d. Dev.
274
-------
if+the assumption is made that the surface OH~ reacts quickly with particulate
NHI+ to produce gaseous ammonia (NH$) which escapes from the sample immediately,
then it can be easily proven that the concentration of OH~ inside the fibers
will be given by Equation (6) below:
= c +
where C is the difference of the initial OH~ and NH concentrations and [tf
is the concentration of W#£ at the time t. Obviously Equation (5) can now
be modified to
=kdt
or
(c
log (C + Mfy; = -kt + const (8)
or
t
=
k C + , (9)
where [o#~ ]' is the initial OH~ concentration in the fibers. Equation (9)
gives the rate expression for the postulated mechanism of degradation of NHi±
on glass fiber filters. Figure 4 gives the plot of the average values of all
the experimental points versus time according to Equation (9) , and Table V
gives the rate constants estimated from the individual experiments. The in-
dividually determined rate constants were calculated by eliminating values
which obviously were outlying the rate expression. From the nine experiments,
the experimental points of only one or perhaps two appear to fit the proposed
rate expression very satisfactorily. In the rest of the runs, some points at
random appear to deviate significantly from the rate expressing curve but the
rest of them fit the rate expression satisfactorily. Also, if a rate constant
close to the one obtained by averaging all the experimental points or that
obtained by averaging the individually calculated rate constants is used for
the reconstruction of each experimental run, the agreement between the ex-
perimental results and the postulated rate expression will be rather good.
Figures 1-3 present the experimental results in relation to the curves
drawn by assuming four different rate constants (k = 0.02; k = 0.03, k = 0.04
and k = 0.05) . Inspection of all the runs shows that a rate constant in the
range of 0.035 - 0.045 day"* may represent the ammonium degradation phenomena.
The apparent scatter of the experimental data can be attributed to several
factors which can be predicted to affect the results. Among them, temperature,
275
-------
KJ
1.2
1.0
r-, 0.8
•l-^t-
Z
+
o
i
o
0.6
0.4
0.2
1 r
-i 1 1 r
i i i l i i l I l I
Oi 2 4 6
8 10 12 14 16 18 20 22 24 26 28 30
TIME (DAYS)
Figure 4. Plot of the average experimental values of log f[off~ .]/fC +
vs time.
-------
TABLE V. REACTION RATE CONSTANTS
ESTIMATED FROM EQUATION (9)
Sample #
1
2
3
4
5
6
7
8
9
K (day'1)
0.053
0.051
0.044
0.038
0.027
0.023
0.024
0.026
0.028
Average k 0.035 ± 0.011
277
-------
humidity, barometric pressure, nature of glass fiber, alkalinity of fibers and
last, but not least, the nature of the particulate and concentration of gaseous
pollutants such as NH^, SC>2, nitrogen dioxide (NO2), etc. should be considered.
In order to investigate the effect of these parameters on the ammonium degra-
dation process, experiments under controlled conditions may be necessary.
Controlled experimental conditions can be obtained by generating aerosol at-
mospheres containing NH^ from which samples on glass fiber filters as well as
on other filtering media can be generated. After collection, the filter sam-
ples should be stored in a controlled atmospheric chamber with adjustable^
humidity and temperature levels. Also, the alkalinity of the filters (OH con-
centration) should be determined independently. This can be done by deter-
mining the OH~ concentration in filter extracts by applying a Gran titration
technique similar to the one currently used for determination of strong acid
in airborne particulates. The same technique is also useful in measuring the
total OH~ in the samples as a function of time. In fact, since the postulated
mechanism is based on the hypothesis of the diffusion of OH~ in the fibers,
determination of the total OH~ in the samples constitutes a direct test of
this theory.
The results presented above clearly indicate that the degradation of
particulate ammonium collected on glass fiber filters is an important artifact
which may affect the data by as much as 89% or more, depending on the loading
and the alkalinity of the filters. Theoretically, a factor based on the rate
constant and the alkalinity of the filters can be applied to correct the data
for degradation occurring during storage, but because of the relatively wide
scatter of the results, this may not produce sufficiently accurate data. For
example, if k = 0.035 day'1 and the initial concentration of OH~ is \_OH~"] ',
then the initial concentration of Nfffa will be given by Equation (10)
= lorl'd-icT0'035*) + Drat], do)
where wtf^ is the concentration of ammonium found on the filter after t days
of storage. If Equation (10) is used for the data presented in Tables I, II,
and III, the calculated initial concentrations agree rather well (within 10%)
with those found when values obtained in the first week after sample collec-
tion are used. The correction of results corresponding to later analyses re-
sulted in values considerably higher than the initial determination. Appar-
ently the problem of degradation of particulate ammonium on glass fiber fil-
ters suggests that other filtering media may be more applicable for determin-
ing airbone ammonium. Preliminary results from a study currently underway
have shown that cellulose ester and Teflon filters do not show significant
loss of ammonium immediately after collection of the samples. Table VI de-
monstrates the effect of one week's storage on particulate ammonium collected
on both Teflon and cellulose ester filters. The observed differences are
rather small, indicating the compatibility of these filters with the analysis
of particulate ammonium. Complete evaluation of these and other filtering
media will be possible after the completion of the study presently underway.
278
-------
XI
TABLE VI. EFFECT OF STORAGE ON THE
ANALYSIS OF PARTICULATE AMMONIUM
COLLECTED ON CELLULOSE ESTER FILTERS
yg/ml NH^ In the Extract
Teflon Cellulose Ester
Immediately After
Collection
0.82
0.96
2.90
0.67
One Week
After Collection
0.73
0.72
2.72
0.93
Immediately After
Collection
3.46
4.27
3.32
2.61
1.22
One Week
After Collection
2.56
4.92
2.51
2.45
1.30
-------
SULFATE CONCENTRATIONS AT TWO LOS ANGELES FREEWAYS
A.H. Bockian, George Tsou, Dennis Gibbons, and Robert Reynolds
California Air Resources Board
El Monte, California
ABSTRACT
The Air Resources Board has sampled for particulates during 1975 and 1976
at the San Diego freeway (and more briefly at the Hollywood freeway) in order
to determine the increase in atmospheric sulfate caused by sulfate emissions
from catalyst equipped vehicles. Twelve 2-hour samples were obtained daily,
three days per week, at upwind and downwind sites at each location. An automo-
tive sulfate contribution is evident at each freeway. The maximum contribu-
tion seems to occur at midday when traffic counts are high, traffic is moving
rapidly, and a moderate wind is blowing across the freeway. Data from both
years are presented together with projections, through 1985, calculated from
the increase in vehicle miles traveled by catalyst equipped cars and the
mandatory phase-down in the sulfur content of unleaded gas. These projections
will be compared with those derived from the GM/EPA Sulfate Dispersion Experiment.
INTRODUCTION
The catalytic converters with which most 1975 and later light-duty
vehicles and trucks are equipped have been found to cause the oxidation of
sulfur dioxide (SO2) to sulfate in exhaust gases. While almost all forms of
sulfate are considered to be deleterious to both health and visibility,
sulfuric acid mist is of particular concern because of its irritant effect on
respiratory tract membranes. In the atmosphere, many sulfates are delique-
scent, so they readily absorb water vapor and grow to light scattering size.
Also, since sulfates are particulates, they must be considered in any control
strategy developed to meet ambient air quality standards for total suspended
particulate matter. Thus, directly or indirectly, sulfates affect three
ambient air quality standards: sulfate, visibility and total particulate.
Additionally, the health hazard of certain sulfates cannot be ignored.
A baseline study was conducted by Rockwell International, Air and Indus-
trial Hygiene Laboratories (California), California Department of Health, and
the Environmental Protection Agency (EPA) along the San Diego freeway during
1974. The study took place before the advent of catalyst equipped vehicles.
It concluded that sulfate emissions from noncatalyst equipped automotive
sources were negligible (1). A brief California Air Resources Board (ARE)
dynamometer study had indicated that sulfate emissions from catalyst equipped
281
-------
cars were many times greater than from previous years' models, although mea-
surable sulfate emissions from pre-1975 vehicles were detected. It thus
became of primary importance to measure sulfate emissions in traffic from
catalyst equipped vehicles and to evaluate the health and atmospheric implica-
tions of any increase in sulfate concentrations. This report briefly presents
the sampling procedures and analytical methods and discusses the increases in
sulfate concentrations that might be attributable to catalyst-equipped vehicles,
METHODS
SAMPLING
The ARE has sampled for sulfate along both the Hollywood and San Diego
Freeways. In May 1975, sites on the east and west sides of the San Diego
Freeway, between Santa Monica and Wilshire Boulevards, were placed in opera-
tion. The "east" site is about 50 feet from the freeway; the "west," or
background site, about 800 feet (Figures 1, 2, 3). During 1976 sampling was
discontinued at the Hollywood Freeway; only the San Diego Freeway site is
currently in operation.
Site selection was governed by several criteria, including: (1) proxi-
mity to freeway so as to measure sulfate emissions as directly as possible
before they diffused completely into the ambient atmosphere; (2) prevailing
wind direction perpendicular to the freeway; (3) low background levels of pol-
lutants due either to direct emissions or to meteorological or photochemical
processes; and, (4) high traffic counts on the freeway for several hours per
day. The San Diego Freeway site meets these criteria well, the Hollywood
Freeway site less well. At the ARB sites, twelve 2-hour particulate samples
are taken every Tuesday, Wednesday and Thursday. The laboratory's usual low
volume procedure (47 mm Gelman triacetate filters, flow rate ca. 4 cfm) is
followed (2) .
ANALYSIS
Sulfate and lead (Pb) are analyzed by X-ray fluorescence (XRF) (3).
Although several other elements are detected by this technique, none is usual-
ly present in sufficient quantity to be reliably estimated on a day-to-day
basis. Analyses have been completed for samples taken through January, 1977.
The laboratory reports as "sulfate" all sulfur compounds detected by XRF,
although the instrument determines sulfur as such regardless of whether it is
present as sulfur, sulfate, sulfide or any other form. A comparison of XRF
results with those obtained by a standard wet chemical sulfate analysis
showed the XRF values to be approximately 30% higher. Data presented here are
corrected for this disparity.
Aerometric data obtained at the San Diego East site consist of oxidant,
carbon monoxide (CO), sulfur dioxide (502), oxides of nitrogen (NO ), wind
speed and direction, hydrocarbons, relative humidity, temperature, UV intens-
ity and b .No aerometric data are obtained at the reference site (San
Diego WescJ.
282
-------
\
K
I \
I \
^
\
\
«
VHollywood Freeway
\ (1-5)
I
\
s
\
X^Santa Monica
~^ """Freeway ) """"• -^
v (1-10) (Los Angeled
. •
v^
I
\
San Diego I
Freeway }
' (1-405) |
\
\
Pacific Ocean
10 miles
Sequential Aerosol Samplers
Figure 1. Sampling sites-ARB freeway sulfate study.
283
-------
Ol
c
Ol
4->
V)
0)
\\
Santa Monica Blvc
\\
\\
\
•P
c
o
-------
San Diego
Freeway
Wilshire Blvd.
Santa Monica
Blvd
1000
Sequential Aerosol
Samplers
Figure 3. Sampling locations San Diego freeway.
285
-------
RESULTS
Diurnal sulfate concentrations at the San Diego Freeway-East and -West
sites for the Fall of 1975 are shown in Figures 4 and 5. The average corrected
monthly differences (East vs. West), as measured and shown in Figure 6, were
1.3 yg/m3 in August; 1.5 in September; 1.1 in October; and 0.9 in November.
There was a substantial diurnal variation in sulfate concentrations. Maximum
monthly differences (East vs. West) for any 2-hour interval of the day, for
example the 10:00 - 12:00 a.m. period, averaged close to 3 yg/m3, usually
during midday. Differences were 2.6 in August; 3.7 in September; 3.0 in
October; and 2.3 in November. The differences seemed to be a function of
traffic count, freeway traffic speed, and wind direction and speed. During
any single 2-hour sampling period, differences exceeding 5 yg/m3 were common,
and differences near or above 7 yg/m3 were not unusual. (The largest single
value was 7.5 yg/m3J. On occasion, differences averaged over an integrated
24-hour period exceeded 3 yg/m3. "Normalized" sulfate differences, incorporat-
ing wind speed and direction, are shown in Figure 7.
Results obtained during the Fall of 1976 are not directly comparable with
the 1975 results because statistical analyses reveal that week-to-week varia-
tions in 1975 were not random. The average monthly differences, west to east,
for August through November, 1976 as shown in Figure 8 were as follows:
August 2.1
September 1.4
October 1.3
November 1.1
These figures take into account the signs of the differences. "Normalized"
values are shown in Figure 9. Although the monthly averages show little
apparent change, 1975 vs 1976, the number of two-hour periods when differences
approximated or exceeded 10 yg/m was substantially greater in 1976. The
largest 24-hour difference found in 1976 was 7.5 yg/m3. Concentrations of
primary pollutants (including SO2) and oxidant were invariably low.
The data obtained at the Hollywood Freeway during 1975 are presented
graphically in Figures 10 through 12. Analyses not presented here suggest
that incremental sulfate values there are somewhat smaller than at the San
Diego Freeway site.
PROJECTIONS
Calculations based on the estimated number of catalyst-equipped vehicles
in Los Angeles County (R.L. Polk and Co. Statistics) and the average number of
miles a car travels annually, depending on its age (4), indicate that catalyst-
equipped cars accounted for about 16.4% of the vehicle miles traveled (VMT)
during the Fall of 1975. If sales of light-duty vehicles in the next several
years equal those for the 1974 model year, and if the current mileage vs. age
figures hold true, the ratio of miles traveled by catalyst-equipped vehicles,
compared to the 1975 mileage, would be about 2.16 by November, 1976; 3.04 by
November, 1977; and 3.78 by November, 1978.
286
-------
Ni
00
CO
E
en
-------
20
00
00
CD
c
o
to
4J
c
(U
o
c
o
-------
3.7
00
VO
•M
0}
-------
50
40
30
20.
10
V V
* M. *
'•' / '••
•• /\ %
,\
\
Aug
... Sept
- Oct
Nov
00 02 04 06 08 10 12 14 16 18 20, 22
Time of Day (local)
Figure 7. "Normalized" diurnal variations in differences in sulfate
concentrations, SDE vs SDK, 1975.
-------
K>
O>
3
it/I
O
«O
01
O
c
O
O
o>
I/I
1
1
0)
3.0 ,
2.0
1.0
£ 0
Overal1
Monthly
Average
August
Sept
Oct
Nov
00 02 04 06 08 10 12 14 16 18 20
Time of Day (local)
2.1
1.9
(1.3)
(1.1)
1975
2.1
2.1
(1.1)
(1.1)
Figure 8. Differences in diurnal variations in sulfate concentrations,
SDE vs SDtf, 1976.
-------
50
40
30
20 '
to
vo
KJ
10 -
00 02 04 06 08 10 12 14 16 18 20 22
Time of Day (local)
Figure 9. "Normalized" diurnal differences in sulfate concentrations
SDE vs SIM, 1976.
-------
U)
20
15 '
10
V x
01 03 05 07 09 11 13 15 17 19 21 23
Time of Day (local)
Overall
Monthly
Average
Sept
Aug
(14.
(14.
Oct (13.6
— Nov (13.0)
Figure 10. Diurnal variations in sulfate concentrations, Hollywood
freeway (HF), 1975 monthly averages.
-------
CO
en
3
c
o
to
o
J
ID
co
20 ,
15
10
••• /
••• -y ,
\
\ x.
\ \
\ „.—- — •
:<"
• \
x •
01 03 05 07 09 11 13 15 17 19 21 23
Time of Day (local)
Overal1
Monthly
Average
September (15.8)
(4 days only)
October (12.1)
November (12.0)
Figure 11. Diurnal variations in sulfate concentrations, Hollywood
freeway reference (HFR), 1975 monthly averages.
-------
NO
VO
Ul
en
4->
-------
If the average sulfur (S) content of lead-free gasoline sold in the Los
Angeles area remains 270 ppm, its reported current weighted average (5) , then
based on the San Diego freeway data, the sulfate contribution from the catalyst-
equipped vehicles would approximate those shown in Table I, assuming no changes
in engine or catalyst performance.
Since about 95% of all VMT is accumulated by cars that are less than ten
years old, and 48% by cars sold in the last three model years, the amount of
VMT by catalyst-equipped cars will eventually be approximately six times the
VMT by the 1975 cars. Long range projections are usually hazardous, but the
current data suggest that in 1979 S contributions would be 4.4 times the 1975
values, assuming that the S content of gasoline remained at 270 ppm. Further
increases in VMT after 1980 would not be offset by a lower S content of gaso-
line unless the limit of 300 ppm after January 1, 1980 were lowered.
In 1980, for example, the S increment from gasoline would be 5.0 times
the 1975 value. In 1981 it would be 5.5 times and in 1982, 6.0 times. From
1985 on, the S contribution would level out (using the conservative assumptions
stated earlier) at about 6.8 times the 1975 level. Projections for each year
are shown in Table II.
These projections agree remarkably well with those which can be derived
from the General Motors (GM) /EPA Sulfate Dispersion Experiment (6). In that
study the traffic count was 10,920 cars every two hours. The S content of
gasoline averaged 300 ppm. The average sulfate increase was 8 yg/m3 and the
maximum increase was 16 yg/m3. The maximum 24-hour traffic count on the San
Diego Freeway averaged 19,250 cars per 2-hour period, and the maximum 2-hour
count was 28,420. Adjusting for traffic count, the maximum 24-hour daily
average increase in sulfate concentrations, based on the GM data, would be
/ 70 250 x 8)
14.1 yg/m3 -. — ' — p. The maximum 2-hour concentration increase would be
( J.U,y^U )
A-, x- / 3 (28,420 x 16) _ . .... ,„ ,-j •>,-., .,
41.6 yg/m° -. — . - j-. Our projections are 14.5 and 36.3, respectively.
During the 14 hour period from 0600-2000 the traffic count on the San Diego
Freeway averages 26,100 cars per 2-hour period; the GM data then project to
.. . . 3 . , . . , (26,100 x 8)
19.1 yg/m3 increase during this period —. — ' — —.
( J.U,y,eU )
SUMMARY
The laboratory has sampled at both the Hollywood and San Diego Freeways
in order to determine the increase in atmospheric sulfate caused by sulfate
emissions from catalyst equipped vehicles. Sampling began in May 1975 but
background data at both sites have been available only since September, 1975.
Data from the San Diego Freeway locations are considered to be more reliable
than those from the Hollywood Freeway.
A freeway sulfate contribution is evident at each site. The maximum
contribution seems to occur at midday when traffic counts are high, traffic is
moving rapidly, and a sea breeze is blowing across the freeway.
296
-------
TABLE I. PROJECTED SULFATE CONTRIBUTION (\ig/m3) FROM CATALYST-
EQUIPPED VEHICLES
Year Monthly Average Maximum in 24 hours Maximum in 2 hours
1976 (November) 2.8 6.4 15.3
1977 (November) 3.8 8.7 21.6
1978 (November) 4.8 11.1 26.8
297
-------
TABLE XL INCREMENTAL SULFATE EMISSIONS. FROM CATALYST EQUIPPED VEHICLES,
1979-85, RELATIVE TO 1975 (S CONTENT OF GASOLINE ASSUMED TO BE 270 PPM)
Year
Catal yst-Eguipped
Vehicle VMT
(Billions)
Ratio to 1975
Catalyst-Equipped
Vehicle VMT
Sulfate Emission
Increase,
(1975=1.00)
1979
1980
1981
1982
1983
1984
1985 on
23.49
26.68
28.94
31.90
34.04
35.77
36.24
4.43
5.03
5.46
6.02
6.42
6.75
6.84
4.4
5.0
5.5
6.0
6.4
6.7
6.8
298
-------
If the currently observed increment in sulfate values continued to be
proportionate to the vehicle miles traveled by catalyst equipped cars, by 1985
the maximum 2-hour concentration at the freeway would increase by about
fourfold, the maximum 24-hour increase would be a significant part of the
ambient air quality standard, and the monthly average would increase by about
7.7 ygr/m3. These predictions agree well with those that may be derived from
the GM/EPA Sulfate Dispersion Experiment.
ACKNOWLEDGMENT
The authors are indebted to Marilyn Smithson, Eric Fujita and Sidney
Shimabukuro for sample collection, and to Arthur Heath and Dr. Donald W. Davis
for analyses.
REFERENCES
1. Appel, B. and Wesolowski, Jr. "Impact of Motor Vehicle Exhaust Catalysts
on Air Quality." Calif. Air Resources Board Contract 3-985. August
1975.
2. Reynolds, R. , Heath, A., and Tsou, G. "Metals and Sulfate Aerosol Data
Report, Summer 1975, SCAB." Calif. Air Resources Board Report DTS-76-25.
3. Reynolds, R "X-ray Fluorescence Data Reduction Program," Calif. Air
Resources Board unpublished report, June 1976.
4. Data Base and Documentation for Estimating Emissions from Motor Vehicles
in California. Calif. Air Resources Board, Technical Services Division
(Emissions Inventory Unit). January 1977.
5. Olson Laboratories, Anaheim, Calif. Private Communication, March 1977.
6. Cadle, S., Chock, D., Monson, P., and Heuss, Jr., "General Motors Sulfate
Dispersion Experiment: Experimental Procedures and Results," J. Air
Poll. Control Assoc. 27:33 (1977).
299
-------
SUMMARY OF LACS CONTINUOUS DATA
Gary F. Evans and Charles E. Rodes
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
Continuous measurements have .been made in the Los Angeles Catalyst Study
for wind direction and speed, temperature, relative humidity, traffic count
and speed, and upwind and downwind concentrations of carbon monoxide, nitric
oxide, nitrogen dioxide, total sulfur and ozone. Each of these variables is
reported in hourly values. The frequency of occurrence of wind directions
and associated speed are discussed along with the diurnal and seasonal patterns
of wind direction, wind speed, temperature and relative humidity- Hour-of-
day patterns in count and speed , presented for Northbound and Southbound
traffic data, have only been collected on-site since September 1976; estimates
of the total volume at the monitoring site for the past five years have been
obtained from the California Department of Transportation. Hourly pollutant
concentrations at upwind and downwind sites and the cross-freeway difference
are examined in relation to wind direction, wind speed, hour-of-day, season,
traffic volume and traffic speed. Emphasis in these comparisons is placed on
carbon monoxide since it is a primary and non-reactive pollutant. Finally,
long-term trends in the concentrations of all continuously monitored pollu-
tants are discussed.
INTRODUCTION
The location of the sampling sites for the Los Angeles Catalyst Study
(LACS) is shown in Figure 1. The four permanent sites are situated along
a line perpendicular to the San Diego Freeway with two sites on either side.
Through this part of the Los Angeles basin, the freeway runs parallel to the
Pacific Ocean (i.e., Southeast to Northwest) at a distance of approximately
four miles inland. The freeway perpendicular, serving as the site line,
occurs at 235° (i.e., in the Southwest octant).
Figure 2 is an aerial view of the Veterans Administration Hospital
property on which the LACS sites are located. The figure indicates the
positions of permanent sites A and B on the ocean (or "upwind") side of the
freeway and C and D on the opposite (or "downwind") side. All of the con-
tinuous monitoring equipment is located at site A (100 feet upwind of the
freeway) and/or site C (25 feet downwind). Also shown in the figure is
temporary site A1 which is located in an open field some 700 feet from the
freeway on the upwind side. A carbon monoxide (CO) analyzer was operated at
301
-------
337.5
NW
292.5°
247.5°
^67.5°
112.5"
SE
Figure 1. Site location in Los Angeles.
302
-------
Figure 2. Aerial view of sampling sites.
-------
site A1 during the summer of 1975 in an effort to determine whether permanent
site A adequately reflects true background concentrations.
Table I lists the variables in the LACS which are continuously monitored
and reported as hourly averages. The sampling location, method and period of
data coverage are shown for each continuously monitored variable.
METEOROLOGICAL DATA
Wind direction and speed have been continuously monitored since the
beginning of the LACS (June 1974) by means of a Climet anemometer located
at site A at an elevation of 10 meters. During the first six months of the
study, simultaneous readings of wind direction and speed were collected with
an MRI anemometer at site D at an elevation of 3 meters. Comparison of the
data from the two systems showed that the wind direction information was
essentially equivalent and that wind speeds recorded at site A, though highly
correlated with corresponding readings at site D, were approximately 3 times
as great due to the difference in elevation. As a result of the favorable
comparison, the anemometer at site D was discontinued in December 1974.
Figure 3 depicts the distribution of occurrences and the associated
average wind speed as a function of wind direction. Wind direction strip
charts are reduced to hourly averages and reported to the nearest ten degrees.
Thus, there are 36 possible values for wind direction plus the classification
'calm' (denoted by zero) which is used for any hour for which the accompanying
wind speed is less than one mile per hour. The lower curve in Figure 3 is a
frequency distribution of all the wind direction data collected at site A since
the beginning of the study. It was plotted by tabulating the number of hours
of occurrence for each possible wind direction, dividing %y the total number
of hours for which valid wind direction data are available (about 18,000 hours)
and expressing the quotients as percentages. The upper curve is the average
wind speed over all hours for which the winds were from the indicated
direction.
Prevailing wind direction was one of the prime considerations involved
in the selection of a suitable site for the LACS. Areas of the Los Angeles
basin which are close to the ocean are characterized by an alternating pattern
of onshore (or "seabreeze") winds during the daylight hours and offshore
(or "landbreeze") winds during the dark hours. Thus, it was felt that a major
freeway running parallel and near to the ocean would provide the desirable
property of prevailing winds which were perpendicular to the freeway for a
large portion of almost every day. Such a wind pattern would enable the
detection of the freeway contribution to ambient levels of air pollutants by
differencing simultaneous measurements taken upwind and downwind of the
freeway.
As shown in the lower curve in Figure 3, wind direction at the LACS
site does follow a bimodal distribution with two distinct patterns: a
seabreeze pattern centered in the Southwest with tails in the South and West
and a somewhat more diffuse landbreeze pattern centered in the North with
tails in the Northwest and Northeast. The upper curve in Figure 3 shows
that the seabreeze is accompanied by the highest wind speeds.
304
-------
TABLE I. INVENTORY OF CONTINUOUS DATA
Variable
Sampling
Location
Sampling
Method
Sampling
Period
Wind Direction/Speed
Wind Direction/Speed
Temperature/Dew Point
Traffic Count/Speed
Carbon Monoxide (CO)
Carbon Monoxide (CO)
Nitric Oxide (NO)
Nitrogen Dioxide
Ozone (O%)
Total Sulfur (S)
Site A (10 m.)
Site D (3m.)
Site A (1m.)
All lanes
Sites A & C ( 1m.)
Site A' (1m.)
Sites A & C (1m.)
Sites A & C ( 1 m.)
Sites A & C ( 1 m.)
Sites A S C (1m.)
Climet Anemometer
M.R.I. Anemometer
Thermistor/LiCl Detector
Inductive Loop System
Nondispersive Infra-red
Nondispersive Infra-red
Chemi1uminescence
Chemi1uminescence
Chemi1uminescence
Flame Photometric
June 74 - present
June 74 - December 74
October 74 - present
September 76 - present
June 74 - present
June 75 - September 75
January 75 - present
January 75 - present
January 76 - present
May 75 - present
-------
Uj
o
8.
6.
a
Ul
s
•» 2.
o
10.
6.
O
ul
E
u.
2.
I I I 1 I ± J L L I I I I I I I I I I I I I I I I I I I I I I I I I I
LANDBREEZE
MODE
I I 1
1 I I I I { I I I 1 1 I i i i I I I 1 I I I I I
°5 8
? I
0 N
8 S S 3
NE
R 88 § ?
£
So o o
m * 35
SE
SO Q Q O
K *
-------
Since the Seabreeze mode (wind directions from the South, Southwest or
West) provides the most favorable conditions for the detection of pollutant
sources on the freeway, it is useful to look at the diurnal and seasonal
patterns in the occurrence of winds from this sector. Figure 4 shows the
percentage occurrence of Seabreeze by hour-of-day for all summer months
(defined as April through September, inclusive). In the early mornincr hours.
winds are typically light and variable and are out of the Seabreeze sector
(S+SW+W) for only about 20% of all summer days. Following sunrise (about
7 a.m.) the wind pattern begins to shift into the Seabreeze mode and from
noon until 7 p.m. winds are nearly always from this favorable sector.
Since the LACS integrated samplers are operated on either a 24-hour
basis (midnight to midnight) or a 4-hour basis (3-7 p.m.), the overall per-
centage occurrence of Seabreeze for each of these intervals is shown in the
figure (61% and 97%, respectively). This suggests that while the 4-hour
afternoon sampling interval should provide a good measure of the freeway con-
tribution to pollutant concentrations, the 24-hour interval will likely be an
underestimate since the diurnal wind pattern is such that all sites will be
impacted upon by the freeway during some portion of the day.
Figure 5 shows the diurnal pattern of Seabreeze for winter months (October
through March). While similar to the summer pattern, the figure shows that the
period of favorable meteorology is of much shorter duration because there are
fewer hours of sunlight. Overall, the Seabreeze accounts for only 35% of
winter hours and from 3-7 p.m. the percentage Seabreeze is only 65%. Thus,
it would appear that cross-freeway differences obtained from integrated
samplers would be notably different between summer and winter seasons. The
LACS integrated data are discussed in a separate report in these proceedings.
The diurnal pattern for temperature is shown in Figure 6 with separate
curves for the summer and winter seasons. Temperature follows a symmetrical
pattern reaching its peak in the early afternoon. The seasonal difference
in temperature is not large (5 F overall average) . Figure 7 provides the
same information for percent relative humidity. Humidity reaches a minimum
in the early afternoon and averages 5% higher in the summer season.
TRAFFIC DATA
The LACS traffic count and speed system became operational in September
1976. The system consists of inductive loops embedded in each of the ten
lanes of the San Diego Freeway, and interface system to convert electric
impulses from the loops to digital counts and a mini-computer located in the
shelter at site C to store traffic count data on magnetic tape. Operating
on a ten minute cycle, the system provides a total vehicle count for each
lane and within each of six speed categories: 0-25, 25-35, 35-45, 45-55,
55-65 and greater than 65 miles per hour (mph). As of December 1976, only
eight of the ten lanes of the freeway were open to traffic (four lanes
Northbound and four lanes Southbound).
Every two weeks, the magnetic tape from the on-site computer is
forwarded to Research Triangle Park, N. C. (RTF) where the data aire reduced
to hourly traffic volume and estimated average speed Northbound and South-
307
-------
% FREQ.
(S+SW+W)
100.
90.
80.
70.
60.
50.
40.
30.
20.
10.
-10.
-20.
-30.
-40. _
SUMMER: APR. - SEP.
OVERALL AVC. = 61%
3-7 PM AVG. = 97%
01 234 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
H
1 3-7
HOUR OF DAY
PM
Figure 4. Frequency of Seabreeze, diurnal pattern.
308
-------
WINTER: OCT. - MAR.
% FREQ.
(S+SW+W)
100.0
90.0
80.0
70.0
60.0 -
50.0
40.0
30.0
20.0
10.0
0.0
-10.0
-20.0
-30.0
-40.0
OVERALL AVG. = 35%
3-7 PM AVG. = 65%
01 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
HOUR OF DAY
3-7 PM
Figure 5. Frequency of Seabreeze, diurnal pattern.
309
-------
100.0
TINT
•0.0
70 A
00 A
BOA
40.0
JOA
20.0
10jO
-10*
-20.0
o o
-a—a
O SUMMER: AM. - SEP. AVO. • M*F
O WINTER: OCT. - MAM. AVO.
• 1 I 3 4 S ^ ? I •
U t» b
»
U U
DAY
8-f PM
Figure 6. Temperature by hour.
310
-------
100.
90.
% R.H.
80.
70.
60.
50.
40.
30.
20.
10.
O SUMMER: APR. - SEP. AVG. = 68%
O WINTER: OCT. - MAR. AVG. = 63%
I i
-10.
0 1 23 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
H H
1 3-7 PM '
HOUR OF DAY
-20.
-30.
-40.
Figure 7. Relative humidity by hour.
311
-------
bound. Figures 8-11 present the diurnal patterns for traffic count and
speed composited over the period for which valid data are available (September
through December 1976).
Since no statistical day-of-week differences appeared for the composited
Monday through Thursday data, these days were pooled into an overall weekday
composite which appears in Figure 8. Traffic volume follows a bimodal pattern
reaching peaks at about 7 a.m. (morning rush hour) and again at about 4 p.m.
(afternoon rush hour). The Southbound lanes exhibit the larger peak during
the morning rush hour while Northbound lanes exhibit the larger peak during
the afternoon rush hour. This phenomenon is consistent with the fact that
the LACS location is Northwest of downtown Los Angeles; thus the weekday
traffic burden shifts from Southbound in the morning to Northbound in the
afternoon. Traffic speed averages about 55 mph but congestion causes it to
dip to about 40 mph Southbound during the morning rush hour and to about 20
mph Northbound during the afternoon rush hour. Average traffic volume for
weekdays was 191,500 vehicles per day for the four month period.
Figures 9, 10 and 11 present the diurnal traffic patterns for composited
Fridays, Saturdays and Sundays, respectively. The Friday pattern is similar
to the Monday-Thursday pattern with the exception that the afternoon rush
hour is of longer duration. Overall traffic volumes for Fridays was 205,000
vehicles per day or 7 percent greater than that for other weekdays. As
expected, the patterns for Saturday and Sunday do not show the characteristic
morning and afternoon rush hour patterns, average speed remains fairly
constant throughout the course of the day and the overall traffic volume is
less than that for weekdays.
The quotient of traffic count and corresponding average speed has units
of vehicles per mile and may be thought of as an instantaneous measure of '
traffic density. Figures 12 through 15 show the diurnal pattern of traffic
density ftfr composited weekdays (Monday through Thursday), Fridays, Saturdays
and Sundays. It is interesting to note that the weekday afternoon rush hour
produces a peak in traffic density which is nearly twice that of the morning
rush hour period. It should be pointed out that the 3-7 p.m. interval for
the operation of integrated samplers encompasses this large peak in traffic
density. This fact, coupled with the very favorable patterns of wind
direction and speed that occur from 3-7 p.m. (as discussed in the previous
section), suggest that the afternoon sampling interval for integrated samp-
lers should present the optimum conditions for assessing the freeway con-
tribution of air pollutants. As expected, traffic density is lower and more
uniform on Saturdays and Sundays.
CONTINUOUS POLLUTANT DATA
Carbon monoxide (CO) has been continuously monitored at the LACS sites
by means of nondispersive infrared analyzers since June 1974. Chemilumi-
nescence instruments for monitoring nitric oxide (NO) and nitrogen dioxide
(N02) were added in January 1975 and instruments for ozone (0$) added in
January 1976. Flame photometric instruments for the determination of total
gas phase sulfur (S) have been operated intermittently since May 1975. All
continuous pollutant instruments are operated in pairs with one each at
312
-------
8000
WEEKDAY COMPOSITE (MON-THURS)
NORTHBOUND LANES
u.
0:00 2:00 4:oo 6:00 8:00
§
10:00 12:00 14:001 16:00 18:001 20:00 22:00 24:00
TIME
3-7 P.M.
SOUTHBOUND LANES
ee
1000
0:00 2:00 4:00 ROD s:oo 10:00 12:00 i4:0o i6.no is:oo 20:00 22:00 24:oo
TIME
SPEED (MPH)
TRAFFIC COUNT (AVE)
Figure 8. LACS traffic data.
313
TOTAL AVERAGE COUNT = 191473
-------
FRIDAY COMPOSITE
NORTHBOUND LANES
0:00 2:00 4:00KM iSo 10:00 12:00 14:00 16:00 18:00 I 20:00 22:00 24:00
3-7 P.M.
SOUTHBOUND LANES
i S
i
2:00 4:00 6:00 B:OO 10:00 12:00 14:00 ie:oo is:oo 20:00 22:00 24:00
TIME
SPEED (MPH)
TRAFFIC COUNT (AVE)
TOTAL AVERAGE COUNT = 204797
Figure 9. LACS traffic data.
314
-------
SATURDAY COMPOSITE
NORTHBOUND LANES
0:00 2:00 4:00 6:00 8:00
10:00 12:00 14:00 [16:00 18:00J 20:00 22:00 24:00
TIME
3-7 P.M.
8000-
SOUTHBOUNO LANES
0:00 2:00 4:00 6:00 8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 24:00
TIME
SPEED (MPH)
TOTAL AVERAGE COUNT = 182828
TRAFFIC COUNT (AVE)
Figure 10. LACS traffic data.
315
-------
8000-1
SUNDAY COMPOSITE
NORTHBOUND LANE*
8000
7000
2:00 4:oo 6:00
10:00 12:00 14:00116:00 18:00
TIME ' 3-7P.M.
20:00 22:00 24:00
SOUTHBOUND LANES
0.00
2:00 4:oo e:oo s:oo 10:00 12:00 i4:oo 16:00 ie:oo 20:00 22:00 24:oo
TIME
TOTAL AVERAGE COUNT = 165682
SPEED (MPH)
TRAFFIC COUNT (AVE)
Figure 11. LACS traffic data,
316
-------
MONDAY - THURSDAY COMPOSITE
700
0:00 2:oo 4:oo e:oo s:oo 10:00 12:00 1*00 15:00 18.00 20:00 22:00 24:00
3-7 P.M.
SOUTHBOUND LANES
NORTHBOUND I AMES
TOTAL (SOUTHBOUND AND NORTHBOUND)
Figure 12. LACS traffic density (count/speed)
317
-------
FRIDAY COMPOSITE
U)
hJ
oa
o:oo 2:00 4:00 e:oo 8:00 10:00 12:00 14:001 ie:oo 18:001 20:00 22:00 24
3-7 P.M.
SOUTHBOUND LANES
NORTHBOUND LANES
TOTAL (SOUTHBOUND AND NORTHBOUND)
Figure 13. LACS traffic density (count/speed).
-------
U)
UJ
_i
(VEH/MII
>
w
Z
Ul
Q
O
u_
u.
(£
1-
700
600
500
400
300
200
100
t
<
/%
SATURDAY COMPOSITE
0:00 2:00 4:00 6:00 8:00 10:00 12:00 14:00
16:00 18:00 20:00 22:00 24:oo
3-7 P.M.
SOUTHBOUND LANES
* NORTHBOUND LANES
-I TOTAL (SOUTHBOUND AND NORTHBOUND)
Figure 14. LACS traffic density (count/speed).
-------
UJ
NO
O
x
111
V)
ut
a
o
DC
H
700-
600-
500-
400-
300*
100
SUNDAY COMPOSITE
0:00 2:00 4:oo
s:oo 10:00 12:00 14:00 16:00 is:oo 20:00 22:00 24:oo
3-7 P.M.
SOUTHBOUND LANES
NORTHBOUND LANES
TOTAL ( SOUTHBOUND AND NORTHBOUND)
Figure 15. LACS traffic density (count/speed).
-------
sites A and C. Inlet probes are at an elevation of about one meter above
freeway grade (i.e., approximately at commuter elevation).
Figure 16 shows CO concentrations at site A, site C and the cross-
freeway (C-A) difference as a function of wind direction. The plots were
constructed using all valid CO data since the beginning of the study,
stratifying by wind direction and computing average concentrations by site
and for the average concentration difference (C-A) for each wind direction.
It is interesting to note that when winds are calm (wind direction = 0) ,
site C exceeds site A by about one part per million (1 ppm) due to a proximity
effect (site C is 25 feet from the freeway while site A is 100 feet from the
freeway). The cross-freeway difference (C-A) curve is slightly (<1 ppm)
negative (A exceeding C) when winds are from the North, Northeast or East.
It crosses the zero line when winds are along the freeway parallels (South-
east or Northwest) and is distinctly positive when winds are from the Seabreeze
mode (South, Southwest or West), reaching a maximum of about 5 ppm when winds
are from the Southwest (i.e., perpendicular to the freeway). This pattern
confirms the utility of the study design in making use of the naturally
occurring Seabreeze to separate emission products from background levels of
automobile-generated pollutants.
In an attempt to investigate the effect of wind speed on pollutant
concentrations, the CO data base was first sorted for all hours when the
prevailing wind direction was from the Southwest (since these essentially
perpendicular winds give rise to the largest cross-freeway differences) .
The restricted data base was then further stratified on the basis of wind
speed intervals, and the average CO concentrations at site A, site C and the
freeway difference (C-A) were computed for each interval. The results are
plotted in Figure 17. Again, it is observed that site C substantially
exceeds site A even at very light wind speeds due to the proximity effect.
As wind speed picks up, the average CO concentration decreases at site A and
increases at site C until wind speed reaches about 4.5 mph. Above this speed,
the average CO concentration at both sites and the cross-freeway difference
decrease due to the effect of increased dilution.
Figure 18 shows the diurnal pattern in CO concentration for each site
and the cross-freeway difference as a composite of all summer months (April
through September). The cross-freeway difference is very small during early
morning hours but begins to build at about 7 a.m. as the Seabreeze mode begins
to predominate and the morning rush hour period begins. It levels off at
about noon when the Seabreeze mode is well established and the morning rush
hour period subsides. Then at about 3 p.m. the cross-freeway difference
increases again in response to the afternoon rush hour period, reaching a
diurnal maximum of about 6 ppm at 5 p.m. Again, it is worth pointing out
that the 3-7 p.m. sampling interval for integrated samplers appears to cover
the optimum period for the detection of emission sources on the freeway.
Figure 19 shows the CO diurnal pattern as a composite of all winter
months (October through March). Though similar in pattern to the summer
months, the winter pattern is strongly affected by the shorter duration of
the Seabreeze mode. By 7 p.m., the cross-freeway difference in CO concen-
tration has practically disappeared for the average winter day.
321
-------
6.99
6.29
5.59
4.87
4.17
3.47
2.75
2.05
1.35
0.55
-0.05
-0.75
I I I I
111 lit
I I I I I I I I I I I I
k"^
1 I I I I I I
a o
2 8
<•
u
§ S 3
SS
S
I II
€M
81688
s <*> w
N
NE
SE
sw
W
NW
N
Figure 16. CO by wind direction (all data).
-------
6.0 -
5.0 -
4.0 -
CO
Ui
NJ
U)
3.0 -i
2.0 -
1.0 -
I I • 1 i I I I 1 T
1.0 2.0 3.0 4.0 5.0 €.0 7.0 8.0 9.0 10.0
WIND SPEED (MPH)
A - AVG. CO CONCENTRATION AT SITE A
C - AVG. CO CONCENTRATION AT SITE C
A - AVG. CO CONCENTRATION DIFFERENCE (C-A)
Figure 27. CO by wind speed for WD = SW.
-------
SUMMER: APR. - SEP.
CONC
(PPM)
10.0 -
9.0 -
8.0 -
1.0
0.0
-1.0
-2.0
-3.0
1.0
I
I 1 I
1
_L
I I
I I I
I
1
01 234 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
h
HOUR OF DAY 3~7 PM
O INDICATES SITE C
• INDICATES SITE A
•A. INDICATES DIFF. (C-A)
AVG- 4.5 PPM
AVG- 1.8 PPM
AVG- 2.8 PPM
Figure 18. Carbon monoxide by hour.
324
-------
WINTER: OCT. - MAR
CONC
(PPM)
10.0 -
-1.0
-2.0
-3.0
3-7 PM
HOUR OF DAY
1.0
D INDICATES SITE C
• INDICATES SITE A
A, INDirATFS niFF 1C-AI
AVG= 4.7 PPM
AVG= 3.4 PPM
AVR= 1 a PPM
Figure 19. Carbon monoxide by hour.
325
-------
The cross-freeway difference (C-A) in CO concentration is plotted again
in Figure 20 to contrast the difference in diurnal pattern between weekdays
(Monday through Friday) and weekends (Saturday and Sunday). The primary
difference between the curves lies in the absence of an afternoon rush hour
effect on weekend days.
As previously mentioned, an attempt was made to determine whether
permanent LACS site A provided an adequate reflection of true background
levels of automobile generated pollutants. To accomplish this, a temporary
site (designated site A') was established in an open field at a distance
of 700 feet from the freeway on the upwind side (see Figure 2). A CO analyzer
was operated intermittently at site A' from June through September 1975. The
diurnal patterns in CO concentration for site A, site A' and the difference
(A-A1) are plotted in Figure 21. Overall, the two sites agreed quite well
throughout the course of the day and the apparent differences probably fall
within the range of calibration error.
Nitric oxide is plotted as a function of wind direction in Figure 22.
As with CO, the cross-freeway difference in NO concentration is maximized
when winds are from the Seabreeze mode (S+SW+W). Unlike CO, the NO concen-
tration at site A practically vanishes under Seabreeze conditions, suggesting
that the background level of NO is very small in the Los Angeles basin.
Figure 23 is a similar plot for M?2- Though not thought to be a primary
pollutant, NO2 is apparently formed in the traverse from site A to site C
through a reaction with the NO emitted on the freeway and the 0$ in the
background. The distinct peak in NO^ concentration occurring at wind direc-
tion of 290 is thought to be related to a Southbound on-ramp which is
oriented at that direction and feeds accelerating traffic onto the freeway
just north of the sampling sites.
Ozone concentration as a function of wind direction appears in Figure
24. It appears that background ozone which is present during the daylight
hours is essentially consumed in the traverse from site A to site C, giving
rise to a negative difference curve.
In Figure 25, the summer season diurnal patterns for the cross-freeway
differences (C-A) in the concentrations of NO, M?2 and 0% are plotted on
the same graph. Emissions of NO become apparent at about 5 a.m. and the
difference curve becomes positive at about 7 a.m. with the commencement of
the seabreeze pattern. A positive cross-freeway difference in N02 concen-
tration appears at about 9 a.m. coupled with a corresponding depletion of
03 (negative cross-freeway difference). Apparently, there is not enough 0%
present in the freeway level background to completely react all of the NO
emitted on the freeway within the traverse of air from site A to site C.
LONG TERM TRENDS
To evaluate trends in pollutant concentrations, it is first necessary
to consider what changes may have occurred in the prevailing meteorological
and traffic conditions at the LACS sites.
326
-------
1
5
PARTS PER I
12.0
11.0
10.0
9.0
8.0
7.0
6.0
5.0
4.0
3.0
2.0
1.0
ok
WEEKDAYS vs. WEEKENDS
01 2 3 4 56 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
h H
HOUR OF DAY 3-7 PM
o TOTAL WEEK DAYS
Q TOTAL WEEK ENDS
Figure 20. CO difference (C-A)
327
-------
(SITE A VS. SITE A')
10.0
OONC
III
-2.0
-3X1
O INDICATES SITE A AVQ- 0.9
6 INDICATES SITE A PRIME AVG- 2.1
A INDICATES DIFF. (A-A PRIME) AVQ- -0.2
Figure 21. Background CO.
328
-------
I I t I I I I I I I I I I I I I I I I I I
-0.05 _
-0.10 _
-0.15
Figure 22. NO by wind direction (all data).
329
-------
0.13 _
0.12 _
0.11 _
0.09 -
0.08 _
0.07 -
0.06 _
0.05 -
0.03 _
0.02 _
0.01 -
0.00 -
-0.01
I I I I I I I I I I I I
Figure 23. NO2 by wind direction (all data).
330
-------
0.09 -
0.07 _
0.06 _
0.04 _
0.03 _
0.02 _
0.01
-0.01 _
-0.02 _
-0.03 -
-0.04 _
-0.06 _
-0.07
I I I I I I I I I I I I
Figure 24. O by wind direction (all data).
331
-------
DIURNAL PATTERN - SUMMER MONTHS
U)
0.30
0.25
CONG Q.20
(ppm)
0.15
0.10
0.05
-0.05
-0.10
1 2 3 4 567 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
HOUR OF DAY
Figure 25. Cross-freeway differences (C-A) for NO, NO , and O .
-------
Wind direction and wind speed data collected since the beginning of the
study are summarized? in Table II. The percent occurrence of favorable wind
directions (i.e., Seabreeze mode) and the accompanying average wind speed
are presented separately for the two sample integrating time intervals em-
ployed in the study (i.e., 3-7 p.m. and 0-24 hours). The data are grouped
chronologically on a seasonal basis (summer and winter) since this classifi-
cation seems to provide reasonable intro-group homogeneity. The summer
seasons, while certainly providing the best meteorological conditions for
the monitoring program, exhabit a slight negative trend across the three
years for both sampling intervals. The winter seasons show a great deal
more variability, but, as pointed out in the footnotes, these seasons are
not balanced in terms of data coverage and thus may not be directly comparable.
The very unfavorable meteorology occurring in the winter of 1976, however,
would be expected to have a notable impact on pollutant concentrations ob-
served at that time.
As previously mentioned, on-site traffic count data are not available
prior to September 1976. However, estimates of the annual average daily
traffic volume at the LACS site for each of the last five years have been
constructed through use of publications from the State of California Depart-
ment of Transportation.1 These data appear in Table III. With the exception
of a slight decrease for 1974, the year of the gasoline shortage, traffic
volume at this location along the San Diego Freeway has remained quite stable
over the five year period.
In Figure 26, the trend in the percentage of Los Angeles County cars
equipped with a catalytic converter is presented for two different methods
of estimation. The lower curve was generated by Rockwell Air Monitoring Center
(RAMC)2 from county sales registration data, while the upper curve was
generated by TRW, Inc.3 by adjusting the Rockwell estimates for actual ve-
hicle miles travelled as a function of model year. This latter curve is
considered more representative of the real vehicle mix on the San Diego Free-
way. It indicates that catalyst equipped vehicles accounted for approximately
30% of the vehicle miles travelled in the Los Angeles basin as of December
1976.
Figures 27-29 are bar graphs showing the seasonal averages at each site
and for the cross-freeway difference (C-A) in the concentrations of CO, NO
and NOz. As previously discussed, the summer season data (shaded bars) pro-
vide the best opportunity to assess long-term trends.
As shown in Figure 27, there has been no discernible change in the
background (site A) concentration of CO since the beginning of the study.
The freeway contribution of CO (C-A), however, increased slightly from the
summer of 1974 to the summer of 1975 (possibly in response to the increase
in total traffic volume) and then decreased by about 25% from summer of 1975
to the summer of 1976. This substantial decrease must be attributed to the
phase-in of catalyst equipped vehicles.
Figure 28 shows very little change in the background level, but a sub-
stantial increase in the freeway contribution of NO between the summers of
1975 and 1976. Similarly, N02 (Figure 29) has exhibited a constant background
333
-------
TABLE II. AVERAGE % FREQUENCY FROM S+SW+W AND THE ASSOCIATED
AVERAGE SPEED BY SEASON AND SAMPLING INTERVAL
Season
Sampling Interval
15-19 hr.
0-24 hr.
% WD Avg. % WD Avg.
(S+SW+W) WS, mph (S+SW+W) WS, mph
Summer, 1974
Winter, 1974
Summer, 1975
Winter, 1975°
Summer, 1976
Winter, 1976
Overall Summer
Overall Winter
93.3 5.6 63.6 3.5
69.7 4.5 34.9 3.7
97.2 6.0 62.3 3.6
78.3 4.5 39.5 3.8
95.9 5.7 58.9 3.5
37.0 3.2 28.1 3.0
97.0 5.8 61.3 3.5
64.8 4.2 34.8 3.6
Summer - April thru September
Winter - October thru March
Season began June 1, 1974
'No data available for December 1975 or January 1976
Season ended December 15, 1976
334
-------
TABLE III. SAN DIEGO FREEWAY TRAFFIC VOLUME @ LACS SITE
FIVE YEAR HISTORY
Year
1972
1973
1974
1975
1976
Average Daily Total
181,500
182,500
180,000
182,500
183,000
Percent Change
+ 0.6%
- 1.4%
+ 1.4%
+ 0.3%
Source: State of California Department of Transportation (estimates)
Note: LACS Avg. Daily Total (Sept. - Dec. 1976) = 188,457
335
-------
30-.
tt)
O - REGISTRATION DATA
D - ADJUSTED FOR ESTIMATED MILES DRIVEN
I i- i 8 i i 8
I i I 1-4— 1-
Figure 26. Percent catalyst cars - LA County.
-------
^s
l-l
SAN DIEGO FREEWAY
B C
|— 25FT-J
L— 25FT-*|
ISEPULVEDA
BLVD
D
| 1
CONC
(PPM)
SITE A
DIF iC-A)
SITE C
Figure 27. CO - average concentrations by season.
-------
n
SAN DIEGO FREEWAY
•100FT-
—I
. r
L— 25FT-4
ISEPULVEDA
BLVD |
D
I 1
j
CONC
(PPM)
U)
U)
CO
0.50
0.40I
0.30
SITE A
DIP (C-A)
SITE C
Figure 28. NO - average concentrations by season.
-------
SAN DIEGO FREEWAY
CONC
(PPM)
II B - "' 1 C 1 SEPULVEDA
II I-2-! . « 1 1 BLVD l-'h
1— ' -I 1
N25FT-4 1
0.07
0.06
0.05
0.04
0.03
0.02
0.01
SITE A
SITE C
Figure 29.
- average concentrations by season.
-------
level and a large increase in the freeway contribution. With only two summer
seasons to compare, it is not realistic at this time to attribute these
increases to the function of the catalyst. Ozone data are available only for
1976 and, thus, inferences about trends cannot be drawn at this time.
Figure 30 consists of plots of the monthly average cross-freeway
differences (C-A) in the concentration of all continuously monitored pollu-
tants. The monthly average percent occurrence of favorable wind directions
is plotted at the bottom of the figure for comparison. It is interesting to note
how closely the cross-freeway difference in CO concentration follows the
wind direction pattern. Sulfur dioxide (SOz) bubbler data is shown in the
figure for comparison with total sulfur. Because of the very low SO^ levels
involved, the data from the total sulfur analyzers is considered somewhat
questionable.
CONCLUSIONS
• The LACS site appears to be a very favorable location in terms of
meteorology and traffic flow for determining the freeway contribution
to ambient concentrations of air pollutants. This is especially so
during summer months and between the hours of 3-7 p.m.
0 Traffic volume estimates from the California Department of Transportation
in conjunction with on-site traffic data indicate that there has been
little change in the traffic flow on the San Diego Freeway at the study
site from 1974 through 1976. The diurnal patterns in traffic count and
speed by day of week are very consistent from week to week. As of
December 1976, the total traffic volume passing the study site averaged
188,000 vehicles per day with an estimated 30% of the vehicle miles
travelled attributed to catalyst equipped vehicles.
• The background levels of CO, NO and N02 have remained essentially
constant since 1974.
• The freeway contribution of CO decreased about 25% from 1975 to 1976
while the freeway contribution of NO and N02 increased substantially.
• There is apparently an inadequate amount of background 03 at freeway
level to convert all of the emitted NO to M>2- The reaction of NO with
03 proceeds rapidly enough so that near zero concentrations of 03 are
observed at the near downwind site.
REFERENCES
1. State of California, Department of Transportation, Traffic Volumes on
California State Highways: 1972, 1973, 1974, 1975 and 1976 (perprint).
2. Rockwell Air Monitoring Center, "Determination of Percentage of Diesel
Trucks and Catalyst Equipped Cars," Proceedings of the LACS Symposium,
April 12 and 13, 1977.
3. TRW, Inc., "A Mobile Source Emission Inventory System for light Duty
Vehicles in the South Coast Air Basin."
340
-------
0.07S .
0.050.
0.025 .
03
0.009
0.006.
0.003
n
0.06
0.04
0.02
0
0.42
0.28
0.14
0
6.C .
\
S02
A. r
NO2
NO
D-l
l i i i i i i i i i i
CO
JUN AUG OCT DEC FEB APR JUN AUG OCT DEC FEB APR JUN AUG OCT DEC
1974 1975 1976
100r-
I I I I I I I I I I 1 I I I I I I I I
% WD
(S+SW+W)
Figure 30. Difference C-A for hours 0-24.
341
-------
NITRIC OXIDE, NITROGEN DIOXIDE, AND OZONE INTERRELATIONSHIPS
ACROSS THE FREEWAY
Robert K. Fankhauser
Environmental Protection Agency
Research Triangle Park, North Carolina
on assignment from the
National Oceanic and Atmospheric Administration
Si 1ver Springs, Maryland
ABSTRACT
The Los Angeles Catalyst Study has produced a combination of basic data
that are available over a period of several years. Among the data are mea-
surement of wind, traffic count with traffic speed, ozone, nitric oxide,
nitrogen dioxide, carbon monoxide, and particulate. Many data are available
at sites of the Freeway- This brief study relates nitric oxide, nitrogen
dioxide, and ozone with wind and traffic. All data are not concurrent since
one or another instrument has been added or taken out of service, but signifi-
cant data exist. Certain data verify chamber studies. Plots of simultaneous
values of the hourly measurements show that changes occur as air acquires
pollutants in passing across the Freeway.
The Los Angeles Catalyst Study (LACS) is producing a unique set of data
because of its location, the completeness of the instrumentation and the
duration of operation. Its design is to sample emmisions from heavy traffic.
Its location is such that meteorological conditions are unusually repetitive.
Its operation has been for an unusually long period compared to other air pol-
lution studies.
For this microstudy, data for periods in May and September 1976 were
plotted along with wind speed and direction. Traffic data were available
beginning with September. For sites A and C the data included hourly values
of ozone (O$), nitric oxide (NO), nitrogen dioxide (NO2) as well as wind
speed and direction. The wind equipment was located at site A.
These sites are on opposite sides of the San Diego Freeway with the Free-
way running from approximately SSE to NNW (145° - 325 ) at a point about 0.5
mile north of the Wilshire Boulevard interchange, which is about 6.4 km C4
miles) from the ocean. The ocean-to-land breeze is established fairly early
in the day (around 0900) during May through September and continues until
1800-2100. Wind speeds during the daytime seldom are over 3 meters per second
(mps) or 7 miles per hour (mph) for an hourly average. During the night when
343
-------
the ocean-to-land breeze has subsided, the average hourly speed is usually
less than 1 mps (2-3 mph). At these lighter speeds the direction is not as
constant so that even though the prevailing wind for an hour is judged to be,
as an example, from the N there are usually excursions from the direction to
either side and even to the exact opposite direction. Thus during any one
hour the opposing sites may share the emissions from the Freeway rather than
receiving them exclusively. The lighter wind speeds would also not disperse
the emissions as rapidly.
A large part of the differences in patterns of the pollutant concen-
trations that are seen between site A and site C is due to the different dis-
tances from the traffic. Site C is a little less than 8 m (25 ft.) from the
northbound side of the San Diego Freeway. Sepulveda Boulevard is about the
same distance on the other side of site C so traffic on it (small compared
to Freeway traffic) also contributes pollution. Site A is about 30 m (100 ft.)
away from the southbound edge of the San Diego Freeway (on the side towards
the ocean). Site C is within the path of the turbulent eddies produced by
fast moving northbound traffic while at site A such eddies have become indis-
tinguishable from the general air flow. Of course the distant edge of the
southbound lanes is farther from site C than it is from site A. The general
effect is that atsite A with winds that have a component moving air from A to-
ward C there is little or no pollution from the Freeway. At C however the
Freeway is always a contributor and there is also a small contribution from
Sepulveda Boulevard with wind intiie C to A direction.
It must be remembered that the sites are surrounded by metropolitan Los
Angeles which supplies a background concentration of these same pollutants.
The background concentration rises and falls hourly in somewhat the same pat-
tern as that due to the Freeway since other activity in the city is geared to
the same phenomena which cause the Freeway traffic to change.
The first graphs of data (Figures 1 and 2) are of the wind directions
and speeds for a period in May and another in September. Since the Freeway
at this point lies along the line of approximately 145° to 325 (SSE to NNW)
these directions have been drawn in to show which winds carry air from site A
towards site C and vice versa. Winds vary direction so continuously that for
directions nearly parallel to the Freeway, sites on both sides of the Freeway
are likely to receive some pollution during such a period.
To show the times when the wind has a component from C to A or from A to
C, a bar graph was drawn in the center of Figures 1 and 2. This portion of
Figures 1 and 2 will be used in other figures to show more simply the effect
of direction on concentrations of pollutants related to Freeway traffic.
At site A (see Figures 3 and 4) during the time of the ocean-to-land
breeze (air movement from site A toward site C) the concentration of 0$ rises
to a peak usually between 1200 and 1500. The concentration pattern follows
relatively closely the pattern of wind speed and insolation. The wind speed
is likely to be related to the mixing height. That is, as the sun's heat pro-
duces rising and falling columns of air the stagnant layer at the surface is
"geared in" with the faster moving air aloft and so is carried along at an
increased speed. Air from a layer above the surface is mixed downward so part
344
-------
(-0
C/J
LU
LU
CC
O
UJ
Q
O
X
DC
LU
Q.
C/J
LU
NOTE: AIR MOVEMENT FROM SITE A TOWARD
SITE C OCCURS WITH WINDS FROM SSE
THROUGH S AND W TO NNW AS SHOWN
BY LINES DRAWN AT 145°and 325°.
AIR MOVEMENT FROM SITE A TOWARD SITE C
AIR MOVEMENT FROM SITE C TOWARD SITE A
16
22
Figure 1. Wind data - Los Angeles Catalyst study - San Diego Freeway.
-------
Ch
AIR MOVEMENT FROM
SITE A TOWARD SITE C
OCCURS WITH WINDS
FROM SSETHROUGH
S AND W TO NNW AS
SHOWN BY LINES
C = CALM
M = MISSING
DRAWN AT 145
325°.
AIR MOVEMENT FROM SITE A TOWARD SITE C
AIR MOVEMENT FROM SITE C TOWARD SITE A
12
13
14
15
16
17
18
SEPTEMBER 1976
Figure 2. Wind data - Los Angeles Catalyst study - San Diego Freeway.
-------
ppm
U)
.150-
.100-
.050-
0
.30-
.20-
.10-
0
.10- -
.05-
1 \
OZONE-SITE A
I
AIR MOVEMENT FROM SITE A TOWARD SITE C
I (
AIR MOVEMENT FROM SITE C TOWARD SITE A
NITRIC OXIDE-SITE A
NITROGEN DIOXIDE-SITE A
AA.
16
17
18
19
MAY 1976
20
21
22
Figure 3. Los Angeles Catalyst study - San Diego Freeway.
-------
U)
*>
CD
ppm
.150-
.100
.050 -|
0
1 r
OZONE-SITE A
AIR MOVEMENT FROM SITE A TOWARD SITE C
AIR MOVEMENT FROM SITE C TOWARD SITE A
NITRIC OXIDE-SITE A
NITROGEN DIOXIDE-SITE A
12
SEPTEMBER 1976
Figure 4. Los Angeles Catalyst study - San Diego Freeway.
-------
of the 03 measured at the surface may be from 03 held over from the day before
and part is from precursors emitted currently upwind of the Freeway (1) .
During the late morning and afternoon hours the usual wind direction is from
the ocean toward the Freeway. Since the distance from the ocean to the Free-
way is about 4 miles (6.4 km) and the wind speed 5 to 8 mph, the time enroute
is from around an hour to as little as 30 minutes from the most distant point.
This is a fairly short "cooking time" for a large amount of O3 to be formed,
yet out of the 14 days considered there were 6 which had concentrations over
the national standard of .08 ppm.
With winds from site C to site A the O3 concentration at A is almost al-
ways zero. ^ This is partly because winds from C to A usually occur at night or
early morning when O3 concentrations are generally low. It is also because
the NO from the exhaust gases reacts with O3 and destroys it before it reaches
A. The NO concentration at site A is very dependent on the wind direction.
Unless the wind direction is from C to A there is seldom any NO recorded at
site A which indicates there is very little background NO. N02 concentrations
also vary with wind direction, following somewhat the same pattern as NO.
However there seems to be some background NO2 so the concentration seldom goes
to zero f"0£ long.
At site C (Figures 5 and 6) the concentrations of NO and NO2 seldom go to
zero, but O3 is usually at zero or close to it. The concentrations of NO vary
with traffic volume, and with wind direction, along with other variables which
are not so evident. Changes is atmospheric stability undoubtedly change the
rate of dispersion. Ozone in the air coming across the Freeway reacts with
some of the NO. There always seems to be an excess of NO at site C except per-
haps during the hours of least traffic around 0300 or 0400. The production of
NO2 by the reaction of NO with O3 causes the concentration pattern of NO2 at
site C to follow the concentration pattern of O3 at site A during the times
of winds from A to C. There seems to be a base value of N02 perhaps indicat-
ing the amount emitted directly in the exhaust gases. The peak concentrations
of NO2 during the middle of the day contrasts with our usual concepts in that
the data from our Continuous Air Monitoring Program (CAMP) stations (for in-
stance) show NO2 as highest shortly after the morning peak of traffic declining
to very low afternoon values.
The low traffic volume (Figure 7) at 0300 to 0400 affects the concentra-
tion of NO at both A and C. The wind direction is often from C to A at that
time of the morning (see September 12, 13, 17, 18) . The dip in traffic is
reflected in the dip in NO as measured at site A. Other variations in con-
centration are probably due to differences in wind speed and differences in
diffusion due to changes in air stability and/or changes in traffic speed.
Changes in traffic speed would surely affect the strength of the eddies in and
around the roadway.
These data should be very interesting to modelers even though the study
was not designed with modelling in mind.
The data show that NO and O3 react quickly and increase the concentration
of NO2 as also occurs in the smog chambers. As the emmissions vary the con-
centrations vary. As the wind changes the concentration pattern shifts. The
349
-------
U)
Ul
o
OZONE-SITE C
AIR MOVEMENT FROM SITE A TOWARD SITE C
AIR MOVEMENT FROM SITE C TOWARD SITE A
NITRIC OXIDE-SITE C
NITROGEN DIOXIDE-SITE C
18
19
20
21
22
MAY 1976
Figure 5. Los Angeles Catalyst study - San Diego Freeway.
-------
ppm
OZONE-SITE C
AIR MOVEMENT FROM SITE A TOWARD SITE C
AIR MOVEMENT FROM SITE C TOWARD SITE A
NITRIC OXIDE-SITE C
NITROGEN DIOXIDE-SITE C
*\4 IS
SEPTEMBER 1976
17
18
Figure 6. Los Angeles Catalyst study - San Diego Freeway.
-------
HOURLY TRAFFIC COUNT-SUM OF N-BOUND AND S-BOUND
U)
12
13
14 15
SEPTEMBER 1976
16
Figure 7. Los Angeles Catalyst study - San Diego Freeway.
-------
traffic count shows that the modeler must definitely allow for different em-
ission patterns on weekends even though the mid-day volume is close to that
of weekdays. All this is shown graphically by the accompanying figures.
REFERENCES
1. Coffey, P. E. and W. H. Stasiuk, "Evidence of Atmospheric Transport of
Ozone into Urban Areas." New York State Department of Environmental
Conservation Publication No. ES4-85, April 1974.
353
-------
APPENDIX
Because it is important that those who use data know what it represents,
a brief description is appropriate of the procedure followed in transforming
the analog wind record to a set of numerical values.
Compass Directions read from O to 360 . The convention has been adopted
that O is used only for calm. North is never O , but is always 360 . Figure
A-l illustrates some of the problems in determining the proper figure to use
in describing the wind direction during a particular hour.
As an example, in figures A-l and A-2, during the hour from 01 to 02 the
wind blew generally from the NW for a little over 30 minutes and then shifted
to WSW and stayed in that approximate direction for 20 minutes. The direction
which should be recorded for the hour is NW or 315 . The reason the data
for the rest of the hour is disregarded is explained below.
Consider pollution as being emitted in short puffs. A puff emitted at
time 0100 would move toward the SE for approximately 35 minutes as shown in
figure A-l. Then it would move toward the ENE for 25 minutes and arrive at
point C. (It is simpler to consider that the wind makes an abrupt shift).
The resultant wind direction (287 ) would be shown by the line AC and the re-
sultant wind speed would be 6.7 mph. This would be an appropriate figure
(rounded off to 290 ) to record if we were only interested in the single puff
of pollutant emitted at 0100. Any other puff during the hour, as for instance
one emitted at 0110 shown by the dotted lines would miss the point C entirely.
Usually in air pollution we are interested in places closer to the source of
pollution and also in continuous emissions. If we consider points within
about 2 miles of the source it can be seen that at D pollution would be blow-
ing past that point for 35 minutes and past F for 25 minutes. Any point along
any direction in between would receive pollution only momentarily as the plume
swung past. Therefore only NW or WSW are readily to be considered and NW must
be chosen since it would have received the pollution for a longer period of
time.
The illustrations in figure A-2 show other examples. In the time period
from 00 to 01 the wind is from the NW for less than 20 minutes, but a receptor
along any other radius would have received pollution for a much lesser time,
so NW is the best choice. We must remember that in this case a sampler which
depends on cumulative collection, even if it was in the most favored direction
would have received pollution for only a third of the hour, so a strong source
might appear weak. Pollution from a point source would be weakened by the
changeability of the wind between 02 and 03 even though the wind stayed in SW
sector.
For a line source such as the San Diego Freeway small changes in the wind
354
-------
direction are not as important unless the wind happens to be nearly parallel
to the line source so that the transport across the Freeway is switched from
one side to the other.
As long as the air transport is from the line source toward the sampling
site any wind direction will blanket the receptor downwind of a line source.
For sites close to the line source there is an increase in concentration
of pollutants when the wind becomes nearly parallel to the line, at least if
the line source is effectively infinite in length. With a line source the
lowest concentrations are received at a receptor site which is on a line per-
pendicular to the line source, however up to 45° deviation from the perpend-
icular makes only a small difference. (1)
REFERENCE
1. D. B. Turner, "Line Source Modeling" Proceedings of the Fifth Netting of
the NATO/CCMS Expert Panel on Air Pollution Modeling, June 4-6, 1974,
Roskilde, Denmark. August 1974
355
-------
AB represents a NW wind at 8 mph for 35 minutes
BC " " VVSW " " 8 mph " 25
AC " the resultant wind direction and speed
AB' represents a NW wind at 8 mph for 25 minutes
B'C' " " WSW " " 8 mph " 25 minutes
AC' " the resultant wind direction and speed
Figure A-l. Wind direction schematic.
356
-------
00
Figure A-2. Hourly wind direction.
357
-------
SUMMARY OF LACS INTEGRATED POLLUTANT DATA
Charles E. Rodes and Gary F. Evans
Environmental Protection Agency
Research Triangle Park, North Carolina
ABSTRACT
This paper is a continuation of the discussion of measurements made at
the Los Angeles Catalyst Study site from June 1974 through December 2976. The
continuous measurements resulting- in hourly averages are discussed in a separ-
ate report, whereas this summary is concerned with measurements integrated
over either 4-hour or 24-hour intervals. With the exception of the sulfur
dioxide bubbler, the integrated measurements in this study are generated from
aerosol sample analyses. The results described are for total suspended parti-
culates, lead, sulfates, sulfur dioxide, and other related measurements. The
aerosol sampler methods discussed include the high-volume sampler, 102 mm
membrane sampler, dichotomous sampler, and the 5-stage Andersen cascade impactor
sampler. The sulfur dioxide bubbler and the high-volume samplers are operated
according to Environmental Protection Agency reference method procedures. The
long-term data trends of each of the pollutants are described as well as the
interrelationships of pollutant combinations. In general the data indicate
decreases in the freeway contributions of total suspended particulates and
lead since 1974, but only a very small positive contribution of sulfates.
INTRODUCTION
The samplers operated at the Los Angeles Catalyst Study (LACS) for the
collection of aerosols and sulfur dioxide (SO^) are operated over selected
intervals. The time intervals utilized are either a 24-hour interval from
midnight to midnight, or a 4-hour interval during the evening rush hour from
1500-1900 hours. The aerosol samplers utilize either filters or impaction
plates to collect the aerosol materials, while the SO£ is collected using the
reference West-Gaeke Method.
Specifications for the aerosol samplers used in the LACS are given in
Table I. The four types of samplers are (1) the standard high-volume (hi-vol)
sampler specified in the Federal Register; (2) the membrane sampler (RAC 2349
or equivalent); (3) the dichotomous sampler (ERC Model 200 or equivalent);
and (4) the 5-stage Andersen cascade impactor with back-up filter. Operation
of a variety of methods provides sample collection over a range of aerosol
size, plus information on method comparability.
359
-------
TABLE I. LACS AEROSOL SAMPLER SPECIFICATIONS
FILTER MATERIAL
FILTER SIZE
mm
FLOW/RATE
m /mm
SIZE RANGE
M
HI-VOL
GLASS FIBER
200 X 250
1.4
MEMBRANE
CELLULOSE ESTER
102
0.14
<100ju
Uj
o\
O
DICHOTOMOUS*
TEFLON
37
0.014
CASCADE BACK-UP
CELLULOSE ESTER
102
0.14
* "FINE" CHANNEL ONLY. SAMPLER ALSO HAS "COARSE" AND "TOTAL" CHANNELS,
2.5 - 10jU AND <10ji RESPECTIVELY.
-------
The hi-vol sampler utilizes specially selected acid-washed glass fiber
filters and is designed to collect Total Suspended Particulates (TSP) defined
as less than 100 microns. The membrane sampler was designed to be an aero-
dynamic equivalent of the hi-vol also collecting TSP, but utilizing a membrane
filter substrate at one-tenth the flowrate. The aerosol collection capabil-
ities of this sampler have not been adequately defined, but it is felt this
sampler also does not quantitatively collect particles greater than 30 microns.
The dichotomous samplers used in the LACS are special prototype versions
of the two-channel virtual (non-impaction) sampler recently made commercially
available. These prototype samplers have a third "Total" channel in addition
to the "Coarse" and "Fine" channels. After the prototypes were placed in the
field, it was subsequently determined that the "Coarse" and "Total" channels
suffered losses of the larger particles making these channels unreproducible.
Fortunately the "Fine" channel, which collects particles in the important size
fraction less than 2.5 microns, operated properly. The analysis information
in this report described as "Dichotomous" refers only to this "Fine" particle
fraction.
Five-stage Andersen cascade impactors are used to provide information
about not only the mass size distribution of the TSP aerosols, but also the
distribution by size of selected analytes. The Andersen sampler is essenti-
ally a membrane sampler with a 5-stage impactor added above the filter.
Therefore, the membrane filter acts as a back-up stage collecting particles
less than approximately 0.7 micron.
The integrated samplers at the LACS as previously mentioned operate over
either a 4-hour or 24-hour period. The results discussed in this report were
compiled from June 1974 thru December 1976 and represent a substantial number
of samples collected. Table II lists the irypes of integrated samples, the
number of valid samples collected, and the analyses performed on each sample.
Note that the membrane and dichotomous samples are also analyzed by X-Ray
Fluorescence Spectrometry (XRF) to obtain selected elemental analyses.
The integrated samplers are located at the sites shown in the elevation
in Figure 1. The majority of the samplers are located at Sites A and C, with
only hi-vols 'and membranes operated at Sites B and D, Sites A and B are
typically "Upwind" of the freeway because of the prevailing wind direction,
and Sites C and D are "Downwind." The freeway contribution of a selected
constituent is determined by subtracting the measurement at Site A from the
measurement at Site C. The samples collected at Sites B and D provide spatial
distribution information and, to a lesser degree, freeway contribution infor-
mation when the wind is blowing opposite to the prevailing direction. Note
that Site A is a "background" site unaffected by the freeway during the 1500-
1900 hour period, but cannot be considered as a background site over a 24-hour
period because of shifting winds.
The integrated samplers are operated according to the schedule shown in
Figure 2. Again, the greatest concentration of measurements are made at Sites
A and C, with daily collection of 24-hours hi-vols and 4-hour hi-vols and
membranes. The SO^ bubblers are also operated at Sites A and C on a 24-hour
361
-------
TABLE II. INTEGRATED SAMPLES COLLECTED AT LACS
#
Sampl es
Collected
2280
3420
570
1270
555
470
1710
Val'id
90.4
94.9
88.7
96.2
76.9
81.5
89.3
#
Val id
Samples
2061
3246
506
1222
427
383
1527
Analyses
TSP, Pb, SO*, 1
TSP, Pb, SO^, 1
Same as Hi-Vol
Same as Hi-Vol
Mass, SO^, NH^
Mass, XRF*
SOo
NO", NHj
NO', NH;
plus XRF*
plus XRF*
, H+, XRF*
24-hr Hi Vols
4-hr Hi Vols
24-hr Membrane
4-hr Membrane
24-hr Dichotomous
24-hr Cascade Backup
24-hr S02 Bubbler
*Elemental analyses by XRF include S, Pb, Br, Ca, Si, Fe, Zn, Al, and Cl.
362
-------
MET
T(
10
)WER
r-V*T!_±r« PREVAILING WIND fr
m
L/V/\/Y/\XY/\/\/\/\
SAMPLER INLETS
1m ABOVE SAN DIEGO
X FREEWAY SURFACE \ FREEWAY SURFACE .
M^ cr 1 n ) M fr
| | fl _/- 2m ABOVE GRADE ^ | | SEPULVEDA BLVD Nl
r -. ^ * — 8m— »|
4 30m » •— -- 3tlm k
SITE A
CO ANALYZER
TOTAL SULFUR (S02) ANALYZER
N0/N02 ANALYZER
03ANALYZER
24-HR S02 BUBBLER
24-HR HI-VOL
4-HR HI-VOL (3-7 PM)
4-HR MEMBRANE (3-7 PM)
24-HR CASCADE
MASSIVE VOLUME AIROSOL SAMPLER
24-HR DICHOTOMOUS SAMPLER
AMBIENT TEMP. AND DEWPOINT
WIND SPEED
WIND DIRECTION
SITE B
24-HR HI VOL
4-HR HI-VOL (3-7 PM)
24-HR MEMBRANE
SITEC
CO ANALYZER
TOTAL SULFUR (S02) ANALYZER
NO/NO?ANALYZER
03ANALYZER
24-HR S02 BUBBLER
24 HR HI-VOL
4-HR HI-VOL (3-7 PM)
4-HR MEMBRANE (3-7 PM)
24-HR CASCADE
MASSIVE VOLUME AEROSOL SAMPLER
24-HR DICHOTOMOUS SAMPLER
TRAFFIC SPEED AND COUNT SYSTEM
SITE D
24-HR HI-VOL
4-HR HI-VOL (3-7 PM)
24-HR MEMBRANE
Figure 1. LACS study site composition and elevation.
-------
V)
24-HR HI VOL
4-HR HI-VOL AND
4-HR MEMBRANE*
24-HR CASCADE
DICHOTOMOUS
4-HR HI-VOL*
24-HR HI-VOL
OJI LID MCMDDAMC
/4-nn WltmBnAnit
III 1
;••.-•. :<'.'••' •"••'• '-"HI i'-l'-'v •'•'•'• • " V*. • r"^l •.'.:•.>':. -.'.••'.•v-v.-v •'Tl
:-.•'.••••.:•:•••••: i \ l{-%- '•••••••• •'"'• •*•••-] {£• :'••* ••'••'• •••••vls»». ^v-'.v i •..--. •.-..::•:.•,. ;i
0 E3 E3 EZ3 23 0 EZ3
i i i i i i
,•:.'*" i''; ; • • .7" f v.'.« •<-,-.•..'..:• .; :. ^\ f^v .•..-•.•;-. •;' •;•*';:
— J t
III 1
1ST DAY
*4-HR SAMPLES COLLECTED 3 TO 7 P.M.
2ND DAY
3RD DAY
4TH DAY
5TH DAY
6TH DAY
7TH DAY
Figure 2. LACS platform sampler schedules.
-------
basis. The every-other-day and every-third-day operation schedule for the
other samplers was dictated by the resources available to provide long-term
trend information.
CONCLUSIONS
The results presented here represent only a portion of the integrated
sample analyses collected at the LACS. This paper describes only those results
which are the most significant in terms of the original study objective to
examine long-term trends. _ The results presented include measurements for TSP,
lead (Pb) and sulfate ~
The most important conclusions based on the LACS integrated measurements
through 1976 are as follows, not necessarily in order of importance:
(1) The background levels of TSP and Pb (approximately 85 yg/m3 and 0.5
yg/zn3, respectively) have changed very little since the beginning of
the study.
(2) The background level of SO^ has decreased 25% since 1974.
(3) The freeway contributions during the evening rush hour (1500-1900) of
TSP and Pb have both decreased 25% since 1975, whereas the contri-
bution of 50i| from the freeway is small and has only recently become
measureable above background. The average 1500-1900 hour contribu-
tions of TSP and Pb during the summer of 1976 were 44 yg/m3 and 7
yg/m3, respectively.
(4) The consensus of methods indicates only 0.5 to 1.0 yg/m3 freeway
contribution of SO^ during periods of favorable meteorology.
(5) Artifact sulfate formation on hi-vol samples does not appear to be
significant in determining 24-hour cross- freeway SOi^ differences, but
apparently contributes a substantial error to 4-hour hi-vol sulfate
measurements .
TOTAL SUSPENDED PARTICULATES AND MASS DATA
The data presented in previous LACS quarterly and semi-annual reports
have been shown graphically on a monthly basis. The freeway contribution for
1975 and 1976 of TSP as measured by the hi-vol at Sites A and C from 1500-1900
hrs is shown in Figure 3 on a monthly basis. This graph can be erroneously
interpreted, however, because of the varying meteorology from month-to-month.
Since the wind speed and direction are not uniform on a monthly basis, the
freeway contribution calculation varies, depending on the magnitude of the
resultant wind vector. Examination of the meteorology by month shows a
definite seasonal! ty in the favorable wind direction sector preferred for
determining cross-freeway differences. Table III summarizes the meteorology
by "Summer" and "Winter" (April thru September and October thru March, respec-
tively) seasons, and shows an average of 95.9% favorable wind directions
during the "Summer" (1500-1900 hour interval) . The percentage during the
365
-------
100
JAN MAR MAY JUL SEP NOV JAN MAR MAY JUL SEP NOV
t 1975 »*4-^ 1976 »•
Figure 3. 1500-1900 hi freeway contribution of TSP, hi-vol (C-A).
366
-------
TABLE III. AVERAGE PERCENT FREQUENCY FROM S+SW+W AND THE ASSOCIATED
AVERAGE SPEED BY SEASON AND SAMPLING INTERVAL
Season6
Sampling Interval
Summer, 1974
Winter, 1974
Summer, 1975
Winter, 1975
Summer, 1976
Winter, 1976°
Average Summer
Average Winter
15-19
% WD
(S+SW+W)
96.6
61.9
96.6
70.0
94.7
32.7
95.9
59.3
hr
Avg
WS, mph
5.1
4.2
5.5
4.2
5.2
3.1
5.3
4.0
0-24
% WD
(S+SW+W)
63.6
34.9
62.3
39.5
58.9
28.1
61.4
35.4
hr
Avg
WS, mph
3.5
3.7
3.6
3.8
3.5
3.0
3.5
3.6
Summer - April thru September;
Winter - October thru March
Season began June 1, 1974
"Season ended December 15, 1976
Weighted
367
-------
"Winters" drops to 59.3%. Note especially that even during "Summer" seasons
the 24-hour interval is only slightly more favorable than the "Winter" 1500-
1900 hour period. These less favorable conditions result in a larger upwind
level at Site A and consequently a smaller freeway contribution based on the C
minus A difference over a 24-hour period. The atypically poor meteorology in
the winter of 1976 makes interpretation of results in this period very difficult.
When the data in Figure 3 are averaged on a seasonal basis to reduce the
effect of meteorology, the results are as shown in Figure 4. This plot also
includes 1974 data, and provides a much clearer picture of the long-term
trends in TSP contribution by the freeway. Note the slight increases in the
summer seasons as compared to the winter seasons caused by the more favorable
meteorology. If the summer seasons alone are examined to virtually eliminate
the effect of meteorology, the results are as shown Figure 5. From this plot
it is apparent that the overall freeway contribution has decreased substan-
tially from 1974 to 1976. The total decrease has been approximately 40% of
which 25% has occurred from 1975 to 1976. Membrane samplers, as mentioned
earlier, are also operated at the LACS to provide comparison information with
other methods. The hi-vol and membrane data agree in trend and on the average
differ by only about 10 pg/m3 with the hi-vol giving larger values. It is
important to remember that each of the points on this graph is an average of
150-180 samples and the error bands, if applied to each point would be very
small.
Examination of the "background" particulate trends as measured upwind at
the LACS shows only a slight decrease in Figure 6 from 1974 to 1976 by a
consensus of 4 different methods. The dichotomous sampler "Fine" fraction is
approximately 1/3 of the TSP values of both the hi-vol and membrane. The TSP
background of approximately 90 pg/m3 shows the effect of meteorology by reduc-
ing during the summer seasons the influence of the freeway on the "background"
sites. During the 4-hour period from 1500-1900 hours the more favorable
meteorology removes the freeway influence on the upwind site.
On a 24-hour basis (Figure 7) the freeway contributions of TSP and mass
are approximately 1/3 those during the 1500-1900 hour interval, because of
less favorable meteorology and lower average emission rates. The effect of
less favorable meteorology becomes very pronounced on a 24-hour basis, showing
the summer values to be approximately twice those during the winter. Figure
8 shows this information on a summer-only basis and indicates only a slight
decrease in TSP by the hi-vol and no change in the dichotomous "Fine" fraction.
The spatial distributions of TSP by site and sampling interval are shown
in Figure 9. This graph indicates that the background TSP on either a 4-hour
or 24-hour basis is a substantial fraction of the downwind measurements, and
the difference between Sites A and B is small. The comparison between Sites
C and D, however, shows a substantial difference between sites and sampling
intervals. The decreases from Site C to Site D is caused by both dilution and
large particle deposition of the freeway contribution to the TSP burden.
Utilization of the information from the 24-hour Andersen cascade results
in graphs plotting particle size (diameter) versus the cumulative percent
368
-------
80
70
60
n
5 50
O
GO
V)
40
30
20
10
I
I
HI-VOL (C-A)
MEMBRANE (C-A)
I
_L
SUMMER
1974
WINTER
1974
SUMMER
1975
WINTER
1975
SUMMER
1976
WINTER
1976
SEASON
Figure 4. 1500-1900 hr freeway contribution of suspended particulates.
-------
80
70
60
50
40
co
3 30
20
10
I
I
I
HI-VOL (C-A)
MEMBRANE (C-A)
I
I
SUMMER
1974
WINTER
1974
SUMMER
1975
WINTER
1975
SUMMER
1976
WINTER
1976
SEASON
Figure 5. 1500-1900 hr freeway contribution of suspended particulates (summer seasons only)
-------
U)
120
n 100
Q
O
in
2
80
60
40
20
\ \
4-hr MEMBRANE (A)
I
24-hr MEMBRANE (B)
24-hr H\-VOL (A)
24-hr DICHOTOMOUS (A)
I
I
I
I
SUMMER
1974
WINTER
1974
SUMMER
1975
WINTER
1975
SEASON
SUMMER
1976
WINTER
1976
Figure 6. Background suspended particulate trends.
-------
U)
X)
50
40
30
M
1 20
Z
Q
V)
V)
10
-10
-20
-30
HI-VOL (C-A)
DICHOTOMOUS C-A)
SUMMER WINTER SUMMER WINTER SUMMER WINTER
1974 1974 1975 1975 1976 1976
SEASON
Figure 7. Twenty-four hour freeway contribution of suspended particulates.
-------
Q
O
C/9
50
40
30
20
10
I
I
I
HI-VOL (C-A)
DICHOTOMOUS (C-A) _
SUMMER WINTER SUMMER WINTER SUMMER WINTER
1974 1974 1975 1975 1976 1976
SEASON
Figure 8. Twenty-four hour freeway contribution of suspended particulates (summer seasons only)
-------
Q-
co
160
140
120
100
80
60
40
20
0
-D
24 hr
SITE
A
7\
SITE
B
FREEWAY
SITE
C
SITE
D
Figure 9. Spatial distribution of TSP by sampling interval (summer seasons)
374
-------
less-than-or-egual-to a given aerosol size. Figure 10 is a composite of data
from samples collected at Sites A and C during the summer of 1975. From this
graph it is apparent that the mass is distributed in a log-normal manner at
both sites, with a slightly smaller mass median diameter (MMD) at Site C than
at Site A. Approximately 50% of the TSP is below 1.0 micron and 80% is less
than 2.5 microns.
If the mass median diameters from both sites are plotted by season as in
Figure 11, it is obvious that there is a strong seasonality in median particle
size. The "Summer" seasons have mass median diameters nearly twice those of
the "Winters" due to a combination of factors including particle growth during
periods of increased photochemical activity. Overall the average MMD is
approximately 0.7 microns.
LEAD DATA
Analysis of Pb data from the aerosol samples collected at the background
sites at the LACS shows (see Figure 12) a very strong influence of the freeway
during the winter months. This again is the result of unfavorable meteorology.
The consensus of methods during the summer months is a background level of
less than 1.0 ug/m3.
The 24-hour freeway contribution shown in Figure 13 shows a seasonal
meteorology effect similar to that of TSP. The "Summer" season values are
approximately 3-5 times as great as the "Winter" levels. Examination of the
"Summer" seasons in Figure 14 shows that there has been a significant decrease
in TSP lead contribution from the freeway as measured by the hi-vol, even as
examined on a 24-hour basis. The dichotomous and cascade back-up samples,
however, showed only a slight decrease. By sampler type the order of decreas-
ing Pb collected is hi-vol, dichotomous "Fine," and cascade backup. Even
though the size of the dichotomous "Fine" fraction (<2.5 microns) is much
larger than the cascade-backup fraction (<0.7 micron), there is only a slight
difference between amount of Pb collected in each fraction. Further examina-
tion of the size distribution of Pb is made in Figure 15 by plotting the
results of the summer season Andersen cascade samples composited and analyzed
by stage. Note that unlike mass, Pb is not log-normally distributed and shows
a strong site dependence. As would be expected the downwind MMD is substant-
ially smaller than the upwind MMD. Also the distribution is somewhat bimodal
with a significant fraction of "large" (>2.5 microns) particle Pb, very little
between 1.0 and 2.5 microns, and 75% less than 1.0 micron.
If the 1500-1900 hour freeway contribution of Pb is examined in Figure
16, it is noted that only a small seasonal effect is apparent with an average
contribution of approximately 7.0 yg/m3 Pb above the background. Examination
of the summer seasons in Figure 17 shows a marked decrease from 1975 to 1976
of approximately 25%. Even though only two points are shown for each line, a
substantial number of data points are averaged to obtain each seasonal value.
The cause of the 25% decrease in Pb was investigated by determining the
fraction Pb in the TSP collected. Since the TSP had also decreased, it was
felt that the Pb decrease may have caused the decrease in TSP. Figure 18
375
-------
10.0
8.0
6,0
4.0
« 2.0
E
1.0
0.8
SUMMER, 1975
m SITE A
• SiTEC
ec
-------
U)
VI
1.2
i.o
0.8
cc
|S 0.6
m
2
0.4
0.2
I
I
I
SUMMER WINTER SUMMER WINTER SUMMER WINTER
1974 1974 1975 1975 1976 1976
SEASON
Figure 11. Comparison of 24-hr mass median diameters.
-------
Co
•Nl
00
E 4
5
_ HI-VOL (B)
MEMBRANE (A)
I
SUMMER WINTER SUMMER WINTER SUMMER WINTER
1974 1974 1975 1975 1976 1976
SEASON
Figure 12. Background lead trends (1500-1900 hr data).
-------
CASCADE BACK-UP (C-A)
DICHOTOMOUS (C-A)
SUMMER
1974
WINTER
1974
SUMMER
1975
WINTER
1975
SEASON
SUMMER
1976
WINTER
1976
Figure 13. Twenty-four hour freeway contribution of lead.
-------
10.0
8.0
6.0
4.0
o
b
I
.- 2.0
oc
UJ
I-
111
5 1.0
m
O
H
DC
0.8
0.6
0.4
0.2
0.1
_ SUMMER 1975
SITE A Pb
SITE C Pb
1 2 5 10 20304050607080 90 95 9899
MASS < PARTICLE DIAMETER, cumulative percent
99.9
Figure 15. Twenty-four hour cascade composites - Pb.
380
-------
Uj
oo
£ 1
Q
01
-i 0
-1
-2
I
DICHOTOMOUS (C-A)
CASCADE BACK-UP (C-A)
I
SUMMER WINTER SUMMER WINTER SUMMER WINTER
1974 1974 1975 1975 1976 1976
SEASON
Figure 14. Freeway contribution of lead (summer seasons only).
-------
U)
00
N)
20
18
16
14
ME 12
f
I 10
Q
UJ
-i 8
I
I
I
I
MEMBRANE (C-A)
I
HI-VOL (C-A)
I
SUMMER WINTER SUMMER WINTER SUMMER WINTER
1974 1974 1975 1975 1976 1976
SEASON
Figure 16. 1500-1900 hr freeway contribution of lead.
-------
U)
CD
u>
20
18
16
14
« 12
10
0
<
SUMMER
1974
I
MEMBRANE (C-A)
HI-VOL (C-A)
WINTER
1974
SUMMER
1975
WINTER
1975
SUMMER
1976
WINTER
1976
SEASON
Figure 17. 1500-1900 hr freeway contribution of Pb (summer seasons only)
-------
Co
09
o.
CO
.Q
Q.
<
DC
50
45
40
35
30
25
20
15
10
5
0
CASCADE BACK-UP (C-A)
_L
4 hr MEMBRANE (C-A)
4hrHI-VOL (C-A)
24hrHI-VOL(C-A)
SUMMER
1974
WINTER
1974
SUMMER
1975
WiNTER
1975
SEASON
SUMMER
1976
WINTER
1976
Figure 18. Pb as fraction of TSP (freeway contribution).
-------
shows that the hi-vol and the cascade back-up samples have shown essentially
no change in the ratio of Pb to TSP since 1974. The disagreement of the 4-
hour membrane reflects the poorer precision (factor of 3) of the 4-hour
membrane in measuring TSP as compared to the other methods. The consensus of
methods in Figure 18 is that both TSP and Pb have apparently decreased. This
interrelationship will be examined in greater detail later in this paper.
The spatial distribution of "Summer" season Pb is shown in Figure 19.
Unlike TSP, Pb is very strongly influenced by sampling interval because of
meteorology and the relatively low background levels. Site B is significantly
higher than Site A, even on a 4-hour basis. An unexpected observation is that
the average downwind Pb level at Site C is essentially the same for both the
4-hour and the 24-hour bases. The C minus A differences are greater for the
4-hour interval because of the low background levels. The decrease of Pb from
Site C to Site D is greater on a 24-hour basis because of less favorable
meteorology.
Comparison of the spatial distribution of 4-hour TSP and Pb in Figure 20
clearly shows the large TSP background as compared to the small Pb background.
Note that TSP decreases more rapidly from Site C to Site D than does lead.
This is probably due to the more rapid dilution of the fine particle lead and
fallout of large particle lead. Because of the very small mass median dia-
meter of Pb particles, they apparently behave much like a gas, and conse-
quently correlate extremely well with measurements of carbon monoxide at Sites
A and C.
The relation of lead and bromine should be well defined, since the
principal lead emission from non-catalyst vehicles is lead bromochloride
(PbBrCl). Figure 21 shows that the two measurements correlate very well
(coefficient of 0.99) and have a mass ratio of lead to bromine of 2.3, which
is approximately stoichiometric.
SULFATE DATA
Sulfate (SO^) measurements are the most important analyses performed on
LACS aerosol samples. Since a reliable ambient sulfuric acid aerosol sampler
is not available, the measurement of water soluable sulfates provides a gross
indication of the sulfur emissions as sulfates or sulfuric acid from the
automobile. Based on the very low sulfate emission rates of non-catalyst
cars, it was expected that during the first year after introduction of the
catalyst, the freeway sulfate contributions would be small. This small
difference (<0.5 yg/m3; unfortunately was overlaid on a background sulfate
level as shown in Figure 22 that was much higher than expected. This plot
shows that the initial sulfate levels by hi-vol and membrane were in the range
of 12 to 16 ygr/m3, especially during the "Summer" seasons. The seasonal
changes in sulfate are not due to freeway influence, but increases in the
sulfate levels in the summer are due primarily to increased photochemical
activity in the LA basin. Examination of the "Summer" season background in
Figure 23 shows a substantial decrease in background sulfates as a consensus
by all methods. This decrease from 1974 to 1976 of approximately 25% is
probably not related to automotive emissions, since these constitute only a
385
-------
8
i6
1 5
Q
LLJ 4
3
2
24~hr
SITE
A
m
FREEWAY
SITE
B
SITE
C
SITE
D
Figure 19. Spatial distribution of lead by sampling interval (summer seasons)
386
-------
CO
O)
3
Q.
Q
HI
1
0
SITE
A
4hrHI-VOLPb
4hrHI-VOLTSP
SITE
B
FREEWAY
SITE
C
160
140
120
100
80
60
40
20
0
SITE
D
O)
a.
h
Q.
CO
Figure 20. Spatial distribution of 1500-1900 hr TSP and lead (summer seasons).
387
-------
UJ
00
CO
10
en
a.
_: 7
h.
CQ
LU
o
CC
CQ
CC
O
3
Q.
Q
<
ill
0
LEAD
R = 0.99
Pb
/Br = 2.3
BROMINE
I
SUMMER
1974
WINTER
1974
SUMMER
1975
WINTER
1975
SUMMER
1976
WINTER
1976
SEASON
Figure 21. 1500-1900 hr freeway contributions of lead and bromine.
-------
22
20
18
16
14
n
o* 12
oo
HI
<
10
24 hr HI-VOL (A)
24 hr MEMBRANE (B)
24 hr DICHOTOMOUS "FINE" (A)
24 hr CASCADE BACK-UP (A)
-SUMMER WINTER SUMMER WINTER SUMMER WINTER
- 1974 4i 1975 H« 1976
SEASON
Figure 22. Background sulfate trends.
-------
very small fraction of the total sulfur emitted daily into the ambient air in
Los Angeles. The proportions of sulfate collected by the various types of
samplers compared to the sizes of the aerosol fractions are consistent with
the distribution of SO^ by size collected by the Andersen sampler. In general
the higher sulfate values for the hi-vol as compared to the membrane suggest
the possibility of artifact sulfate formation from S02 conversion, however,
the previous agreements for TSP and Pb between these various types of samplers
have also been less than ideal.
The dashed line in Figure 23 represents sulfate data obtained from the
Southern California Air Pollution Control District at a site located less than
a mile from the LACS sites. These data using a 24-hour hi-vol agree very well
with our hi-vol background data. Comparison in Figure 24 of the monthly
sulfate data from the two monitoring programs indicates the very favorable
comparison, even with the expected spatial differences in sulfate levels.
The size distribution of sulfate particles on a 24-hour summer season
basis is shown in Figure 25. These plots again show that TSP sulfate is not
log-normally distributed, and the size distribution upwind and downwind is
essentially the same. It is interesting to note, however, contrary to the
originally assumption that the sulfate would be found only in very small
particles, there is a significant fraction greater than 1.0 micron. According
to the plots approximately 30% is greater than 1.0 micron and 2% greater than
2.5 microns.
The freeway contribution of sulfate on a 24-hour basis as measured by a
variety of samplers is shown in Figure 26. Notice that only by expanding the
sulfate scale can the consensus difference of approximately 0.5 yg/m^ in the
summer of 1976 be demonstrated. Because the C minus A differences showed such
a small contribution from the freeway, the difference in hi-vols between Sites
D and B was also computed (even though B is not a true background) to examine
the negative sulfate differences in 1975. The "Summer" season averages shown
in Figure 27 indicate that even though the dichotomous "Fine" fraction and
cascade back-up sulfate differences were positive, the freeway contribution of
both sets of hi-vols was inexplicably negative in the "Summer" of 1975.
The previous sulfate contribution data presented in the LACS semi-annual
reports were based only on hi-vol measurements. The 4-hour freeway contribu-
tion of sulfate shown in Figure 28 not only gives the hi-vol results but adds
the 4-hour membrane differences. This graph shows a substantial difference in
both absolute level and trend between the two methods. Whereas the 4-hour hi-
vol has indicated as much as 5.2 yg/m3 sulfate contribution of sulfate, the 4-
hour membrane is averaging only 1.0 yg/m3. The summer season plot in Figure
29 also shows the nearly four-fold difference between the hi-vol and membrane
results. Again the problem of artifact formation on glass fiber substrates,
appears to be relevant. The ratios of SO^ to TSP for several data sets are
shown in Figure 30. These plots dramatically indicate that the proportion of
sulfate collected on the 4-hour hi-vols is significantly greater than the
other methods, suggesting artifact sulfate formed by conversion of sulfur
dioxide. However, the fraction of artifact sulfate formed in the 24-hour hi-
vol appears to be an insignificant portion of the total.
390
-------
U)
VO
20
18
16
14
12
n
8* 10
D
oo
24-hr HI-VOL (A)
24 hr MEMBRANE (B)
SCAPCD DATA
(24-hr HI-VOL)
24-hr DICHOTOMOUS
24 hr CASCADE BACK UP
I
I
I
I
SUMMER
1974
WINTER
1974
SUMMER
1975
WINTER
1975
SEASON
SUMMER
1976
WINTER
1976
Figure 23. Background sultate trends (summer seasons only).
-------
- SOUTHERN CALIFORNIA
AIR POLLUTION
CONTROL DISTRICT
(WESTWOOD)
O LACS (A) LOS ANGELES
CATALYST STUDY
JAN MAR MAY JUL
L^ 1975-
SEP NOV JAN MAR MAY JUL
* 1976
SEP
NOV
Figure 24. Comparison of background sulfate data - LACS/SCAPCD (24-hr hi-vols)
392
-------
10.0
8.0
6.0
4.0
I
I 2.0
of
UJ
5 0.8
UJ
O 0.6
cc
2 0.4
0.2
™ SUMMER 1975
0.1
SITE A SO4
SITE C SO4
1 2 5 10 20304050607080 90 95 9899 99.9
MASS < PARTICLE DIAMETER, cumulative percent
Figure 25. Twenty-four hour cascade composites - SO. ,
393
-------
E
1
O*
J2 0
CO
-1
-2
HI-VOL (C-A)
HI-VOL (D-B)
o —
ICHOTOMOUS (C-A
CASCADE BACK-UP (C-A)
SUMMER
1974
WINTER
1974
SUMMER
1975
WINTER
1975
SEASON
SUMMER
1976
WINTER
1976
Figure 26. Twenty-four hour freeway contribution of sulfate.
-------
U)
<£>
01
UJ
D
CO
-1
-2
CASCADE BACK-UP (C-A)
HI-VOL (D-B)
SUMMER
1974
WINTER
1974
SUMMER WINTER SUMMER WINTER
1975 1975 1976 1976
SEASON
Figure 27. Twenty-four hour freeway contribution of sulfate (summer seasons only)
-------
U)
VO
E
f
O*
c/3
<
LL
t/5
HI-VOL (C-A)
MEMBRANE (C-A)
-1
-2
-3
I
SUMMER WINTER SUMMER WINTER SUMMER WINTER
1974 1974 1975 1975 1976 1976
SEASON
Figure 28. 1500-1900 hr freeway contribution of sulfate.
-------
to
*o
VI
o* 1
co '
LLJ
D
CO
-1
-2
-3
HI-VOL (C-A).
MEMBRANE (C-A)
I
SUMMER WINTER SUMMER WINTER SUMMER WINTER
1974 1974 1975 1975 1976 1976
SEASON
Figure 29. 1500-1900 hr freeway contribution of sulfate (summer seasons only)
-------
03
S?
a."
GO
o
00
<
DC
10 -
8 -
6 -
4 —
1
0
24 hr HI-VOL (C-A)
4 hr MEMBRANE (C-A)
SUMMER
1974
WINTER
1974
SUMMER
1975
WINTER
1975
SUMMER
1976
WINTER
1976
SEASON
Figure 30. SO as fraction of TSP (freeway contribution).
-------
The sulfur contents of gasolines in the Southern California area since
1974 are shown in Figure 31 where the average weight percent sulfur in leaded
and unleaded regular gasolines is weighted by consumption. Unleaded gasoline
contains approximately one-half the sulfur of the leaded gasoline. Also, it
is apparent that there has been no significant increase or decrease in the
sulfur contents since 1974. Since the majority of sulfur in gasoline is
emitted as sulfur dioxide from automobiles, the freeway contribution of S(?2 is
shown in Figure 32. This graph shows a seasonal pattern because of meteorology,
even though sulfur in the fuel is not seasonal. Overall there appears to be a
slight decrease in SC>2 from 1974 to 1976, if only the summer_seasons SC>2
values are considered. Comparison of the 24-hour S02 and soil levels obtained
from Site A plotted in Figure 33 shows the freeway influence on SO^, and the
seasonal change in S0i+ resulting from the increased photochemical conversion
of S02 to 50j+ in the "Summers."
To further examine the discrepancy between the 4-hour hi-vol and the
other methods for sulfate, led to comparison with other parameters such as
S02t total sulfur, temperature., and relative humidity. The only parameter
which, so far shows a significant correlation is relative humidity. Figure 34
shows the hi-vol C minus A SOi+ difference and the background 4-hour hi-vol
SO^ as compared with the relative humidity as measured at Site A. The peak in
the summer of 1975 and subsequent decline of all three measurements tends to
support the contentions in the literature that humidity is an important
factor. In order for the sulfate difference measurements to be affected by
artifact formation, there must be a difference in conditions at the two sites.
There is certainly more SO^ at Site C than at Site A, however preliminary
comparisons show little correlation of the measured sulfate contribution with
the SC>2 level. Because automobiles emit water vapor and heat, there could be
a slight increase in relative humidity and temperature downwind, and measure-
ments to quantify these expectations are being made.
One problem area which became apparent from examining freeway S0i+ contribu-
tion information is shown in Figure 35. Comparison of SQi+ by the Methyl
Thymol Blue (MTB) method to sulfate from XRF sulfur measurements was very good
(correlation coefficient - 0.92) for cellulose ester membrane filters, and no
problem was foreseen with the Teflon dichotomous filters. As can be seen the
two methods agree in Figure 35 within 1.0 yg/m3, Jbut do not indicate the same
trend'in freeway contribution. Since each of the points represents an average
of many analyses, it is felt that the difference is real.
Because X-ray Fluorescence is a surface measurement technique as compared
to the MTB method which extracts the entire filter, one could expect XRF
measurements to be slightly lower if there was a significant amount of material
embedded in the filter. In this case, however, the XRF differences are
greater and this rationale apparently does not apply. The most probable
explanation is that the MTB method is not as sensitive as XRF in this range (<
1.0 pg/m3;.
RELATED DATA
Because the most prevalent sulfate compounds in the ambient air are
probably ammonium salts, aerosol samples are also analyzed for ammonium ion
399
-------
o
o
ss
I-
T
LU
5
of
D
CO
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
LEADED REGULAR
UNLEADED REGULAR
SUMMER WINTER SUMMER WINTER SUMMER WINTER
1974 1974 1975 1975 1976 1976
SEASON
Figure 31. Sulfur content of gasoline in Southern California (major brands weighted by consumption)
-------
20
18
16
-------
CO
j:
5
20
18
16
14
12
8 10
cc
o
o" 8
VI
Hi-VOL SO4~(A)
I
I
SUMMER WINTER SUMMER WINTER SUMMER WINTER
1974 1974 1975 1975 1976 1976
SEASON
Figure 33. Comparison of 24-hr SO and hi-vol SO
-------
18 10
17 9
16 8
15 7
.£ 14 6
3
T
^" UJ *|o til K
O _i I0 _i °
""" O O
IU CO CO
< 12 4
_i
D
11 3
10 2
HI-VOL (A)
90
RELATIVE
HUMIDITY
(A)
HI-VOL (C-A)
SUMMER WINTER SUMMER WINTER SUMMER WINTER
1974 1974 1975 1975 1976 1976
SEASON
80
70
D
X
HI
60 <
LLI
cc
50
40
Figure 34. Relation of 4-hr hi-vol sulfate to relative humidity.
-------
o
Ift.
.E
a.
O
S2 0
<
LL.
-I
D
-1
-2
SO4~FROM XRF S (C-A)
SO4"BY MTB (C-A)
SUMMER WINTER SUMMER WINTER SUMMER WINTER
1974 1974 1975 1975 1976 1976
SEASON
Figure 35. Freeway contribution of sulfate - dichotomous samples.
-------
The data in Figure 36 shows_ that both the hi-vol and membrane exhibit
a small freeway contribution of NHi+, even though the trends do not agree in
the summer of 1976, and the measured hi-vol NH^ levels are known to be erron-
eously low because of sample degradation. The size distributions of NH* at
both sites are shown in Figure 37, and indicate that like SO^ there is little
difference between sites and the relations are not log-normal. If the sulfate
and ammonium distributions for each site are plotted together as in Figures 38
and 39, it is obvious that below 1.0 micron the two species are in the same
size fraction. Above this size they apparently are not related to one another,
but must be combined with other anions and cations. This would also be true
for as much as 20% of the aerosol collected by the less than 2.5 micron
dichotomous "Fine" fraction, but none on the cascade back-up filters.
Nitrate (NO$) analyses have been performed on aerosol samples in all size
ranges, but there appears to be no significant contribution of nitrate from
the freeway. The size distribution of NO^ in Figure 40, however, suggests
that there is a difference in aerosol size across the freeway. There appears
to be agglomeration of NO$ particles across the freeway, perhaps due to react-
ion with emission products.
In 1975 dustfall buckets, which were added at all sites collect non-
suspended (settleable) particulates greater than 100 microns. Compositing the
summer season results for mass in Figure 41 shows the comparison with hi-vol
TSP. Both show an increase across the freeway, although dustfall is more
severely influenced by proximity to the freeway than is the hi-vol. Note also
that the dustfall drops off from Site C to Site D faster than does the TSP.
Analysis of the dustfall samples for Pb gave the interesting conclusion shown
in Figure 42 that hi-vol Pb correlates spatially very well with dustfall lead
near a freeway. This relation may be useful in providing estimations of total
ambient lead burdens.
DISCUSSION
It is apparent from the data presented in this report that use of the
catalyst is reducing emissions to the ambient air. The freeway contributions
of TSP and Pb have both decreased since the beginning of the study. The
nearly constant ratio of Pb to TSP since the beginning of the study was puzzl-
ing until the information shown in Table IV was obtained from Dr. Ronald Bradow
of the Mobile Sources Emissions Research Branch (EPA). These numbers repre-
sent typical emission rates from catalyst and non-catalyst cars operated in
driving cycles similar to Los Angeles Freeway conditions. For a non-catalyst
equipped (pre-1975) vehicle, the typical Pb emission rate of 40 mg/mi is 50%
of the total mass emitted. The balance of the emissions are primarily carbon-
aceous materials and fuel additives such as bromine. By contrast the 20 mg/mi
emissions from a catalyst equipped auto consist almost entirely of a sulfuric
acid aerosol in equilibrium with water. The carbonaceous aerosol material of
the non-catalyst auto is converted to gaseous forms, by the catalyst, and Pb
and its related additives are omitted from the fuel. Assuming the catalyst to
total vehicle mix based on estimated vehicle miles travelled is 30%, the
estimated mass emission rate would be 62 mg/mi compared to a lead emission
rate of 28 mg/mi, resulting in a ratio of Pb/mass of 45%. This number is not
405
-------
_- 2
I
z
5: 1
z
o
MEMBRANE (C-A)
H\-VOL (C-A)
-1
-2
-3
I
SUMMER WINTER SUMMER WINTER SUMMER WINTER
1974 1974 1975 1975 1976 1976
SEASON
Figure 36. 1500-1900 hr freeway contribution of ammonium.
-------
10.0
8.0
6.0
4.0
2
u
1 2.0
•k
CC
LU
LU
| 1.0
£ 0.8
_i
2 0.6
cc
£ 0.4
0.2
0.1
SUMMER 1975
SITE A IMH4
SITE C NH4
1 2 5 10 20304050607080 90 95 9899 99.9
MASS < PARTICLE DIAMETER, cumulative percent
Figure 37. Twenty-four hour cascade composites - NH .
407
-------
10.0
8.0
6.0
4.0
c
E
I 2'°
*
QC
LU
LU
1.0
Q 0.8
LLI
o 0.6
cc
< 0.4
0.2
0.1
SUMMER 1975
SITE A NH4
SITEASO4 ~
1 2 5 10 20304050607080 90 95 98 99 99.9
MASS < PARTICLE DIAMETER, cumulative percent
Figure 38. Twenty-four hour cascade composites - NH /S0~ upwind.
408
-------
10.0
8.0
6.0
4.0
I I TTT
_ SUMMER 1975
in
O
1 2.0
*
DC
UJ
UJ
§1.0
5 0.8
UJ
o 0.6
h-
cc.
£ 0.4
0.2
0.1
"i n n H
SITE C NH4
SITE C SO4 ~
1 2 5 10 20304050607080 90 95 9899 99.9
MASS < PARTICLE DIAMETER, cumulative percent
Figure 39. Twenty-four hour cascade composites - NH /SO ~ downwind.
409
-------
10.0
8.0
6.0
4.0
_ SUMMER 1975
v>
I 2.0
CC
LU
H
LU
1.0
Q 0.8
LU
0.6
< 0.4
0.2
0.1
SITE C NOs
SITE A NO3
I I I I I I I I I I I I I I I
1 2 5 10 20304050607080 90 95 9899
MASS < PARTICLE DIAMETER, cumulative percent
99.9
Figure 40. Twenty-four hour cascade composites - NO
410
-------
7
£
c
1 6
CM
E
1 5
co"
CO
I 4
_l
if 3
co
D
Q
2
1
o
C
E
-
-
—
-
IB
—
3 DUSTFALL MASS
3 24hrHI-VOLTSP "
•i
••
I
//
'/,
//
I
mm
I
//
/i
/ /
'/I
•^p"
//
//
%
//
'/.
//
\
'/i
/,
/i
\
//
mm
-
t
— i
•••
I
/ /
/i
'/i
I
_
—
120
110
CO
C
100 ~|
.
Q.
CO
90 ^
O
80 x
70
60
50
40
FREEWAY
SITE SITE SITE SITE
A B CD
Figure 41. Summer season dustfall - mass.
-------
N>
12
11
1
"| 10
_r
g 9
Q
ai
i
j 8
<
1 1
Uw
Q
6
5
4
[
"E
—
—
"•
-
D
3
•B
DUSTFALL LEAD
24hrHI-VOLLEAD
n
'/f
SB^
7\
'/''
/ /
\
I l_
MB
^pv
/
'//
//
//
//
' ,, /
• ;
////
/ /
* /
^M
"*• _
—
I
//,
','t
\
-
SITE SITE SITE SITE
A
8
7
00
6 S
o.
5 Q
LU
4 i
3 X
2
1
B CD
Figure 42. Summer season dustfall - lead.
-------
TABLE IV. PB AND SO, AUTOMOBILE EMISSION TRENDS
4
MG/MI
MASS. Ea Sffij PB/MASS SOJJ/MASS
NON-CATALYST 80 40 <1 50 ^0
CATALYST 20(15) M3 10 ^0 50(67)
30% CATALYST 62 28 3 45 5
-------
substantially different than the initial Pb/mass ratio of 50%, and explains
why the ratio should have remained essentially constant.
A similar analysis of the SO^ contribution from a fleet of 30% catalyst
cars shows that an estimated 5% of the mass should be SO^. This number is
reasonably consistent with the ratios shown in Figure 29 of 2-3%. Because the
sulfuric acid/water aerosol jsmitted from catalyst cars remains as sulfuric
acid during analyses for SO^ when performing the emission measurements, the
equivalent SO^ reported is typically one-half of the mass measured. The
balance is accounted for by the water in equilibrium. In that the H^SOi^
aerosol is very reactive and probably has degraded into a sulfate form by the
time an ambient sample is analyzed, the numbers in parenthesis in Table IV
represent the equivalent masses if the 10 mg/mi of SOi^ was combined as (NH^) 2
The measured decrease in mass emissions from the freeway was unexpected
in that this was not a design function of the catalyst. Any beneficial
effect from reduction of the total fine particle aerosol emissions must be
considered carefully, as the composition of the reduced emission is substant-
ially different (i.e., lead has decreased but sulfate (H2SO^) has increased.
ACKNOWLEDGMENTS
The authors wish to acknowledge Franz J. Burmann and Gerald G. Akland for
their assistance in reviewing the data and interpreting the results.
414
-------
STATISTICAL ANALYSIS OF THE LOS ANGELES CATALYST STUDY DATA -
RATIONALE AND FINDINGS
George C. Tiao and Steven C. Hillmer
University of Wisconsin
Madison, Wisconsin
ABSTRACT
This discussion presents an analysis of pollutant data gathered in the
Los Angeles Catalyst Study. The ultimate purpose is to use the data to deter-
mine any increase or decrease in the observed emissions since the introduction
of the catalytic converter. An empirical model describes selected observed
pollutant concentrations as a function of wind speed, wind direction, traffic
counts, and the speed of traffic. Therefore, the analysis takes into account
varying traffic patterns and meteorological effects upon the observed pollutant
concentrations. The empirical model helps determine any trends in the data
after adjusting for changes in the traffic and meteorological conditions.
INTRODUCTION
This paper presents a statistical analysis of the Los Angeles Catalyst
Study (LACS) data covering the period June 1974 through November 1976. Our
main objective is to assess the effects on the atmospheric concentration of
various pollutants due to the introduction of the catalytic converter.
The catalytic converter was adopted by American automobile manufacturers
and used on new cars since the 1975 model year. It was designed to reduce the
emission of carbon monoxide (CO) and hydrocarbons (EC). Also, new cars
equipped with the converter must run on unleaded gasoline. Its use should
lead to reduction in lead (Pb) emissions. However, tests have shown (1) that
the converter increases the emissions of sulfuric acid which reacts in the
atmosphere to form sulfate (SO^) . To study the environmental impact of the
catalyst, four air monitoring sites (A,B,C and D) , two on each side of the San
Diego Freeway (Figure 1) , were established by the • Environmental Monitoring and
Support Laboratory (EMSL) of the Environmental Protection Agency (EPA). Data
on a number of pollutants including SOi+, CO and Pb and meteorological variables
such as wind speed and wind direction have been collected since June 1974 (2) .
Also, traffic counts and speeds have been measured since September 1976.
This paper reports our preliminary findings on the following: (1) daily
4 hour afternoon (3-7 p.m.) SOi+ readings (from two different types of sampling
equipment (the high-volume (hi-vol) and the membrane samplers); (2) 24 hour
hi-vol S0t+ readings; (3) hourly CO readings; and (4) the 4-hour afternoon hi-
vol Pb readings. Hourly wind speed and wind direction data for the entire
415
-------
ws,
A Site D
Site A
Figure 1. Simplified map of the air monitoring sites.
416
-------
period, and the available hourly traffic counts and traffic speed for the
period September 1976 through November 1976 will be used in our analysis.
In the discussion of analysis of SOi+ data, a preliminary examination of
4 hour hi-vol, 4 hour membrane and 24 hour hi-vol S0i+ data is presented. The
next discussion concerns CO data and, in particular, presents an empirical-
mechanistic model relating the CO readings at site C to wind speed, wind
direction, and traffic counts and speeds. The model is used to assess the
trend in CO emissions since June 1974. Next, the 4 hour hi-vol Pb data are
analyzed. In particular, the model developed for CO is extended to the Pb
data for assessing the trend.
PRINCIPAL FINDINGS
Our principal findings are as follows:
1.
(i) The background level (from sources other than the freeway), of
decreased from 1974 to 1976.
(ii) The across the freeway difference of the 4 hour hi-volume read-
ings decreased from 1975 to 1976.
(Hi) The across the freeway differences for both the 4 hour membrane
and the 24 hour hi-volume readings increased from 1975 to 1976.
(iv) The 4 hour hi-volume data appear to be measuring something
different than the 4 hour membrane data. Further research seems
necessary to determine the cause of this discrepancy.
2. CO
(i) There appears to have been a decrease in CO emissions from 1975
to 1976.
3. Pb_
(i) There was a steady decrease in Pb emissions from 1974 to 1976.
(ii) At site C the average 3-7 p.m. concentration level is about 60%
higher on the weekends than that for the weekdays.
ANALYSIS OF S0t+ DATA
Data on SO^ consist mainly of daily 4 hour afternoon (3-7 p.m.) readings
from the hi-vol sampler and the membrane sampler, and the daily 24 hour hi-vol
readings. The 4 hour readings are of particular interest since, during the
afternoon hours, the wind is usually blowing in a direction roughly perpendi-
cular to the freeway (2,5) with the result that the difference in readings
between sites across the freeway should reflect contributions from the freeway
traffic.
417
-------
As an initial step in the analysis of the S0i+ data, we have employed
daily readings to calculate monthly averages at various sites. In computing
monthly averages, extreme observations were eliminated according to the_
following procedure. For a particular month, let Y be an observation, Y be
the average and s the estimated standard deviation of Y. If \Y -?|>3s, Y\
will be excluded and then Y and s ^recalculated. The process will be repeated
until all observations fall with Y ± 3s.
ANALYSIS OF MONTHLY MEANS
The monthly averages of the 4 hour hi-vol readings a't-aite^A to D are
shown by the solid lines in Figure 2(i) to (iv), respectively. From these
plots we make the following observations.
(1) At each location there appears to have been a decrease in the SO^
readings from 1974 to 1976.
(2) The monthly means at the sites C and D downwind from the freeway, are
higher than those for the upwind sites A and B, indicating a contri-
bution of S0i+ from the freeway traffic.
(3) The freeway contribution is, however, very small when compared to the
average level of S0i+ across these four sites.
Next, the monthly means of the 4 hour membrane data for sites A and C are
given by the dotted lines in Figures 2(i) and (Hi). Similar to the 4 hour
hi-vol data, at both sites the level decreased from 1975 to 1976. Finally,
the monthly means of the 24 hour hi-vol readings at sites A and C are shown in
Figures 2(v) and (vi). Again the level of SO^ at both sites decreased during
the data period.
COMPARISON OF THE HI-VOLUME AND MEMBRANE READINGS
It is of interest to compare the 4 hour hi-vol and membrane readings.
Figures 2(i) and (iii) show that while the seasonal patterns of these two
readings are similar, the hi-vol means are consistently higher than the
membrane means. For further comparison, Figures 3(i) and 3(ii) show, respec-
tively for sites A and C, the scatter plots of the daily 4 hour membrane vs.
hi-vol readings. We observe that for site A the points fall close to a line
with a slope approximately equal to 1 and an intercept of about 4\ig/m^. On
the other hand, for site C the points are again close to a line but having a
slope of about 0.8 and an intercept of 7.7 yg/m^. These figures strongly
suggest that the membrane and the hi-vol readings are systematically biased
and that the nature of the bias is different at different sites.
ACROSS THE FREEWAY DIFFERENCES
We next consider the monthly means of the across the freeway difference
(C-A) of SOi, for the 4 hour hi-vol, the 4 hour membrane and the 24 hour hi-vol
readings. These means are plotted in Figures 3(iii)-(v). We observe that:
(1) The 4 hour hi-vol readings indicate a decrease in the across the freeway
418
-------
yg[m-
30.
20.
10.
— hi-vol.
— membrane
1/75
1/76
(i) Site A.
Figure 2. Monthly means of 4-hr hi-vol. and membrane afternoon
readings of SO. for various sites.
419
-------
yg|m;
30.
20.
10.
1/75
1/76
(ii) Site B.
Figure 2. Monthly means of 4-hr hi-vol. and membrane afternoon
readings of SO. for various sites.
420
-------
pg|m;
30.
20.
10.
— hi-vol.
— membrane
X.J
1/75
1/76
(Hi) Site C.
Figure 2. Monthly means of 4-hr hi-vol, and membrane afternoon
readings of SO. for various sites.
421
-------
30.
20.
10.
1/75
1/76
Civ; Site D.
Figure 2. Monthly means of 4-hr hi-vol. and membrane afternoon
readings of SO for various sites.
422
-------
30.
20.
10.
1/75
1/76
(v) Site A.
Figure 2. Monthly means of 24-hr hi-vol readings of SO..
423
-------
30.
20.
10.
1/75
1/76
(vi) Site C.
Figure 2. Monthly means of 24-hr hi-vol. readings of SO..
424
-------
hi-vol
jug/m3
48.
36.
24.
12.
xx
xx32x2x
x x2344*3
2x32xxx2
x5xx22x x
2x233xx x
33322
x23
x x
x xxxx
x222x
4- 3*
x
j
0.
12. 24.
(i) Site A,
36.
membrane
48.
Figure 3, Scatter plots of hi-vol."vs. membrane SO .
425
-------
hi-vol
jug/m3
48.
36.
24.
X
X
xx
x 2
3 x
xx 2
xx3 x xx
3 2x3 2 x
12.
33x3 2322
X227352 2
x423x x
X2x x
x 3
x
0.
12. 24.
(ii) Site C.
36.
membrane
48.
Figure 3. Scatter plots of hi-vol. vs. membrane SO
4
426
-------
yg|mc
5.0
2.5
0.0
1/75 1/76
(Hi)
Figure 3. Monthly means of 4-hr hi-vol. SO(C-A)
427
-------
i3 U
5.0
2.5
0.0
J 1 1 I
1/75
1/76
(iv)
Figure 3. Monthly means of 4-hr membrane SO. (C-A)
428
-------
yg|m;
5.0
2.5
0.0
1/75 1/76
(v)
Figure 3. Monthly means of 24-hr hi-vol. SO (C-A)
429
-------
differences from 1975 to 1976. (2) On the other hand, the 4 hour membrane
readings indicate that there was an increase in the across the freeway differ-
ence from 1975 to 1976. (3) The 24 hour hi-vol readings also show an increase
from 1975 to 1976. These observations are partially confirmed by the t-values
given in Table I for comparing the overall means of the across the freeway
differences (C-A) for the months June through October between 1975 and 1976.
Specifically, the t-values indicate that the increase in the 4 hour hi-volume
readings and the decreases in the other two measurements are real.
The sample means and standard deviations given in the table also show
that for each year, the across the freeway differences in the 4 hour hi-vol
readings are considerably larger than those for the other two. Thus, both in
terms of the level and the trend, the 4 hour membrane measurements and the 24
hour hi-vol measurements are in close agreement, but they are incompatible
with the 4 hour hi-volume readings.
A Conjecture
The preceding analysis raises a number of perplexing questions. First,
how can we account for the discrepancies between the different methods of
measuring S0i+? Second, which, if any, of the methods are giving correct
pollutant readings? These questions must be answered before any conclusions
about trends in the across the freeway difference can be made.
According to one researcher (6) , SO2 can react on filters used in the hi-
vol sampler to form SOq., and this may be a reason for the observed discrep-
ancies. One possible theory that could explain what we have observed runs as
follows:
• The 4 hour hi-volume difference (C-A) is mostly SO^ formed by SO2
reacting on the filter. This artifact formation progresses more
quickly at site C in the afternoon because of exposure to additional
SO2 from the freeway.
• The artifact formation occurs to a lesser extent, if at all, on the
membrane filter.
• After a sufficiently long time the artifact formation of SOi+ will be
about equal at sites A and C so that the 24 hour hi-volume difference
represents largely contributions from the freeway.
Although this theory could explain to a certain extent the observed discrep-
ancies, further research seems necessary to determine the precise nature of
these two sampling methods.
Conclusions
We can tentatively conclude from our analysis that (1) there is a sub-
stantial difference between the 4 hour hi-vol and membrane SOi+ readings, and
(2) from the available data there is a small amount of SOi+ being contributed
from the freeway traffic.
430
-------
Table I. COMPARING THE MEANS OF THE ACROSS
THE FREEWAY DIFFERENCES BETWEEN 1975 AND 1976
June-Oct .
4 hour
4 hour
24 hour
hi-volume
membrane
hi-volume
mean
5.11
- .40
-1.32
s
4
1
1
.d.
.45
.82
.54
1975
no. of
obs.
143
36
123
June-Oct. 1976
mean
2.22
1.00
.45
s.d.
3.74
1.56
1.80
no. of
obs.
133
119
125
Difference 1975-1976
t- value
5.
-4.
-8.
86
15
28
sig-level
.0000
.0001
.0000
fcj
-------
ANALYSIS OF THE CO DATA
CO data consist of hourly readings at sites A and C. As a preliminary
step in our analysis, monthly means of the daily 4 hour (3-7 p.m.) averages
were determined and they are shown in Figures 4(i) and (ii) for sites A and C,
respectively. From these figures we can make the following observations: (1)
The behavior of CO is markedly seasonal, being higher in the winter months
than in the summer months. Thus, any trend analysis must make proper allow-
ance for the seasonal effect. (2) The level at site C is considerably higher
than that at site A. (3) At site A the level is particularly low in the
summer months, about .5 ppm. Since we are mainly interested in the freeway
contribution of CO, in what follows we will concentrate our analysis on the
downwind site, C.
Two major factors influencing the behavior of CO are traffic and the wind
(3,5,7). These two factors must be taken into account in studying the effect
of the catalytic converter. We have available hourly readings of (i) wind
speed and wind direction for the entire period, and (ii) traffic counts and
traffic speeds for the period September through November 1976. In what
follows we built an empirical-mechanistic model (8) to describe the observed
CO concentrations at site C as a function of the wind and traffic. The model
will then be used to assess the trends in CO emissions.
MOTIVATION FOR THE MODEL
Figure 4(iii) gives the average diurnal curves of CO at site C computed
from the following three data periods:
(i) summer weekday: Mondays-Thursdays, June-October 1975
(ii) summer weekend: Sundays, June-October 1975
(Hi) winter weekday: Mondays-Thursdays, December 1974-April 1975.
These three diurnal curves illustrate that the behavior of CO changes from
weekday to weekend and from summer to winter. The summer weekday curve has
two peaks which roughly correspond to the morning and afternoon rush hours.
In contrast, the summer weekend plot fails to have these peaks. The diurnal
behavior of CO also changes from summer weekday to winter weekday, the most
notable change being the absence of a peak corresponding to the morning rush
hours in the winter.
In Figure 4(iv) the dotted line gives the diurnal pattern of the average
counts of vehicles passing the monitoring sites on weekdays. The figures
shown are average counts for each hour of weekdays (Monday-Thursday) obtained
during the period September 1976 through November 1976. Since this is the
only period for which traffic data are available and the traffic patterns in
Los Angeles are not expected to vary much over time, we will for the moment
assume that the figures shown are applicable to the entire period (June 1974-
November 1976) under study.
432
-------
ppm
6.
4.
2.
1/75 1/76
(i) Site A.
Figure 4. Monthly means of 4-hr (3 p.m. - 7 p.m.) CO.
433
-------
ppm
6.
4.
2.
1/75
1/76
(ii) Site C.
Figure 4. Monthly means of 4-hr (3 p.m. - 7 p.m.) CO.
434
-------
ppm
7.5
5.0
2.5
— summer weekday
— summer weekend
—-- winter weekday
8 a.m. 4 p.m.
(Hi)
Figure 4. Diurnal plot of CO, Site C.
435
-------
veh/mi
300
200
100
•••• counts
— density weekday
— density weekend
counts
15000
10000
5000
8 a.m. 4 p.m.
(iv)
Figure 4. Diurnal plot of traffic.
436
-------
Comparing the diurnal pattern of the weekday counts with that of the
summer weekday CO, we observe that the counts do not seem to reflect the
magnitude of CO during the evening rush hours. A possible reason for this is
the increase in traffic congestion during that time. The solid curve in
Figure 4(v) shows the diurnal pattern of the traffic speed for weekdays.
Notice the two dips during the morning and the evening rush hours and the
difference in magnitude between them. Thus, there is a large volume of
traffic moving very slowly during the evening rush hours, and the vehicle
counts do not accurately reflect the density of the traffic.
TRAFFIC DENSITY
To more appropriately reflect the situation, we consider the traffic
density (TD) defined as
TD _ number of vehicles passing the sites per hour
average speed per hour
which is a measure of the intensity of traffic around the sites (in units of
vehicles per mile). The solid line in Figure 4(iv) shows that the diurnal
pattern of TD for weekdays seems to match the CO pattern much better than the
vehicle counts. In addition, for the weekends, the diurnal pattern of TD
shown by the dashed line in Figure 4(iv) also correlates with that of CO given
in Figure 4(Hi) . From the above discussion it appears that the observed
concentrations of CO are approximately proportional to TD.
THE WIND VECTOR
Another major factor influencing the pollutant readings is the speed and
direction of the wind. Data on wind speed and wind direction are available on
an hourly basis. Following (5) we have decomposed the observed hourly wind
vector at site A into two components: one which is perpendicular to the
freeway (WSjJ and the other parallel to it (WS^). The positive directions for
the two components are shown in Figure 1. Specifically, let WS be the wind
speed and WD the wind direction. (WD = O when the direction is due north) .
Then, since the freeway is situated at approximately 235 from the north, we
have that
WSJ.= WS'cos(WD-235°)
= WS*sin(WD-235°) .
The diurnal patterns of WSJ. for the summer months (June-October 1975) and
the winter months (December 1974-April 1975) are plotted in Figure 4(vi). We
observe that the WSj^in the winter is lower than that in the summer. In
particular, for the winter, from midnight to 9 a.m. the WSj^is negative so
that the wind is blowing away from site C. The WSj. for the same hours in the
summer months is nearly zero. This change in the WSJ, from the summer to the
winter could explain the absence of a peak in CO during the morning rush hours
in the winter (see Figure 4(Hi)). Therefore, it appears that seasonal
changes in WSJ,, may account for the observed changes in the diurnal behavior of
CO from the summer to the winter.
437
-------
mph
60.
40.
20.
— weekday
— weekend
j i
8 a.m. 4 p.m.
(v)
Figure 4. Diurnal plot of traffic speed.
438
-------
mph
4.
2.
0.
— summer
— winter
8 a.m.
4 p.m.
(vi)
Figure 4. Diurnal plot of
439
-------
In order to determine more precisely the effect of WSj_on CO, we have
divided all the available hourly WS£,. readings into intervals of .5 mile per
hour. For each interval we have calculated the average of the corresponding
CO readings. These averages are plotted in Figure 5(i). Observe that a
convenient way to characterize the dependency of CO on WS is to take CO
proportional to the dispersion factor
where b and W are two appropriate constants.
THE CO MODEL
The above considerations have led us to consider the following tentative
model for the diurnal behavior of CO at site C:
CO = a-f K' (T&^'e — + a
where
CO.: the observed CO for site C at hour t.
TD : the traffic density at hour t.
•J-t: the perpendicular component of the wind vector at hour
t.
a : an added error term for hour t.
a.- a parameter measuring the background CO.
K: a parameter proportional to emissions.
e "" : a diffusion factor involving the perpendicular wind
component and two parameters b and W .
o
Given a set of data on CO, WS|_ and TD, the parameters can be estimated by
employing standard non-linear procedures. Specifically, on the assumption
that the errors a 's are independently and normally distributed with mean zero
and variance a2(a,K,b, and W ) are estimated by minimizing the sum of squares
of the a 's with respect to these parameters simultaneously.
Testing the Model
The model is useful for our purposes due to the following: (1) It is a
relatively simple model with only four parameters to be estimated from the
data. (2) It takes into account the effects of the traffic and the wind. (3)
440
-------
ppm
8.
6.
4.
2.
2.0 4.0 6.0 8.0
(1)
Figure 5. Plot of CO vs. WS±
mph
441
-------
The parameter K has an important interpretation. It is proportional to emis-
sions, and hence it will play a crucial role in determining if there has been
significant changes in emissions over time.
However, before we use the model for trend assessment, it is important to
verify that it can in fact account for the changes in the observed behavior of
CO from weekday to weekend and from summer to winter. For this purpose, we
have employed the three sets of CO average readings shown earlier in Figure
4(iii), the corresponding WS± averages and the TD values given in Figure
4(iv) .
The parameter estimates are shown in Table II. The actual CO readings
and the predicted values are plotted in Figures 5(ii)-(iv). It seems that the
model produces very close agreement between the actual and the predicted in
all three cases. Notice from Table II that for the weekend the estimated
emission constant K is considerably lower than those for the weekdays. This
is perhaps due to the fact that CO emissions may also depend on the driving
model (in particular, accelerations and decelerations) which is very different
between weekdays and weekends.
Evaluation of Trend
Our main objective is to assess the effect of the catalytic converter on
ambient CO concentrations. This can be done in terms of the model ideally by
allowing the parameter K to depend on the fraction of vehicles equipped with
catalytic converters. Since precise estimates of the fractions of catalyst
cars are not available, we shall proceed with a trend analysis by investi-
gating possible changes in the value of K over different time periods.
As a first attempt, the model has been used for evaluation of trends as
follows. We divided the entire time span of the data into the following five
periods: (1) June 1974 through October 1974, (2) November 1974 through April
1975, (3) May 1975 through October 1975, (4) November 1975 through April 1976
and (5) May 1976 through October 1976. This division roughly corresponds to
the "summer" and the "winter", and each period is of sufficient length to
allow for an appreciable increase of catalyst equipped cars. The emission
parameter K was allowed to vary from period to period but the other parameters,
a,b and W , were constrained to be the same for all five periods. The weekday
data were employed in estimating the parameters. Within each period we calcu-
lated the 24 hourly averages of CO and WS± for each of the three consecutive
two month segments. (For the first period, we took averages for June, then
July and August, finally September and October.) The CO and WSj. averages for
all five periods together with the averages of the available weekday TD were
then employed to estimate the parameters a,b,W ,jq , ... ,K5 where K. is the
emissions parameter in period j for j = 1,...,?. In addition, we3used a
weighted least squares estimation procedure to allow for possible changes in
the variance from period to period. The parameter estimates are given in
Table III.
Consider the changes in the emission parameter K over these 5 periods.
We see from Table III that the estimated K decreased between the first and the
442
-------
Table II. CO MODEL TEST FITS
June 1975-October 1975 Weekdays (M-R)
Parameter Estimate Std. Err.
a
b
w0
K
1.90
.00035
3.27
.0241
.14
.00005
.13
.00096
June 1975-October 1975 Sunday
Parameter
a
b
w
K
Estimate
1.79
.00013
2.54
.0190
Std . Err .
.23
.00011
1.11
.0016
December 1974-April 1975 Weekdays
Parameter Estimate Std. Err.
a 1.42 .17
b .00029 .00006
W0 2.89 .26
K .0228 ' .0010
443
-------
Table III. ASSESSMENT OF CO TREND
Period
Parameters
Estimate
Std. Err.
Summer 74
Winter 74-75
Summer 75
Winter 75-76
Summer 76
a (ppia)
b
Wo (mph)
Kl
K2
2J
2J
1.71
.00026
2.93
.0195
.0235
.0245
.0196
.0209
.68
.88
.66
.90
.63
.076
.00002
.098
.0006
.0008
.0007
.0008
.0006
444
-------
second winters, and also between the second (2975; and the third (1976)
summers. This can be associated with the increase in the number of catalyst
cars. On the other hand, the increase in K between the first and the second
summers, from .0195 to .0245, is somewhat surprising. This anomaly may very
well be due to the fact that the same TD data were used for all five periods.
While the traffic pattern is not expected to vary much over time in Los
Angeles, because of the energy crisis, the (estimated) total traffic volume
did show a slight dip in 1974, from 182,500 in 1973 to 180,000 in 1974 and to
182,500 in 1975*. Thus, there was an increase of about 1% from 1974 to 1975.
In the meantime, a major change in the speed limit occurred around January,
1975, from 70mph to 55mph, representing roughly a 20% decrease. Since TD is
the ratio of traffic counts over average speed, the density figures used in
the estimation contain a considerable upward bias for the first period and
could account for most, if not all, of the increase in K from 1974 to 1975.
Conclusions
We have developed a model which relates the observed behavior of CO at
site C to traffic density and the perpendicular wind component. By using this
model to assess the trend in CO, we conclude that CO emissions decreased from
1975 to 1976 by about 15%.
ANALYSIS OF THE Pb DATA
This section presents a preliminary analysis of the lead readings. The
Pb data consist mainly of the 4 hour afternoon hi-volume readings from June
1974-October 1976. Hi-volume readings for 4 morning hours were also collected
during the period June 1974-January 1976**. To begin with, we consider the
monthly means of the weekdays (Monday through Thursday) afternoon readings.
These monthly averages for sites A through D are plotted in Figures 6(i)
through (iv) , respectively. It is clear that the levels at sites C and D are
much higher than those at A and B. Also, in the summer months there is very
little Pb at these latter sites. As in the case of CO we shall concentrate
our analysis on the downwind sites, C and D.
For both site C and site D the level of Pb in the summer months appears
to have steadily decreased from 1974 to 1976. Table IV shows that these
decreases are statistically significant.
EXTENSION OF MODEL TO Pb DATA
The model developed for hourly CO data can be readily extended to the 4
hour Pb readings. This is achieved by summing the right hand side over the 4
afternoon hours to obtain
*These are estimated daily traffic totals near the sampling sites obtained
from the State of California Department of Transportation publications
"Traffic Volume on California State Highways" for 1973, 1974 and 1975.
**Samplers operated 6-10 a.m. from June 1974 to January 1975 and 8-12 a.m.
from January 1975 through January 1976.
445
-------
ppm -
8.
6.
4.
2.
8 a.m. 4 p.m.
(ii)
Figure 5. CO model summer weekday fit (x predicted - observed)
446
-------
ppm
8.
6.
4.
2.
8 a.m. 4 p.m.
(Hi)
Figure 5. CO mddel winter weekday fit (x predicted - observed)
447
-------
ppm
8.
6.
4.
2.
8 a.m. 4 p.m.
(iv)
Figure 5. CO model summer weekday fit (x predicted - observed).
448
-------
yg m;
6.
4.
2.
1/75 1/76
(i) Site A.
Figure 6. Monthly means of 4-hr hi-vol. afternoon readings of Pb for various sites.
449
-------
yg|m:
6.
4.
2.
1/75 1/76
(ii) Site B.
Figure 6. Monthly means of 4-hr hi-vol. afternoon readings of Pb for various sites.
450
-------
ug|m:
6.
4.
2.
\
1/75 1/76
(Hi) Site C.
Figure 6. Monthly means of 4-hr hi-vol. afternoon readings of Pb for various sites.
451
-------
6.
4.
2.
I I
1/75 1/76
Site D.
Figure 6. Monthly means of 4-hr hi-vol. afternoon readings of Pb for various sites,
452
-------
Table IV. SIMPLE T-TESTS TO COMPARE THE SUMMER MEANS OF Pb, 2974-2976
{MAY-OCTOBER}
Site D
Year mean
1974 5.30
1975 4.49
1976 4.11
Years
1974-1975
1975-1976
1974-1976
s.d. no.
1.20
.79
.92
t-statistic
4.75
2.79
6.59
of observations
68
84
76
sig. Level
.000
.006
.000
Site C
Year
1975
1976
Years
1975-1976
mean
6.76
5.80
s.d. no.
1.40
2.44
t-statistic
4.28
of observations
84
76
sig. Level
.000
453
-------
:' if
Pb , = a' + K' It (TD) e + a
t t=16 C C
t' stands for a particular day,
(TD)1" is traffic density at hour t on day t',
WS£ is perpendicular wind speed at hour t on day t',
t
a' is the error term for day t',
a' is a parameter measuring the background,
K' is a parameter proportional to Pb emissions,
b and W are, as before, parameters measuring effect of the wind.
o
Evaluation of the Trend in Pb Emissions
Model for lead has been used to assess the trend in Pb emissions in a
manner similar to that for the CO data. Specifically, we divided the time
span of the data into the same five periods as was done in the case of CO.
The daily 4 hour (3-7 p.m.) afternoon Pb readings from June 1974 through
October 1976, the corresponding hourly WSj, for the four afternoon hours,
together with the (same) average TD figures for these four hours, were then
employed to estimate the parameters a',b,W ,K',...,K'. Here, K'. measures the
Pb emissions for period j, j=l,...,5. In an initial fitting a'^was found to
be insignificant and hence ignored in our final estimation. The parameter
estimates of b,W ,K ,... ,K for sites C and D are given in Table V.
We observe that for both sites there was a decrease in Pb emissions.
This is probably due to an increased use of unleaded gasoline. Also, the
estimated values of the emission constant at site D are lower than those at
site C. This seems reasonable because site D is situated farther from the
freeway and a part of the lead particles will settle on the ground before
reaching D.
Comparison of Weekday and Weekends
As reported earlier (5), we have found that the afternoon Pb readings are
substantially higher in the weekends than during the weekdays. The monthly
means of both the weekday and weekend afternoon Pb readings for sites C and D
are plotted in Figures 6(v) and (vi), respectively. It is clear from these
figures that there is a considerable difference; for site C the weekend values
are about 60% higher than the weekday values. This increase on the weekend is
somewhat puzzling since there is a decrease in traffic density from weekday to
weekend [see Figure 4(iv)].
Now, recall from Figure 4(v) that the speed of the traffic decreases
drastically during the weekday afternoon rush hours. However, we see from the
dotted line in the same figure that such a decrease does not occur on the
454
-------
Table 7. ASSESSMENT OF Pb TREND
Site C
Period
Summer 75
Winter 75-76
Summer 76
Parameters Estimate
b .00013
W0 (mph) 2.28
K3 (\ig/m2) .00640
K^ fyg/m2; .00514
K5 (]ig/m2) .00514
03 1.25
vk 1.19
05 1.44
Std. Err.
.00002
.40
.00023
.00025
.00025
Site D
Period
Summer 74
Winter 74-75
Summer 75
Winter 75-76
Summer 76
Parameters Estimate
b .00020
W0 (mph) 2.45
KI (\ig/m2) .00507
K2 (Vg/M ) .00428
K$ (]ig/m2) .00454
K^ (\ig/m2) .00416
KS (vg/m2) .00378
D! 1.00
02 1-53
03 .50
Oij . 76
05 .99
Std. Err.
.00002
.19
.00013
.00021
.00013
.00016
.00016
455
-------
15.
10.
5.
\
— weekday
— weekend
1/75
1/76
(v) Site C.
Figure 6. Monthly means of afternoon Pb.
456
-------
ug|m3
15.
10.
5.
— weekday
weekend
1/75 1/76
(vi) Site D.
Figure 6. Monthly means of afternoon Pb.
457
-------
weekend. This would explain the observed weekday to weekend difference if
emissions of Pb varied inversely proportionally with the speed of traffic.
This theory is partially supported by a comparison of the weekday vs. weekend
monthly means of the morning Pb readings shown in Figure 6(vii). Specifi-
cally, we observe from Figure 4(v) that, in the morning, the difference be-
tween weekday and weekend speed is smaller than that in the afternoon. At the
same time Figure 6(vii) indicates that there is no real difference between the
weekday and weekend morning Pb readings. In fact, it has been reported in (4)
that Pb emissions do tend to vary proportionally with the traffic speed.
Conclusions
we have illustrated for the CO data that a simple model involving the
wind speed and traffic density can, to a good approximation, describe the
behavior of the pollutant near the freeway. This model has been used in
assessing changes in CO and Pb emissions since the introduction of the catalytic
convex car. We have found a decrease in both CO and Pb emissions, which can
probably be attributed to the increased number of catalyst equipped cars on
the freeway. We have shown that there is an increase in Pb emissions from the
weekday to weekend. This increase is probably due to the increased speed of
the traffic on the weekend. In our analysis of the SOi+ data we have found
L jat there is a discrepancy between the 4 hour hi-volume readings and the 4
hour membrane readings. A plausible explanation seems to be that there is an
a -.ifact formation on the hi-volume filter causing the discrepancy, but much
further research is needed to determine the precise nature of these two
sampling techniques.
REFERENCES
1. Seltzer, M., Campion, R.J., and Peterson, W.L. "Measurement of vehicle
particulate emissions," Society of Automotive Engineers Paper 740286,
February, 1974.
2. Evans, G., "Summary of continuous pollutant data." Presented at the Los
Angeles Catalyst Study Symposium April 12-13, 1977, Raleigh, North
Carolina.
3. Gifford, F.A. and Hanna, S.R., "Technical note: Modelling urban air
pollution," Atmos. Environ., 7_, 1973, 131.
4. Hirschler, D.A., Gilbert, L.F., Lamb, F.W., and Niebylski, L.M., "Parti-
culate lead compounds in automobile exhaust gas." Industrial and.
Engineering Chemistry, 49, 1957, 1131.
5. Phadke, M.S., Tiao, G.C., and Hillmer, S.C. "Statistical evaluation of
the environmental impact of the catalytic converter," to be presented at
1977 annual meeting of Air Pollution Control Association, Toronto, June
1977.
6. Rodes, C.E., "Summary of integrated pollutant data." Presented at The
Los Angeles Catalyst Study Symposium April 12-13, 1977, Raleigh, North
Carolina.
458
-------
15.
10.
5.
• weekday
weekend
1/75 1/76
(vii) Site D.
Figure 6. Monthly means of afternoon Pb.
459
-------
7. Tiao, G.C., Box, G.E.P. and Hamming, W.J., "A statistical analysis of Los
Angeles ambient carbon monoxide data 1955-1972," J. Air Pollution
Control Assoc., 25, 1975, 1129.
8. Tiao, G.C., Phadke, M.S. and Box, G.E.P., Some empirical models for the
Los Angeles photochemical smog data, J. Air Pollution Control Assoc. May
1976, 26, 485-490.
460
-------
DISCUSSION
Dr. William Pierson
Ford Motor Company
Dearborn, Michigan
The following observations and estimates (surrogate calculations) were
made by Dr. William Pierson of Ford Motor Company at the LACS Symposium
utilizing data presented:
In general the sulfur consumption (from the fuel) in automobiles is
mg/mi, including all sizes of SOi^ plus SOZ. This number may even be higher,
especially if diesels are included.
Pb emissions from non-catalyst cars (Ron Bradow's estimate) are ^40
mg/mi airborne, all sizes.
Therefore one would expect the ratio of contributions from the freeway at
the LACS to be:
AIS
Pb
50
. 2
From LACS 24-hour ambientjneasurements ASC>2 = ^9 ygr/m3 30% = ^4.5 pgr/m3 S,
(dichot.) - ^0.5 vg/m3 SOi+ = ^0.17 wgr/m3 S, and ApJb = °»4 yg/zn3.
Summing the sulfurs and dividing by the Pb:
f ASOk + ASOo) 4.5 + 0.17
- a- *
4 '
Examining the ambient moleratio of dichotomous SO^ to total sulfur:
I
^0.04
[A504 -f- AS02J 0.17 + 4.5
Assume a 30% catalyst car mix, then the SO^ conversion from catalyst cars
calculated from ambient measurements would be - 0.04/30% = 0.13 or 13%.
If the ratio of sulfate to Pb contribution
^=^= 0.125,
API? 4
and if the Pb emission rate is typically = 40 mg/mi, then the calculated
emission rate of S0~ from all cars would be = 40 x 0.125 = ^ 5 mg/mi.
461
-------
Again assuming a 30% catalyst car mix: 5/30% = VZ7 rag SO^/mi calculated
emission rate from catalyst cars based on LACS ambient measurements. Ron
Bradow offered 10 mg/mi from actual emission tests.
Conclusions: The preceding estimates indicate that the LACS ambient SOi+
and Pb measurements are consistent with those predicted from typical emission
factors. One could have used CO instead of Pb as the surrogate.
The above estimates are crude and ignored differences in fuel sulfur
between leaded and unleaded gasoline. Also I just picked averages off the
slides, but it all seems to "hang together."
If the wind-speed dependence of APJb is unequal to that of (hso~^) ,
then something is seriously wrong.
462
-------
LOS ANGELES CATALYST STUDY SYMPOSIUM
CONCLUSIONS AND RECOMMENDATIONS
Thomas R. Hauser
Environmental Protection Agency
Research Triangle Park, North Carolina
ACHIEVING OBJECTIVES
At the outset of Los Angeles Catalyst Study (LACS) in June 1974, four
prime objectives for the study were established.
The first objective is to determine the impact of catalyst equipped cars
on air quality for automobile related pollutants. At the present time, it has
been demonstrated that the LACS site is a very adequate site to determine the
effect of catalyst emissions on air quality particularly during the summer
months between 1500 and 1900 hours. The LACS study has collected a large
amount of air quality data with simultaneous meteorology and traffic informa-
tion, all measured at the same station, accompanied with very extensive quality
assurance procedures. Hence this body of data provides a multitude of possible
comparisons among various collection methods, analysis methods, and pollutant
interrelationships.
The second objective of LACS is to provide an active data base to contrast
roadside levels of automobile related pollutants caused by pre- and post-
catalyst equipped cars. The study was initiated in June 1974, prior to the
introduction of catalysts on cars, and has continued to date. The rigorous
quality assurance procedures that have been applied to the LACS attest to the
fact that a valid, high quality data base is available from June of 1974 to
the present time to make judgments concerning the effect of the addition of
catalysts on 1975 and later model year automobiles.
The third objective is to utilize the measured ambient levels of carbon
monoxide (CO), lead (Pb), and sulfur dioxide (802) to determine the surrogate
(predicted) sulfuric acid ambient levels. Based upon the recent reports that
conversion of sulfuric acid (H2SO^) to sulfate occurs very rapidly in the
ambient air and that the measurement levels of sulfates attributable to the
freeway are very low, it appears that surrogate projections of sulfuric acid
levels from roadways need to be reevaluated. The methodology for continuous
ambient measurement of #2S'O4 has not been available; however, the measured
sulfate contribution (and probably H^SO^ levels) from the freeway are substant-
ially less than was expected and do not appear to constitute a serious problem
at this time.
463
-------
The fourth objective is to evaluate and improve field methodology for
detection and measurement of air quality changes attributable to catalyst
emissions. Based upon the presentations at the LACS Symposium, it is obvious
that this objective is being met by evaluating and improving field methodology
as the need arises and resources are available. Several intermethod compari-
sons have been made to solve existing problems and add credibility to the data
base.
SPECIFIC CONCLUSIONS
RESULTING FROM CONTINUOUS MEASUREMENTS
m The LACS site has been shown to be a very favorable location in terms
of meteorology and traffic flow for determining freeway contributions,
particularly during the summer from 1500-1900 hours.
• Traffic flow estimates by the California Department of Transportation
coupled with Environmental Protection Agency (EPA) data indicate that
there has been little change in the traffic volume on the San Diego
Freeway at the study site from 1974 through 1976. The diurnal
traffic patterns by day of the week are very consistent. The total
number of automobiles passing the catalyst site averages about 188,000
vehicles per day, and by the end of 1976 about 30% of the vehicular
miles travelled were attributed to catalyst equipped cars.
• The freeway contribution of CO has decreased about 25% since the
inception of the study; from 1975 to 1976 the contributions of nitric
oxide (NO) and nitrogen dioxide (N02) from the freeway have increased
more than 50%.
• There is an inadequate amount of 03 at freeway level to convert all
of the NO to NO2. The reaction of NO and ozone (O^) is rapid enough
to result in almost zero concentrations of 03 at the near downwind
site.
• The background levels of CO, NO, and NO2 have remained essentially
constant since 1974.
• An empirical model relating diurnal behavior of CO to traffic density
and perpendicular wind speed has been developed and used to assess
the trend of freeway contribution of" CO during 74-76.
RESULTING FROM INTEGRATED MEASUREMENTS
• The background levels of total suspended particulates (TSP) and Pb
have changed very little since the beginning of the study.
• The background levels of sulfate have decreased 25% since 1974.
• The freeway contributions of TSP and Pb have both decreased
since 1975.
464
-------
• Previous LACS summary reports indicating freeway contributions of
sulfate as high as 5 yg/m3 were based on four hour high volume
sampler (hi-vol) measurements. These short term measurements by this
technique are now thought to be adversely affected by artifact
sulfate formation, resulting in erroneously large cross-freeway
difference estimates. The concensus of all methods indicates only
0.5 to 1.0 yg/m3 freeway contribution of sulfate during periods of
favorable meteorology.
• The apparent artifact sulfate formation on the hi-vol samples does
not appear tc> be significant in determining 24 hour cross-freeway
differences, but seems to contribute a substantial error to four-hour
sulfate measurements.
• The model for CO has been extended to assess the trend in Pb (4-
hour).
• During the most favorable meteorology, weekend Pb concentration is
""60% higher at the downwind site than weekday level. This appears to
be related t:o the slow traffic speed during the weekdays.
FUTURE LACS MONITORING PROGRAM
The following observations are pertinent concerning the utility of
continuing the LACS.
There is a need to continue to evaluate the impact of catalysts on air
quality. This will bts increasingly important as new catalyst-type vehicles
and additional diesel traffic are added to the traffic mix. Our present data
base and site are extremely well suited for assessing effects of transportation
control strategies on air quality data. It is EMSL's position that only
through a long-term monitoring study under appropriate quality control can
theoretical and laboratory findings be verified from which Agency regulatory
decisions can be based. However, monitoring frequency could possibly be
reduced. It is our opinion that at least two more summer seasons of data are
necessary to establish
-------
It is intended that time series models will be developed to concurrently
employ data from all sites to estimate spatial and temporal variability across
the freeway. This will include fixed in-roadway and mobile monitoring data.
Time series models for other pollutants including one for NO-NO2-O$ will be
required to judge what effect future oxides of nitrogen (NO^) control strateg-
ies may have on ambient concentrations off the roadway.
The present data base contains information to assist EPA in examining a
variety of other equally important issues associated wzth transportation
control. Some examples are: (a) establishment of the proposed Pb standard;
(b) input to the formulation of an oxidant control strategy; (c) examination
of the cross-freeway relationships of the NO^ and 03; and (d) verification of
the relationships between auto emission measurements and actual ambient
levels at the roadside. It should be noted that an additional two summer
seasons of data will greatly augment this data base.
At present, EPA is concerned with environmental imtult due to toxic
substances, some of which are emitted from mobile sourc&s, e.g., PAN, benzene,
etc. Maintaining surveillance of air quality in proximity of major roadways
could serve to identify and determine trends of air quality due to pollutants
from mobile sources for the purpose of determining the need for and effect of
control measures. Such a monitoring program could also be used to measure the
presence of toxic substances identified by the Fuel and Fuel Additive Regis-
tration Program.
466
-------
' TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO. '2
EPA-600/4-77-034
4. TITLE AND SUBTITLE
WS ANGELES CATALYST STUDY SYMPOSIUM PROCEEDINGS
7. AUTHOR(S)
Thomas R. Mauser , Editor
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Environmental Monitoring and Support Laboratory
12. SPONSORING AGENCY NAME AND ADDRESS
15. SUPPLEMENTARY NOTES
16. ABSTRACT
3. RECIPIENT'S ACCESSION-NO.
5. REPORT DATE
6. PERFORMING ORGANIZATION CODE
8. PERFORMING ORGANIZATION REPORT NO.
10. PROGRAM ELEMENT NO.
1AA601
11. CONTRACT/GRANT NO.
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
The Proceedings consisted of 21 technical papers covering such areas
as sampling and analytical methods for aerosols, including NH. , SO. , Pb,
and NO?; quality assurance procedures; statistical analysis of LACS data; and
trends in ambient air pollution concentration measured just off a heavily
traveled roadway in Los Angeles.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
Ambient Air Monitoring
Air Pollution
Sul fates, Lead, Carbon Monoxide,
Aerosols
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
b.lDENTIFIERS/OPEN ENDED TERMS
Los Angeles Freeway
Catalytic Converter
Impact on Ambient
Air
19. SECURITY CLASS (This Report)
Unclassified
20. SECURITY CLASS (This page}
Unclassified
c. COSATI Field/Group
13B
21. NO. OF PAGES
476
22. PRICE
EPA Form 2220-1 (9-73)
467
*U.S. GOVERNMENT PRINTING OFFICE.-19 77 -?itO -110/307 REGIONNO.4
------- |