EPA-650/4-75-023
June 1975
Environmental Monitoring Series
COMPARISON OF METHODS
FOR DETERMINATION
OF NITROGEN DIOXIDE
IN AMBIENT AIR
U.S. Environmental Protection Agency
Office of Research and Development
National Environmental Research Center
Research Triangle Park. N. C. 27711
-------
EPA-650/4-75-023
COMPARISON OF METHODS
FOR DETERMINATION
OF NITROGEN DIOXIDE
IN AMBIENT AIR
by
L.J. Purdue, G.G. Akland, and E.G. Tabor
Quality Assurance and Environmental Monitoring Laboratory
ROAP No. 22ACK
Program Element No. 1HA326
U.S. ENVIRONMENTAL PROTECTION AGENCY
National Environmental Research Center
Office of Research and Development
Research Triangle Park, N.C. 27711
June 1975
-------
EPA REVIEW NOTICE
This report has been reviewed by the Office of Research and Development, EPA, and
approved for publication. Mention of trade names or commercial products does not
constitute endorsement or recommendation for use.
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into series. These broad categories were
established to facilitate further development and application of environmental
technology. Elimination of traditional grouping was consciously planned to
foster technology transfer and maximum interface in related fields. These 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
9. MISCELLANEOUS
This report has been assigned to the ENVIRONMENTAL MONITORING series. This
series describes research conducted to develop new or improved methods and
instrumentation for the identification and quantification of environmental pollutants
at the lowest conceivably significant concentrations. It also includes studies to
determine the ambient concentrations of pollutants in the environment and/or the
variance of pollutants as a function of time or meteorological factors.
Copies of this report are available free of charge to Federal employees, current
contractors and grantees, and nonprofit organizations - as supplies permit -
from the Air Pollution Technical Information Center, Environmental Protection
Agency, Research Triangle Park, North Carolina 27711; or, for a fee, from the
National Technical Information Service, 5285 Port Royal Road, Springfield, Virginia
22161.
11
-------
ACKNOWLEDGMENTS
Many staff members of the Quality Assurance and Environmental Monitoring Laboratory
were actively involved in planning and conducting this study. Constructive suggestions
in regards to the different aspects of the study were contributed by numerous individuals.
Because of the number of people involved, it is not possible to recognize individual
contributions to the success of the study. However, special acknowledgment is made of
the assistance provided by Mr. Vinson Thompson in designing and constructing the samp-
ling manifold, planning the sampling system, and maintaining and operating the instru-
mentation. Recognition is also given to Mr. Van Wheeler who performed the chemical
analyses of the samples and to Messrs. Gary Evans and Jack Suggs who performed the
statistical analyses of the data.
111
-------
CONTENTS Page
LIST OF FIGURES v
LIST OF TABLES vi
ABSTRACT vii
SUMMARY ix
1. INTRODUCTION 1
Objectives 2
2. STUDY DESIGN 3
Facilities 3
Air Distribution and Monitoring Systems 3
Methods 3
Operating Schedule ..... . 4
Calibration 6
Data Acquisition 6
3. EXPERIMENTAL 8
Methods and Instrumentation 8
Sampling System 10
Test Atmospheres 12
Start Up and Calibration 12
Sailing and Analysis Procedures 13
Quality Control 17
4. DATA SUMMARIES 20
Phase I 20
Phase II 24
Phase III 24
5. STATISTICAL ANALYSIS 25
Phase I 29
Phase II 30
Phase III 31
Statistical Observations 32
6. REFERENCES 34
APPENDIX A-l. Tentative Method (Chemiluminescence Principle) for
Continuous Measurement of Nitrogen Dioxide in
Atmosphere 35
APPENDIX A-2. Tentative Method (Colorimetric Principle) for
Continuous Measurement of Nitrogen Dioxide
in Atmosphere 40
APPENDIX A-3. Tentative Method (Sodium Arsenite Procedure) for
IV
-------
Determination of Nitrogen Dioxide in Atmosphere 45
APPENDIX A-4. Tentative Method (TGS-ANSA Procedure) for Determination
of Nitrogen Dioxide in Atmosphere 52
APPENDIX B. Statistical Evaluation of Methods Tested 60
Figure
1.
LIST OF FIGURES
Phase II N02 Addition Schedule
Page
5
2. Sampling System 11
3. Example of Strip Chart for Chemiluminescence NO- Analyzer. . . 16
A-l Automated NO, NO,, NO Chemiluminescence Analyzer 39
L* X
A-2 Sampling Train for Sodium Arsenite Method 47
A-3 Sampling Train for TGS-ANSA Method 54
B-l Phase I - N0? Methodology Study - Hourly Averages
for Continuous Methods 71
B-2 Phase I - Diurnal Patterns (Excluding Intermittent
Spike Days) 72
B-3 Method Difference Versus Interferent Concentrations (Excluding
Intermittent Spike Days) 73
B-4 Phase II - Design of Experiment 77
B-S Phase II - Method Difference 80
B-6 Ozone Interference - 75 ug/m
83
B-7 Ozone Interference - 150 yg/m 83
-------
Table LIST OF TABLES Page
1. Pollutant Measurement Methods 9
2. Zero Drift and Span Drift Data 18
3. NO, Methods Comparison Data - Phase I 21
4. N02 Methods Comparison Data - Phase II 22
5. NCL Methods Comparison Data - Phase III 23
6. Statistical Analysis of Method Difference 26
7. Cumulative Frequency Distributions of Absolute
Difference between Duplicate Daily Measurements 27
8. Intermethod Concentration Ratios by Concentration Interval . . 28
B-l Intramethod Comparisons - Bias and Precision 67
B-2 Intermethod Comparisons - Bias 68
B-3 Diurnal Pattern of Intermethod Differences 69
B-4 Cumulative Frequency Distribution of Hourly Intermethod
Differences by Concentration Grouping 69
B-5 Analysis of Variance - Difference from True 75
B-6 Analysis of Variance - Method Difference 76
B-7 Analysis of Variance - Phase II 78
B-8 Summary of Duncan's Test - Phase II 78
B-9 Analysis of Variance - Phase III 82
B-10 Summary of Duncan's Test - Phase III 82
VI
-------
CHEMICAL NAMES AND FORMULAS
Acetic acid
Ammonia
8-Anilino-l-naphthalene
sulfonic acid ammonium
salt (ANSA)
Carbon dioxide
Carbon monoxide
Formaldehyde
Guaiacol (see o-methoxyphenol)
Hydrochloric acid
Hydrogen peroxide
Methanol
p_- Me thoxypheno 1
2-Naphthol-3,6-disulfonic
acid disodium salt
N-(l-naphthyl)-ethylene
diamine dihydrochloride
(NEDA)
Nitric oxide
Nitrous oxide
Ozone
Phenol
Phosphoric acid
Sodium arsenite
Sodium hydroxide
Sodium metabisulfite
Sodium nitrite
Sulfanilimide
Sulfanilic acid
Sulfur dioxide
Tartaric acid
Triethanolamine
CII3COOH
N1I3
8-C6HsNl I- 1-C101 I6S03-NH4+
C02
CO
HCHO
HC1
CH30H
CI130-C6H4-OH
HO-C10Hs{S03Na)2
[Nll-C10H7(Cll2)2NU2]2HCl
NO
N20
°3
I13P04
NaOH
Na2S203
NaN02
[4-(Il2N)C6N4S02NII2]
M12-C6H4-S03H
S02
(CIO1-C021Q2
ABBREVIATIONS
Peroxyacetyl nitrate PAN
Total sulfur TS
Total suspended particulates TSP
VII
-------
ABSTRACT
A study of four methods for the measurement of nitrogen dioxide (NC^) in ambient
air to determine the intramethod and intermethod comparability of the procedures under
a variety of carefully controlled conditions is presented. Two of the methods were
automated continuous methods and two were manual methods (24 hour integrated). The
two automated methods were chemiluminescence and colorimetric (Lyshkow modification
of the Griess-Saltzman method) procedures. The two manual methods were the arsenite
and triethanolamine, guaiacol, sodium metabisulfite, 8-amino-l-naphthalenesulfonic acid,
ammonium salt (TGS-ANSA) method (recently developed).
The study was conducted in three phases. Simultaneous nitrogen dioxide concentra-
tion data were obtained from a common source using duplicate N0£ analyzers for each of
the automated methods and quadruplicate samples were collected for each of the manual
methods. Phase I was a 22-day experiment in which additonal N02 was added to the
ambient air sampled in order to provide a wide range of N02 concentrations. Phases
II and III were limited experiments to determine the effect of variable NO? concentra-
tion fluctuations during a sampling period and to investigate suspected ozone interfer-
ence with the automated colorimetric method.
The most important conclusion drawn from the study was that the four N02 methods,
when used by skilled technicians under carefully controlled conditions, are capable
of producing data that are in remarkably good agreement.
The effect of variable NC»2 concentration fluctuations during a sampling period on
the results produced by the methods was minor. Suspected interference of ozone on the
automated colorimetric method was verified.
Vlll
-------
SUMMARY
The most important conclusion that can be drawn from the results of this study is
that the four methods for Pleasuring the NC>2 content of the ambient atmosphere, vihcn used
by skilled technicians under carefully controlled conditions, are capable of producing
data that are in remarkably good agreement.
The variability within a given method was least with the chcmiluminescence method,
followed by the arsenite, the TGS-ANSA, and the continuous colorimctric methods. In
the worst case, the colorimetric method, there was a small bias of 7.5 micrograms of
N02 per cubic meter (pg NC>2/in3) of air (0.004 part per million) between the data from
the two analyzers. This bias was less than 1 percent of the full-scale range of the
analyzers.
Internicthod comparisons of data obtained when sampling ambient air showed that the
average difference between any of the methods was never greater than 7.5 yg NC>2/i"3
(0.004 ppm). This maximum difference occurred between the arsenite and the TdS-ANSA
methods. In all cases, variability of the differences among the four methods was only
slightly higher than the variability within each method. The correlation coefficients
for the intcnnethod comparisons were greater than 0.985 in all cases.
No intcnnethod differences could be related to the concentrations of nitric oxide
(NO), carbon .monoxide (CO), carbon dioxide (CD2), ozone (03), total sulfur (TS), and
total suspended particulatc matter (TSP) in the ambient air sampled. This finding is
significant because the procedure for the arsenite method (Appendix A-3) states that
NO is a positive intcrfcrcnt. At NO concentrations as high as 302 njj/nr (0.246 ppm),
no interference was detected.
The effect of fluctuations in the N02 concentration during a sampling period on
the results produced by the four methods was relatively minor. The method differences
in almost all cases were within the 95 percent confidence interval determined for the
intermethod comparisons of Phase I of the study.
Suspected interference by Oj on the automated colorimetric method was verified.
Significant negative interference was found at NO2 concentrations of 75 and 100 ug/m
in combination with 03 concentrations of 353 and 667 ug/nr5. At an 03 concentration
of 100 ue/in-5, no interference was detected. The effect of high 03 concentrations
on the performance of the other methods tested was minor relative to its effect on
the colorimetric method.
The performance of all methods was monitored closely throughout the study. I-'or
the automated continuous measurement methods, the performance of the chemilumines-
cencc analyzers was better than that of the colorimetric analyzers in terms of
zero drift, span drift, response times, and overall operation. Of the two methods,
IX
-------
the arsenite method is less susceptible to analytical variables such as reagent
addition time or blank problems.
The good agreement between the ambient data generated by the continuous and
manual methods validates the calibration procedures and is an indication that the
methods tested are capable of producing accurate N02 data, with the exception of
potential 03 interference in the continuous colorimetric method.
-------
COMPARISON OF METHODS FOR DETERMINATION
OF NITROGEN DIOXIDE IN AMBIENT AIR
1. INTRODUCTION
On June 14, 1972, the Administrator of the Environmental Protection Agency (liPA)
stated that the reference method for measurement of nitrogen dioxide (NO 2) concentra-
tions in ambient airl was suspected of being unreliable^; results of laboratory testing
and N0£ measurements made over a period of several months at a large number of locations
suggested apparent deficiencies associated with routine field use of the method.
Accordingly, the Administrator announced that the reference method would be reevaluated.
The problems relating to the routine field use of the reference method for monitor-
ing ambient concentrations of N02 were addressed in some detail in the rcderal Register
oE June 8, 1973-*, in which the Administrator stated that it had not been possible to
resolve the variable collection efficiency problem and that this method could no
longer serve as the official reference method. Based on statements in the Federal
Register of June 8, 1973, it was concluded that at least three other N02 measurement
tedmiques were worthy of consideration as a replacement. Two of these were automated
continuous measurement methods -- a chemiluminescence method (described in Appendix A-l)
and a colorimetric method (described in Appendix A-2); the third was a manual, integra-
ted (bubbler)method, using sodium arsenite (described.in Appendix A-3). A fourth
method, the TGS-ANSA (triethanolamine, guaiacol, sodium metabisulfite, 8-anilino-l-
naphthalenesulfonic acid ammonium salt) method, was included in the study reported
here on the basis of additional test data. The TGS-ANSA method was not listed in
the Federal Register^ because, at the date of that publication, it had not been
sufficiently tested to determine whether it merited further consideration.
Because of the need for an adequate measurement method for ambient N02, an
intensive study of these four methods was carried out by liPA to compare the results
obtained by use of these methods under a variety of conditions.
-------
OBJECTIVES
The basic objective of the study was to determine the intramethod and intermethod
comparability of each of the four ambient N02 measurement techniques when sampling the
same atmosphere under a variety of carefully controlled conditions. The study was conr
ducted in three phases.
Phase I was designed to investigate intramethod and intermethod variability over
N02 concentration ranges expected to occur in ambient air. Measurements of the ambient
concentrations of nitric oxide (NO), carbon monoxide (CO^, carbon dioxide (C02), ozone
(03), total sulfur (TS), and total suspended particulate matter (TSP) were made concur?
rently during this phase of the study to provide data for use in determining whether
any intra- or intermethod differences could be attributed to interference by any of these
pollutants.
Phase II was a limited experiment designed to determine the effect, if any, of N02
concentration fluctuations during a sampling period on the comparability of the methods.
To minimize the variability caused by other pollutants, this phase of the study was con-
ducted by sampling clean air with controlled NO- additions, rather than by sampling
ambient air.
Phase III was designed to investigate suspected interference of 0, on the automated
4
colorimetric method and to determine the effect of 0, on the other methods.
-------
2. STUDY DESIGN
Because of the complexity of the study and the magnitude of effort required for
its conduct, a detailed study plan was developed. To minimize variability resulting
from operator errors and procedural variations, the study was conducted under the most
favorable conditions. Meticulous attention was paid to the condition of equipment and
the instructions to operating personnel regarding adherence to sampling procedures and
quality control.
FACILITIES
All the instruments and equipment were located 'in the EPA Durham (North Carolina)
Air Monitoring and Demonstration Facility (DAMDF), a building specially designed to permit
the evaluation of air monitoring methodology. Ambient concentration data for NC>2 and
other pollutants are routinely obtained at the DAMDF, which is equipped with a special
glass manifold for the distribution of ambient air. The ambient air at this location
was expected to contain at least moderate concentrations of N02 as well as many of the
potentially interfering pollutants to the methods under comparison.
AIR DISTRIBUTION AND MONITORING SYSTEMS
A rather complicated physical system possessing considerable flexibility was
required for this study. The basic air distribution system in the DAMDF was modified
to permit the installation of the necessary instrumentation and the incorporation of
the various subsystems used for the different phases of the study. The complete system,
which is described in greater detail in Chapter 3 was comprised of the following sub-
systems: (1) the DAMDF air distribution system that supplied ambient air for the N0£
methods under test and for the measurements of ambient pollutant concentrations; (2)
a pollutant generation system for the addition of NC>2 and 03 to clean air; and (3) a
sampling manifold for the distribution of either ambient, clean, or spiked air to the
intakes of the analyzers and samplers under test.
METHODS
Two of the methods tested are "automated methods" in which the sample collection
and the analysis are performed by means of a continuous analyzer. One of the automated
methods is based on the gas-phase chemiluminescent reaction between NO and 03 at reduced
or near-atmospheric pressure; N0£ is measured as NO after quantitative conversion, to
NO. The other automated N02 method (continuous colorimetric) is based on a specific
reaction of nitrite ion (N02) with diazotizing-coupling reagents to form a colored
azo dye that is measured colorimetrically.
The other too methods are "manual colorimetric methods" in which the sample collec-
tion and analysis are performed separately. Samples are collected by passing air through
3
-------
an absorbing solution at a constant flow rate for a prescribed time period. The samples
are then transported from the field location to a central laboratory for analysis, which
is based on a specific reaction of nitrite ion (NC>2) with diazotizing-coupling reagents
to form a colored dye that is measured colormetrically. Each' method requires different
absorbing solutions and diazotizing-coupling reagents.
Comparisons were made by simultaneously analyzing air from a common manifold for
22-hour periods using each of the methods. To determine the variability within a given
method, data were collected by two identical continuous analyzers for each of the auto-
mated methods. Duplicate multiple gas bubbler samplers were employed for the collection
of samples for the manual methods. Each sampler collected duplicate samples so that,
for each sampling period, four concentration values were obtained for each of the manual
methods and two values were obtained for each of the automated methods.
OPERATING SCHEDULE
All sampling periods were planned to be of approximately 22 hours duration (12
noon to 10 a.m.). The 2-hour period between 10 a.m. and 12 noon was needed for calibra-
tion of the continuous analyzers and for completion of other necessary tasks. All instru-
ments used in a given phase were operated concurrently for the full sampling period.
In the event of instrument breakdown or other problems, the day's operation was consider-
ed void for all methods and was repeated on the following day.
Phase I
Because of the relatively low atmospheric NC>2 concentrations encountered at the
DAMDF, it was necessary to conduct this phase of the study by sampling ambient air to
which N02 was added continuously to assure that the comparisons would be made over a
wide concentration range. A 20-day sampling schedule was planned during which the NC^-
addition concentration was varied randomly from day to day. The following N02 addition
concentrations were used: 0 ug N02/m3 (0 ppm), SO ug N02/m3 [0.027 ppm), 100 ug N02/m3
(0.053 ppm), 200 ug N02/m3 (0.106 ppm), and 800 yg N02/m3 (0.424 ppm). The schedule
allowed for four daily periods of sampling at each of these N02-addition concentrations.
Addition of 800 ug N02/m3 was limited to 3 hours because concentrations at this level
are not likely to occur for 22-hour periods.
In addition to the N02 measurements by the four methods under comparison, simulta-
neous analyses were made of the ambient air for N02, NO, 03, TS, CO, and C02- High
volume samples of TSP were collected outside near the ambient air intake of the DAMDF
manifold during each period.
Phase II
The second phase of the study was conducted by sampling clean air to which known
concentrations of N02 were added intermittently during the 22-hour sampling periods.
The N02-addition concentrations were selected such that the resulting 22-hour-average
concentration was either 100 pg N02/m3 (0.053 ppm) or 200 vg N02/m3 (0.106 ppm).
The 8-day sampling schedule given below and shown graphically in Figure 1 was followed.
4
-------
OAYt
o
X
M
i
cT
UJ
a
733 Afl/
733 (jg/m3
'367 ;jg/rr
DAYZ
CONTINUOUS N02 ADDITION-22 hr
DAY 3
CONTINUOUS N02 ADDITION-6 hr
DAY 4
CONTINUOUS N02 ADDITION-3 hr
DAY 5
INTERMITTENT N02 ADDITION - 3 hr
INTERMITTENT NO, ADDITION • 4 hr
I I I I
~275/jg/m3
DAY 7
CONTINUOUS N02 ADDITION-8hr
J
avg.
DAY 6 —
>-100^N02/m3avg.
2 4 6 8 10 12 14 16 18 20 22~
TIME k DAY 8
TIME, hours
Figure 1. Phase II N02 addition schedule.
-------
Day 1 (0 ug N07/m3) - Clean, NO,- free air.
3 3
Day 2 (200 ug N02/m ) - Clean air spiked continuously at 200 ug NCL/m .
Day 3 (200 ug N0,/m3) - Clean air spiked at 733 ug N0,/m3 for 6 hours and
£» L
clean air for 16 hours;
Day 4 (100 ug N0,/m ) - Clean air spiked at 733 ug NO./tn for 3 hours and
clean air for 19 hours.
Day 5 (100 ug N0,/m3) - Clean air spiked at 733 ug N0,/m3 f°r 1-5 hours and
3
clean air for 2 hours; clean air spiked at 733 ug NO^/m
for 1.5 hours; and clean air for 17 hours.
Day 6 (100 ug N0,/m3) - Clean air spiked at 733 ug NCL/m3 for 1 hour; clean
3
air for 1 hour; clean air spiked at 366 ug N0,/m
for 1 hour, clean air for 1 hour; clean air spiked at
733 ug NCL/m for 1 hour; clean air for 1 hour, clean
3
air spiked at 366 ug N0,/m for 1 hour, and clean
air for 15 hours.
Day 7 (100 ug N02/m3) - Clean air spiked at 275 ug N02/m3 for 8 hours, and
clean air for 14 hours.
Day 8 (0 ug NC^/m3) - Clean, NCyfree air.
Phase III
The possibility of 0, interference was investigated by sampling known mixtures of
NO, and 0, in clean air. A 6-day sampling schedule (22-hour sampling periods) was
planned to obtain data for each of the four methods using clean air containing the
following mixtures of NO, and 0,:
Nitrogen dioxide, Ozone,
Day ug/m
1 75
2 75
3 75
4 150
5 150
6 150
CALIBRATION
Initially, each of the chemiluminescence and colorimetric instruments was
subjected to a 5-point calibration using the gas phase titration procedure described
in the Federal Register of June 8, 1973 . Calibration of the manual bubbler methods
was accomplished statically as described in the procedures (Appendices A-3 and A-4).
Recalibrations were repeated as necessary.
DATA ACQUISITION
Responses of the automated analyzers to the NO, in the air sample were recorded on
individual matched strip chart recorders. The strip chart traces were processed by
the manual optical averaging procedure to obtain hourly average concentrations of N02
from which average concentrations for each 22-hour sampling period were computed. This
-------
process is described in detail in Chapter 3. Concentration data from the manual
bubbler methods were obtained by the process described in the procedures (Appendi-
ces A-3 and A-4).
-------
3. EXPERIMENTAL
METHODS AND INSTRUMENTATON
The study was designed to simultaneously obtain N02 concentrations from a common
source using duplicate analyzers for each of two automated N02 methods and collecting
quadruplicate samples by each of two manual methods. Ambient concentrations of NO, CO,
C02, 03, TS, and TSP were also measured. Table 1 lists all the methods used for the
study, an identification code for each, a designation as to whether the method is auto-
mated or manual, the operating range, and the sample flow requirements.
Comparison Measurements
Two model 8101-B chemiluminescence NO-M^-NOx analyzers (Bendix Corporation) and
two Air Monitor IV colorimetric analyzers (Technicon Instruments Corporation) were
used to represent the automated N02 methods. The operational instructions and the
calibration procedures for each of these methods are described in Appendices A-l and
A-2, respectively. Two gas bubbler samplers (Research Appliance Company), each capable
of collecting five samples simultaneously, were utilized to collect two sets of dupli-
cate samples by each of the manual methods.
Ambient Measurements
A third Bendix model 8101-B NO-N02-NOX analyzer, a Bendix model 8002 chemilumines-
cence Oj analyzer, a Bendix model 8301 flame photometric TS analyzer, a model LIRA
200 nondispersive infrared (NDIR) CO analyzer (Mine Safety Appliance Company), and a
model 215AL NDIR C02 analyzer (Beckman Instruments, INc.) were used to make ambient
pollutant measurements. A standard high volume sampler (General Metal Works) was
used to collect ambient TSP samples. The C02 analyzer was modified by mechanically
offsetting the zero to 300 ppm C02 to provide a measurement range of 300 to 700 ppm.
An integral strip chart recorder was used with each of the continuous analyzers to
record the sensor response.
-------
Table 1. POLLUTANT MEASUREMENT METHODS
Method
Comparison Measurements
Chemi luminescence (N0?)
Chemi luminescence (N00)
c
Colorimetric (NO-)
Colorimetric (NO-)
Sodium arsenite (NOp)
TGS-ANSAb (N02)
Analyzer
CM-1
CM- 2
C-l
C-2
ARS-1
ARS-2
ARS-3
ARS-4
TGS-1
TGS-2
TGS-3
TGS-4
Chemi luminescence (O.J ; 0,
Ambient measurements
Chemi luminescence (NO+NO-
Chemi luminescence (0.,)
j
Flame photometric (TS)
Nondispersive
infrared (CO)
Nondispersive
infrared (COp)
Hi-Vol sampler (TSP)
Automated
X
X
X
X
X
CM- 3 X
°3
TS
CO
X
X
X
co2
TSP
X
Manual
X
X
X
X
X
X
X
X
X
Range,
ug or
ppm
0-0.5
0-0.5
0-0.5
0-0.5
0.01-0.4
0.01-0.4
0.01-0.4
0.01-0.4
0.01-0.4
0.01-0.4
0.01-0.4
0.01-0.4
0-0.5
0-0.5
0-0.5
0-1.0
0-50
300-750
-
Flow,3
ml/min
139
137
290
290
200
200
200
200
200
200
200
200
1000
135
1000
130
2000
2000
1.7e
Sample flow rate in millilieters per minute.
TGS-ANSA - Triethanolamine, guaiacol, sodium metabisulfite, 8-anilino-
1-naphthalene sulfonic acid ammonium salt.
This analyzer was used to monitor 0- in the N02 sampling manifold during
Phase III.
Total Sulfur (TS) calibrated as Sulfur dioxide.
Cubic meters per minute.
Range in pg or ppm.
-------
SAMPLING SYSTEM
The general sampling system schematic is shown in Figure 2. The manifold
system was arranged so that ambient air could be drawn from the DAMDF ambient air
manifold into a specially constructed N02- sampling manifold for sampling by the
four methods under comparison. The DAMDF manifold is a multi-sampling port,
all-glass system through which an excess of ambient air is drawn at a rate of
approximately 10 liters per minute (liter/min). The intake of this manifold is an
inverted glass funnel located 1.2 meters above the roof of the building (6 meters
from ground level). The special NCy sampling manifold for the four methods under
test was constructed from 2.5-centimeter (cm) internal diameter (I.D.) glass tubing
and was designed to allow the uniform addition of known concentrations of gaseous
pollutants to a clean or ambient air stream prior to introduction into the sampling
manifold.
Ambient air was drawn through the NCL-sampling manifold using a vacuum pump
with a flow control valve and vacuum regulator that allowed the ambient air flow
to be varied from 0 to 20 liters/min. A mass flow meter placed downstream from
the sample ports provided an accurate measurement of this flow. The total flow
through the system was the flow rate determined by the mass flowmeter plus the total
air flow required for each of the test methods. The sample flow requirements for
each of the analyzers and each of the bubbler samplers is given in Table 1. Air
from each of the manifold sampling ports was analyzed for NO, to assure that the
clean or ambient air and the added NCL were uniformally mixed and distributed
throughout the N0?-sampling manifold.
As shown in Figure 2, the sample intakes of the two chemiluminescence NCL
analyzers were connected to a single 500-milliliter (ml) mixing flask, which was
connected to the NO.,-sampling manifold. This mixing flask was used to minimize
the occurrence of negative NO, responses caused by rapid changes of NO and N02 con-
centrations in the ambient air and to compensate for response time differences between
the chemiluminescence and the colorimetric analyzers. The 0, analyzer shown connected
to the NO., sampling manifold was used only during Phase III.
Pollutant Generation
The pollutant generation system shown in Figure 2 was used for the addition
of N02 and Oj (Phase III) to the ambient air or the clean air in the NO.-sampling
manifold. Gravimetrically calibrated permeation tubes were used for generating
NO, in a system similar to that described by Scaringelli, O'Keefe, Rosenberg, and
5 3
Bell . Nitrogen dioxide is introduced at a constant rate (100 on /man) into the
air stream, which is passed through the mixing chamber and then through the sampling
manifold. The N02 concentration in the sampling manifold is the sum of the amount
added and the amount present in the ambient air. The amount added was varied to
10
-------
OAHDF AIR DISTRIBUTION SYSTEM
AMBIENT
AIR IN
TRAP
&
1
1 1
1 ^-T=5T N
^•J S Cfr A
Jk
& ^=
BLOVI
H02 SAMPLING SYSTEM
CM -3
—
AMBIENT
AIR IN
^-\
TS
1
fr
co M
II
03
GENERATOR '
PRESSURE
REGULATOR
I
I POLLUTANT GENERATION SYSTEM |
Figure 2. Sampling system. (See Table 1 for identification codes.)
-------
obtain different NO^ levels for each sampling period by either changing the total
ambient air flow rate through the sample manifold, the number of permeation tubes
in the chamber, or both. Permeation tubes with rates ranging from 1 to 2 ug NCL/min
were used in combination with ambient air dilution rates from 5 to 20 liters/min
to allow the addition of NCL at concentration levels ranging from 0 to 800 pg/m .
The flow rate through the NCL-sampling manifold was never lower than 5 liters/min
to assure that an excess of air was always available for all the methods under
test, which required a total of approximately 2.5 liters/min.
Ozone was generated by passing clean air through an ultra-violet irradiated
quartz tube into the NCL-sampling manifold. The concentration of 0^ in the sampling
manifold was varied by increasing or decreasing the surface area of the quartz tube
exposed to the ultra-violet source.
TEST ATMOSPHERES
In Phase I, the N0~-sampling manifold was connected to the DAMDF ambient air
manifold as shown in Figure 2. The NO- was added to the ambient air stream from
the DAMDF manifold in accordance with the planned random sampling schedule at the
four N02 addition concentrations. Simultaneous ambient air measurements were made
from the DAMDF manifold.
In Phase II, the NO, sampling manifold was disconnected from the DAMDF ambient
air manifold, and clean dilution air was introduced into the ICU-sampling manifold
through the 0, generator with the ultra-violet source turned off. The N02 was
added intermittently to this clean air stream in strict accordance with the planned
sampling schedule (Figure 1).
Phase III was conducted using the same sampling system used in Phase II. The
N02 was added continuously to the clean air stream. Ozone was added continuously
to this air stream by turning on the 0, generator and adjusting the ultra-violet
exposure of the air stream to obtain the desired 0, concentrations. The 0, con-
centrations in the sampling manifold were determined by the chemiluminescence 0,
analyzer shown in Figure 2.
START UP AND CALIBRATION
The chemiluminescence NO-N02 analyzers and the automatic colorimetric N0?
analyzers were thoroughly examined by representatives of the respective analyzer
manufacturers to assure that the instruments were operating properly and in con-
formance with the recommended procedures prior to initiation of the study. The
chemiluminescence analyzers were operated on the 0 to 0.5 ppm (0 to 940 ug NO^/m )
full-scale range, which is the most sensitive and normally recommended operating
range for ambient measurements. The operating range of the colorimetric analyzers was
adjusted from the manufacturer's recommended operating range of 0 to 1.15 ppm to
0 to 0.5 ppm by a simple span control adjustment. This adjustment was made on the
output of the colorimeter and did not affect the performance characteristics of the
analyzers.
12
-------
Initially, each continuous analyzer was set up and started in strict accordance
with the manufacturer's operating instructions. Each analyzer's zero control was
adjusted, using zero air, to obtain a * 5 percent offset on the recorder chart in
order to facilitate observing negative drift. Each analyzer's span control was
adjusted using dynamic standards to obtain a full-scale response equivalent to the
range given for each analyzer in Table 1. A 5-point calibration curve was constructed
to verify linearity for each of the analyzers. The chemiluminescence NO- analyzers
I and the colorimetric analyzers were calibrated using dynamic NO, standards prepared
3
by the gas-phase titration technique. A high pressure steel cylinder containing
approximately 100 ppm of NO in nitrogen was used as the standard. This cylinder
was assayed by gas-phase titration with known concentrations of 0, from an 0,
3
generator calibrated with the neutral buffered potassium iodide (KI) procedure.
A Bendix Model 8851 calibration system was used to perform the gas-phase titrations.
Following the initial calibration of each of the chemiluminescence and colorimetric
N02 analyzers, the instruments were zeroed and spanned daily throughout the ex-
periment. The continuous ambient 0,, TS, CO-, and CO analyzers were zeroed daily
and spanned at the beginning and end of the study.
The colorimetric analyzers were also statically calibrated using sodium
nitrite standard solutions in water as recommended by the manufacturer. This
calibration is based on the empirical observation that 0.72 mole of sodium nitrite
produces the same color as 1 mole of NO,. The static calibration produced a
calibration curve with a slope that was 94 percent of the slope of the calibration
curve obtained by using the gas-phase titration technique. Dynamic gas-phase
titration calibration was used for this study to be consistent with the procedure
followed for the chemiluminescence analyzers.
The manual methods were calibrated statically using sodium nitrite standards
made up in the respective absorbing solutions for each method. Dynamic calibra-
tion of the manual methods was not required because the NO, collection efficiencies
£ rt £•
for each method has been predetermined, " and appropriate correction factors
are included in the procedures.
The flame photometric TS analyzer was calibrated with a gravimetrically
calibrated sulfur dioxide (S02) permeation tube. The NDIR CO and CXL analyzers
were calibrated using standard calibration gases in high pressure cylinders. The
0, analyzers were calibrated using an 0, generator calibrated by the neutral
3
buffered KI procedure.
SAMPLING AND ANALYSIS PROCEDURES
The four methods were compared for 22-hour periods consistent with the
recommended sampling period for the manual methods. Daily samples were collected
for each of the manual methods for approximately 22 hours between 12:00 noon and
10:00 a.m. the following day. The automated methods, including the ambient
analyzers, were operated concurrent with this period. During the 2 hours from
13
-------
10:00 a.m. to 12:00 the automated NO, analyzers were zeroed and spanned and
readjusted to the conditions of the initial calibration, if necessary; the NC^
addition concentration was changed for the next day's experiment; and the eight
absorption bubblers for the manual method were changed.
Prior to beginning each sampling period, the continuous NC^ analyzer's sample
inlet lines were connected to the calibration system containing clean zero air.
The zero response for each analyzer was adjusted to 5 percent of chart, and the
zero adjustment settings were recorded. A known NCL concentration between 60 and
90 percent of full scale (0.3 to 0.45 ppm) was generated using the gas-phase
titration technique. After allowing the analyzers to reach a stable reading, the
span control for each analyzer was adjusted to obtain a response equal to the
response at the initial calibration. The appropriate permeation tubes were placed
in the permeation tube chamber, and the air flow through the sampling manifold
was adjusted to obtain the desired NCL addition concentration for the day's
experiment. During Phase I of the study, the N02 was added to the ambient air
stream. During Phases II and III, the NCL was added to a clean air stream. All
sample inlet lines of the NO- analyzer were then transferred from the calibration
system to the sampling manifold.
Two absorption tubes for each of the manual methods were placed in each of
the two gas samplers, and the initial flow measurements for each of the bubbler
trains were made and recorded. The absorption tubes for each method were
positioned alternately in the gas samplers as shown in Figure 2. These positions
were reversed daily to eliminate the possibility of manifold effects within the
gas samplers.
A clean, preweighed, 8- x 10- inch glass fiber filter was placed on the
high volume sampler. When the colorimetric analyzers stabilized at the new N02
concentration, the sampling period was initiated by the start of the gas samplers
and the high volume sampler. An initial air flow measurement for the high volume
sampler was taken immediately after the start of the sampling period.
At the end of the sampling period, the final flow measurements for each
bubbler train and the high volume sampler were made, the samplers were shut off,
and the time was recorded. The absorbing tubes were removed from the gas sampler
and stored for subsequent analysis. The unadjusted zero and span control settings
for each analyzer were recorded and compared with the settings at the beginning
of the sampling period to verify that they were not changed during the sampling
period.
All analyzer inlet lines were transferred from the NCL-sampling manifold to
the calibration system. An unadjusted zero was recorded for each analyzer and
then readjusted, if necessary, to 5 percent of chart. An NCU concentration
between 60 percent and 90 percent of full scale was generated, and after allowing
the analyzers to reach a stable reading, the unadjusted span readings were
recorded. The span controls for each analyzer were then adjusted, if necessary,
14
-------
to obtain a response equal to the response at the initial calibration. The new
zero and span control settings were recorded. All NCL analyzer inlet lines were
transferred from the calibration system to the NCL'S^ling manifold, containing
the N02 addition concentration for the next test day. The absorbing tubes for
the next sampling period were placed in the gas samplers, the appropriate flow
measurements were made, and the next sampling period was initiated.
Figure 3 is an example of a typical strip chart from a day's experiment for a
chemiluminescence NCU analyzer. Between 10:00 a.m. and 12:00 noon before the start
of the Day 2 test, the analyzer was zeroed and a known concentration of N0? was
measured by the analyzer (unadjusted span); the analyzer response was then adjusted
to the initial calibration response (adjusted span). The analyzer sample inlet
was then put back on the NCL-sampling manifold. At 12:00 noon the experiment
was initiated by the start of the gas samplers (manual methods). At 10:00 a.m.
the following day, the gas samplers were stopped, and the analyzers were zeroed
(unadjusted zero) and readjusted if necessary (adjusted zero). An unadjusted
span was taken and readjustment was made, if necessary, and preparations for
the next day's experiment were made.
Strip chart values from the automated methods were converted to hourly
averages for each day's experiment and recorded on data forms for submission for
statistical analysis. Only hourly averages for time periods concurrent to the
manual methods sampling periods were submitted for analysis. A base line for
zero corrections was made by drawing a straight line between the adjusted zero
at the beginning of each day1s experiment and the unadjusted zero at the end of
the day's experiment. The average reading in chart divisions for each hour was
determined visually by drawing a straight line parallel to the chart division
lines such that the area bounded by the trace and the two hour lines was equal on
both sides of the line. This is illustrated in Figure 3 by hours 1200-1300 and
1400-1500. The baseline reading in chart divisions is subtracted from the hourly
average reading, and this reading is converted to NCL concentration in ppm by
reference to the calibration curve. Concentration values in ppm were converted
to ug NCL/m by multiplying the ppm value by 1880.
Manual Sample Preparation and Analysis
Absorbing reagent for 3 weeks of operation (IS liters) was prepared and stored
in polypropylene carboys. One week's supply of absorption tubes was prepared
at the beginning of each week and delivered to the DAMDF for sampling. Poly-
propylene sampling tubes were scrubbed with soap and water solution and were
rinsed with tap water followed by three individual distilled water rinsings.
The sampling tubes were segregated by absorbing reagent to dry. Orifice bubblers
were cleaned in a 1:3:6 HC1-HNO,- water solution and rinsed in a distilled water
bath followed by forcing distilled water through the bubbler at least three times.
The inside diameter of each orifice bubbler was checked with jewelers bits, and
only orifice bubblers within 0.35 ^ 0.05 millimeter diameter were accepted for the
15
-------
ro>
o
o
o
o
o
o
o
o
o
CORRECTED
-—BASELINE
10
30 40 50 60 70 80 90 100
START MANUAL METHOD (DAY 2)
ADJUSTED ZERO
UNADJUSTED ZERO
STOP MANUAL METHOD (DAY 1)
Figure 3. Example of strip chart for chemiluminescence N02 analyzer.
16
-------
study. Assembly of the absorption tubes required adding 50 ml of the respective
absorbing reagent, capping the tube, and marking the day, sampler number, and
sampler port position on the tube.
The collected bubbler samples were returned weekly to the laboratory for
analysis. The absorption tubes were uncapped, the absorbing solution was forced
from the orifice bubbler, and each orifice bubbler was rinsed with distilled water.
The rinsings plus additional distilled water were added to the collection tube to
bring the tube to the original 50-ml volume.
Each week's samples were analyzed in accordance with the procedures given in
Appendices A-3 and A-4. Minor modifications to the methods were required. Sulfa-
nilamide solution, coupling reagents (ANSA, NEDA), and nitrite standards were made
up in volumes twice that specified in order to conveniently accommodate the large
number of samples to be analyzed. The nitrite standard concentrations for both
methods were altered to 0.1, 0.2, 0.5, 0.8, 1.0, and 1.4 ug N02"/ml in order to
improve the accuracy of analysis at lower concentrations. Absorbance measurements
were made on a Gary 14 spectrophotometer at 540 nanometers (nm).
Nitrogen dioxide concentrations in pg/m were recorded on data cards, which were
initiated when the sample was taken. The date, the sample identification code, the
starting and ending time, and the starting and ending flow rates were also recorded
on these cards.
QUALITY CONTROL
All of the procedures were subjected to rigorous quality control checks through-
out the study. The automated NO- analyzers were dynamically zeroed and spanned daily
and, if necessary, readjusted to initial calibration conditions. For the duration of
the study, unadjusted zero and unadjusted span data were recorded daily, allowing the
computation of the zero and span drift data given in Table 2. Critical parameters
for each automated method -- such as flow measurements, flow meter readings,
vacuum and pressure gauge readings, and span and zero knob settings -- were moni-
tored daily, and observations were recorded on analyzer check sheets. Flow
measurements and leak checks were made prior to each sample period using a mass
flow meter. Strip chart readings were periodically redetermined by different strip
chart readers. Hourly averages were estimated to the nearest +_ 0.005 ppm, and
two readers were in agreement within + 0.005 ppm. Twenty-two-hour averages
determined by two readers were in agreement within +_ 0.002 ppm.
Several routine precautions were taken to assure the integrity of the samples
collected by the manual methods. Collection tubes were alternated in the manual
gas samplers to minimize any possible manifold effects. Five unknown quality
control samples over a 0.20 to 1.40 ug NCL/ml range supplied by an independent
laboratory were analyzed and determined to be well within the expected variation
of each unknown concentration. Reanalysis of 10 percent of the samples demonstrated
that the average relative differences between the two analyses for the arsenite
17
-------
Table 2. ZERO DRIFT AND SPAN DRIFT DATA
Zero driftb
Number of recordings
Minimum value
Maximum value
Mean
Standard deviation
Span drift0
Number of recordings
Minimum value
Maximum value
Mean
Standard deviation
Analyzer*
CM-1
35
-33
30
n
8
33
-5. *
4.8
-e.20
2.20
CM- 2
34
-10
?.B
3
7
35
-4.6
f.3
P. 24
2.0?
CM- 3
21
-2
1
0
1
20
-3.0
d.5
0.22
2.27
C-l
34
-17
38
10
12
34
-10.7
13.1
0.23
A. 95
C-2
34
-5
19
3
4
34
-6.5
18.7
1.H2
1.38
a CM-1, CN-?, and CM-3 are cheroiluminescence instruments; C-l and C-2
are cotorimetric instruments (see Table 1).
Zero drift = adjusted zero - unadjusted zero (gg NO./m ).
i span-unadjust
adjusted span
° Span drift = adjustedjpan-unadjusted span x 1COj at fl(J ± ]Q percent
of full scale.
method was 1 percent with a range of 0 to 3.4 percent; for TGS-ANSA the average
was 2.7 percent with a range of 0.18 percent. Absorbance values derived from each
standard were compared between analysis periods and within each analysis period;
old standards were retained for comparison with newly prepared sets. Sets of five
standard solutions covering the analytical range of each method were analyzed at
the beginning, middle, and end of each analysis period. Computer-fitted, linear
least-squares calibration curves were constructed for each method for each analysis
period. A time schedule for addition of sample and reagents was devised in order
to minimize the time of the analysis period and to guarantee a consistent time
schedule for the analysis of both standards and samples. Thirty-five percent of
the samples were reanalyzed by an independent laboratory. The average difference
between laboratories was 4 percent for both manual methods throughout the study.
One group of shared arsenite method samples differed by 20 percent; however,
a close examination of the data generated for that sample period indicated that
an anomaly had occurred during the analysis by the independent laboratory.
18
-------
Two different types of blanks were analyzed with the samples for the purposes
of establishing the integrity of the analytical reagents and the absorbing solu-
tions. Analytical blanks were determined by analysis of the stock absorbing
solutions at the same time collected samples were analyzed. The TGS-ANSA method
analytical blank ranged from 0.005 to 0.020 ug NO^/ml; the arsenite method
analytical blank was equal to or less than 0.005 ug N02/ml. Field blanks were
determined by the analysis of the absorbing solution in tubes that were filled,
capped, sent to the sampling site, and treated in the same manner as all the
absorption tubes except that they were not connected to the sampler.
The field blanks ranged from 0.00 to 0.02 vg NO"/ml and from 0.00 to
0.21 pg NOZ/ml for the arsenite and TGS-ANSA procedures, respectively. The
TGS-ANSA field blank was lowered to 0.02 pg NO^/ml after extreme measures for
cleaning the collecting tubes and caps were taken. Subtracting the TGS-ANSA
field blank only disrupted the otherwise excellent agreement between the arsenite
and TGS-ANSA data. Considering the good agreement between the TGS-ANSA and the
arsenite data without field blank corrections and the fact that a field blank of
this type is not required in the TGS-ANSA procedure, field blank corrections were
not made.
19
-------
4. DATA SUMMARIES
Average NIL concentrations were computed for each method for each sampling period.
The method of computation depended on the method used and has either been discussed
in the previous chapter or is included in the method write-up. For convenience in
comparing results, all data acquired during the study are presented in this section.
Data for each phase are listed separately in Tables 3, 4, and 5. Twenty-two hour N02
concentrations obtained by the automated analyzers and by each of the manual methods
are given in these tables. Hourly averages were used for the computation of the
average NO, concentration for the period. Complete data sets from these tables were
subjected to statistical analysis. Incomplete sets are included, however, because
they provide additional information on the comparability of results produced by
some methods.
PHASE I
The average NO, concentrations for each of the methods tested are presented in
Table 3. The ambient concentrations of NO, CO, C02, 0,, TS, and TSP are also given.
All data represent concurrent sampling. For each period, two values are given for
the continuous automated methods and four values for each of the manual methods. An
estimate of total NO, concentration, which was obtained by adding the NO- addition
concentration to the ambient NO, measurement, is given for each period. This value
is designated as the total concentration. This estimate was not used for determining
accuracy of the methods because the gravimetric calibration history of the permeation
devices had not been sufficiently established, because of the possibility of reactions
of the added NO, with unknown constituents in the ambient air, and because the ambient
NO, measurements were made using one of the NO, methods under test. Twenty-two days
were required to complete this phase because 2 days were voided because of malfunc-
tions of one or more of the automated analyzers.
20
-------
TABLE 3. N02 METHODS COMPARISON DATA - PHASE I (22-hour averages)
(UQ NOo/ffl )
Day
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Zl
zz
Ambient Measurements
TSP
136
127
53
55
154
70
106
66
44
144
121
50
48
81
51
47
76
89
70
190
123
34
cou
3.9
3.3
6.2
1.4
1.6
1 .4
2.4
1.3
2.3
6.9
5.6
1 .3
1 .3
l.b
2.5
1 .3
2.1
2.3
1 .4
6.4
4.5
1.1
CO/
583
572
702
585
615
601
610
559
682
593
610
639
586
586
622
617
603
714
646
581
TS<|
28
10
5
29
24
8
26
3
5
18
8
5
0
8
10
8
26
13
24
31
16
0
°1
20
22
61
26
14
45
20
55
22
16
22
55
28
12
49
39
16
20
22
16
20
57
NO
146
125
12
18
242
20
48
2
45
302
192
2
15
33
1
7
40
59
10
264
130
0
N0?
103
85
34
54
66
58
85
51
58
62
81
28
36
51
24
3T
38
66
56
64
86
17
N02
Additions*
SpUr
75
0
45
83°
207
0
207
83
51
51
94
51
94
0
207
107°
102d
207
94,.
102tf
10Zd
0
TOTAL
178
85
79
137
273
58
292
134
109
113
175
79
130
51
231
141
140
273
150
166
188
17
"
Automated
Chemilu-
rinescence
CM-1
173
88
73
148
273
58
290
128
102
107
175
73
128
53
235
147
164
263
150
169
192
17
CM-2
165
86
73
152
276
58
288
132
100
109
179
62
126
49
233
145
165
265
150
147
194
13
Colon -
metric
C-l
164
81
66
147
274
49
286
147
103
103
173
94
128
53
252
-
148
261
169
W
197
15
C-2
164
75
62
143
267
47
284
130
98
98
169
68
132
43
237
-
145
259
147
.
182
8
Method
_ __ .
s
Manual
Arsenlte
Sampler A
ARS-1
172
96
80
146
" "273'
57
2>5
133
106
129
184
81
132
52
227
157
156
251
140
170
197
17
ARS-Z
169
94
68
155
262
64
281
13?
107
127
191
75
133
48
220
163
154
249
141
167
193
17
Sampler B
ARS-3
180
"87
75
144
Zed
66
286
143
109
izz
195
74
141
50
"~233
160
158
240
143
196
198
17
A1S-4
178
9b
73
150
270
65
28fr
133
lUb
1 26
198
74
144
53
'"221"
147
Ibl
254
140
187
196
18
--
T6S-ANSA
Sampler A
TGS-1
'164
y;;
64
141
2/4
61
Z/9
128
y:>
ZUJ
Mi
74
Mi
49
""'219
\Vi
141
240
134
164
1/3
20
TGS-2
160
84
50
IJO
K/b
63
Wi
IJb
IUU
1 \i
175
72
122
49
215
162
150
245
138
163
174
21
. Sampler B
TGS-3
161
83
73
136
281
64
282
126
117
108
187
70
124
46
215
137
144
233
138
160
176
19
TGS-4
156
81
62
146
271
68
276
126
115
108
161
74
119
45
226
134
147
247
133
174
189
18
8 Ambient H02 measurement (N02) plus N02 added spike • estimate of N02 in sampling manifold (TOTAL).
b«1lligram/cubic meter (mg/m3).
c Total Sulfur as S02.
dResultant from 3-hour addition at 60p to 800
-------
Table 4. N02 METHODS COMPARISON DATA - PHASE II (22-hour averages)
(wg N02/m3)
Day
1
2
3
4
5
6
7
8
N02
Solke3
0
200
733
733
733
733/367
275
0
Hours .
SoikedD
0
22
6
3
3
2/ '2
8
0
Targetc
0
200
200
TOO
100
100
100
0
Automated
Cheml lu-
minescence
CM-1
0
152
201
-
90
88
79
0
CM- 2
0
150
197
115
90
92
- 77
0
Colori-
metric
C-l
0
156
211
117
90
85
75
0
C-2
0
158
209
117
88
94
85
0
Manual
Arsenite
Sampler A
ARS-1
0
162
219
115
95
98
87
4
ARS-2
0
170
211
120
94
99
83
3
Sampler B
ARS-3
0
174
221
120
93
97
88
2
ARS-4
0
167
218
116
97
98
87
2
TGS-ANSA
Sampler A
T6S-1
5
163
200
109
90
91
78
6
TGS-2
3
161
200
107
87
92
76
5
Sampler B
TGS-3
4
137
202
104
89
89
78
5
TGS-4
4
138
199
108
87
88
76
k 5
aN02 addition concentration.
Hours of N0? addition to clean air.
Estimated NO- concentration for the 22-hour sampling period.
-------
Table 5. N02 METHODS COMPARISON DATA - PHASE III (22-hour averages)
( ua N02/m3)
Nitrogena
dioxide
75
150
Ozone
98
353
676
98
353
667
Automated
Chemi lu-
minescence
CM-1
66
66
78
141
128
128
CM- 2
71
56
71
130
120
128
Col or i-
metric
C-l
73
41
36
139
105
56
C-2
64
34
32
141
107
51
Manual
Arsenite
Sampler A
ARS-1
69
73
87
152
142
141
ARS-2
73
70
88
159
144
139
Sampler B
ARS-3
72
71
85
151
139
138
ARS-4
72
72
85
155
140
138
TGS-ANSA
Sampler A
TGS-1
72
73
82
147
143
144
TGS-2
70
72
82
147
141
141
Sampler B
TGS-3
70
72
84
145
141
144
TGS-4
70
71
79
140
142
143
added to clean air.
Ozone added to clean air containing N0?.
-------
PHASE II
The average N02 concentrations as measured by each of the four methods
under a variety of controlled NCL concentration variations are given in Table 4.
The data from the manual methods were not corrected for clean air values (days 1
and 8) because no justification for doing so could be found. The NO- addition
schedule and computed average NCL concentration values (target) are also given.
As in Phase I, these target values were intended only as an estimate of NC^ con-
centrations. The poor agreement between the target values and the measurements
by the four NCL methods was attributed to a change in rate of one of the permeation
tubes used for generation of the NCL that was added to the sampling manifold.
PHASE III
The average NCL concentrations for each period as determined by the four
methods are presented in Table 5. The NCL and 0, concentrations introduced into
the sampling manifold for each of the six sampling periods are also listed.
24 /
-------
5. STATISTICAL ANALYSIS
The data collected during the study and shown in Tables 3 (Phase I),
4 (Phase II), and S (Phase III) were statistically analyzed to provide a basis
for comparing both the intramethod and intermethod performances of the four methods.
Data analysis consisted of the application of the following statistical procedures:
descriptive summaries by method, correlation analysis, multivariate regression
analysis, and analysis of variance. Results of these analyses are summarized and
presented as: (1) differences (Table 6), (2) cumulative frequency distributions
of differences (Table 7), and (3) intermethod ratios by concentration grouping
(Table 8).
Table 6 shows the mean difference, the standard deviation about the mean, the
95 percent confidence interval about the mean, and the correlation coefficient for
each method and method combination. Determination of statistical significance
was based on the observation that the confidence interval did not include zero.
Table 7 shows the distribution of individual daily differences within a
specific method and among combinations of methods. For example, the intramethod
comparison in Phase I for the chemiluminescence method shows that 90 percent of
the differences are 4 ug NCL/m (0.002 ppm) or less.
Table 8 lists the ratios of the average results of one method compared with
the average of another method. The ratios are arranged in concentration intervals
from which general observations about concentration effects on the ratios may be
easily seen. Each entry in the table is the average of concentration ratios
within the interval. For example, the first entry, 1.16, is an average of the
two (the number in parentheses) chemiluminescence/colorimetric concentration
ratios within the concentration range 0 to 50 ug NCL/m (0.027 ppm). The last
column is the average of all individual ratios across the five concentration
intervals.
The general results from the application of these four methods will be
discussed below. Details of these statistical analyses, when appropriate, are
provided in Appendix B.
Only complete data sets without missing observations for each method for a
given sampling period were considered for analysis. In addition, all complete data
sets were included in the analysis without regard to outliers. Essentially, the
judgment was made that if no valid reason could be found that would justify the
elimination of the outlier from the data set, then it must be considered to be
representative of the instrument/method capability. Mean differences in approxi-
mate percentages may be obtained by dividing the mean difference in pg NCL/m by
3
the average concentration in Phase I (ISO ug NCL/m ) and by the average concentra-
25
-------
Table 6. STATISTICAL ANALYSIS OF METHOD DIFFERENCE
(ug/m3)
OrrDarisor
Cheiail/Chenil
Color/Color
;-v --• (A)
•'< -" (B)
TGS/TGS (A)
TGS/TGS (B)
. i
"-V (B)
TGS (A)/
T3S (B)
!•" *£• = t "C
Chemil/Color
Chemil/occ
Cheml/TGS
Col or/ -"-.S
Color/TGS
A-S/TGS
P-?.se I
Stand.
Fairs r'ean Dev. r
22 1.3 5.6
20 7.5 7.5
22 0.6 5.6
22 0.2 5.6
20 -0.9 3.8
22 0.2 9.4
22 -2.6 5.6
20 0.6 7.5
20 3.8 7.5
22 -1.9 9.4
20 5.6 9.4
20 -3.8 11.3
18 3.8 11.3
20 7.5 7.5
C.I/
Lower Unoer
-1.1 3.8
3.8 11.3
-1.9 +3.8
-1.9 1.9
-1.9 .8
-3.8 +3.8
-5.G +0.1
-3.8 +3.8
0.0 +7.5
-5.6 +1.9
+0.9 9.4
-9.4 1.5
-1.9 9.4
+3.8 11.3
Corr.
Coeff."
0.999
0.995
C.997
C.996
0.999
0.992
C.997
C.996
C.994
C.991
0.990
r.989
0.985
C.994
Pairs
7
S
8
8
8
8
8
8
7
7
7
8
8
8
Phase II P."-e£e HI
Stand, 955 Corr. Stand 95S Corr-rf
fear3 Dev. b C.I.- Coeff? Pairs .«eana Dev. c C.I.C Coeff?
Lower Upoer Lower Uooer
0.6 2.5 -1.7 2.9 C.999 6 5.2 6.3 -1.4 11.8 0.984
-2.1 4.7 -6.0 1.8 0-998 6 3.5 4.6 -1.3 A. 3 0.998
0.0 5.0 -4.2 4.2 0-998 6 -1.5 3.7 -5.4 2.4 0.964
1.2 3.4 -1.6 4.0 C-999 6 -1.0 1.5 -2.6 0.6 0.999
1.4 1.3 0.3 2.5 C.999 G 1.3 1.2 0.0 2.6 0.999
0.4 2.2 -1.4 2.2 r.999 6 1.8 2.6 -0.9 4.5 0.998
-1.4 2.3 -3.3 0.5 c.999 6 1.7 1.8 -0.2 3.6 0.999
3.9 8.3 -3.0 10.8 C.993 6 1.3 2.0 -0.8 3.4 0.998
-4.0 4.6 -8.2 0.3 n.999 6 25.2 28.8 -5.0 5J.4 0.730
-8.4 6.8 -14.7 -2.1 C-999 6 -12.3 5.2 -17.8 -5.8 0.995
- .9 2.7 - 3.4 1.6 r.999 6 -10.3 5.7 -16.3 -4.3 0.992
-4.0 4.9 - 8.1 0.1 0.998 6 -37.5 2"3.8 -67.7 -7.3 0.743
4.0 6.5 - 1.4 9.4 r.998 6 -35.5 32.0 -69.1 -1.9 0.679
S.O 7.9 1.4 14.6 0.999 6 3.3 3.7 -0.6 7.2 0.985
Signed difference.
Standard deviation.
95 percent confidence interval of mean difference.
Correlation coefficient between paired values in calculating mean difference.
-------
Table 7. CUMULATIVE FREQUENCY DISTRIBUTIONS OF ABSOLUTE DIFFERENCE
BETWEEN DUPLICATE DAILY MEASUREMENTS
(vg N02/3)
Method
Intramethod
Phase I
Chemil
Color
ARS
TGS-ANSA
Phase II
Chemi 1
Color
ARS
TGS-ANSA
Phase III
Chemi 1
Color
ARS
TGS-ANSA
Intermethod
Phase I
Chemil /Col prd
Chemi 1/ARS°
Chemi l/TGSb
Color/AHS
Color/TGS
AP.S/TGS
Phase II
Chemil/Color
Chemil /ARS
Chemi 1/TGS
Color/ARS
Color/TGS
ARS/TGS
Phase III
Chemil/Color
Chefnil/ARS
Chemi 1/TGS
Color/ARS
Color/TGS
ARS/TGS
N
18
18
36
36
7
8
1C
16
6
6
12
12
20
20
20
20
20
20
7
7
7
7
7
7
6
6
6
6
6
6
Min.
0
0
0
0
0
0
0
0
0
2
0
0
1
0
0
2
0
0
0
0
0
0
0
2
0
4
2
4
2
0
10
0
2
2
0
0
0
0
0
0
2
0
0
3
1
1
2
1
1
0
0
0
0
0
2
0
4
2
4
2
0
20
2
2
2
2
0
0
1
1
5
2
0
0
3
3
2
3
3
2
0
3
1
3
1
4
4
11
8
14
4
0
30
2
4
2
2
0
0
1
1
5
2
1
1
4
5
4
5
5
4
0
5
1
6
3
7
4
11
8
14
4
0
Percent! les
40
2
4
4
4
0
2
1
1
7
4
1
1
4
6
5
7
5
6
0
5
1
6
3
7
18
11
8
34
34
2
50
2
6
4
4
2
2
1
2
7
4
1
1
5
6
6
8
7
7
1
8
1
6
4
8
18
11
8
34
34
2
60
2
6
4
4
2
2
4
2
8
5
2
2
8
7
9
11
8
8
2
8
2
7
5
9
23
12
11
36
36
4
70 80
2 4
9 17
6 8
6 9
2 4
2 9
4 5
2 2
10 10
7 7
-3 4
2 3
8 9
8 9
10 14
12 15
9 11
11 15
2 6
8 17
2 4
7 8
5 7
9 17
40 40
18 18
15 15
52 52
48 48
4 4
9U
4
23
11
13
4
10
8
3
11
9
4
5
10
15
15
10
19
17
11
18
5
11
10
18
74
18
18
85
89
10
Max.
8
26
13
90
4
10
8
4
11
9
7
5
18
18
23
26
25
18
11
18
5
*J
1 1
1 1
10
18
74
18
18
85
89
10
Duplicate daily measurements were made by the chemiluminescence
[.and colorimetric methods.
Quadruplicate daily measurements were made by the arsenite and
TGS-ANSA methods.
27
-------
Table 8. INTERMETHOD CONCENTRATION RATIOS BY CONCENTRATION INTERVAL3' b
Method
Concentration Interval, ngNO_/in
50 50-99 100 - 149 150 - 199 200
Avr.
Phase I
Chemil/Color
Chemil/ARS
Chemil/TGS
Color/Cheinil
Color/ ARS
Color/ TGS
ARS/Chemil
ARS/Color
ARS/TGS
TGS/Cheinil
TGS/Color
TGS/ARS
Phase 11
Chemil/Color
Chemil/ARS
Chemil/TGS
Color/Chemil
Color/ ARS
Color/ TGS
ARS/Chemil
ARS/Color
ARS/TGS
TGS/Chemil
TGS/Color
TGS/APS
Phase III
Chemil/Color
Chemil/ARS
Chemil/TGS
Color/Chemil
Color/ ARS
Color/ TGS
ARS/Chemil
ARS/Color
ARS/TGS
TGS/Chemil
TGS/Color
TGS/ARS
1.16 (2)
0.92
0.92
0.86
0.80
0.81
.09
.25
0.99
.09
.23
.01
1.00(2)
-t.
-^
1.00
_c
--
~\,
-c
_c
_c
-C
_c
1.08(4)
0.93
1.01
0.93
0.88
0.96
1.08
1.14
1.08
0.99
1.04
0.93
1 .00(3)
0.92
1,01
1.00
0.93
1.02
1.09
1 .08
1.10-
0.99
0.98
0.91
1.59(3)
0.88
0.91
0.63
0.62
0.64
1.13
1.61
1.03
1.10
1.57
0.97
1.00(6)
0.96
1.03
1.00
0.96
1.03
1.04
1.04
1.07
0.97
0.97
0.93
1.50(3)
0.89
0.90
0.66
0.68
0.70
1.12
1.47
1.01
1.11
1.43
0.99
1.05(4)
0.38
1.08
0.95
0.93
1.02
1.02
1.08
1.10
0.93
0.98
0.90
0.95(2)
0.91
1.00
1.05
0.95
1.05
1.10
1.05
1.10
1.00
0.95
0.91
1.00(4)
1.04
1.05
1.00
1.04
1.05
0.96
0.96
1.01
0.95
0.95
0.99
1.04
0.97
1.03
0.96
0.94
1.00
1.03
1.07
1.06
0.97
1.01
0.94
0.99
0.92
1.01
1.01
0.94
1.03
1.09
1.07
1.10
0.99
0.97
0.91
1.54
0.88
0.90
0.64
0.65
0.67
1.12
1.54
1.02
1.10
1.50
0.98
a Ratio shown is average of the first method concentration divided by the second
method concentration.
Number in parenthesis is number of days within interval.
c Ratio is undefined or zero.
28
-------
tions in Phase II and III (100 pg N02/m3).
PHASE I
Intramethod Comparison
The results of the analysis of data collected during the intramethod variabi-
lity study of Phase I (Table 3) are summarized in Tables 6 through 8. In all cases,
the averages shown are arithmetic averages and the correlations are product-moment
correlations.
It can be observed in Table 7 that, in general, each of the methods when used
in duplicate under controlled field conditions can yield data that agree remarkably
well. Based on the intramethod differences shown in Table 7, the best agreement
between duplicate data was shown by the chemiluminescence method followed by the
arsenite, TGS-ANSA, and colorimetric methods. The strong degree of association
was further evidenced by the fact that the correlation was always better than
0.99. The worst case was shown by the automated color imetric method. The average
differences between 22-hour NCL concentration values generated by the two
3
analyzers was 7.5 ug NCL/m (0.004 ppm), and the 95 percent confidence interval
3
of the mean differences was 3.8 to 11.3 ug NCL/m (0.002-0.006 ppm). The frequency
distribution of absolute differences between duplicate daily measurements given
in Table 7 shows that the maximum intramethod difference occurred with the
TGS-ANSA method. Although this difference of 90 yg NCL/m (0.047 ppm), which
occurred only once, was considered an outlier, it was not excluded from the
analysis because no justification for rejection could be found. Excepting this
value, the next highest maximum difference was 26 pg NCL/m (0.014 ppm), which
occurred in the comparison of the continuous colorimetric analyzers.
Intermethod Comparison
A study of the intermethod data (Table 6) showed a statistically significant
difference between concentration data produced by the TGS-ANSA and the chemi-
luminescence methods, with the TGS-ANSA values being lower on the average by
5.6 vjg NCL/m (0.003 ppm). TGS-ANSA method data are also significantly lower
3
than the arsenite method data on the average by 7.5 pg NC^/m (0.004 ppm). Again
the correlations are strongly positive, always above 0.98.
The intermethod comparisons shown in Table 7 indicate that the daily average
intermethod differences (based on the median or 50th percentile value) are from
5 to 8 ug NCL/m (0.003 - 0.004 ppm), which is slightly higher than the intra-
7
method differences of 2 to 6 ug NCL/m (0.001 - 0.003 ppm). The frequency distri-
butions in Table 7 show that the maximum differences between methods were greater
than 20 ug NOj/m (0.011 ppm) in these cases. These occurred with comparisons
of chemiluminescence versus arsenite, chemiluminescence versus TGS-ANSA, and
automated colorimetric versus TGS-ANSA.
29
-------
It can be seen in Table 8 that the average ratio over all concentration intervals
is always within +_ 10 percent. However, ratios involving colorimetric method con-
centrations below 50 ug NCL/m (0.027) differ from the others by as much as 20 percent.
Several trends can be seen in Table 8, such as (1) the ratio of the chemilumi-
nescence to arsenite data increases slightly with increasing NCL concentration,
and (2) all comparisons with the chemiluminescence and colorimetric methods show
an increase in ratios with increasing concentration up to 150 pg NCL/m .
The hourly average data produced by the automatic instruments were also
examined and are discussed in Appendix B. The concurrent ambient data collected
during the study for NO, CO, C02, 03, TS, and TSP were added to a multivariate
regression equation to examine the influence of these pollutants on the four NCL
measurement systems. The analysis showed that these pollutants, at the ambient
levels found, did not account for the small differences in results from the
various methods. This observation is consistent with the remarkably close agree-
ment among the results produced by the four methods.
PHASE II
The data collected during Phase II (Table 4) were analyzed for effects of
variable NCL addition on the results produced by the four methods.
Intramethod Comparison
From Table 6 it can be seen that there is a slight bias, amounting to
1.4 ug N02/m3 (0.001 ppm), in the TGS-ANSA sampler A, but this effect is not
sufficient to result in a sampler bias for the TGS-ANSA method. Based on duplicate
samples (Table 7), the best intramethod agreement is with the chemiluminescence
and TGS-ANSA methods, but the arsenite and colorimetric methods are almost as
self-consistent. As in Phase I, the correlation coefficients were always above
0.99. Comparing these results with those obtained from Phase I, we find that
the two continuous colorimetric instruments yielded results that agreed more
closely with each other; so, the instrument bias seen in Phase I was not observed
in Phase II. The small bias of 1.4 pg NCL/m (0.001 ppm) between the two samples
from TGS-ANSA sampler A is in contrast to Phase I results, but the practical
significance is minimal. In Table 7, a slight intramethod improvement can be
noted. This may be attributable to the fact that this portion of the study was
run on clean air to which NCL had been added instead of ambient air as was used
in Phase I.
Intermethod Comparison
Two comparisons, both involving the arsenite method, provided significant
data in this phase of the experiment. First, the arsenite method yielded values
significantly higher than those produced by the chemiluminescence method by an
average of 8.4 pg NCL/m (0.004 ppm). Similarly the arsenite data were higher
than those generated by the TGS-ANSA method by 8.0 yg N02/m3 (0.004 ppm). Based
on Phase I results, it is not surprising that the arsenite method produced higher
30
-------
values than the TGS-ANSA method, but the significant difference between the
chemiluminescence and arsenite methods is a surprise because this was not observed
in Phase I. Furthermore, it appears, when comparing the mean differences of
Phase I with Phase II (Table 6), that the chemiluminescence method behaved more
erratically than other methods in Phase II. For example, the mean difference
in Phase I with chemiluminescence/colorimetric changed from 3.8 to -4.0 vg N02/m
(0.002 to -0.002 ppm).
The intermethod differences (Table 7) observed in Phase II are quite similar to
those seen in Phase I. All comparisons of data involving the colorimetric method
show better agreement in Phase II than in Phase I, not only for the average but also
for the maximum differences observed. For example, the chemiluminescence/colorimetric
comparison shows a drop from 5 to 1 vg N02/m3 and 18 to 11 vg N02/m for the median
and maximum, respectively, when Phase I results are contrasted to Phase II results.
Again referring to Table 8, it can be seen that the average ratios are all within
^10 percent. The ratio comparisons involving the colorimetric method are now all
nearer unity.
The detailed procedure for statistically comparing the effects caused by the
N0? addition procedure on method performance is presented in Appendix B. The analysis
shows that the NCU addition procedure had an effect. If it can be assumed that a
desired concentration can be generated consistently from day to day, then the data
from this experiment would indicate that the shorter the NO, addition time, the higher
the relative results obtained. Perhaps the most significant fact is that each of
the four methods behaves similarly to changes in the NCL addition procedure. Hence,
if there were difficulties in generating a specific NCL concentration, all methods
reflected this problem.
PHASE III
The data collected during Phase III (Table 5) were analyzed for effects of
different concentrations of 0, on the measurement of NO- in a clean air-I^-O^ system.
Intramethod Comparison
No biases between the data generated by any of the methods resulting from the
addition of 0, were observed (see Table 6), Like Phases I and II, the correlation
coefficients are always above 0.9 indicating that there is a similarity of response
within methods. The data in Table 7 indicate that intramethod differences in
Phase III are much like those seen in Phases I and II.
Internetliod Comparison
Whereas in Phases I and II the average mean differences were always less than
10 pg N07/m (0,005 ppm), now the intermethod differences are greater than
20 ug N02/m3 (0.011 ppm), in three instances, each occurring when the colorimetric
method is compared with other methods. (The colorimetric concentration data are
lower than data obtained by the other methods.) In addition, the correlation co-
efficients in Phase III show less agreement than in Phases I and II. For example,
31
-------
the chemiluminescence/colorimetric correlation coefficients decreased from 0.99 in
in Phase I and II to 0.75 in Phase III. Statistically significant differences
are found in four instances, two involving the automated colorimetric method
(Color/ARS, Color/TGS-ANSA) and two in which the chemiluminescence method results
are compared with others (Chemi/ARS, Chemi/TGS-ANSA).
The intermethod differences shown in Table 7 dramatically demonstrate that
high concentrations of 0, interfered with the performance of the colorimetric
3
methods. Whereas average (median) intermethod differences of 5 to 8 yg NC^/m
(0.003 - 0.004 ppm) were seen in Phases I and II, average differences of 18 to
34 yg NCL/m (0.01 to 0.018 ppm) can be seen in Phase III data, with a maximum
T
difference of 89 yg NO^/m (0.047 ppm) between the colorimetric and TGS-ANSA
methods.
Referring once again to Table 8, it is obvious that something influenced
the performance of the colorimetric method. Whereas on the basis of Phases I
and II the ratios might be expected to be somewhere near unity (+_ 10 percent),
in Phase III the ratios involving the colorimetric method deviate from the norm
by around +_ 50 percent. The statistical analyses comparing the methods (Appendix B)
confirm that the colorimetric method differs significantly from the other methods
and that the higher the 0, concentrations, the greater the error introduced. This
conclusion is only valid for the commercial version of the colorimetric method
tested (Air Monitor IV, Technicon Instruments Corporation). Other commercial
versions of this method may or may not exhibit this interference from 0,.
STATISTICAL OBSERVATIONS
The primary observations that result from the statistical analyses are
summarized below.
Intramethod Comparisons
The overall agreement within each method throughout all phases of the study
is good, as characterized by a mean difference in results from instrument or
bubbler pairs of no worse than 7.5 yg N0_/m (0.004 ppm).
Intramethod data correlations for the four methods are better than 0.95.
This means that each method responded similarly to changes in NO. concentrations.
A small ( <10 percent), but statistically significant, bias exists between
the data generated by the two colorimetric instruments.
Fifty percent of the absolute differences between duplicate measurements by
any of the four methods is less than 10 yg N02/m (0.005 ppm) for each of the
four methods.
Intermethod Comparisons
The overall agreement among NO- concentration values generated by the
different methods is outstanding, as reflected by mean differences of less than
10 yg N02/m (0.005 ppm), except when the measurement systems were deliberately
challenged by elevated 0, concentrations (Phase III).
32
-------
Intermethod data correlations are all better than 0.95 in Phases I
and II.
Statistically significant mean differences were observed as follows
(Table 6):
1. In Phase I, the TGS-ANSA data were lower than the chemiluminescence
data by 5.6 ug NCL/m (0.003 ppm) and the arsenite data by 7.5 ug NC^/m
(0.004 ppm).
2. The arsenite data were higher than the chemiluminescence data by
8.4 ug NCL/m3 (0.004 ppm) in Phase II and by 12.3 ug N0-/m3 (0.007 ppm)
3
in Phase III and were higher than the TGS-ANSA data by 8.0 ug N02/m
(0.004 ppm) in Phase II.
When clear air spiked with both NCL and 0, was analyzed for N02, the inter-
method mean NCL concentration differences of the colorimetric method with respect
to those of the other methods were significantly lower at 0, concentrations of
3
350 vjg/m , indicating that the added 0, resulted in a negative NO, bias for the
3
colorimetric method. Average differences greater than 25 ug N0?/m (0.013 ppm),
over six times those found during the first two phases, were observed.
Fifty percent of the absolute differences between average NO, concentrations
3
found by use of the four measurement systems was less than 10 ug N02/m (0.005 ppm)
for all of the six possible comparisons. However, absolute differences in the
colorimetric method data in Phase III were 2 to 3 times higher, that is, 20 to
30 ug N02/m3 (0.011 - 0.016 ppm).
Average concentration ratios of results produced by the four methods for
all NO, concentration groupings are generally within +_ 10 percent, except in the
case of the colorimetric method when challenged with added 0,. In ambient air at
3
the concentration interval of 50 ug N02/m (0.027 ppm), ratios involving the
colorimetric method differ by as much as 20 percent.
33
-------
6. REFERENCES
1. Title 40 - Protection of Environment, National Primary and Secondary Ambient
Air Quality Standards, Appendix F - Reference Method for the Determination of
Nitrogen Dioxide in the Atmosphere (24-hour Sampling Method). Federal Register.
36(84):8200-8201, April 30, 1971.
2. Title 40 - Protection of Environment, National Ambient Air Quality Standards,
Approval and Promulgation of State Implementation Plans. Federal Register.
37^(115): 11826-11848, June 14, 1972.
3. Title 40 - Protection of Environment, National Primary and Secondary Ambient
Air Quality Standards, Reference Method for Determination of Nitrogen Dioxide.
Federal Register. 38(110): 15174-15180, June 8, 1973.
4. Baumgardner, R. E., T. A. Clark, J. A. Hodgeson, and R. K. Stevens.
Determination of an Ozone Interference in the Continuous Saltzman N02 Procedure.
Analytical Chemistry. 4^(3):515-521, March, 1975.
5. Scaringelli, F. P., A. E. O'Keefe, E. Rosenberg, and J. P. Bell. Presentation
of Known Concentrations of Gases and Vapors with Permeation Devices Calibrated
Cravimetrically. Analytical Chemistry. 42:871, 1970.
6. Beard, M. E. and J. H. Margeson. An Evaluation of the Arsenite Procedure
for the Determination of Nitrogen Dioxide in Ambient Air. Environmental Pro-
tection Agency, Research Triangle Park, North Carolina. Publication Number
EPA-650/4-74-048. November 1974.
7. Mulik, J. D., R. G. Fuerst, J. R. Meeker, M. Guyer, and E. Sawicki. A
Twenty-four Hour Method for the Collection and Manual Colorimetric Analysis of
Nitrogen Dioxide. (Presented at 165th American Chemical Society National Meeting,
Dallas, April 9-13, 1973.)
8. Fuerst, R. G. and J. H. Margeson. An Evaluation of the TGS-ANSA Procedure
for the Determination of Nitrogen Dioxide in Ambient Air. Environmental Pro-
tection Agency, Research Triangle Park, North Carolina. Publication Number
EPA-650/4-74-047. November 1974.
34
-------
APPENDIX A - 1.
TENTATIVE METHOD (CHEMIIUMINESCENCE PRINCIPLE)
FOR CONTINUOUS MEASUREMENT
OF NITROGEN DIOXIDE IN ATMOSPHERE
35
-------
CHEMILUMINESCENCE METHOD
1. Principle and Applicability
1.1 Atmospheric concentrations of nitric oxide (NO) can be measured by the
chemiluminescent reaction of ozone (0,) with NO at reduced or near atmospheric
i J
pressure . Nitrogen dioxide (NO,) is measured as NO in the system after conversion
•y "I +•
of NO- to NO. ' Air samples are drawn directly into the analyzer to establish a
NO response; then a switching valve directs the sample air through the converter
where the NO, is converted to NO. The photomultiplier measures the light energy
resulting from the chemiluminescent reactions of NO and 0^. By subtracting the
NO signal from the NO+N02(NOx) signal, the amount of N02 is determined. The sub-
tractive process is accomplished electronically. Total time for both measurements
is less than 1 minute.
1.2 The method is applicable to the measurement of NO, at concentrations
7 '
in the atmosphere ranging from 9.4 to 18,800 ug/m (0.005 - 10 ppm).
2. Range and Lower Detectable Limit
Z.I A wide variety of ranges can be used in the measurement of N02> Recom-
mended ranges are 0 to 376 ug/m3 (0 - 0.2 ppm), 0 to 990 ug/m (0 - 0.5 ppm),
0 to 1,880 Pg/m3 (0 - 1 ppm), 0 to 3,760 yg/m3 (0 - 2 ppm), and 0 to 18,800 ug/m3
(0 - 10 ppm). Separate ranges should be made available for NO, N02, and NOX, if
possible. These higher ranges are included because NO concentrations often
Jv
exceed 1 ppm.
2.2 The lower detectable limit of the chemiluminescence method for the measure-
ment of N02 at the 0 to 376 ug/m3 range is 9.4 ug/m (0.005 ppm).
3. Interferences
3.1 The chemiluminescent detection of NO with 0, is not subject to inter-
ference from any of the common air pollutants, such as 0^, N02, carbon monoxide
(CO), ammonia (NH?), or sulfur oxides (SO ).
3.2 When the instrument is operated in the NO mode, any compounds that
A
may be oxidized to NO in the thermal N02 converter are potential interferents.
The principal compound of concern is ammonia; however, this is not an interferent
for converters operated at less than 330°C. Unstable nitrogen compounds, such
as peroxyacetyl nitrate (PAN) organic nitrites, decompose thermally to form NO
and may represent minor interferences in some polluted atmospheres.
4. Apparatus
4.1 General Description. Most analyzers consist of a particulate filter, a
thermal converter , an 0, generator, a reaction chamber, an optical filter, a
photomultiplier tube, and a vacuum pump. See Figure A-l for a general schematic
of the chemiluminescence analyzer.
5. Reagents
5.1 Oxygen. A cylinder of extra-dry oxygen is recommended as a source for
the generation of 0,.
36
-------
6. Calibration
6.1 Permeation Tube Method. Atmospheres containing accurately known amounts
of NCU at levels of interest can be prepared using permeation tubes. In the system
for generating these atmospheres, the permeation tube emits NCL gas at a known
constant rate, provided the temperature of the tube is held constant (*_ 0.1°C) and
provided the tube has been accurately calibrated at the temperature of use. The
NOp gas permeating from the tube is carried by a low flow of dry inert gas to a
mixing chamber in which it is accurately diluted with dry NCL-free air to the
level of interest. Systems for the preparation of standard atmospheres have been
described in detail by O'Keefe and Ortman ; Scaringelli, O'Keefe, Rosenberg, and
Bell , and Scaringelli, Rosenberg, and Rehme . Commercial calibration systems
using the permeation tube technique are now available.
6.1.1 Preparation of Standard Atmospheres. Permeation tubes may be prepared
or purchased. Scaringelli, O'Keefe, Rosenberg, and Bell give detailed, explicit
directions for permeation tube calibration. Tube permeation rates from 0.2 to
3.0 micrograms per minute, inert gas flow of about 50 milliliters per minute
(ml/min), and dilution flow rates from 1 to 20 liters per minute (liter/min),
conveniently give standard atmospheres containing desired levels of NO, (9.4 to
3
1880 gg/m ). The concentration of N02 in any standard atmosphere can be calculated
as follows:
(A-l)
where: C = Concentration of N02, ug/m at reference conditions
P = Tube permeation rate, ug/min
Rj = Flow rate of dilution air, liter/min at reference conditions
R, = Flow rate of inert gas, liter/min at reference conditions
6.1.2 Precautions for NO, Permeation Tube Use. When using W32 permeation
tubes the following precautions should be taken:
1. Tubes must be prepared from a pure, dry N02 source.
Precautions should be taken to assure that condensation is not
introduced as the tube is filled.
2. All dilution gases must be clean and dry.
3. Tubes should not be subjected to temperatures above 30°C or
below 20°C.
4. Tubes should be stored when not in use in a dry atmosphere with
continuous purging with about 50 ml/min of clean dry purge gas (nitrogen
or air].
5. The gravimetric calibration of the tube should be carefully checked
periodically during the lifetime of tube.
6.2 Gas Phase Titration Method. The gas phase titration method is described
in Reference 8.
37
-------
7. Procedure
7.1 Ambient air sampling is accomplished by following the procedure
described in 6.1 The sample is pulled into the instrument by a sample pump or
by vacuum from the vacuum pump used in the detector. Ozonized oxygen at a
constant flow is drawn into the detector; Figure A-l shows a typical flow diagram..
For exact operating procedures, refer to the manufacturer's instruction manual.
8. Calculation
8.1 Concentrations of NO and NO- found in the atmosphere can be obtained by
referring directly to the individual calibration curves.
9. References for Appendix A-l
1. Contijn, A., A. J. Sabadell, and R. J. Ronco. Homogeneous Chemilumi-
nescent Measurement of Nitric Oxide with Ozone. Anal. Chem. £2^6, 575,
1970.
2. Hodgeson, J. A., J. P. Bell, K. A. Rehme, K. J. Krost, and R. K. Stevens.
Application of a Chemiluminescent Detector for the Measurement of Total
Oxides of Nitrogen and Ammonia in the Atmosphere. In: Proceedings of the
Joint Conference on Sensing of Environmental Pollutants. Palo Alto, Calif.,
November 8, 1971. (American Institute of Aeronautics and Astronautics,
New York, N. Y.) Paper No. 70-1067.
3. Hodgeson, J. A., K. A. Rehme, B. E. Martin, and R. K. Stevens. Measure-
ments for Atmospheric Oxides of Nitrogen and Ammonia by Chemiluminescence.
(Presented at the Air Pollution Control Association Meeting, Miami,
June 1972, Paper No. 72-12).
4. Stevens, R. K. and J. A. Hodgeson. Application of Chemiluminescent
Reactions to the Measurement of Air Pollutants. Anal. Chem. 45{4):443-449,
April 1973.
5. O'Keefe, A. E. and G. C. Ortman. Primary Standards for Trace Gas Analysis.
Anal. Chem. 38_:760, 1966.
6. Scaringelli, F. P., A. E. O'Keefe, E. Rosenberg, and J. P. Bell. Permeation
of Known Concentrations of Gases and Vapors with Permeation Devices Cali-
brated Gravimetrically. Anal. Chem. 42_:871, 1970.
7. Scaringelli, F. P., E. Rosenberg, and K. A. Rehme. Comparison of Per-
meation Tubes and Nitrite Ion as Standards for the Colorimetric
Determination of Nitrogen Dioxide. Environ. Sci. Tech. £:924-929, 1970.
8. Title 40-Protection of Environment. Tentative Method for the Continuous
Measurement of Nitrogen Dioxide. Addendum C - Method for the Calibration
of NO, N0_, and N0x Analyzers by Gas-Phase Titration. Federal Register.
38(110):15178-15180, June 8, 1973.
38
-------
SAMPLE INLET
SOLENOID
VALVES
03
GENERATOR
I I
DETECTOR
HOUSING TEMPER-
ATURE CONTROL
OPTICAL FILTER
N
HIGH VOLT
AGE POWER
SUPPLY
THERMO-ELECTRIC
COOLED HOUSING
NO
TREND
N02
TREND
NOX
TREND
• NONOx
«GROUND
EXHAUST
PUMP
Figure A-l. Automated NO, NO2, NOx chemiluminescence analyzer.
-------
APPENDIX A - 2.
TENTATIVE METHOD (COLORIMETRIC PRINCIPLE) FOR CONTINUOUS
MEASUREMENT OF NITROGEN DIOXIDE IN ATMOSPHERE
40
-------
COLORMETRIC l^ETHOD
1. Principle and Applicability
1.1 This method is based on a specific reaction of nitrite ion (NO-) with
diazotizing-coupling reagents to form a deeply colored azo-dye that is measured
colorijnetrically. The nitrogen dioxide (NO-) in the ambient air is converted to
nitrite ion (NCL) upon contact with an absorbing solution containing the
diazotizing-coupling reagents. The absorbance of the azo-dye is directly pro-
portional to the concentration of NO- absorbed.
1.2 This method is applicable to the measurement of NO, at concentrations
3
in the ambient air from 18.8 to 1880 ug/m (0.01 - 1 ppm).
2. Range
2.1 A wide variety of ranges is possible. A nominal range of 0 to 80 ug/m
(0 to 1 ppm) with non-linear response is quite common for ambient monitoring.
Recently developed instruments are capable of giving linear response in ranges
of 0 to 376 ug/m (0 - 0.2 ppm) and 0 to 940 ug/m3 (0 - 0.5 ppm).
3. Interferences
3.1 Interferences from other gases that might be found in the ambient air
have been reported to be negligible ; however, most interferent studies have been
done on manual procedures and may not be applicable to continuous methods. Recent
studies indicate that ozone (03) produces a negative interference as follows:
ratio of N02 to 03 1:1 = S.S percent, 2:1 = 19 percent, and 3:1 = 32 percent.
4. Apparatus
4.1 General Description. Sample air is drawn through a gas/liquid contact
column at an accurately determined flow rate counter-current to a controlled flow
of absorbing reagent. All sample inlet lines prior to the absorber column should
be constructed of either glass or Teflon. The absorber must be carefully designed
and properly sized because N02 is somewhat difficult to absorb. Sufficient time is
allowed for full color development; then, the colored solution is passed through
a colorimeter in which the absorbance is measured continuously at about 550 nano-
meters.
4.2 Installation. Instruments should be installed on location and demon-
strated, preferably by the manufacturer, to meet or exceed manufacturer's
specifications and those described in this method.
4.3 Absorbing Solution. The two most widely used absorbing solutions for
this procedure are the Griess-Saltzman reagent and the Lyshkow modification
of the Griess-Saltzman reagent. Either of these is acceptable. The composition
of these solutions is as follows:
1. Griess-Saltzman. 0.5 gram(g) of sulfanilic acid, 50 milliliters (ml)
glacial acetic acid, and 50 ml of 0.1 percent N-(l-napthyl)-ethylene
diamine dihydrochloride diluted to 1 liter with deionized water.
2. Lyshkow. 1.50 g of sulfanilamide; 15 g of tartaric acid; 0.05 g of
41
-------
N-(l-napthyl)-ethylene diamine dihydrochloride; 0.05 g of 2-napthol 3,
6 disulfonic acid disodium salt; and 0.25 ml of Kodak photoflow (as
a wetting agent) diluted to 1 liter with deionized water.
5. Calibration
5.1 Permeation Tube Method. Atmospheres containing accurately known amounts
of NO, at levels of interest can be prepared using permeation tubes. In the
system for generating these atmospheres, the permeation tube emits NO, gas at a
known constant rate, provided the temperature of the tube is held constant
(^ 0.1°C) and provided the tube has been accurately calibrated at the temperature
of use. The NO, gas permeating from the tube is carried by a low flow of dry
inert gas to a mixing chamber in which it is accurately diluted with dry NO,-free
air to the level of interest. Systems for preparation of standard atmospheres
have been described in detail by O'Keefe and Ortman ; Scaringelli, O'Keefe,
Rosenberg, and Bell ; and Scaringelli, Rosenberg, and Rehme . Commercial
calibration systems using the permeation tube technique are now available.
5.1.1 Preparation of Standard Atmospheres. Permeation tubes may be prepared
or purchased. Scaringelli, O'Keefe, Rosenberg, and Bell give detailed, explicit
directions for permeation tube calibration. Tube permeation rates from 0.2 to
3.0 micrograms per minute (ug/min), inert gas flow of about 50 milliliters per
minute (ml/min), and dilution flow rates from 1 to 20 liters per minute
(liter/min) conveniently give standard atmospheres containing desired levels of
N02 (9.4 to 1880 ug/m ). The concentration of NO, in any standard atmosphere
can be calculated as follows:
c = LJUO3 (A.2)
Rd + R
where: C = Concentration of NO,, Pg/m at reference conditions
P = Tube permeation rate, ug/min
PI, = Flow rate of dilution air, liter/min at reference conditions
R, = Flow rate of inert gas, liter/min at reference conditions
5.1.2 Precautions for NO, Permeation Tube Use. When using NO, permeation
tubes the following precautions should be taken:
1. Tubes must be prepared from a pure, dry NO, source.
Precautions should be taken to assure that condensation is not introduced
as the tube is filled.
2. All dilution gases must be clean and dry.
3. Tubes should not be subjected to temperatures above 30°C or
below 20°C.
4. Tubes should be stored when not in use in a dry atmosphere with continuous
purging with about 50 ml/min of clean, dry purge gas (nitrogen or air.
5. The gravimetric calibration of the tube should be carefully checked
periodically during the lifetime of tube.
42
-------
5.2 Gas Phase Titration Method. The gas phase titration method is described
in Reference 7.
5.3 Preparation of Calibration Curve. A series (usually six) of standard
atmospheres containing NCU levels covering the operating range of the instru-
ment is prepared. Each atmosphere is sampled using exactly the same conditions
as will be used during atmospheric sampling. The concentration of NO, in
micrograms per cubic meter (x - axis) is plotted against instrument response
(y - axis), and the line of best fit is drawn.
6. Procedure
6.1 Instrument calibration is described in 5 above. For specific operating
instructions, refer to the manufacturer's manual. The instrument should be
calibrated dynamically at least once per month. Static calibration checks are
recommended daily or at least once per week. Most instruments have a static
calibration mode through which sodium nitrite standard solutions can be
introduced.
7. Calculations
7.1 The concentration is determined directly from the calibration curve.
No calculations are necessary.
7.2 NCL concentrations in micrograms per cubic meter are converted to
ppm as follows: ug ^3
ppm = (A-3)
1880
(where: 1880 represents the concentration of NC^ in pg/m equivalent to
1 ppm NCL by volume. The figure was generated for 25°C and 760 mm Hg).
8. References for Appendix A-2
1. Saltzman, B. E. Colorimetric Micro Determination of Nitrogen Dioxide
in the Atmosphere. Anal. Chem. 2^:1949, 1954,
2. Saltzman, B. E., Modified Nitrogen Dioxide Reagent for Recording
Air Analyzers. Anal. Chem. 32^135, 1960.
3. Lyshkow, N. A. A Rapid Sensitive Colorimetric Reagent for Nitrogen
Dioxide in Air. J. Air Pol. Control Assoc. 15;10, 481, 1965.
4. O'Keefe, A. E. and G. C. Ortman. Primary Standards for Trace Gas
Analysis. Anal. Chem. 38_:760, 1966.
5. Scaringelli, F. P., A. E. O'Keefe, E. Rosenberg, and J. P. Bell.
Permeation of Known Concentrations of Gases and Vapors with Permeation
Devices Calibrated Gravimetrically. Anal. Chem. £2_:871, 1970.
6. Scaringelli, F. P., E. Rosenberg, and K. A. Rehme. Comparison of
Permeation Tubes and Nitrite Ion as Standards for the Colorimetric
Determination of Nitrogen Dioxide. Environ. Sci. Tech. 4_:924-929,
1970.
-------
7. Title 40-Protection of Environment. Tentative Method for the Continuous
Measurement of Nitrogen Dioxide. Addendum C - Method for the Calibra-
tion of NO, N02, and N0x Analyzers by Gas-Phase Titration. Federal
Register. 38(110):15178-15180, June 8, 1973.
44-
-------
APPENDIX A - 3.
TENTATIVE METHOD {SODIUM ARSENITE PROCEDURE)
FOR DETERMINATION OF NITROGEN DIOXIDE IN ATMOSPHERE
-------
SODIUM ARSENITE PROCEDURE
1. Principle and Applicability
1.1 Nitrogen dioxide is collected by bubbling ambient air through a sodium
hydroxide-sodium arsenite solution to form a stable solution of sodium nitrite.
The nitrite ion (NO^) produced during sampling is reacted with phosphoric acid,
sulfanilamide, and N-l-(naphthyl) ethylenediamine dihydrochloride to form an azo
dye and then determined colorimetrically.
1.2 The method is applicable to the collection of 24-hour samples in the
field and their subsequent analysis in the laboratory.
2. Range and Sensitivity
2.1 The range of the analysis is 0.04 to 2.0 ug NOj/ml. Beer's law is
obeyed through this range (0 to 1.0 absorbance units). With SO ml of absorbing
reagent and a sampling rate of 200 cm /min for 24-hours, the range of the method
is 20 to 750 vg/m (0.01 to 0.4 ppm) nitrogen dioxide ,
2.2 A concentration of 0.04 yg NOl/ml will produce an absorbance of
approximately 0.02 with 1-cm cells.
3. Interferences
3.1 Nitric oxide (NO) is a positive interferent . The presence of NO can
increase the NO, response by 5 to 15 percent of the N02 sampled .
3.2 The interference of sulfur dioxide is eliminated by converting it to
sulfate ion with hydrogen peroxide before analysis .
4. Precision, Accuracy, and Stability
4.1 The relative standard deviations for sampling NO- concentrations of 78,
3 *
105, and 329 ug/m are 3, 4, and 2 percent respectively.
4.2 No accuracy data are available.
4.3 Collected samples are stable for at least 6 weeks.
5. Apparatus
5.1 Sampling. A diagram of a suggested sampling apparatus is shown in
Figure A-2.
5.1.1 Probe. A Teflon, polypropylene, or glass tube with a polypropylene
or glass funnel at the end.
5.1.2 Absorption Tube. Polypropylene tubes, 164-x 32-mm, equipped with
polypropylene two-port closures. Rubber stoppers cause high and varying blank
values and should not be used. A glass-tube restricted orifice is used to
disperse the gas. The tube, approximately 8 mm O.D. and 6 mm I.D., should be
fal
152 mm long with the end drawn out to 0.3 to 0.8 mm I.D.1 J The tube should be
positioned so as to allow a clearance of 6 mm from the bottom of the absorber.
5.1.3 Moisture Trap. A polypropylene tube equipped with two-port closure.
The entrance port of the closure is fitted with tubing that extends to the bottom
of the trap. The unit is loosely packed with glass wool to prevent moisture
entrainment.
5.1.4 Membrane Filter. A filter of 0.8 to 2.0 micrometers porosity.
46
-------
HVPODERMIC
NEEDLE
AIR
PUMP
BUBBLER
TRAP
Figure A-2. Sampling train for sodium arsenite procedures.
-------
5.1.5 Flow Control Device. Any device capable of maintaining a constant
flow through the sampling solution between 180 and 220 on /min. A typical flow
control device is a 27-gauge hypodermic needle, 3/8 inch long. (Most 27-gauge
needles will give flow rates in this range.) The device used should be protected
from particulate matter. A membrane filter is suggested. Change the filter
after collecting 10 samples.
5.1.6 Air Pump. Capable of maintaining a pressure differential of at
least 9.6-0.7 of an atmosphere across the flow control device. This value
includes the minimum useful differential, 0.53 atmospheres, plus a safety
factor to allow for variations in atmospheric pressure.
S.I.7 Calibration Equipment. A flowmeter for measuring airflows up to
275 cm /min within +_ 2 percent, a stopwatch, and a precision wet test meter
(1 liter/revolution).
5.2 Analysis
5.2.1 Volumetric Flasks. 50, 100, 200, 250, 500', and 1000 ml.
5.2.2 Graduated Cylinder. 1000 ml.
5.2.3 Pipets. 1, 2, 5, 10, and 15 ml volumetric; 2 ml, graduated in
1/10 ml intervals.
5.2.4 Test Tubes, approximately 20 x ISO mm.
5.2.5 Spectrophotometer. Capable of measuring absorbance at 540 nm.
6. Reagents
6.1 Sampling
6.1.1 Sodium Hydroxide. ACS reagent grade.
6.1.2 Sodium Arsenite. ACS reagent grade.
6.1.3 Absorbing Reagent. Dissolve 4.0 g of sodium hydroxide in distilled
water, add 1.0 g of sodium arsenite, and dilute to 1000 ml with distilled water.
6.2 Analysis
6.2.1 Sulfanilamide. Melting point, 165-167°C.
6.2.2 N-(1-Naphthyl)-ethylenediamine dihydrochloride (NEDA).
Best grade available.
6.2.3 Hydrogen Peroxide. ACS reagent grade, 30 percent.
6.2.4 Sodium Nitrite. Assay of 97 percent NaNO, or greater.
6.2.5 Phosphoric Acid. ACS reagent grade, 85 percent.
6.2.6 Sulfanilamide Solution. Dissolve 20 g of sulfanilamide in 700 ml
of distilled water; add, with mixing, 50 ml of concentrated phosphoric acid;
and dilute to 1000 ml. This solution, if refrigerated, is stable for 1 month.
6.2.7 jNEDA Solution. Dissolve 0.5 g of NEDA in 500 ml of distilled water.
This solution, if refrigerated and protected from light, is stable for 1 month.
6.2.8 Hydrogen Peroxide Solution. Dilute 0.2 ml of 30 percent hydrogen
peroxide to 250 ml with distilled water. This solution, if protected from light
and refrigerated, may be used for 1 month.
48
-------
6.2.9 Standard Nitrite Solution. Dissolve sufficient desiccated sodium
nitrite and dilute with distilled water to 1000 ml so that a solution containing
1000 ug NOl/ml is obtained. The amount of NaNO- to us® is calculated as follows:
G = _ x 100 (A-4)
A
where: G = Amount of NaN02» grams
1.500 = Gravimetric factor in converting NO- into NaN02
A = Assay, percent
7 . Procedure
7.1 Sampling. Assemble the sampling apparatus as shown in Figure A- 2.
Components upstream from the absorption tube may be connected, where required,
with Teflon or polypropylene tubing; glass tubing with dry ball joints; or
glass tubing with butt-to-butt joints with tygon, Teflon, or polypropylene.
Add exactly 50 ml of absorbing reagent to the calibrated absorption tube
(8.1.3). Disconnect the funnel, insert the calibrated flowmeter, and measure
the flow before sampling. If the flow rate before sampling is not between 180
and 220 cm /min, replace the flow control device and/or check the system for
leaks. Start sampling only after obtaining an initial flow rate in this range.
Sample for 24 hours and measure the flow after the sampling period.
7.2 Analysis. Replace any water lost by evaporation during sampling by
.adding distilled water up to the calibration mark on the absorption tube.
Pipet.10 ml of the collected sample into a test tube. Pipet in 1 ml of the
hydrogen peroxide solution, 10 ml of the sulfanilamide solution, and 1.4 ml of
the NEDA solution with thorough mixing after the addition of each reagent.
Prepare a blank in the same manner using 10 ml of the unexposed absorbing reagent.
After a 10-minute color-development interval, measure the absorbance at 540 nm
against the blank. Read ng NO^/ml from the calibration curve (Section 8.2).
Samples with an absorbance greater than. 1,0 must be reanalyzed after diluting
an aliquot (less than 10 ml) of the collected sample with unexposed absorbing
reagent.
8. Calibration and Efficiencies
8.1 Sampling
8.1.1 Calibration of Flowmeter. (See Figure 2). Using a wet test meter
and a stopwatch, determine the rates of air flow (cm /min) through the flowmeter
at a minimum of four different ball positions. Plot the ball positions versus
the flow rates.
8.1.2 Flow Control Device. The flow control device results in a constant
rate of air flow through the absorbing solution. The flow rate is determined
in Section 7.1.
8.1.3 Calibration of Absorption Tube. Calibrate the polypropylene
absorption tube (Section 5.1,1) by first pipeting in 50 ml of water or absorbing
149
-------
reagent. Scribe the level of the meniscus with a sharp object, go over the
area with a felt-tipped marking pen, and rub off the excess.
8.2 Calibration Curve. Dilute 5.0 ml of the 1000 ug NC^ /ml solution to
200 ml with absorbing reagent. This solution contains 25 ug N02 /ml. Pipet 1,
1, 2, 15, and 20 ml of the 25 ug NOj /ml solution into 100-, 50-, 50-, 250-, and
250-ml volumetric flasks and dilute to the mark with absorbing reagent. The
solutions contain 0.25, 0.50, 1.00, 1.50 and 2.00 ug NO^/ml, respectively.
Run standards as instructed in 7.2, including the blank. Plot absorbance versus
ug NOl/ml. A straight line with a slope of 0.48 +_ 0.02 absorbance units per
Mg NOl/ml, passing through the origin, should be obtained.
8.3 Efficiencies. An overall average efficiency of 82 percent was
obtained over the range of 40 to 750 ug/m N02>
9. Calculation
9.1 Sampling
9.1.1 Calculate volume of air sampled:
Fi + F? *
V = — ^ X T X 10"° (A-5)
2 3
where: V = Volume of air sampled, m
F, = Measured flow rate before sampling, cm /min
3
F, = Measured flow rate before sampling, on /min
T = Time of sampling, min
10" = Conversion of cm to m
9.1.2 Uncorrected Volume. The volume of air sampled is not corrected to
standard temperature and pressure because of the uncertainty associated with
24-hour average temperature and pressure values.
9.2 Calculate the concentration of nitrogen dioxide as pg N02/m
using:
„ XTI-I / 3 (|jg NO^/ml) X 50 ,. -,
ug N0,/m = •*-« 3 ' (A-6)
L V X 0.82
where: 50 = Volume of absorbing reagent used in sampling, ml
V = Volume of air sampled, m
0.82 = Collection efficiency
9.2.1 If desired, the concentration of nitrogen dioxide may be calculated
as parts per million N02 using:
ppm N02 = (ug N02/m3) X 5.32 X 10"4 (A-7)
10. References for Appendix A-3
1. Christie, A. A. et a^. Field Methods for the Determination of
Nitrogen Dioxide in Air. Analyst. 95_:S19-524, 1970.
2. Environmental Protection Agency, Research Triangle Park, N. C. 27711.
Unpublished results.
3. Merryman, E. L. et al. Effects of NO, O>2, CH4, H20 and Sodium
Arsenite on NO, Analysis. (Presented at 2nd Conference on Natural
SO
-------
Gas Research and Technology, Atlanta, June 5, 1972.)
4. Jacobs, M. B. and S. Hochheiser. Continuous Sampling and Ultramicro-
determination of Nitrogen Dioxide in Air. Anal. Chem. 30_:426, 1958.
5. Lodge, J. P. e_t al_. The Use of Hypodermic Needles as Critical Orifices
in Air Sampling. J. Air Pol. Control Assoc. 1^:197-200, 1966.
51
-------
APPENDIX A - 4.
TENTATIVE METHOD (TGS-ANSA PROCEDURE)
FOR DETERMINATION OF NITROGEN DIOXIDE IN ATMOSPHERE
52
-------
TGS-ANSA PROCEDURE
1. Principle and Applicability
1.1 Nitrogen dioxide (N07) collected by bubbling air through a solution of
^ 1
.triethanolamine (T), o-methoxyphenol (guaiacol) (G), and sodium metabisulfite CS). The
nitrite ion (NOl) produced during sampling is determined colorimetrically by
reacting the exposed absorbing reagent with sulfanilamide and 8-anilino-l-
naphthalenesulfonic acid ammonium salt (ANSA).
1.2 The method is applicable to the collection of 24-hour samples in the
field and subsequent analysis in the laboratory.
2. Range and Sensitivity
2.1 The range of the analysis is 0.025 to 4.0 gg NCL/ml. Beer's law is
obeyed throughout this range. With 50 ml of absorbing reagent and a sampling
rate of 200 cm /min for 24 hours, the range of the method is 20 to 700 wg N02/m .
2.2 A concentration of 0.025 yg NCL/ml will produce an absorbance of approx-
imately 0.025 using 1-on cells.
5. Interferences
3.1 At a N0? concentration of 100 pg/m , the following pollutants, at the
levels indicated, do not interfere: ammonia, 205 wg/m , carbon monoxide,
154,000 ug/m ; formaldehyde, 750 pg/m ; nitric oxide, 734 \ig/m ; phenol,
150 ug/rn ; ozone, 400 vg/m ; and sulfur dioxide, 439 vg/m .
3.2 A temperature of 40°C during collection of sample had no effect on
recovery.
4. Precision and Accuracy
4.1 On making measurements from standard nitrogen dioxide atmospheres, prepared
by using permeation devices, a relative standard deviation of 2 percent and a
collection efficiency of 93 percent were determined throughout the range of the
method.
4.2 Stability
4.2.1 The absorbing reagent is stable for 3 weeks before sampling,
and the collected samples are stable for 3 weeks after sampling.
5. Apparatus
5.1 Sampling. A diagram of a suggested sampling apparatus is shown
in Figure A-3.
5.1.1 Probe. A Teflon, polypropylene, or glass tube with a polypropylene
or glass funnel at the end.
5.1.2 Absorption tube. Polypropylene tubes 164-x 32-mm equipped with
polypropylene two-port closures. Rubber stoppers cause high and varying blank
values and should not be used. A glass tube restricted orifice is used to
disperse the gas. The tube, approximately 8 mm O.D. and 6 mm I.D., should
be 152 mm long with the end drawn out to 0.3 to 0.6 mm I.D. The tube should
be positioned so as to allow a clearance of 6 mm from the bottom of the absorber.
53
-------
BUBBLER
Figure A-3. Sampling train for TGS - ANSA method.
-------
5.1.3 Moisture Trap. A polypropylene tube equipped with a two-port closure.
The entrance port of the closure is fitted with tubing that extends to the bottom
of the trap. The unit is loosely packed with glass wool to prevent moisture
entrainment.
5.1.4 Membrane Filter. A filter of 0.8 to 2.0 micrometers porosity.
5.1.5 Flow Control Device. Any device capable of maintaining a constant
flow through the sampling solution between 180 and 220 cm /min is acceptable.
A typical flow control device is a 27-gauge hypodermic needle, 3/8 inch long.
(Most 27-gauge needles will give flow rates in this range.) The device used
should be protected from particulate matter. A membrane filter is suggested.
Change the filter after collecting 10 samples.
5.1.6 Air Pump. A pump capable of maintaining a pressure differential of
at least 0.6 - 0.7 of an atmosphere across the flow control device. This value
includes the minimum useful differential, 0.53 atmospheres , plus a safety
factor to allow for variations in atmospheric pressure.
5.1.7 Calibration Equipment. A flowmeter for measuring airflows up to
275 on /min within +_ 2 percent, a stopwatch, and a precision wet test.meter
(1 liter/revolution).
5.2 Analysis
5.2.1 Volumetric Flasks. Two 250-ml flasks, two 1000-ml flasks, three
200-ml flasks, seven 100-ml flasks, and one 500-ml flask.
5.2.2 Volumetric Pipets. One each, 2-, 3-, 9-, 10-, 20-, and.50-ml.pipets.;
seven 5-ml pipets.
5.2.3 Serological Pipets. One each 1-, 5-ml pipets graduated in l-/10.ml
divisions.
5.2.4 Test Tubes. Each approximately 20 x 150 mm.
5.2.5 Spectrophotometer. Capable of measuring absorbance at SSO nm.
5.2.6 Graduated cylinder. One 50-ml cylinder.
6. Reagents
6.1 Sampling
6.1.1 Triethanolamine [N(C2H4OH)3], Reagent grade.
6.1.2 o-Methoxyphenol (o-CH,OC6H4OH]. Also known by its trivial name,
guaiacol. Reagent grade. Melting point 27-28°C. (CAUTION: Technical.grade
material will not meet this specification and should not be used).
6.1.3 Sodium Metabisulfite (Na^SJD,.). ACS reagent grade.
6.1.4 Absorbing Reagent. Dissolve 20 g of triethanolamine, 0.5 g of
o-methoxyphenol, and 0.250 g of sodium metabisulfite consecutively in 500 ml of
distilled water. Dilute to 1 liter with distilled water. Mix thoroughly. The
solution should be colorless. This solution, if kept refrigerated, is stable for
3 weeks.
55
-------
6.2 Analysis
b.2.1 Hydrogen 1'croxLdc [ILO^J. ACS reagent grade, 30 percent.
6.2.2 Sulfanilamide [4-(II2N)C6H4S02N1I2]. Melting point 165-167°C.
6.2.3 8-Anilino-l-naphthalenesulfonic acid ammonium salt (ANSA)
(8-C,HJ-ilI-l-C10ILSO--ML+). Minimum analysis, 98 percent.
6.2.4 Sodium Nitrite, [NaN02]. ACS reagent grade, assay of 97 percent
NaN00 or greater.
6.2.5 Methanol, absolute [CI 1,011]. ACS reagent grade.
6.2.6 Hydrochloric acid, [I1C1]. Concentrated. ACS reagent grade.
6.2.7 Hydrogen Peroxide Solution. Dilute 0.2 ml of 30 percent hydrogen
peroxide to 250 ml with distilled water. This solution, if protected from
light and refrigerated, can be used for 1 month.
6.2.8 Sulfanilamide Solution (2 percent in 4N HC1). Dissolve 2.0 g of
sulfanilamide in 35 ml of concentrated 11C1 and dilute to 100 ml with distilled
water. Mix. This solution, if refrigerated, can be used for 2 weeks.
6.2.9 ANSA Solution. 0.1 percent weight per volume (wt/v) Dissolve
0.1 g ANSA in 50 ml of absolute mcthanol. Dilute to 100 ml with absolute
methanol in a volumetric flask. Mix. Keep stoppered when not in use to
minimise evaporative losses. Prepare fresh daily. (CAUTION: Older reagent may
result in lower ubsorbance).
6.2.10 Standard Nitrite Solution. Dissolve sufficient desiccated sodium
nitrite and dilute with distilled water to 1000 ml to obtain a solution con-
taining 1000 ug NOr /ml. The amount of NaNO, to use is calculated as follows:
G = ii|2£ x 100 (A-8)
where: G = Amount of NaN02, g
1.500 = Gravimetric factor inconverting ML into NaNO,,
A = Assay, percent
7. Procedures
7.1 Sampling. Assemble the sampling apparatus, as shown in Figure A-3.
Components upstream from the absorption tube may be connected, where required,
witli Teflon or polypropylene tubing; glass tubing with dry ball joints; or
glass tubing with butt-to-butt joints with tygon, Teflon, or polypropylene.
Add exactly 50 ml of absorbing reagent to the calibrated absorption tube
(8.1.3). Disconnect the funnel, insert the calibrated flow meter (8.1.1) into
the end of the probe, and measure the flow before sampling. Denote rate as F,.
3 *-
If the flow rate before sampling is between 180 and 220 cm /min, replace the
flow controlling device and/or check the system for leaks. Start sampling only
after obtaining ;ui initial flow rate in this range. Sample for 24 hours and
measure the flow after sampling by again inserting a calibrated flowmeter into
the probe, after removing the funnel. Denote rate as I;2.
7.2 Analysis. Replace any water lost by evaporation during sampling by
-------
adding distilled water up to the calibrated mark on the absorption tube. Mix
well. Pipet 5 ml of the collected sample into a test tube, add 0.5 ml of the
peroxide solution, and mix vigorously for approximately 15 seconds. Add 2.7 ml
of sulfanilamide solution and mix vigorously for about 30 seconds. Then pipet
3 ml of the ANSA solution; mix vigorously for about 30 seconds. The ANSA solu-
tion must be added within 6 minutes of mixing the sulfanilamide solution.
(CAUTION: Longer time intervals will result in lowered absorbance values).
Prepare a blank in the same manner using 5 ml of unexposed absorbing solution.
The absorbance of the blank should be approximately the same as the y- intercept
in the calibration curve (Section 8.2). Determine absorbance at 550 nm with
distilled water in the reference cell using 1-on cells. The color can be read
anytime from 1 to 40 minutes after addition of the ANSA. Read ug NOl/ml from
the calibration curve (Section 8.2).
7.3 Spectrophotometer cells must be rinsed thoroughly with distilled water
and acetone, and dried; otherwise a film will build up on the cell walls.
8. Calibration and Efficiencies
8.1 Sampling
8.1.1 Calibration of Floivmeter. (See Figure 2). Using a wet test meter
and a stopwatch, determine the rates of air flow (on /min) through the flow-
meter at a minimum of four different ball positions. Plot the ball position
versus the flow rate.
8.1.2 Flow Control Device. The flow control device provides a constant
rate of air flow through the absorbing solution and is determined in 7.1.
8.1.3 Calibration of Absorption Tube. Calibrate the polypropylene
absorption tube, (Section 5.1.2) by first pipeting in 50 ml of water or
absorbing reagent. Scribe the level of the meniscus with a sharp object, go
over the area with a felt- tipped marking pen, and rub off the excess.
8.2 Calibration Curve. Dilute 5.0 ml of the 1000 ug NOl/ml solution to
250 ml with absorbing reagent. This solution contains 20 ug NO./ml. Dilute
5.0 ml of the 20 ug NCL/ml standard to 200 ml with absorbing reagent. This
solution contains 0.50 ug NO^/ml. Prepare calibration standards by pipeting
the indicated volume of the standard into volumetric flasks and diluting to
the mark with absorbing reagent.
Volume of Standard .. nFijial . Concentration
- Volume, ml ug N
10 ml of 0.50 ug NC>2 /ml 100 0.05
20 ml of 0.50 ug NO^ /ml 100 0.10
2 ml of 20 ug NC>2 /ml 200 0.20
Use 0.50 ug/ml Standard Directly - 0.50
5 ml of 20 ug NOZ /ml Standard 100 1.00
9 ml of 20 ug NO /ml Standard 100 1.80
57
-------
Run standards, plus a blank, as instructed in 7.2. Plot absorbance versus
ug NO,/ml. A straight line should be obtained with a slope of approximately
0.50 absorbance units per ug NCL/ml, and a y-intercept (i.e., zero ug NC^/ml)
of approximately 0.01 absorbance units. The absorbance is linear up to a
concentration of 4.0 ug NCX/ml, absorbance of 1.9. Therefore, if the samples
exceed the absorbance of the highest calibration standard and the above absorb-
ance is within the range of the spectrometer, the calibration curve can be
extended by including higher concentration standards. If a higher absorbance
range is not available, samples mist be diluted with absorbing reagent until
the absorbance is within the range of the highest standard.
8.3 Efficiencies. An overall average efficiency of 93 percent was
obtained from test atmospheres having a nitrogen dioxide concentration of
20 to 700 ug/m3.
9. Calculation
9 , 1 Sampling
9,1.1 Calculate volume of air sampled:
x T x 10'° (A-D)
where: V = Volume of air sampled, m^
F, = Measured flow rate before sampling, cm /min
•L T
F- = Measured flow rate after sampling, on /min
T = Time of sampling, min
10 = Conversion of on to m
9.1.2 Uncorrected Volume. The volume of air sampled is not corrected to
standard temperature and pressure because of the uncertainty associated with
24- hour average temperature and pressure values.
9.2 Calculate the concentration of nitrogen dioxide as pg N07/m :
, (ug NOl/ml) X SO
(jg NCm'5 = - ~ - CA'10)
where: 50 = Volume of absorbing reagent used in sampling, ml
V = Volume of air samples, m
0.93 = Overall efficiency of method
9.2.1 If desired, the concentration of nitrogen dioxide may be calculated
as ppm X>2'.
ppm = (ug N02/m3) X 5.32 X 10'4 (A- 11)
10. References for Appendix A- 4
1. Mulik, J. D., R. G. Fuerst, J. R. Meeker, M. Guyer and E. Sawicki.
A Twenty-Four Hour Method for the Collection and Manual Colorimetric
Analysis of Nitrogen Dioxide. (Presented at 165th American Chemical
Society National Meeting, Dallas, April 8-13, 1973.)
58
-------
2. Lodge, J. P., Jr., J. B. Page, B. E. Amnons, and G. A. Swanson. The
Use of Hypodermic Needles as Critical Orifices in Air Sampling. J. Air
Pol. Control Association. 16:197-220, 1966.
59
-------
APPENDIX B.
STATISTICAL EVALUATION OF METHODS TESTED
60'.
-------
PHASE I
Introduction
Concurrent measurements of nitrogen dioxide (NO,) concentrations in ambient
air and in ambient air injected with additional amounts of NO- were made over a
period of 22 days by the four methods being tested. Duplicate automated instruments
(chemiluminescence and colorimetric) were operated, and quadruplicate samples were
collected daily for subsequent analysis by each of the two manual methods (arsenite
and TGS-ANSA). The daily (22-hour periods beginning at noon) average NO- con-
centrations obtained by the four methods are examined and compared in this section.
Comparison of the methods was based on the determination of difference in
bias and the comparison of precision of simultaneous duplicate measurements of NO-.
In general, we are interested in how well two instruments (or duplicate
measurements by a manual method) compare and also in the agreement between
concentration data generated under the same conditions by the four methods under
test. It is obvious that the terms bias and precision take on specific meanings
for this discussion. The best way to convey the meaning of bias and precision
is to consider the simultaneous NO, measurements as indicated by the quantities:
Yli = 61 + Xi + eli (B-D
Y2i = B2 + Xi «• £2i (B-2)
where: i = 1, 2,...22 days.
Bj and 62 = The biases (unknown) of instruments 1 and 2, respectively,
which are made up of drift, calibration error, etc.
x^ = The "true" value (unknown) that is being measured.
EJ^ and ^2i = The random errors (unknown) associated with instruments 1 and 2,
respectively. They are composed of measurement errors, the
inability of expressions B-l and B-2 to totally describe a measure-
ment, and other unexplained errors.
For purposes of hypothesis testing, the c,.'s and e,-'5 are assuraed
to be normally and independently distributed of each other and independent of X..
The assumption of normality is not unreasonable because the measurement errors are
composed of a number of individual errors, such as calibration errors, which
although possibly not distributed normally as individual observations, tend to
be distributed normally as the sum of individual components increases. Before
an analysis of the data was made, a plot of the variances of the instruments for
each method against the corresponding target value showed that a transformation
was not necessary.
The error variance oj: is a measurement of the precision of instruments 1 and
2, or in other words, a measurement of the closeness of repeated measurements
taken by instruments 1 or 2 (or pairs of results obtained using the manual methods).
The relative degree of precision attained by an instrument is determined by
comparison with other measures of precision on other instruments. For two differ-
61
-------
ent instruments of the same type, we are interested in 6. - e_.
1
These notions, which conform (by design) to methods given by Grubbs , will
be employed rigorously in the following analyses.
Analysis of Results
Intramethod Comparisons
Chemilumlnescence method- Two identical chemiluminescence instruments were used
to make simultaneous NO- measurements. To compare the bias (unknown) of the two
instruments, we will take advantage of the natural pairing of the measurements
resulting from simultaneous readings of a series of 22-hour averages. In doing
this, estimates of precision that are free of variability in the day-to-day true (X^
value can be obtained. This is done by obtaining the difference in the paired
measurements by applying the following:
vi = Yli - Y2i = "I - 62 + Eli - £2i (B-3)
and then calculating S^(V), which is an -estimate of ati-^2- To test the
hypothesis that the biases (equal means) are the same (that is, HQ : 0^ = 62) >
we calculate.
22
V = 2 Vj = 0.0007 (B-4)
(In this discussion, units are expressed in parts per million. Conversion to
ug/m is made by multiplying ppm by 1880. J This t-value is distributed as a
Student's t with 21 degrees of freedom (n-l=21). Comparing t (calculated) with
t (theoretical) at a = 0.05 level of significance and 21 degrees of freedom,
we find that t=1.163 and t (a =0.05, 21 df) =2.08. [t (theoretical with the
corresponding level of significance and degrees of freedom is underscored through-
out this appendix to differentiate it from t (calculated).] Therefore, the data
are in concordance with the hypothesis that the biases (unknown) are equal.
To compare the precision (equal variance) of the two instruments, o^ and a^ ,
we first find the sum of the paired readings:
. Wi = ^li * *2i = (Bl + 62) + 2Xi + (e^ + t^ (B-6)
then calculate the correlation coefficient of V and W denoted by r(vw). To
test the hypothesis of equal precision (H0 : 0^= o^2) we must calculate:
Q.1375Q2\£0~ f
~[l-(-0.137504)2]0.5"°-62 ^^
62
-------
This t-value, according to Grubbs, is distributed as Student's t with n-2 = 20
degrees of freedom, resulting in the following expression:
t = -0.62 >-tC"=Q.Q5, 20 df) = -2.086. Therefore, we conclude that the two
chemiluminescence instruments are of equal precision in their measurement of NO,
and that they measured ambient NO, with the same bias and precision.
Colorimetric method - The expressions developed in the discussion of the chemilumi-
nescence method also apply here. To determine if there is a difference in the bias
(unknown) of the two different colorimetric instruments, we compute t with the aid
of formula B-5:
t VvfT" -Q.QQ4V26"
0.004 - '
Because this value is less than -t(a =0.05, 19 df) = -2.093, we conclude that
there is a significant difference in bias between the two instruments. To
determine whether the two instruments have equal precision (that is, l
we use formula B-7 to compute:
- _ n
-0-40
and similarly for replicates 3 and 4:
t „
_
[l-(0.025994)2]0.b ~
and conclude that both sets of replicates (1 and 2) and (3 and 4) show the same
precision of measurement.
The next step is to extend the discussion to compare the average bias of
samples 1 and 2 (sampler A) with the average bias of samples 3 and 4 (sampler B.) .
To do so requires calculations using the following expressions:
(Bi * B2) (63 * 84) . (ei. * e2) (e3 + e4) ,R a-,
2 "^ ~^Z - 1-B-8J
(B-9)
2 2
22
vi = Y3i - Y4i = (63 - 64) = foi - E4i5 (B-10)
2 2
along with estimates of variance S (P) and S (V) .
To test the hypothesis that the average bias of samples 1 and 2, (Sj + e^/Z, is
equal to the average bias of samples 3 and 4, (63 + 84) /2, we calculate another
t-value.
Because this value is greater than -t Ca =0.05, Zl df) = 2.080, we conclude
that the average bias of arsenite samples from sampler A is the same as the
average bias of those from sampler B.
63
-------
To ascertain whether the average of the variances of sampler 1 and 2,
C°?i + °c7)/2 is equal to the average of the variances of sanplers 3 and
9 9
4, (o + of )/2, we make a slight departure from Grubb's theory and form the
e3> e4
ratio:
F=
2
which has an F distribution with degrees o£ freedom for both the numerator and
2
denominator determined according to Satterthwaite . In general, let r denote the
desired degrees of freedom for either the numerator or denominator of F. Let r,
77
and r2 denote, respectively, the degrees of freedom for c£ and o£ . Then:
r = 04 04 . (B-13)
Fl F2
In our case, r, = T~ and n-1 = 21. Therefore, according to Cochran r lies between
28 and 14. Comparing the calculated F=0.000010/0.000009=1.11 with the theoretical
F with numerator degrees of freedom between 28 and 14 and denominator degrees of
freedom between 28 and 14, we must conclude that the average variance of measurements
1 and 2 is the same as the average variance of measurements 3 and 4. The range of
theoretical F values for a = 0.05 and for this range of degrees of freedom is not less
than 1.87 nor greater than 2.46.
Comparing this value with -t (a = 0.05, 18 df) = 2.101, we conclude that the two
instruments have equal precision.
Arsenite method - Evaluation and comparison of the manual methods (arsenite
and TGS-ANSA) are more involved because quadruplicate samples were collected
and subsequently analyzed by each of the two methods to provide the NO 2 data.
Duplicate absorption tubes (bubblers) for each of the two manual methods were
connected to the manifold in each of the two manually operated gas samplers
(sampler A and sampler B). The same expressions (B-l and B-2) employed in the
chemiluminescence section will be used to denote measurements made on arsenite
method samples 1 and 2 (sampler A), respectively, and measurements on arsenite
samples 3 and 4 (sampler B) by Yj^ = 63 + X-^ + £3^ and ¥4^ = 64 + X^ + £4^,
respectively. To determine whether arsenite samples 1 and 2 have equal bias, we
must use formula B-S to compute
-003 -
and because this is much greater than -t (a = -0.05, 21 df) = -2.080, we must
conclude that the two samples 3 and 4 were at the same level of bias.
64
-------
To compare precision between arsenite sample replicates 1 and 2, we
compute with formula B-7:
«. _ (0.016449 V2Q
[1-(0.016449) <=]u-
0.073
TGS-ANSA method - Expressions developed in the analysis of the arsenite method
data also apply in the evaluation of the TGS-ANSA method.
To determine if the biases (unknown) of TGS-ANSA samples 1 and 2 are the same,
we use formula B-5 to calculate
f _ -0.0005V2T_
t 67002
Likewise, we calculate for samples 3 and 4
o.oooivgr =
r 0.005
Because neither of these t values is greater than t (a =0.05, 21 df) = 2.093,
we must conclude that there is no difference in the level of bias for samples 1 and
2, and likewise, no difference in bias for samples 3 and 4. To test the hypothesis
that samples 1 and 2 were measured with equal precision (HO : aei = ac7^'
we use formula B-7:
_ CO. 103887 V"
"
Also for R, : Or = o| we calculate:
31
t _ r-0.014182W5Q"
r " [i.c-oroiiiiz)2]"-* " °-063
Because neither of these values fall outside the limits specified by
^ t (a =0.05, 20 df) = +_ 2.093, we must conclude that samples 1 and 2 were
measured with equal precision as were those for samples 3 and 4.
To test the hypothesis that the average bias of samples 1 and 2 equals the
average bias of samples 3 and 4, we calculate with the aid of formula B-ll:
, . 0.0005^0" . ...
t 0042-= °'319
Because t
-------
differ from the average variances of 3 and 4. It is difficult to attribute
this to the different manifolds to which these instruments were connected because
this was not the case for the arsenite method. As far as bias goes, no differences
were found among the TGS-ANSA replicate results.
Intermethod Comparisons - A comparison between methods was made by taking the average
of the measurements of instruments (or method) of the same type and comparing these by
way of independent t tests. Because the formulas and computations for intermethod
comparisons are the same as those applied in the preceding paragraphs, they will
not be repeated here. Detailed statistical comparisons of the data generated by
the four methods were made following the pattern established for intramethod data
analyses.
Summary - The results of the tests for intramethod biases and precision have been
presented in detail above and are summarized in Table B-l. A similar summary
for intermethod comparisons is presented in Table B-2. The data in Tables B-l and
B-2 indicate that there is a significant difference between the data from the
chemiluminescence and TGS-ANSA methods, and between the arsenite and TGS-ANSA
methods.
Comparison of Hourly Average Concentrations
The average NO- concentrations as determined by the two chemiluminescence
instruments (OT) and the two colorimetric instruments (CT) were calculated for each
.hour of study period for which data were available. A difference function
(D = CM-C) was computed and analyzed to assess the agreement between the two
methods.
The intermethod differences were averaged on an hour-of-day basis over the
study period to see whether any systematic diurnal pattern existed that might be
suggestive of "method drift." These averages are shown in Table B-3.
Although these intermethod differences appear to vary systematically with
time, the relationship is not of the simple linear form that would be suggestive
of a relative linear drift between the two methods. Rather, the difference
builds to a positive maximum (CM > M) in midafternoon and then subsides to a
relatively constant negative value (CR < R) throughout the evening and morning
hours.
Frequency distributions of the entire data set have shown that larger method
differences are apt to occur as this N02 concentration increases. This apparent
tendency may well be confounded with the evidence just seen that method difference
is positively correlated with ozone (63) concentrations. In order to test this,
the cumulative frequency distributions shown in Table B-4 were constructed after
deleting the data for hours 10 through 17 when the Oj concentration is significant-
ly higher.
66
-------
Table B-l. INTRAMETHOD COMPARISONS - BIAS AND PRECISION
Method
Chemi luminescence
Colorimetric
Arsenite
Box A
Arsenite
Box B
TCS-ANSA Box A
TGS-ANSA Box B
No. of
Instruments
or samples
2
2
2
2
2
2
no. of
pairs
22
20
22
22
22
22
r
(corre-
lation)
0.999
0.995
0.997
0.996
0.99?
0.992
Mean
difference
0.0007
-0-004
0.0003
-0.001
-0.0005
0.0001
Standard
deviation
of
difference
0.003
0-OC4
0.003
O.CC3
0.002
0.005
Null
hypothesis h
(Ho) t-test"
Equal means 1 .163
Equal variance -0.62
Equal means -4.472
Equal variance O.'O
Equal means 0.469
Equal variance 0.214
Equal means 0.156
Equal variance 0/35S
Equal means -1.17
Equal variance-0.467
Equal means 0-093
Equal variance-0-063
Decision
Accept Ho
Accept Ho
Reject Ho
Accept Ho
Accept Ho
Accept Ho
Accept Ho
Accept Ho
Accept Ho
Accept Ho
Accept Ho
Accept Ho
Conclusion
The chemi luminescence
instruments measure with
the same level of bias
and precision .
The colorimetric method
measures with the same
precision but one instru-
ment measures with a bias
ofO.004 ppm higher than
the other.
In Box A, the arsenite
samples are measured with
same level of bias and
precision.
In Box B, the arsenite
samples are measured with
the same level of bias
and precision.
In Box A.TGS-ANSA samples are
measured with the same
level of bias and precision.
In Box B.TGS-ANSA samples are
measured with the same level
of bias and precision.
jlsigned difference,
Significant at a =0.05 significance level.
-------
Table B-2. INTF.WETHOn COMPARISONS - BIAS
' Ho. of ' '
instru- Standard
nents or No. of Mean deviation of
Methods samples pairs difference difference
Chemi luminescence 2 22 0.0016 0.0038
versus
Colorimetric
Chemilumirtescence 2 22 -0.0014 0.0050
versus
Arsenite
Chenilutn'nescence 2 20 0.0028 0.0047
versus
TGS-ANSA
Coloriraetric 2 20 -0.0022 0.0062
versus
Arsenite
Colorimetric 2 18 0.0022 0.0058
versus
TGS-ANSA
Arsenite 4 20 0.0039 0.0037
versus
TGS-ANSA 4
Arsenite (A) 2 22 -0.0014 0.0035
Arsenite (B) 2
TGS-ANSA (A) 2 20 0.0003 0.0042
TGS-ANSA (B) 2
1 ! i - •
Null , . '
hypothesis . •'
(Ho) t-test ' Decision
Conclusion
Equal means 1.844 ' Accept Ho ; Chemi luminescence and Colorimetric methods
measure with equal levels of bias.
Equal neans -1.321 Accent Ho Chemiluminescence and arsenite methods
measure with the sane level of Mas.
Equal means 2.654 Reject Ho Chenriluninescence bias level exceeds the
TGS-ANSA bias level by 0.002B ppm.
Equal means -1.601 . Accept Ho Coloriraetric
measure with
Equal means ' 1.588 ' Accept Ho Colorimetric
with the same
i '
Equal means 4.719* Reject 'Ho Arsenite bias
i i bias level by
and Arsenite methods
the same level of bias.
and TGS-ANSA methods measure
level of bias .
level exceeds TGS-ANSA
0.0039 ppm.
Equal means -1.876 Accept Ho Arsenite bias level in sampler A is equal
: : to arsenite bias level in sampler B.
, i
Equal means ' 0.319 ; Accept Ho i TGS bias level in sampler A is equal
{ i to TGS bias level in sampler B.
1 1
aSiqned differences, ppm.
Significant at the o=.05 significant level.
-------
Table B-3^ DIURNAL PATTERN OF INTERMETHOD DIFFERENCES a
Hr
1
2
3
4
5
6
Dm
-1.5
-3.0
-2.1
-3.2
-3.2
-0.9
Hr Dm
7
8
9
10
11
12
-2.1
+0.4
+1.1
+8.6
Calibrate
Cal ibrate
Hr
13
14
15
16
17
18
Dm
+9.2
+8.8
+16.5
+19.0
+15.8
+5.6
Hr
19
20
21
22
23
24
Dm
+1.3
+0.8
+1.1
+0.4
-0.4
-1.1
- CM-C
Table B-4. CUMULATIVE FREQUENCY DISTRIBUTION OF HOURLY INTERMETHOD
DIFFERENCES BY CONCENTRATION GROUPING
(Percent)
NO, concentration,
L. <*
Range of D a, yg/m
pg/m
<94
94-188
188-282
>282
$9
87.9
94.2
82.1
100
98.4
<19
100
98.9
100
99.7
<28
100
100
100
Dm = CM"C>
69
-------
On the basis of the foregoing it appears that in this case the inter-
method (tM-C) difference is not dependent on NO- concentration, but is, rather,
a direct function of 0, concentration, an inverse function of nitric oxide (NO)
concentration, or some combination of both 0, and NO.
In an attempt to find possible explanations for the above observation,
diurnal NO, N02> and 0, concentrations were plotted (Figure B-l). nata were
not recorded for hours 11 and 12 during the study period because this time
interval was used each day to calibrate instruments and to perform other tasks
in preparation for the following 22-hour sampling period. For the majority of
the days in the study period, either no N02 or a constant-level N02 spike was
administered over the course of the day, which accounts for the constant differ-
ence between ambient NO- and the total NO- curves during the morning and evening
hours. On 5 of the 21 days, a 3-hour N0? spike was administered (from hour 13
to 16 or hour 14 to 17), which accounts for the mid-.day peak in total NO-.
The difference between the chemiluminescence and colorimetric hourly
averages (G^-C) between 1300 and 1900 hours when the ambient 0, concentrations
are highest indicates that colorimetric N0? values are indeed depressed by the
higher levels of 0, in the atmosphere. However, the situation is somewhat
clouded by the fact that the 3-hour spiking also was performed during this
portion of the day and this spiking itself might be responsible for the greater
V
To separate the possible effects of 3-hour NO- spiking from those caused
by 0,, the diurnal averages were recalculated after deleting data for those
5 days when 3-hour N02 spiking was done. These results appear in Figure B-2.
Now the curves for total N02 follow the ambient curve very closely; the near
constant difference being the result of a uniform (day-long) spiking rate.
The amount that the chemiluminescence method values exceeded those of the
colorimetric method is shown as the shaded area on the upper graph.
In Figure B-3, the hourly average method differences are plotted against
the hourly average concentrations of 0, and NO. The correlation coefficients
of N02 intermittent differences (Dm) with O.j and NO are +0.87 and -0.78,
respectively. This supports the hypothesis that there is a negative inter-
ference of Oj and a positive interference of NO with the colorimetric NO- method.
Effects of Other Ambient Pollutants
A multiple regression equation that incorporates the concurrent ambient
data collected on six common pollutants (independent variables) -- total sus-
pended particulate matter (TSP), carbon monoxide (CO), carbon dioxide (C02),
total sulfur (TS), ozone (0,), and nitric oxide (NO) -- was developed. Another
variable, the ^2/^3 rati°> was added. These data were incorporated into a
linear model to find answers to the following questions:
1. Will including these data improve the comparability of a method so
70
-------
20
m 1C
E 16
I I I I I
I
"
u -
> rji
£ 8
o
4
0
-4
280
I I
CHEMILUMINESCENCE-
COLORIMETRIC
I II I I I I I
I I I I I I 1 I I I I I
I I III I I I
N02 TOTAL CHEMILUMINESCENCE
I I I I I I
I I I I I I
NO
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
TIME, hours
Figure B-1. Phase I - N02 methodology study • hourly averages for continuous methods.
-------
--J
K>
CJ
CM
o
240
200
160
120
80
40
0
1 I I I I T
AVG. OF COLORIMETRIC INSTRUMENTS
, AVG. OF CHEMILUMINESCENCE INSTRUMENTS
1 \ I T
I I I I I
I I I I I
I I I
I I I I I
I I 1 I
£UU
240
E
^200
CJ
| 160
K-
Z
Ul
cc 120
Ul
cc
£ 80
z
40
n
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
—
- \
V ^ •*
\ .• *. .•
**"'v .*'"*"* ""^"* \^ /
•• ••"" "• /
*^ •••"**" * *
• •
•n
~~ ^^V •
*• ^^^W *
~" ^ ^*X
1 I I 1 1 1- — S r-T 1 1 1 L-J...-J -I-' 1 \
1 /\\ \
•* **»*»fc ^"*
^» *•*.»* •••••
•
X
—
^—
—
03 ~~
- 1 1 1 1
8
TIME, hours
Figure B-2. Phase I - diurnal patterns (excluding intermittent spike days).
24
-------
-3.8 —
-7.5
38 75
NO CONCENTRATION, pg/m3
Figure B-3. Method difference versus interferent concentrations (excluding intermittent spike days).
-------
that the response is now closer to the true (target) value?
2. Will including these data improve the inter-relatability of the
four methods? Thus, if one could adjust the measurements made by one method,
the effect caused by 0,, for example, and thereby improve the intermethod
agreement, then it can be concluded that the results from one method relative
to another are affected (interfered with) by that pollutant.
Analysis of Data -In answer to the first question, 22-hour concentrations were averaged
over the number of instruments or analyses for each method for each day. For example,
the 22-hour averages from the two chemiluminescence instruments were averaged
and that one result was taken as the estimate of NO- concentration for that day
for that method. Similarly one average for each day was recorded for the
colorimetric method (2}, arsenite method (4), and TGS-AMSA method (4). Each
method average was then subtracted from the target N0? value for the day. This
array of differences was then considered as the dependent variable by which the
data from the independent variables were to be added in an attempt to "explain"
this residual difference.
The analysis of variance table (Table B-5) in all cases indicates that the
F value for the overall regression equation is not significant. This means that
the independent variables could not significantly account for any variability
in the dependent variable. It can be observed that the best fit with added variables
2
(as measured by R ) is for the colorimetric method (nearly 50 percent of the variation).
The largest coefficient for all methods except the chemiluminescence method is
associated with the C02 variable. (CO is second to the TS variable for the
chemiluminescence method).
A similar procedure was followed for testing the inter-relatability of data
obtained by each of the methods. In order to derive the dependent variable array,
differences between method averages were calculated for each day. The equation
incorporating independent variables was fitted to this array of differences through
step-wise multiple regression techniques.
The analysis of variance table (Table B-6) in all cases indicates that the
F value for the overall regression equation is also not significant. This affirms
the conclusion that the independent variables could not significantly account for
any variability in the array of differences between methods. It can be observed
that the best fit with added variables (based on R values) involves the chemi-
luminescence/ arsenite methods.
However, even in this best case comparison, only 37 percent of the variation
is explained. Consistent with Table B-5 results, the largest coefficient for all
models, that is, all comparisons, is associated with CO.,. However, this variable
is still not significant in any of the models, nor are any of the other variables.
Therefore, the overall conclusion must be that, during Phase I at the ambient
(22-hour average) levels of the pollutants measured concurrently with NO-, no
significant interferences were detected.
74
-------
Table B-5. ANALYSIS OF VARIANCE - DIFFERENCE FROM TRUE
(ppm)a
Measurement
method
Chemiluminescence
Colori metric
Arsenite
TGS-ANSA
Source
NO
C02
°3
TS
CO
Ratio
Intercept
NO
C02
03
TS
CO
Ratio
Intercept
NO
C02
03
TS
CO
Ratio
Intercept
NO
C02
°3
TS
CO
Ratio
Intercept
Sequential
F value
0.03
0.18
2.21
3.96
0.05
0.14
-
1.68
1.07
0.08
4.56
1.11
0.79
-
2.22
0.04
0.15
0.07
0.11
0.16
-
3.44
0.17
1.67
0.21
0.06
1.53
0
B values
0.030
-0.047
0.002
-0.068
-0.001
-0.000
+0.018
0.083
-0.263
-0.028
-0.063
0.010
0.000
0.092
0.024
-0.149
-0.010
-0.019
-0.005
-0.000
0.054
-0.007
-0.119
-0.029
-0.034
-0.034
-0.000
0.051
T for
Ho: B=0
- 0.48
- 0.24
0.08
- 2.03
0.15
0.37
0.26
1.32
- 1.33
- 1.41
- 1.72
1.23
0.89
1 .35
0.23
- 0.46
- 0.29
- 0.34
- 0.40
- 0.40
0.48
- 0.08
- 0.44
- 1.04
- 0.72
- 0.02
- 1.24
0.55
Standard
error of B
0.063
0.196
0.020
0.034
0.008
0.0001
0.067
0.063
0.198
0.020
0.037
0.008
0.001
0.068
0.105
0.326
0.034
0.057
0.014
0.0002
0.112
0.086
0.269
0.028
0.047
0.011
0.000
0.092
Overall significance
of regression equation
F R2
1.095 0.34
1.547 0.46
0.460 0.18
1.180 0.35
Units of ppm apply to B Values and the standard of B. Other values are unitless.
-------
Table B-6. ANALYSIS OF VARIANCE - METHOD DIFFERENCE (ppm)c
••vti-.oO
: 1 ji:ii nes. pn- .'/ '
n'i luriinosconco/
sen;te
mlluminescence/
rsenHo
ColoriMCtric/
Arscnite/
TCS-A:IS/I
uurce
no
C02
03
TS
CO
Ratio
Intercept
:io
C02
03
TS
CD
Ratio
Intercept
110
;o2
03
TS
CO
Ratio
Intercept
!IO
;02
0,
it
CO
Ratio
Intercept
:io
;o2
0-,
ll
CO
Ratio
Intercept
NO
C02
°3
TS
CO
Ratio
Intercept
Sequential
F value
0.02
2.23
t'.OO
P. 4 3
0.03
0.09
-
1.03
f .17
1.29
1 .96
C.72
0.23
•
1 .51
0.52
0.31
0.90
0.95
0.00
-
0.62
0.43
0.65
0.31
0.50
0.02
-
1.23
2.13
0.23
0.17
0.84
0.01
-
0.44
2.20
0.20
0.03
0.22
0.01
I Value
0.5H
-0.177
0.009
0.024
0.002
-0.000
-0.058
0.005
-0.055
0.005
0.054
-0.008
-0.000
0.010
-0.035
0.232
0.013
0.046
-0.013
-0.000
-0.074
-0.052
1.123
••11.004
0.029
-0.011
-0.000
-0.040
-0.143
0.409
0.004
0.021
-0.015
-0.000
-0.132
-0.091
0.280
&.008
-0.008
-0.005
-0.000
.-0,022.
1 for
rn: P=P
'.Undard
rrror
of il
0.74 0.079
-0.72 , 0.245
0.39 0.024
0.57
0.043
0.22 0.010
-0.30
0.004
0.69 0.084
0.06
-0.20
0.10
l.!4
-0.78
-0.18
0.19
-0.75
0.65
0.087
0.272
0.026
0.047
0.011
0.0005
0.093
0.114
0.354
0.38 0.034
0.74 , 0.062
-0.93 0.014
-0.30 , 0.0006
-0.62 0.121
i
-0.42
0.31
-0.12
0.43
-0.68
-0.14
-0.30
-1.05
0.96
0.09
0.29
-0.90
-0.08
-0.91
-1.14
1.16
0.34
-0.19
-0.47
0.10
;1.09_ ..
0.126
0.392
0.033
0.068
0.016
0.0007
0.134
0.137
0.427
0.041
0.074
0.017
0.0008
0.145
0.079
0.248
0.024
0.043
0.010
0.0004
0.084
Overall r,i<:rif icanct
0.46(1
0.900
0.37
0.720-
0.422
0.32
0.22
0.900
0.37
0.515
0.26
'Unit:, of |>pn apply only to B Value", and tlr- slandard error of U. Other values are uniLless.
76
-------
PHASE II
Introduction
In Phase II, NO, measurements were made using four different NO. measurement
methods. In this phase, the NO, being measured had been added to a clean air
3 3
system at three fixed levels -- 0 ug N07/m (clean air), 100 ug N0,/m and
3
200 pg N02/m . The NO- levels were varied according to a predetermined addition
schedule. To analyze the data, a nested factorial experimental design was
assumed. The tree diagram given in Figure B-4 gives the underlying design of
the experiment.
METHOD:
NO, LEVELS,
3
CHEMILUMINESCENCE
SPIKING
SCHEDULED
\
28
Same Design for
Colon'metric
Arsenite
TGS-ANSA
Methods
1A 3A 3B 3C 3D ZA
Figure B-4. Phase II - Design of experiment.
In the analysis, each different type of instrument (or manual method) was
considered as giving a replicate measurement for that particular method. The
purpose of the analysis was to compare different methods on the basis of their
abilities to measure different levels of NO- in clean air.
Analysis of Results
An analysis of variance was performed with subsequent tests of hypotheses
about the underlying model
Yijkl
where: Y. .. = The 1
i]KJ.
Mi
. \
Nj
SN
k(j)
J.T.
NO, measurement by the i method on the j
th
+,\
(B-14)
N02 level
M.
administered according to the k schedule
= A constant appearing in all measurements
- The effect of the i method on the measurement
= The effect of the j NO, level on the measurement
+"L i.U
= The combined effect of the i method and the j level of N0?
= The spiking schedule for a certain NO, level (nested effect)
on the measurement
el(ijk) = The random error of measurement
For purposes of hypothesis testing, the E, ,. -^'s are assumed to be
normally and independently distributed about zero and with constant variance.
Results of this analysis are presented in Table B-7.
SN,
77
-------
Table B-7. ANALYSIS OF VARIANCE - PHASE II
Source
Method
NO 2
Method x N02
Schedule (N02)
Error
Total
DFa
3
2
6
4
79
94
Sum of squares
0.00040238
0.11012932
0.00027575
0.00593331
0.0003399
0.11518236
Mean square
0.00013412
0.05550646
0.000045958
0.00148333
0.00000493
F
27.21
11,222.30
10.06
30.09
DF = degree of freedom.
All F values were significant at the a= 0.01 level; consequently, the
hypothesis was rejected.
Using Duncan's Multiple Range Test (Table B-8), the means are ranked from
the lowest to highest. Any two means shown that are not underscored by the same
line are significantly different.
Table B-8. SUMMARY OF DUNCAN'S TEST - PHASE II
Methods;'
Means :
NO,
CM TGS COLOR , ARS
__„ L ..... .„_... 4 .... ..... -| -----
r ........ - -I
C-0 S-0 A-0 T-C C-100 T-.05 S-100 A-200 T-200 C-200 S-200 A-200
0~o" 0~0 1~3 4~9 88~9 90.6 93.8 99.3 174.8 175.4 ""183.3 192.5
I ............ H I- ......... H I- ......... - ....... H
I— - ..... - ...... — -H
Schedule (N02) 1A (0.0) 3D (100) 3B (100) 3C (100) 3A (100) 2A (200) 2B (200)
Means: 2.1 89.7 91.4 91.9 113.0 157.4 207.2
aChemi luminescence (CM), TGS-ANSA (TGS), colorimetric (COLOR), arsenite (ARS).
Conclusions
The results of Duncan's Test indicate that, on the average, a significantly
higher (a = 0.05) reading was given by the arsenite method as compared with those
of the chemiluminescence, colorimetric, and TGS-ANSA methods, which gave
essentially the same readings. At the 0 ug N02/m3 level, the average chemilu-
minescence, colorimetric, and arsenite readings did not differ significantly
(a =0.05). However, each of these methods gave readings that differed
significantly from those given by the TGS-ANSA method. At the 100 ug N02/m3
level, the colorimetric and arsenite methods, on the average, gave the same
78
-------
results but both gave significantly higher results than either the chemilumi-
nescence or the TGS-ANSA method. Also, the TGS-ANSA method gave significantly
higher results than the chemiluminescence method. At the 200 yg N02/m level,
the TGS-ANSA, chemiluminescence, and colorimetric methods agreed on the average,
but the arsenite method gave significantly higher results. At the 100 vg N02/m
level, the 3-hour intermittent spike (3B) gave the same results as the 4-hour
intermittent spike (3C). However, all the other spiking schedules had
different effects on the results with the 9-hour continuous (3D) giving the
lowest readings and the 3-hour continuous (3A) giving the highest readings at
the 100 yg N02/m average level. For the 200 ug N02/m concentration, the 22-hour
continuous spike (2A) gave results that were significantly lower (a=0.05) than
the 6-hour continuous spike (2B).
In a statistical sense, the methods do not give the same results under similar
conditions and do not respond to substantial changes in NO, levels in the same
manner. The length of time N02 remains in the air in the absence of possible
interfering compounds has a definite effect on daily averages of NO, measurements.
Aside from a strictly quantitative analysis of Phase II results, graphical
displays of the data are interesting. Figure B-S illustrates the response of
different methods when N09 is added into the clean air system at a level of
3
100 ug/m on the different spiking schedules. The abcissa of the graph represents
the spiking schedule. In general, for the straight 3-hour spike (3B), the values
were somewhat closer to the 100 gg/m horizontal line; and for the 8-hour con-
tinuous spike C3D) all values were below the line. The two continuous methods,
chemiluminescence and colorimetric, were very close and almost parallel except for
the 4-hour intermittent spike (3C). The two manual methods, arsenite and TGS-ANSA,
gave parallel responses that were separated in magnitude,with TGS-ANSA always reading
lower by approximately 8.6 vg/m , The general pattern of the graphs is probably
caused by variations in the addition of N02.
PHASE III
Introduction
During Phase III, N02 measurements were made using the same NO, methods as
in Phases I and II. However, in Phase III, three fixed levels of 0, were
combined with two fixed levels of N02 in a clean air stream. The purpose of
Phase III was to answer the following questions:
1. Do different methods, when used under similar environmental conditions,
give the same response within limits of experimental error?
2. Does changing the level of Oj have any effect on the measurement of
N02?
3. Does each method respond in the same manner to a change in the
03 level?
Generalizations beyond those stated above, such as extending the conclusions
79
-------
COLORIMETRIC
— —___„ TGS-ANSA
ARSENITE
CHEMILUMINESCENCE
3B 3C
SPIKING SCHEDULE
Figure B-5. Phase II - Method difference.
to cover a wider range of NCL or 0, levels, would be misleading. Therefore, we
are dealing with a fixed linear model of the form:
Yijkl
N.J
MNO
ijk + Eijkl
where:
Yijkl
The 1 observation 1 + 1, 2, . . . n. .. on the i method
for the jth N0 level and the kth O level: i=l, 2, 3;
k = 1, 2, 3
v = A constant that is common to all observations and is estimated
by averaging over observations in the experiment
M. = The effect of the i method on the observation. For example,
•f-Vi
u + M. is the mean of the i method taken over all mixtures
of N02 and
.th
N. = The effect of the j level of NCL on the observation. For example,
80"
-------
y + N. is the mean of the j level of NCL taken over all levels
of 0, and all methods
«.u
(X = The effect of the k level of 0^ on the observation. For example,
v + 0, is the mean of the k level of 0^ taken over all levels of
N02 and all methods
MN.. = A measurement of non-additivity of the ij cell. For example, the
mean of the i method and j N02 level taken over all 03 levels
may not be M + Nh + N. (additive), but v + M^ + N- + MNi.
MO-, = A measurement of non-additivity of ik cell
NO., = A measurement of non-additivity of jk cell
MNO.., = A measurement of non-additivity of ijk cell. For example, the mean
of the 1th method, the jth level of N02, and the k1^ level of 03
may not be P + t^ + N. + Ok (additive) but u + H^ + N. + 0^. + MNOj.^
There are i = 4 methods, j = 2 levels of NO-, k = 3 levels of ozone, and
1 = observations.
Analysis of Results
To find answers to the questions previously stated, the questions are
reformulated in terms of the hypothesis concerning linear combinations to the
elements of the linear model and subsequently tested by the appropriate statistical
techniques.
An analysis of variance was performed on the data to provide tests of these
hypotheses, and the results are summarized in Table B-9.
Each line in the table (excluding error) corresponds to a specific hypothesis.
Rejection of a hypothesis, as indicated in the table, means that the probability
of making a wrong statement based on the calculated F statistic is a very low 0.01.
The data in Table B-9 indicate that the methods in general give different N^
readings, and that 0, affects the measuring capabilities of the different methods.
To find out which methods gave different readings and also which methods
were affected by the presence of 0,, a Duncan's Test procedure was used. This
is summarized in Table B-10. Any two means not underscored are significantly
(a=0.05) different; whereas, any two means underscored by the same line are not
significantly different. Figure B-6 shows how the different methods compare in
measuring N0? at approximately 75 vg NCL/m at three different 0, levels.
Figure R-7 is similar graph for comparing methods as to their ability to
measure approximately 150 yg/N02/m in the presence of three different concentra-
tions of O.
81
-------
Table B-9. ANALYSIS OF VARIANCE-PHASE III
Source
Method
°3
N02
Method X 03
Method X M02
03 X N02
Method X 03 X N02
Error
Total (corrected)
DFa
3
2
1
6
3
2
6
47
70
Sum of Squares
0.0037885
0.00034285
0.2124812
0.00185973
0.00017399
0.00053495
0.00022950
0.00012842
0.02830592
Mean Square
0.001222
0.0036737
0.1779864
0.000312165
0.00005391
0.00027840
0.00003825
0.00000273
F
462. 17b .
62.74b
7776. 73b
113. 44b
21.22b
97.90b
14.00b
DF = degrees of freedom.
Hypothesis rejected at the 0.01 level
Table B-10. DUNCAN'S TEST-PHASE III
Methods3:
Means
Ozone:
Means
COLOR
73.1
353
98.5
i ._ —
CM
98.3
667
99.6 ,
TGS
110.2
I
i
100
107.9
ARS
110.4
Method X O2one:
C-667 C-353 CM-353 CM-667 CM-100 CL-100 TGS-353 AS-353 TGS-100 AS-667 AS-100 TGS-667
43.2 71.8 92.5 100.6 101.9 104.3 106.6 106.6 107.5 111.9 112.8 117.1 ,
I-
-\
Colorimetric (COLOR), chemiluminescence (CM), TGS-ANSA (TGS), arsem'te (ARS).
82
-------
150
2 113
3
800
Figure B-7. Ozone interference- 150/ug/m3.
-------
Conclusions
The results of Duncan's Test indicate that, on the average, the arsenite and
the TGS-ANSA methods gave the same NO, values but that the cliemiluminesccnce method
values were slightly lower. However, the colorimetric method was subject to the
greatest interference; it gave significantly lower readings at both the 353 and
667 uji/m levels. Also, the colorimetric method readings were significantly
lower at the 667 Mg/m level than at the 353 ng/m level. '(Tie chemiluminescence,
arsenite, and TGS-ANSA methods did not seem to be affected by changes in the 0^
level.
84'
-------
TECHNICAL REPORT DATA
(Please read laarvclions on the reverse before completing/
. REPORT NO,
EPA 650/4-75-023
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
Comparison of Methods for Determination of Nitrogen
Dioxide in Ambient Air
5. REPORT DATE
June 1975
5. PERFORMING ORGANIZATION CODE
7. AUTHOR
------- |