MONITORING FROM AIRBORNE PLATFORMS FOR AIR QUALITY ASSESSMENT
(Selected Papers)
March 26-27, 1975
Held At The
NATIONAL ENVIRONMENTAL RESEARCH CENTER
U. S, ENVIRONMENTAL PROTECTION AGENCY
Las Vegas, Nevada
*
Printed July 1975
-------
epft owl
FOREWORD
The Monitoring Applications Laboratory of the National
Environmental Research Center in Las Vegas is very pleased
to have sponsored this meeting on Monitoring from Airborne
Platforms for Air Quality Assessment. In the past few years
there has been a rapid increase in both the number and the
objectives of aerial sampling missions. Concern over
generalized environmental degradation and the potential
impact of the accelerated energy resource development activ-
ities has contributed to the growth of aerial monitoring
programs. With this in mind, the primary focus of the meeting
was in exchanging information among the scientists and groups
involved in performing and/or evaluating aerial platform mea-
surements for environmental quality assessment. These pro-
ceedings should serve to promote uniformity in the application
of technology that has been developed over these past few years.
The technical papers are arranged in the order that they ap-
pear in the Meeting Program Agenda. The program and agenda
were prepared by a Program Committee with Roy B. Evans as
Chairman.
David N. McNelis
General Chairman
May 8, 1975
RXDOnnob7TB
-------
MONITORING FROM AIRBORNE PLATFORMS FOR ENVIRONMENTAL QUALITY ASSESSMENT
NATIONAL ENVIRONMENTAL RESEARCH CENTER-LAS VEGAS
Wednesday and Thursday, March 26 and 27, 1975
AGENDA
D. N. McNelis, General Chairman
NERC-Las Vegas
March 26
Welcome J. R. McBride
Acting Director
NERC-Las Vegas
SESSION I
APPLICATIONS OF AIRBORNE PLATFORM MONITORING:
KNOWN AND POTENTIAL
R. Neligan, MDAD, EPA
Session Chairman
Utilization of an Airborne Platform
in an Eastern, High-Ozone Concentra-
tion Study
Measurement and Characterization of
the St. Louis Urban Plume
Airborfie Pemote Monitoring in Air
Enforcement and Regulatory Programs
The Los Angeles Reactive Pollutant
Program
Assessment of Three-Dimensional
Pollutant Variability With Airborne
Platforms
J. J. B. Worth
Research Triangle
Institute
R. Husar
Washington University
C. Ludwig
SAI
W. A. Perkins
Metronics
D. L. Blumenthal
MR I
Measurements of Power Plant Plumes
With Light Aircraft
Assessment of the Impact'of Halo-
carbons on Stratospheric Ozone
Dr. Paul Harrison
MRI
F. S. Rov/land
University of
California-Irvine
-------
SESSION II
EXPERIMENTAL DESIGN
R. B. Evans, NERC-LV
Session Chairman
Sample Size Requirements for D. Mage
Statistical Validity NERC-LV
Meteorological Considerations J. McElroy
for Statistical Validity NERC-LV
SESSION III
MEASUREMENT TECHNIQUES
R. Stevens, NERC-RTP
Session Chairman
March 27
Determination of Particulate
Chemical Composition
Ron Draftz
IITRI (Chicago)
Analysis of Size Fractionated
Aerosols
T. Dzubay
NERC-RTP
Particulate Size Measurement
Methods
Minnesota
K. Whitby
University of
Methods of Analysis of Halocarbons
Air Pollutants
N. Hester
NERC-LV
Instrumentation to Measure Gaseous
and Particulate Pollutants From
Airborne Platforms
R. Stevens
NERC-RTP
DATA VALIDITY AMD DATA HANDLING
J. J. B. Worth, RTI
Session Chairman
Development of Standard Reference E. Hughes
Materials for Air Quality Measurement NBS
-------
SESSION IV (Cont'd)
Measurement of Ozone and Oxides
of Nitrogen in the Lower Atmo-
sphere From an Airborne Platform
J. B. Tommerdahl
RTI
Instrument Time Response and Its
Implications
D. Mage
NERC-LV
Data Processing Related to Field
Data Acquisition Systems
R. Browning
NERC-RTP
-------
TABLE OF CONTENTS
Page
Airborne Remote Monitoring In Air
Enforcement And Regulatory Programs,
C.B. Ludwig &. Michael Griggs
.Science Applications, Inc. .
1-10
Assessment Of Three-Diminsional
Pollutant Variability With Airborne
Platforms
D.L. Blumenthal
.Meteorology Research, Inc.
11-37
Sample Size Requirements For David T. Mage
Statistical Validity NERC-LV 38-53
Analysis Of Size Fractionated T. G. Dzubay
Aerosols NERC-RTP 54-72
Methods Of Analysis Of Halocarbon Norman E. Hester
Air Pollutants NERC-LV 73-79
Instrumentation To Measure Gaseous
And Particulate Pollutants From Robert K. Stevens
Airborne Platforms - A Review NERC-RTP 80-84
Development Of Standard Reference
Materials For Air Quality Measure- Ernest E. Hughes
ment National Bureau of Standards 85-98
Measurement Of Ozone And Oxides Of J. B. Tommerdahl, J. H. White,
Nitrogen In The Lower Atmosphere R. B. Strong
From An Airborne Platform Research Triangle Institute 98-124
Instrument Time Response And Its David T. Mage
Implications NERC-LV
125-136
-------
AIRBORNE REMOTE MONITORING
IN AIR ENFORCEMENT AND
REGULATORY PROGRAMS
Claus B. Ludwig
Michael Griggs
Science Applications, Inc.
La Jolla, CA 92038
The work upon which this presentation is based
was performed pursuant to Contract No. 68-02-2137
with the Environmental Protection Agency.
-------
OVERVIEW
REMOTE SENSING TECHNIQUES CAN BE USED IN:
STACK EMISSION MONITORING
PERIMETER MONITORING
AMBIENT AIR QUALITY MONITORING
-------
OVERVIEW OF REMOTE MONITORING USES
-------
APPLICATIONS
ENFORCEMENT
COMPLIANCE MONITORING (Evidence Presented to Plant Owner)
EVIDENTIARY MONITORING (Evidence Presented to Court of Law)
ASSIST IN DEVELOPMENT OF REGULATIONS FOR
STANDARDS OF PERFORMANCE OF SOURCES
PERFORMANCE SPECIFICATIONS OF MONITORING EQUIPMENT
SURVEILLANCE
ASSIST IN ENFORCEMENTS
WIDE AREA COVERAGE
2
-------
ADVANTAGES
COST EFFECTIVE
INSTRUMENTS MORE OBJECTIVE IN OPACITY MEASUREMENTS
MONITORING CLANDESTINE VIOLATORS
NON-INTERFERING
THE ONLY PRACTICAL WAY TO MONITOR WIDE GEOGRAPHICAL AREAS
RAPID RESPONSE MONITORING
DISADVANTAGES
POSSIBLE HIGH INITIAL CAPITAL COST
LIMITED APPLICATIONS IN CERTAIN ATMOSPHERIC CONDITIONS
CALIBRATION PROCEDURES AND EQUIVALENCY MORE COMPLICATED
POSSIBLE SAFETY HAZARD FOR SOME LASER SYSTEMS
-------
ENFORCEMENT MONITORING —
o COMPLIANCE
Presenting evidence to plant owner
e EVIDENTIARY
Court acceptance depends upon:
Is the scientific principle underlying the instrument's
operation valid?
Does the instrument successfully embody and apply
this underlying principle?
Was the instrument in proper working order at the time
of the test?
Was the person conducting the test qualified to do so?
Did the person conducting the test use the proper
procedures?
If different from the person conducting the test, is the
person interpreting the test's results qualified to do so?
4
-------
RECOMMENDED ACTION PLAN FOR
REMOTE TECHNIQUES
make candidate technique available to field personnel;
field personnel satisfied that technique produces reliable data;
data introduced into court and test cases established;
EPA develops and proposes performance specs (guidelines)
for different source industries;
performance specs promulgated;
EPA develops and proposes performance criteria
(equivalency) for different source industries;
performance criteria promulgated.
5
-------
NEW LEGAL ISSUES
RECENT RULINGS BY THE APPELLATE COURT IN WASHINGTON:
• There are no requirements for the use of a method of
testing to observe possible violations of a standard
(surveillance). (Portland Cement Association)
• However, if a method of testing is used as a standard,
and if violations can result in enforcement actions,
this method must be consistent with the statute and
congressional intent (Clean Air Act), i.e., well-
conducted and documented tests, response to public
comments and specific citations in literature.
(Portland Cement Association)
• The techniques used in arriving at the standards and
in determining compliance with the standard should
not differ significantly. (AMOCO Oil)
® There is no fixed requirement to promulgate standards
and test methods at the same time. (AMOCO Oil)
• Statistical error must be smaller than the deviation
from the standard. (AMOCO Oil)
6
-------
WESTERN ENERGY DEVELOPMENTS
Page
Major new coal mine developments 1973-1983, West of the 1
Mississippi Valley.
Coal resources and mines in the Western United States. 4
Coal conversion plants. 19
Existing capacity (Electrical Generating) Resources as 21
of December 31, 1974-
Thermal generating facilities in excess of 300 MW under 49
construction within next two years.
Summary of significant power plant additions, 1975-1984. 80
Civilian electric power nuclear reactors: Operating, being 112
built, and planned for Regions VI, VIII, IX, and X.
Oil shale projects. 116
Tar sand developments in Utah. 128
Major uranium mining areas, milling companies, and nuclear 130
industrial capabilities.
Geothermal developments - on line, commercial feasibility 137
demonstrations, and R&D projects.
Outer continental shelf oil and gas lease areas and lease 140
planning schedule.
-------
STATEMENT BY RICHARD J. DENNIS, JR. (GENERAL COUNSEL, EPA)
"Courts are looking much more closely at test methods and technical
support for standards than they once did. Therefore logical
quantitative relationships between standards and tests used for
compliance must be developed and clearly articulated in support
of all EPA test methods. Mere reliance on agency "expertise"
will no longer be adequate to withstand legal attacks on EPA
standards and test methods. "
7
-------
NEW ADDITIONAL REQUIREMENTS
o Establish "Equivalency" between Reference and Remote Method
9 Establish Stack Emission and/or Clean Air Standards by Remote Techniques
8
-------
PERFORMANCE SPECIFICATIONS OF REMOTE" MONITORS
Existing for Continuous Monitors
Additional Requirements
e Span (lower detectable limit to upper level
which must be greater than the emission
standard by a given factor)
o Noise (usually 0. 5 of the lower detectable
limits)
e Accuracy (percentage difference between
values measured by the [remote] sensor and
the applicable reference method)
9 Calibration error (absolute mean value plus
95 percent confidence interval)
o Zero and calibration drift (absolute mean value
plus 95 percent confidence interval)
• Response time (rise and fall time to 95 percent)
e Interference equivalent (influence on pollutant
signal by interfering gases in plume)
o Range (minimum and maximum distance
between source and observer, in which the
instrument must give data with the specified
accuracy)
e Field-of-view (should be small enough for
the source plume to fill the fov)
® Interference equivalent (influence on pollutant
signal by the interviewing atmosphere as well
as sky radiation beyond the plume)
© Sampling time (involves integration time
which will depend upon the plume turbulence)
® Supporting plume data (velocity, temperature,
etc.)
9
-------
REMOTE vs. ON-SITE
• Supreme Court Decision in Western Alfalfa Case
— Remote monitoring does not violate the
4th Amendment
o Possible New Argument
— Since enforcement action can involve
heavy penalties, on-site inspection without
search warrant conflicts with the right of
privacy guaranteed by the 4th Amendment
(even though the Clean Air Act provides
the right to enter). The use of remote
sensing could provide the tool to prove
probable cause for obtaining search warrant
(J. B. Zimpritch)
10
-------
Assessment of Three-Dimensional Pollutant Variability
with Airborne Platforms
by
D. L. Blumenthal
Meteorology Research, Inc.
464 West Woodbury Road
Altadena, California 91001
For presentation to Conference on Monitoring from Airborne Platforms for
Environmental Quality Assessment; National Environmental
Research Center - Las Vegas, March 2b and 27, 1975.
-------
Introduction
The formation, distribution, and transport of air pollutants are
three-dimensional phenomena. To properly understand them,
measurements must be made throughout and above the surface mixing
layer as well as at the surface. Instrumented light aircraft are
effective, versatile, and relatively inexpensive sampling platforms
for making these measurements.
This paper presents excerpts of data obtained and analyses
performed during a two-year airborne sampling program in the Los
Angeles basin, sponsored by the California Air Resources Board.
The results presented here are intended to demonstrate the versati-
lity of the airborne sampling technique and to show the types of
analyses which can be performed using three-dimensional data. More
extensive discussions of the data are contained in reports by
Blumenthal et al. , 1974. 2
Sampling Program
The sampling program utilized two light aircraft, each of which
made morning, midday, and afternoon flights over a preselected
route on each sampling day. For reference, the three sampling
routes normally used and a map of the L. A. basin are presented
in Fig. 1. Only two of the three routes were flown on any one day.
A typical flight consisted of a sawtooth pattern with about 8 to 10
vertical traverses (spirals). The aircraft would depart from an
airport, climb to the top of the mixing layer over the next traverse
point, and then spiral down (see Fig. 2). When the traverse point
was at an airport, the aircraft would descend to within 10 feet of the
ground; elsewhere, it descended to within 500 feet of the ground.
11
-------
Doth aircraft were equipped to measure continuously; O3, CO,
NOx> scattering coefficient, condensation nuclei count, temperature,
turbulence, relative humidity, altitude, and position. All data were
recorded in digital form on magnetic tape. The standard data format
produced from the traverses was a series of graphs of pollutant con-
centration versus altitude (vertical profiles).
In addition to the aircraft data, extensive surface pollutant and
meteorological data, as well as upper air wind data, were obtained
for each sampling day. Analysis of the data yielded a great deal of
information about both the chemistry and the transport processes at
work in the L. A. basin. In addition, the combination of meteorological
and pollutant parameters measured by the aircraft contributed to a
better understanding of the meteorology of the basin than could be
'obtained by meteorological measurements alone.
Mixing Layer Structure
A key element in forecasting and understanding air pollution
phenomena is an understanding of the mixing layer structure. Figure
3 is a vertical profile which clearly shows how pollutants are trapped
beneath a slight temperature inversion. Above the inversion, the
pollutant concentrations, as well as the turbulence indicative of the
mixing, drop off. In the particular profile shown in Fig. 3, the
temperature inversion clearly defines the top of the mixing layer; in
other cases, however, temperature alone cannot be used to define the
mixing depth, yet the pollutant parameters themselves still
indicate a confined layer.
From the combination of all the vertical profiles from a flight,
a mixing layer isopleth map can be constructed. These maps are
useful in explaining the distribution of pollutant concentrations at the
surface. Figure 4 is a mixing layer map showing the structure in
12
-------
the L. A. basin during a mild "Santa Ana" condition. The region of
low mixing layers is influenced by the sea breeze and contains the
highest pollutant concentrations. The marine air is trapped under
warmer air aloft. The regions of high mixing height indicate the
influence of the hot desert winds from the north and east. In these
regions, the sea breeze and desert winds converge, pollutants are
mixed through a deeper layer, and surface concentrations are lower.
Pollutant Budgets
Surface data alone can give an indication of health effects and of
human exposure, but to develop control strategies, the sources, total
pollutant budgets, and transport patterns should be understood. In a
location such as the L. A. basin, the mixing depth can change rapidly
and three-dimensional data are required to estimate pollutant budgets.
A useful technique for studying the total pollutant budget within
an air mass is to integrate the pollutant concentration through the
depth of the mixing layer, thus obtaining the total pollutant burden
over a point on the ground. This technique lets one follow the growth
of the pollutant burden in the air mass as the air moves along a
trajectory and yet takes into account changes in depth of the mixing
layer along the way. This method was used to analyze the data
within the envelope of possible trajectories of air arriving at Redlands
at 1800 PDT on July 25, 1973 (Fig. 5). The envelope shown in
Fig. 5 was constructed from both surface and upper air winds.
Pollutants arriving at Redlands at 1800 originated somewhere
within the envelope. When aircraft data were available for a given
point within the envelope, an actual integration was performed. In
some locations, however, ground data were used to supplement the
aircraft data. In these cases, the ground concentration was assumed
constant throughout the mixing height and was merely multiplied by
the prevailing height in the area. Mixing layer heights were determined
from the aircraft soundings and interpolated between soundings.
13
-------
Figure 6 shows how integrated contaminant loadings within the
surface mixed layer changed along the trajectory to Redlands (RED).
The aircraft data showed that pollutants were well mixed within the
mixing layer and that the top of the surface mixing layer was always
well defined.
Although there is considerable scatter in the data, the trends
are quite clear. There are large increases in the loadings of CO,
03, and visibility-reducing particulates between the coast and the
boundary between the western and eastern portions of the basin (near
Brackett [BRA]) across an area which includes most of the emission
sources in the Los Angeles basin. East of Brackett (BRA), loadings
of CO actually declined, as local emissions in the eastern basin were
apparently not sufficient to offset the effects of dilution or reaction.
The emissions in the western basin were sufficient to account for the
CO seen along the trajectory in the eastern basin. Ozone loadings east
of Brackett remained relatively constant, possibly due to continued
formation of ozone by previously emitted precursors and to the lack of
fresh emissions to scavenge it.
Nearly all the CO in the Los Angeles atmosphere comes from
the exhaust of motor vehicles. Motor vehicles are also the source of
about 75 percent of the NOx and 87 percent of the reactive hydro-
carbons. Using CO as a tracer for ozone precursors which were not
measured directly, we deduced that a large fraction of the ozone
precursors introduced into the air mass sampled at Redlands was
emitted in the western basin.
Another type of analysis which can be performed using 3-D data,
but which is impractical with surface data alone, is the estimation of
pollutant fluxes from one area to another. Both the budget and the flux
type analyses are useful for developing control strategies. Figure 7
14
-------
presents the result of a flux calculation performed for about 1700
PDT, July 25, 1973. The ozone flux from the western to the eastern
portion of the L. A. basin was calculated using vertical profiles and
wind data from pilot balloon soundings. The figure shows that enough
ozone was being transported across the boundary between the two
sections of the basin to cause the federal ambient air oxidant standard
to be exceeded in the eastern basin, even with no local emissions.
Figure 8 is still another example of the use of 3-D data in deter-
mining pollutant budgets. Isopleths of scattering coefficient integrated
through the mixing layer are shown for midday, September 20, 1972.
Basically, the plot is an indication of the mass of light scattering
aerosol (0. 1 < aerosol diameter < 1 (im) within the mixing layer.
This large concentration of aerosol was later shown to be transported
eastward into the convergence zone shown earlier in Fig. 4. The
transport was masked in the surface data, however, because the
deepening mixing layer to the east caused a decrease in surface
concentration, but not in total budget.
Layers Aloft
Pollutants which are trapped in layers aloft and are isolated
from fresh surface emissions may have important effects on surface
concentrations. For example, these upper layers can impinge on
elevated terrain, or be transported long distances and then be
reentrained within the surface layer when the mixing depth deepens
enough to include them. Slow chemical reactions such as the conver-
sion of SOs to sulfate or the reactions of nitrogen oxides and ozone to
form nitrates can take place aloft without interference by fresh
emissions or scavenging of reactants by the surface. The chemical
and transport processes occurring within these layers can be most
easily studied by an airborne sampling system. Some examples of
layering phenomena follow.
15
-------
One major type of mochnnium for the creation of layers aloft
la buoyancy. Warm point source plumes can penetrate into a stable
layer and become trapped aloft. Likewise, solar heating of mountain
slopes can warm nearby air and drive it aloft to an equilibrium
elevation. Figure 9 is a vertical profile which shows a layer formed
by upslope flow. Air from the surface layer which was normally
confined by the slight inversion based at around 2500 feet was carried
up the mountain slopes north of Rialto (RLA in Fig. 1) and then
carried back over the basin. The heated air reached an equilibrium
in the stable isothermal layer aloft. A layer of clean air can be seen
between the two polluted layers.
Another mechanism for the creation of upper layers is shown
in Fig. 10. This figure contains vertical cross-sections of scattering
coefficient from the coast at Torrance (TOA) inland to Redlands
(RED). Higher numbers indicate poorer visibilities or higher
concentrations of particulates. This figure shows the effect of
convergence between the sea breeze (westerly) and a mild Santa Ana
wind (easterly). The high aerosol concentrations seen near Fullerton
(FUL) in the midday cross-section were transported east by afternoon
N
and then carried aloft in the convergence. Between 4000 and 5000 ft
msl, some of the aerosol was actually carried back toward the coast
by the upper winds.
A common type of mechanism in the Los Angeles basin for the
formation of layers aloft is the undercutting of a portion of a previously
well mixed layer either by an intrusion of cleaner and cooler sea air
or by the formation of a nighttime radiation inversion. Figures 11
and 12 show the effect of undercutting by the sea breeze. Figure 11
is a vertical profile obtained at El Monte (EMT) after the passage of
the sea breeze "front". Sea air containing relatively fresh emissions
is seen up to about 1400 ft msl. The air between 1500 feet and 2500
16
-------
feet was previously part of the surface mixing layer, but was isolated
from the surface by the onset of the cooler marine air below. The
pollutant ratios in the upper layer are indicative of an air mass in
which the pollutants are relatively well aged compared to the surface
layer. Figure 12 is a vertical b . cross-section from the coast at
SC3.U
Santa Monica (SMO) to Redlands (RED) similar to the one in Fig. 10.
The undercut layer from Fig. 11 can be seen to extend from the coast
across the basin to the Pomona area (BRA), a distance of about 40
miles. This layer aloft was not detected by surface monitors, and the
ozone concentration within the layer exceeded 0. 5 ppm even near the
coast.
An example of a plume trapped aloft is shown in Figs. 13 and
14'. Figure 13 is another bscat vertical cro ss-section for the
northern portion of eastern L. A. basin. A relatively dense layer
of aerosol was spread out over much of the eastern basin at about
1800 ft msl. This layer resulted from the plume from a power
generation and steel mill complex near Rialto (RIA). The plume was
buoyant enough to penetrate the surface based inversion and stable
layer, but was trapped within the upper subsidence inversion. From
Fig. 14, one can see that the plume and the surfaced based radiation
inversion layer contained large amounts of primary pollutants which
were photochcmically "young". These data were obtained at night and
the ozone was virtually completely scavenged within the plume and the
surface layer. Other layers aloft above the plume, left over from
the day before, still contained relatively high ozone concentrations
and were well aged. These upper layers as well as the plume were
reentrained as surface heating deepened the surface mixing layer.
A combination of many of the phenomena mentioned so far is
shown in the vertical profile in Fig. 15. In this figure, the tempera-
ture profile by itself reveals little, if anything, about the structuring
17
-------
of the pollutant layers, but the combination of parameters taken as
a whole is quite revealing of both the chemistry and the meteorology.
Using the pollutant parameters a surface mixing layer can be clearly
defined into which fresh pollutants were being emitted and in which
some photochemical ozone production was starting to occur. Above
that layer, a large buoyant plume of fresh pollutants can be distinguished.
The high NO, condensation nuclei, and CO and the deficit of ozone
identify this layer as containing relatively fresh emissions. Above the
plume, a third layer can be seen which contained very well aged
pollutants. The ozone and bscaj. concentrations were well correlated
and the primary pollutant concentrations were very low. This layer
was still separated from fresh surface emissions and contained
pollutants left over from the day before.
A look at Fig. 16 shows what happened later in the day. Surface
heating raised the mixing layei; and all the layers up to 3000 ft msl
were entrained. Above 3000 ft msl, a remnant of the layer left over
from the previous day can still be seen. Without understanding the
three-dimensional morning layered structure, the afternoon surface
concentrations would be difficult to interpret and understand.
Special Effects
During the sampling program in the L. A. basin, several
special experiments were performed to look at specific phenomena.
During one of these, a series of north-south horizontal traverses
was made at varying altitudes over Upland (CAB). Upon examination
of the data, we saw that several bulges had occurred in the mixing
layer and that in specific areas, pollutants were penetrating through
the inversion for several hundred feet. A vertical cross-section of
ozone concentration along one north-south route is shown in Fig. 17.
The wind at the time was east-west and the bulges were shown to
18
-------
coincide with the major roadways in the area. We believe that solar
heating of the roads rather than auto exhaust is primarily responsible
for generating the buoyancy necessary to distort the top of the mixing
layer. The bulge near San Antonio Dam on Fig. 17 is probably due
to surface heating of the mountain slopes adjacent to the dam.
Conclusions
From the data presented, one can see that the 3-D distribution
of pollutant concentrations and the meteorological structure of the
atmosphere can be very complex. In order to understand the evolution
of surface concentrations and to device effective control strategies,
three-dimensional effects must be taken into account. Airborne
monitoring systems have proven to be versatile and cost effective
tools to obtain an under standing of the 3-D distribution and transport
of pollutants.
19
-------
REFERENCES
1. Blumenthal, I). L. , et al. , "Three-dimensional pollutant gradient
study - 1972- 1973 program, " MRI 74 FR-1262 report to the Cali-
fornia Air Resources Board, prepared under Agreement Nos.
ARB-631 and ARB 2- 1245 (1974).
2. Blumenthal, D. L. , et al. , "Determination of the feasibility
of the long-range transport of ozone or ozone precursors, "
EPA - 450 / 3-74-061 , report to the Environmental Protection
Agency under Contract No. 68-02- 1462 (1974).
20
-------
Fig. ] . SAMPLING ROUTES IN THE LOS ANGELES BASIN
-------
Fig. 2. VERTICAL PROFILE FLIGHT PATH
22
-------
5,000
I6C0
1400
4.00C —
UJ
Q
3
3,000
to
0J
2,000
1,000
1200
1000
o
- 800
600
400
200
irrT-
< ? f
: >
i • i
i-U L.
J
st : t
*-2.-1-
—7
2
i
i
; t
? t
—7
? t
7 £
* f
* t
<
»
r
0 I
0.2
0 J
04
10
15
20
25
30
35
40
45
0 5
i
50
6
¦
¦ 6
t
5
¦
10
¦
15
i
20
25
i
30
35
40
1
45
i
0
10
1
20
i
30
40
i
50
¦
1
K &
o
70
80
90
1
100
0
1
2
3
4
5
6
7
8
9
10
Ground Elevation
96 ft (30 m)
MIXING LAYER
Parameter
NO.
Oj
CO
Temperature
Relative Humidity
b>cot
CN
TuRSyvCHCt
Unit
ppm
ppm
°C
%
I0"4m"'
cm"3 1104
S*fc33L
N
Z
c
T
H
B
X
E
Fig. 3. VERTICAL PROFILE OVER FULLERTON (FUL)
SEPTEMBER 20, 1972, 1240 PDT
-------
Fig. 4. MIXING LAYER HEIGHTS (ft) - AFTERNOON
September 20, 1972
-------
Fig. 5. TRAJECTORY ENVELOPE FOR AIR ARRIVING OVER REDLANDS AT 1800 PDT, JULY 25, 1973
-------
E
"V
O*
o
u
WESTERN
EASTERN
BASIN
^REO
°RE0
1200 1400 1600 1800
TIME (POT) ALONG TRAJECTORY
fl (DIMFNSI0NLES5 OPTICAL QUALITY
CALCUL AT CO FROM b,(0|)
Fig. 6. POLLUTANT LOADINGS IN SURFACE MIXED LAYER
ALONG TRAJECTORY ARRIVING AT REDLANDS IN
AFTERNOON OF JULY 25, 1973. (Aircraft measure-
ments plus estimates made from ground data and
mixing height. )
26
-------
WESTERN BASIN
NOxEMlSSlON RATE
95 METRIC TONS/hr
r'-y
&
O3 FLUX THR0U6H,
O SHADED BOUNDARY;
EMT (85 ~ 60 METRIC,.
TONS/hr
.(HOOR-
(Zf
juJjci
EASTERN BASIN
NO* EMISSION RATE
10 METRIC TONS/hr
l5h
EDLANDS 18=00 PDT
MEAN TRAJECTORY
FUL
*3?
BOUNOARY-
BETWEEN
WESTERN
EASTERN
BASINS
y Rv
o
SNA
(m)
CNO WINDS ALOFT
r29
"•51
1000 -
AVERAGE CONCENTRATION IN
BOX (IF STEADY STATE)
IS 25 pphm
O n
RIA
sfc
_C J
lSr
92
(km/hr)
iSl
T
Oil
5«»
'V2?
N?J
O
RIV
__ "l
Fig. 7. ESTIMATED OZONE FLUX FROM WESTERN TO EASTERN BASIN, JULY 25, 1973,
1700 PDT
-------
Fig. 8. bBcat INTEGRATION - MIDDAY
September 20, 1972
-------
5,000
4,000 -
3,000
a
3
CO
!£>
2,000
1,000
1600
1400
1200
1000
u
o
3
- 800
600
400
.200
LAYER CAUSED by
UPSi-0pE FLOW
SURFACE MIXING LAYER
Ground Elevation
1434 ft (435 m)
Parameter Umit Stmbol
A
0
1
0 1
i
¦
0 2
•
i
0.3
>
04
>
>
05
1
0
3
>
10
1
(5
¦
20
•
23
¦
30
35
40
'
45
50
1
-IS
0
5
10
15
20
¦
25
30
35
40
45
¦
0
10
20
30
40
i
50
¦
60
•
70
80
¦
90
100
¦
0
1
2
3
4
6
6
7
8
9
10
N0X
Oj
CO
T[UI>CRATU»C
Relative Huwoity
buai
CN
TURBULtHCt
ppm
ppm
°C
%
10-Vi-'
em-s ¦ I04
Cff^'sac*1
N
2
C
T
H
B
x
E
Fig. 9. VERTICAL PROFILE AT RIALTO (RIA) JULY 19, 1973, 1738 PDT,
SHOWING LAYER CAUSED BY UPSLOPE FLOW
-------
Midday
. . Afternoon
all cross section numbers » 10
'indicates top op mixing layer
Fig. 10. VERTICAL CROSS-SECTION OF bscat MIDDAY
AND AFTERNOON, OCTOBER 24, 1972
30
-------
1600
5,000
1400
c £ ,t
C • KB
C •
—«yit
T t*
4,000
1200
Vc
-V* *
.* % t i
T
> C
1000
•jj
o
H
_J
<
3,000
o
o
_ 800
<
2,000
1,000
Fig. II.
600
400
200
0
1
•
0.1
•
i
0 2
<
0.3
i
04
•
0 5
•
0
1
5
1
10
1
15
•
20
>
25
i
30
i
35
i
40
i
45
i
50
¦
-6
•
0
¦
5
¦
10
i
15
i
20
i
25
i
1
.
O
35
•
40
45
i
0
1
10
i
20
•
30
40
<
50
i
60
<
70
•
80
•
90
¦
100
•
0
1
2
3
4
5
6
7
8
9
10
Ground Elevation
290 ft (88 m)
SUBSIDENCE INVERSION
LAYER-AGED POLLUTION
WITH HIGH bteo1 AND
OZONE, MODERATE
CONDENSATION NUCLEI
COUNT
MARINE LAYER"FRESH
EMISSIONS WITH HIGH
CONDENSATION NUCLEI
COUNT, MODERATE b»c
-------
— •SUBSIDENCE HVERStON
U4RINE INVERSION
GROUND level
Fig. 12. VERTICAL CROSS-SECTION OF b FOR AFTERNOON OF JULY 25, 1 973
scat
(Units are 10" 4 m" '. j
-------
EARLY MORNING
50001
4000"
ft. (MSL)
3000-
2000"
1000"+
.3JS.-V
(bscat I0'4-"1)
1
BRA
0421 PDT
1
ONT
0549 PDT
I
RIA
0535 PDT
1
RED
0519 PDT
Fig. 13. VERTICAL CROSS-SECTION OF b
26-27 July 1973
scat
-------
5,000
4,000
3,000
o
3
t-
_l
<
CO
2,000
1,000
0«-
1600
1400
1200
1000
- 800
6C0
400
200
AGED LAYERS ALOFT
POINT SOURCE PLUME
SURFACE MIXED LAYER
Ground Elevation 1430 ft (43 5 m)
0
5
i
10
i
15
'
20
1
25
>
30
¦
35
t
40
•
45
¦
50
-B
0
5
10
15
20
25
30
35
40
45
i
0
K)
20
30
40
50
*
60
70
60
90
1
too
i
6
1
2
3
4
&
6
7
8
9
10
P«IUUgTEB
NO,
03
CO
Temperature
b.ct
CH
Turbulence
Umit
Symbol
N
ppm
Z
ppm
c
°C
T
%
H
IO-4m-'
B
CITT3 1104
X
ern^W
E
Fig. 14. VERTICAL PROFILE OVER R1ALTO (RIA) 0535 PDT JULY 27, 1973
-------
w
Cn
5000
4000
IU
o
z>
:3000
2000
1000
1600 r
1400 -
1200
1000 -
LU
O
3
h 800 -
600 -
400 -
200 .
N ZX
.11 1 III
0
1
¦
0.1
1
¦
0.2
i
i
0.3
i
i
0.4
i
i
0.5
i
"5
i
0
1
5
10
•
15
i
20
25
i
30
i
35
i
40
<
45
i
0
1
1
'
2
>
3
—1_
4
L.
5
i
6
i
7
•
8
i
9
i
10
i
AGED LAYER FROM
PREVIOUS DAY
POINT SOURCE
PLUME
1
SURFACE MIXING
LAYER ^
GROUNO ELEVATION
950 ft (290 m)
PARAMETER UNIT SYMBOL
10
15
20 25
30 35
40
45
50
NO,
ppm
N
°3
Ppm
Z
TEMP
°C
T
^scot
10"4 m"'
B
CN
cm-3* I04
X
CO
ppm
C
Fig. 15. VERTICAL PROFILE OVER ONTARIO (ONT) 0953 PDT JULY^ 26, 1973
-------
5000
4000
1600 r
1400
1200
3000
6 1000
2000
1000
ui
o
3
_J
<
800
600
400
200 "
1 1 1 1 1 1
0
0.1
0.2
0 3
0.4
0.5
1
i
1
1 1
i *
i i
i
-5
i
0
1
5
1
10 15
• •
20 25
30
35 40
' i
45
*
T
AGED POLLUTANTS
FROM PREVIOUS
DAY
1
J
SURFACE MIXING
LAYER. POINT SOURCE
PLUME SEEN IN
MORNING SOUNDING
HAS BEEN ENTRAINED
i
GROUND ELEVATION
950 f 1 (290 m)
PARAMETER UNIT SYMBOL
8
10
N0X
°3
TEMP
bsco»
CN
ppm
ppm
.4 . I
10 m
cm'3* I04
N
Z
T
B
X
Fig. 16. VERTICAL PROFILE OVER ONTARIO (ONT) 1318 PDT JULY 26, 1973
-------
CO
<1
s
V
3
5000
4500 -
4000
3500
3000 -
2500
2000 -
0 2
San
Bernardino
F reeway
Ozone (ppm)
Highway
$
Ozone and Temp-
erature Vertical
Profile
Fig. 17.
VERTICAL, CROSS-SECTION OF OZONE CONCENTRATIONS ALONG
EUCLID AVE. , UPLAND, OCTOBER 4, 1973, AFTERNOON
-------
SAMPLE SIZE REQUIREMENTS FOR STATISTICAL VALIDITY
Dr. David T. Mage
RAPS Project Officer
Air Monitoring Branch
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
NERC-LV
-------
I. INTRODUCTION
Aircraft sampling produces almost instantaneous measurements of air
pollution at points in space connected by the flight path. These data
can be used to estimate the time average concentration at these points in
space or in given surrounding volumes. However, air quality standards and
diffusion models usually relate to time averages on the order of one hour.
The instant measurement (i.e. once every five seconds) represents a sample
of size 1 from the population of 720 samples which compose the full hour.
The accuracy with which the single sample represents the entire distribu-
tion depends primarily on the variance of the population being sampled.
We can express this accuracy by use of a confidence level that the measure-
ment is within a given value of the true mean of the distribution. To achieve
a greater confidence level or to minimize the uncertainty a larger sample
size is necessary. Based upon the desired accuracy and confidence level of
the measurements to be made, and an estimate of the variance of the
population to be sampled one can design the sampling process. Because these
data are highly correlated,' a one-minute flight producing 12 consentive
samples every five seconds, will not be the same statistically as 12 separate
flights during the hour each producing one data point, with the spacing
between flights chosen randomly. Standard statistical techniiues may be
misleading if independence is assumed. This paper discusses the problem from
the point of view of the experimental design and presents the areas which must
be considered in the process.
38
-------
II. STATEMENT OF THE PROBLEM
At the inception of the project, a goal should be clearly stated for
the measurement in terms of
1. how accurately it is to be made, and
2. the desired level of confidence for the measurement.
As an example, the estimation of the hourly average of a pollutant C
might be the objective of the experimental measurements, where:
p _ S S SSCdxdydzdt
^
The specification may be that this measurement is to be within 2 ppm of
the true value with a confidence level of 90%.
Inhernnt Limitations
Measurements from airborne platforms are subject to greater uncertainties
than the identical measurement made on a bench in an air conditioned-temper-
ature controlled laboratory. For a given instrument the errors to be consid-
ered are:
a. Instrumental
The instrument itself has a fundamental accuracy due to variability
of manufactured components, nonlinearities, sensitivity to contaminants,
minimum detectable levels, hysteresis, etc. These uncertainties are usually
stated by the manufacturer, for example, as a concentration or per cent
uncertainty; i.e. + 1 ppm or -f 1% of full scale, whichever is greater.
b. Calibration gas
The impurities in a zero gas, and the accuracy of a span gas
analysis and its subsequent change with time in the cylinder add additional
uncertainty to the measurement.
c. Environment
In non-pressurized aircraft, variations of altitude (pressure)
39
-------
can produce an effect as well as variations in temperature. Vibration
may also cause problems unanticipated in equipment designed for laboratory
use .
d. Response Time
In an aircraft moving at high speed, the finite response time
of an instrument will spread the measurements out in space and time, damping
out maxima and raising minima. The details of this type of error are discussed
in a paper to be presented by the author later in this meeting. Each of these
four categories of uncertainty introduce an error into the measurement.
Assuming these errors are independent and normally distributed one can describe
the uncertainty as a standard deviation in terms of the four standard deviations as
• h T • Vj
'a *(->,_ t
-------
The only conditions it requires is that samples be statistically independent.
There is ample evidence that consecutive samples are correlated and are not
statistically independent. With random sampling, however, the samples are
no longer consecutive, and the time between samples varies in a random fashion.
If the samples are statistically independent, the central limit theorem should
apply, leading to a normal distribution.
The normal distribution (Figure 1) has a probability density function
given by equation (1):
,1'o) ¦ wr exp
(x-M)2
—71
(1)
2cT
The two parameters u and 0" are the location and scale parameters,
respectively, of the distribution. Although the function cannot be
integrated directly, tables are widely available to calculate the probability
associated with particular ranges of x. The shaded area under the curve in
Figure 1 represents 5 percent of the total area and corresponds, approximately,
to the probability of occurrence of values of x outside the interval x 4- 2o*
(actually x-f 1.96cr).
According to the central limit theorem, the means of individual sample
groups, each comprised of n random samples, must have the same distribution
as in Figure 1. Furthermore, the average of the sample group means must
approach the actual population mean u , while the dispersion of these means,
as determined by cr, is related to n and cr , the standard deviation of the
J P
original population:
cr = (2)
N/tf-
Thus, as n becomes large, the dispersion of the means of sample groups
approaches 0, which is what would be expected; as n becomes large, the sample
group more closely resembles the original population. If the standard
41
-------
0.4
0.3
u.
Q
a.
0.2
0.1
4$r~
NORMAL PROBABILITY
DENSITY FUNCTION
(PDF)
FOR THE CASE OF !
M =0.0 j
&
cr = 1.0
\
\
\
^A5% of total area
\ \
\ \
;¦ *
i
+1 +2 *3 <-4
-------
deviation is known confidence levels can be determined prior to the
experiment.
A more valid approach from a statistical standpoint is to calculate
2
the sample variance for the group of n samples obtained from each locatior
i and then use this variance to determine the confidence interval. The
2 2
sample variance s^ is only a rough estimate of the true variance at
location i, and we must, therefore, introduce the Student's t-distribution
to calculate the confidence interval. Tables are available giving the value
of the factor t„T , for any confidence level CL, and the mean is said to
CLi) n-¦ 1.
lie within the following confidence interval:
si
*1 ± ^CL.n-l' ~ <3)
-------
12.0
ii
10.0
8.0
I
c
6.0
U
4.0
2.0
PLOT OF THE FACTOR
*CL,n-1
AS A FUNCTION OF
DEGREES OF FREEDOM
n-1
-CL=0.99
/
CL=0.95
VCL=0.90
40
n-1
60
80
FIGURE 2
100
-------
Sampling Statistics - With Correlation
To discuss non-independent sampling statistics we can look at different
pollutant distributions each with a mean (C) of 0.5 during a 1-hour period,
where C = Jcf (C)dC, shown in Figure 3-
\
' IT'
V
T ' f 7 * • r •"
I- \ .
I: ¦
C ^
M V ¦
C
i I
A ...
o
time
time
time
C
Case I
/ i
Z..:
c
CaseH
c
Case HE
Figure 3. Different Types of Distributions
The bottom diagrams correspond to the frequency distribution of the means of
consecutive samples over time /it taken from the distributions at random,
where ^t is large compared to the frequency of data fluctuations. The first
distribution is one which is not stationary, i.e. the mean is decreasing with
time. The second distribution is composed of random fluctuations between
0 and 1, corresponding to flips of a coin. The distribution has a stationary
mean of 0.5 but a rather high standard deviation. The third distribution is
the nicest to sample from, i.e. the mean is stationary at 0.5 and the random
fluctuations about it are of small amplitude. The frequency distribution
44
-------
of samples will have a small standard deviation and a small sample will have
a high probability of success. It is quite obvious that one sample of size
At is not equivalent to n samples of size (^t/n) chosen randomly throughout
the hour. Indeed the last two cases shown, where consecutive values fluctuate
about the mean, corresponding to the flip of a coin, are the only cases
shown where the central limit theorem as previously discussed would apply
directly to an increase in ^Xt. We may therefore look at the previous case
of independent consecutive observations as the best possible situation
and an upper limit on the confidence for the sample.
Examples of Variability
Figures 4 and 5, represent two spirals taken as part of a plume study
on February 19 by a RAPS helicopter 10 miles downwind of the Labadie Power
Plant, west of St. Louis. These spirals down from 3000 msl to 1000 msl
took 5 minutes each. An additional three minutes were necessary to climb
from 1000 to 3000 ft. so one could treat them both as samples separated by
an eight minute interval. Neglecting the correction for response time, it
is readily seen that there are significant differences as seen on Figure 4.
The first spiral measured a maxima of 1.05 ppm SO2 at 1200 ft. and 0.4 ppm
SC>2 at 2400 ft. The second spiral measured 0.40 ppm SC>2 at 1200 ft. and a
maxima of 0.J ppm S0„ at 2400 ft. Figure 5 shows the NO data measured
I — x
simultaneously in the same spirals. At 1200 ft. the first spiral had a maxima
of 0.24 ppm NO compared to 0.10 ppm NO at 1200 ft. found in the second
X x
spiral. The difference between the two spirals is probably due to the fact
that the wind may have shifted slightly in the eight minute interval between
the samples. A shift of 5.7° would be sufficient to move the plume center
line by 1 mile at a distance of 10 miles from the source. It is of interest
to note that Turner in his Workbook of Atmospheric Dispersion Estimates onlv
45
-------
46
-------
2500-
22000
Q)
73
D
+*1
<
1500u
1000,
0.10 0.15 0.20
(NOx), ppm
47
-------
gives atmospheric dispersion coefficients which "represent time periods of
about 10 minutes'". Thus these spirals, under the influence of an elevated
source, may be only valid estimates of the mean of an eight minute period,
where consecutive eight minute mean values are changing with time and their
sequence is non stationary. To put it more bluntly, if we assume the NO^
samples at 1200 ft. are random and independent the sequence 0.24, 0.10 would
probably fluctuate about 0.17 if the spirals were continued. If the data
are correlated the next sample would be expected to be below 0.10 and the
mean of the series would be closer to 0.10 than O.I7. Even if we have 2
samples we are in a quandary. When we only have one small sample the uncer-
tainties may be so great that an estimate of an hourly average may only be
made within a very large confidence interval.
Figure 6 shows the burden of SO2 in ppm-Meters at Washington University,
St. Louis, Missouri measured by a correlation spectrometer on 2 August 197^
reported by EMI at the January 1975 RAPS conference at St. Louis. Assuming
SO2 does not penetrate the inversion capping the mixing layer, and the mixing
layer is 1000 meters deep, a burden of 100 ppm-M corresponds to an average
concentration of 0.10 ppm SO2 in the 1000 meter layer. The average concentra-
tion measured by the helicopter in a spiral would approximate this value.
It is evident that these burden data are highly correlated since high
values cluster together and low values cluster together. For the (0 one-
minute values between 1500 and 1559 the average C = 85.4 ppm-Meters and
the standard deviation s = 88.1 ppm-Meters. By the central limit theorem,
if this hour were sampled by 5 independent, or randomly chosen samples, an
infinite number of times, the averages of these 5 minute samples would form
a normal distribution with a mean C = 85.4 and a standard deviation
-------
500-
400
(M
8
2-300f
l
e
a 20O-
100
<£>
r,
0
O
0
1500
fpcccoo
0Q
©
•v-'
a
CFXT
£G
0
!r-
©
J C-.
C--
c-
r
©
©
2
0
fv>
©
e
§
1530
TIME
-lees'- 1
1600
.DK
G)
-------
125
.100
.075 -
.050
.025
FIGURE 7
BURDEN
RANDOM
SAMPLES
0 L
0
100 200
ppm-Meters
300
50
-------
.150
.125 i
.100
.050
.025
.075
x\\^
FIGURE 8
BURDEN
CORRELATED
SAMPLES
100
\ \
\x
-V-
iv
¦V
. \ >
N*
200
ppm -Meters
51
300
-------
QO "I
j.y. = 39.4 ppm. This distribution is shown in figure 7 as a histogram of
the fraction of the population occurring in each interval of 10 ppm-M. The
helicopter spirals could start at any time during this hour, and for the
purpose of this paper, let us allow 60 possible starting points each separated
by one minute. These 60 flights would produce 60 5-minute averages of
consecutive data which are not independent, but on the contrary, highly
correlated. These data were averaged consecutively 5 at a time. The resulting
distribution of these data is shown as figure 8. The mean is still 85.4
ppm-Meters but the standard deviation is J2 ppm-Meters, still relatively high,
and almost double the standard deviation which would occur if the 5 one-minute
samples were independent. Faced with these observations how can one give a
confidence level to an arbitrary 5-minute sample taken by an aircraft in a
region where elevated plumes may or may not be present. The following section
describes a technique by which confidence levels for non-independent samples
may be obtained.
Development of Estimates of Confidence Levels for Non-independent Samples
1. Obtain several data sets, such as figure 6, for an entire 1-hour
period, for various stability classes and wind directions, from the ground
or from the air by continuous spiraling.
2. Prepare an autocorrelation diagram showing the autocorrelation
coefficient as a function of time lag ^= n A t for the cases of interest.
3- Develop an autoregressive (Markov) process + t = at ^t * ^t
where a and b are random variables uncorrelated with C . Choose the
t t t
distributions of a and b such that the series C , C , C „ ,
t t + At t + 2at'
C - , C , , . . . , C
t + 3 A t t + 4 A t t + n^t
has a similar autocorrelation diagram to the data. Develop the computer
52
-------
model for this analysis.
4. Run the computer model to generate a time series of 60 values from
through + 59^ t' 3n<* ran^om^y choose a value for a starting point and
/
compute the average of the 5 consecutive values at that point , comparing
this to the mean of the distribution C,n. Repeat this procedure a large
bu
number of times.
5. From these realizations of the process of generation and sampling
Qr
develop a histogram of frequency of —- and from this diagram obtain the
C60
confidence intervals for the 5-minute sample.
Summary and Conclusion
Each airborne measurement should be treated as a random sample from
a distribution of non-independent data. Standard sampling theory for
independent observations can lead to quite erroneous results if used for
these data. Proper experiment design requires knowledge of confidence levels
for various sample sizes and sampling schemes. A procedure is outlined by
which confidence levels may be obtained for non-independent samples. Without
a careful analysis, sample sizes may be too small to give any reasonable
confidence to the results of comparisons between the measurement and models
which predict the measured parameters for a significant larger averaging
t ime.
53
-------
ANALYSIS OF SIZE FRACTIONATED AEROSOLS*
T. G. Dzubay
Chemistry and Physics Laboratory
National Environmental Research Center
Environmental Protection Agency
Research Triangle Park, N.C. 27711
*
Presented at conference on "Applications of Airborne Platform
Monitoring," NERC-Las Vegas, March 27, 1975
-------
introduce ain
The measurement of aerosols must be given a high priority m
studies to characterizc the urban plume. Aerosols can not only contain
toxic substances, but because of the ability of particles to discolor
a blue sky and to degrade visibility, aerosols cause one of the most
obvious and objectionable aspects of air pollution.
In a modern aerial platform study a variety of chemical and phy-
sical parameters need to be measured. Measurement of only the total
mass concentration would be insufficient because the size distribution
of an aerosol has a large effect on its ability to affect visibility or
to lie inhaled into a human respiratory system.
Previous measurements by Whitby et al have demonstrated the exis-
tence of a himodal distribution of particle volume ( and hence mass)
in the atmosphere (1). In a recent experiment, particles in the two
modes were spcarately collected in a dichotomous sampler and analyzed
by x-ray fluorescence spectroscopy (2). This measurement confirmed
that the fine and coarse particles differed completely in chemical com-
position. The fine particles contained elements which were character-
istic of combustion products; the coarse fraction contained elements
which are commonly found in the earth's crust. A comprehensive program
to monitor the composition of size fractionated aerosol could provide
a valuable indication of the atmospheric burden of particles due to
various energy and transportation sources. A set of measuring methods
for such a program Ls briefly described in the text that follows.
54
-------
AEROSOL SAMPLING
Cascade Impactors: The cascade impactor has come into widespread use
for collecting suspended particles according to aerodynamic size.
Subsequent chemical and gravimetric analysis can be performed on the
particles which arc collected in each size range.
For quantitative sampling of aerosols using a cascade impactor, it
js necessary that the impaction surface have an adhesive coating. If
the adhesive coating is not used, particles can bounce off of their in-
tended impaction surfaces and be collected on subsequent stages or on
the backup filter. This leads to severe distortion of the apparent
size distribution. An example of this distortion is shown in Figure 1,
which compares size distributions made using two identical cascade im-
pactors which were coated with dry and with grease-coated impaction sur-
faces [3). In a scries of measurements, it was found that the apparent
mass median diameter of the size distribution, which one deduces, is
erroneously decreased by factors ranging from 2 to 5 when the grease-
coated impaction surface is replaced by a dry aluminum surface. Although
the use of an adhesive coating is desirable for quantitative sampling,
that coating often interferes with chemical procedures for analyzing
the deposits.
DIC110T0M0US, VIRTUAL IMPACTORS: A virtual impactor can be used as a
means of eliminating the particle bounce problem. In a virtual impactor,
particles are impacted into a slowly-pumped void and then drawn onto a
filter where they arc collected. There is no impaction surface for the
55
-------
particles to bounce off of. A virtual impactor which collects
particles in two size ranges is commonly referred to as a dichotomous
sampler. For most monitoring applications, the size separation diameter
for the dichotomous sampler should be set at about 2 to 3.5 pm. A se-
paration at 2 iun would correspond to the typical location of the mini-
mum in the bimodal distribution (1). A separation at 3-5 pm would cor-
respond to aerosol size separation made by the upper and lower passages
of the human respiratory system.
An automated dichotomous sampler which operates on the principle
of virtual impaction has been developed by Loo of the Lawrence Berkeley
Laboratory (4). The device is illustrated in Figure 2. When operated
at a flow rate of 50 liters per minute, the separation between fine and
coarse particles occurs at a 50 percent cutpoint diameter of 2.5 urn.
The complete size fractionation characteristics and loss curves are shown
in Figure 3. The cutpoint could be shifted to 3.5 pm by operating the
device at a lower flow rate, which is estimated to be about 42 liters
per minute.
Ten of the automated dichotomous samplers have been deployed in St.
Louis as a part of the Regional Air Pollution Study. An additional
sampler will soon be deployed near a copper smelter in Utah. These
devices have the capacity to collect 36 consecutive samples for time
intervals adjustable between 1 and 100 hours. A regulator maintains the
flow rate at the inlet at 50 liters per minute. In heavily polluted
56
-------
atmospheres, the maximum sampling duration may have to be limited to
4 to 12 hours in order to prevent excessive loading of the filters.
DETERMINATION OF COMPOSITION
In order to fully characterize an aerosol, it is important to deter-
mine the chemical and elemental composition. Carbon compounds, chlorine,
bromine and lead are characteristically present in auto exhaust. Sul-
furic acid or sulfate compounds are characteristically present in the
primary and secondary particles resulting from the burning of fossil
fuels to generate electricity. Al, Si, K, Ca and Fe are characteris-
tically present in suspended soil dust. A combination of analytical
methods is needed to characterize these species.
Chemical Analysis: A summary of several methods for measuring the
chemical composition of aerosols is given in Table l. In the past, there
has been a reliance on wet chemical techniques for making chemical
measurements. Information on interferences and accuracy for these methods
has not generally been available. For the four wet chemical methods for
sulfate, given in Table I, a study of interferences, intercomparabi1ity
and agreement with x-ray fluorescence methods for total sulfur is
being conducted (5). Recently, Dr. J. liusar has utilized a flash vol-
atilization technique in which samples are rapidly heated to decompose
the sulfate. The resulting SO^ is detected using a flame photometric
detector (FPD) . A related technique is being optimized for determining
ll^SO^ which is collected on teflon filters. Sampling times are limited
to a few minutes in order to minimize reactions taking place between
57
-------
particles on the filter. The filters are gently heated in a dry atmos-
phere containing methanol in order to selectively volatize SO^, which
is analyzed using a FI'D (6).
Brosset of Sweden has developed a program for determining three
species which are associated with the fate of sulfuric acid aerosol. For
particles which are collected on Fluoropore filters, sulfate is determined
using a wet chemical method, and ammonium ions are analyzed using an
ion-seiective electrode. Strong acid is determined using a Gran tit-
ration procedure, which discriminates against weak acids. With these
measurements, the amount of sulfuric acid and ammonium sulfate can be
deduced, whenever the concentration of other strong acids (NHO^ or HC1)
is low.
Patterson (7) has developed a combustion technique for determining
total carbon in collected aerosol (as well as hydrogen and nitrogen).
A technique is presently being developed for determining volatile carbon
compounds as well as total carbon (8). For the purpose of determining
mass balance in aerosols, these techniques may be somewhat more useful
than the methods which only measure benzene-soluble organics.
lilemental Analysis: During the past two years, energy dispersive x-ray
fluorescence analysis (XRF) has become widely used for determining the
elemental content of aerosol samples. It is nondestructive and requires
only a few minutes for analysis. The XRF methods is generally applicable
to elements which have atomic numbers above 12. In Table II, the detection
limits are given for a modern energy dispersive XRF spectrometer, which
58
-------
uses secondary fluorescers (9). The detection limits in Table II are
2
expressed in units of ng/cm assuming that the particles are collected on
2
a membrane filter which has a mass per unit area of 5 mg/cm . The de-
tection limits are also plotted in Figure 4. Also shown in Figure 4
are typical ranges of concentrations of trace elements which have been
measured in urban atmospheres (12).
The detection limits of Table II and Figure 4 can easily be ex-
3
pressed in units of ng/m . This is done by dividing the tabulated values
3 2
by the volume sampled per unit area of filter (in m /cm ). This manner
of expressing detection limits is very convenient because it allows the de-
signer of an experiment to consider a variety of filter media, flow
rates and sampling Intervals.
Detection limits for trace clement analysis by neutron activation
analysis (NAA), by atomic absorption (AA) and by. emission spectroscopy
have been reported by Zollcr and Gordon (10) and by Thompson et al (11).
The detection limits were reported by those authors for 24-hour samples
using specific filter media. The values in Table II are expressed in
2
the more general units of ng/cm . From this table it is clear that XRF
compares favorably with the other methods. For certain elements the
other methods are definitely more sensitive, but the uniformly good de-
tection limits of XRF give it an advantage for many applications. If
only one or a few elements are to be analyzed by AA, then, according to
Thompson et al , the procedure can be optimized to obtain much better
detection limits (11).
59
-------
Filter Media: In many applications it is desirable to use very short
sampling intervals. In such cases, two criteria must be satisfied.
Each element to be analyzed must exceed the blank level in the filter.
Also, a sufficiently large sample must be collected so that the detection
limits of the analytical method are exceeded. Figure 5 shows a comparison
of XRF spectra for glass fiber and Fluoropore (Teflon with polyethylene
grid backing) filters. The impurities in the glass fiber filter are
far too high for use with XRF or NAA, but the Fluoropore is excellent
in this regard. Table III shows typical impurity levels for Millipore
AA (mixed esters of cellulose) and Whatman 41 (cellulose) filter media (13).
The ability of a filter to pass air at high flow rates is important
when brief sampling intervals are required. A summary of the pro-
perties of a wide variety of filter media is given by Lockhart et al (14),
and some of these properties are listed in Table IV. Included in Table
IV arc the flow rates which occur when the pressure drop is 1/3 atm. In
a certain sense, these flow rates are the optimum maximum values. Some-
what higher flow rates are possible, but it becomes increasingly dif-
ficult to regulate the flow rate as the pressure drop increases above 1/3
atm. Tabic V shows the volume per unit area which can be sampled through
various filter media during a ten-minute sampling interval. The values
in Table V can be used to convert the detection limits of Table II or
2 ^
Figure 4 from units of ng/cm to units of ng/m .
The use of IPC 1478 filter media allows one to sample air-at ex-
tremely high flow rates. This filter medium has been designed for sampling
60
-------
by aircraft in the upper atmosphere. According to the measurements by
Lockhart ct al, the collection efficiency is poor at low flow rates (13).
The efficiency is expected to improve as the flow rate is increased.
However, it is questionable whether the collection efficiency for sub-
micron aerosol would be adequate. (The efficiency would be expected
to be very good for sampling in the upper atmosphere where the air
pressure is very low).
AA Millipore filter has been commonly used for XRF analysis because
aerosols arc collected close to the filter surface. This minimizes
self absorption problems for light elements. However, the ability to
analyze 10-minute air samples is limited by the rather low flow rate.
Higher flow rates are possible using Whatman 41 filters, but the ap-
plicability to elements which have atomic numbers below 20 is questionable
due to the greater thickness and greater penetration of aerosol into
the filter during sampling.
CONCLUSIONS
The foregoing discussion is a brief outline of some of the capa-
bilities and limitations of presently available instrumentation for
chemical and elemental analysis of aerosols. This information could be
helpful in the design of an airborne platform system for aerosol analysis.
If very short sampling periods are required, then there may be difficulty
in collecting enough material for analysis if the most desirable mem-
brane filter were used. If analysis of the light elements is not needed,
then Whatman 41 filter material can be used. This problem can be
61
-------
alleviated as the detection limits on XRF analyzers improve. A new pulsed
XRF spectrometer developed at the Lawrence Berkeley Laboratory does
achieve significantly improved detection limits. Also, a new XRF
analyzer for total sulfur is now under development and should be useful
with brief sampling intervals. Finally, a new evaluation of filter
media is being conducted, in order to identify a filter which is suitable
for XRF analysis and which has flow and efficiency characteristics that
arc optimum for high speed aerosol sampling(15).
62
-------
1.
2,
3
4
5
6
7
8
9
10
1
REFERENCES
K.T. Whitby, R.B. Husar, B.Y.H. Liu,J. Colloid and Interface Sci.,
39, 177 (1972).
T.G. Dzubay, and R.K. Stevens, "Ambient Air Analysis with Dichoto-
inous Sampler and X-Ray Fluorescence Spectrometer", Submitted to
Environ. Sci. and Tech. in June (1974).
T.G. Dzubay, L.E. Mines and R.K. Stevens, "Particle Bounce Errors in
Cascade Lmpactors", to be submitted to Atmos. Environ, in April (1975).
Billy Loo, to be published.
E.L. Kothney, B.R. Appel and J.J. Wesolowski, "An Intercomparative
Study of Wet Chemical and Instrumental Methods for Sulfate De-
termination in Atmospheric Aerosols", EPA Contract #68-02-1660,
to be completed in December, 1975.
P.S. Mudgett, L.W. Richards and J. Roehrig, "A New Technique to
Measure Sulfuric Acid Ln the Atmosphere", in Analytical Methods
Applied to Air Pollution Measurements , (eds.) R.K. Stevens and
W.F. lierget, Ann Arbor Science Pub., Inc. (1974)
R.K. Patterson, Anal. Chcm 45_, 605 (1973).
Field Methods Development Section project, Chemistry and Physics
Laboratory, EPA, NERC-RTP (1975).
F.S. Goulding and J.M. Jaklevic, "X-Ray Fluorescence Spectrometer for
Airborne Particulate Monitoring", EPA Report tl EPA-R2-73-182 (1973)
W.L. Zoller and G.E. Gordon, Anal. Chem. 42, 257 (1970).
R.J. Thompson, G.B. Morgan and I,.J. Purdue, Atomic Absorption News-
letter 9, 53 (1970) .
-------
12. J.A. Cooper, Battelle Pacific Northwest Laboratory Report
BNIVL-SA-4690, June ], (1973).
13. R. Dams, K.A. Rahn and J.W. Winchester, Environ. Sci. and Tech.
6, 441 (1972).
14. L.B. Lockhart, R.L. Patterson and W.L. Anderson, "Characteristics of
Air Filter Media Used in Monitoring Airborne Radioactivity", NRL
Report 60S4 (1963).
15. B.Y.ll. Liu, private communication
-------
TABLE I.
Parameter
Sulfate
Sulfuric Acid
NH3
Nitrate
Carbon
Method
NASN colorimetric
A1HL microchemical
Brossett colorimetric
Turbidimetric
Flash volatilization
Gran titration
Volatilization at 100° C.
Colorimetric
ion selective electrode
Colorimetric
Ion selective electrode
Pyrolysi s-Reduction
63
-------
TABLE II.
DETECTION LIMITS FOR ELEMENTAL ANALYSIS
NAA
a
EXPRESSED IN UNITS OF ng/cm
b c
XRF ES
aa'
Na
1
A1
1
180
Si
100
S
38
CI
200
30
Ca
1000
15
V
0.5
30
15
50
Mn
1
25
50
5
Fe
1
28
800
50
Cu
20
16
10
5
As
5
100
Se
0.02
5
Br
2
8
Cd
50
50
1
Ba
0.1
170
100
Pb
16
50
10
These values are calculated from the detection limits given in Reference
(20) assuming that 25 of air is sampled through a 57-mm diameter
membrane filter.
Present work using Ti, Mo and Sm secondary fluorescers. See also
Reference (9).
Calculated from the detection limits given in Reference(l1) assuming
that 2000 of air are sampled through a 407 cm^ area of glass fiber
filter.
-------
TABLE III.
IMPURITIES IN FILTER MEDIA
MILLIPORE AA W 41
Na 400 - 600 ng/cm^ 150 ng/cm^
A1 6-15 12
CI 1000 - 1700 100
K 100 - 200 15
Ca 200 - 700 140
Ti 5 - 15 10
Fe 40 - 80 40
Ni <20 <10
Cu 20 - 90 <4
Zn 10 - 30 <25
From Reference 13
65
-------
TABLE IV.
SUMMARY OF PHYSICAL PROPERTIES OF VARIOUS FILTER MEDIA
FILTER TYPE
COMPOSITION
AA ESTERS
OF CELLULOSE
W41 IPC 1478
CELLULOSE CELLULOSE
TYPE A GLASS
FIBER
Pore Dia., nm 0.8
Thickness, mm 0.15
2
Mass/area, mg/cm 4.8
Face velocity'3, m/scc 0.94
b 2
Flow rate/area , 1/min-cm 5.6
% efficiency @ 0.3 [ini DOP 100
0. 25
8.9
3.7
22
100
0.56
14.8
>12C
>72c
0.46
9.4
4.2
25
100
J These data are from Reference (14)
h For these flow rates, the pressure drop is 250 Torr (1/3 atm)
Private communication, John Cooper
66
-------
TABLE V-
VOLUME THROUGHPUT FOR TEN MINUTE SAMPLING INTERVALS
FILTER VOLUME/AREA (m3/cm2)
AA Millipore/Fluropore 0.056*
W 41 0.22*
IPC 1478 >0.72*
Type A, glass 0.25*
HiVol Sampler 0.038
*Pressure drop = 1/3 atm
67
-------
40
I I I
1 1 1 I I I
.GREASED SURFACE
ALUMINUM FOIL
30
CO
.E
o>
a.
z"
o
h
<
cc
H
Z
ill
o
z
o
a
CO
to
<
2
20
10
x-
±
f
i
J I I I
10
DIAMETER, jam
Figure 1: An illustration of particle
bounce errors for a cascade impactor.
-------
INTAKE
I
Figure 2: Dichotomous Sampler which
uses two substages of virtual im-
paction. (Ref 4).
69
-------
PARTICLE SIZE ( w )
Figure 3: Fractionation characteris-
tics of dichotomous sampler. (Ref 4).
-------
105
104
n
"1 103
c
^ 102
GC
S!
« 101
5E
2 100
s
10-1
10 2 r
TYPICAL VALUE
M
S»
60
99
55 M
Is4 0 ni>
RAW Dn
-I 67ta
a
70
Pt _
82
Hi
A :
u
71 75
K "t.
74 Oi If T
k" J T
79 i
X
90 92
10 3
105
10*
103
102
101
100
10-1
102
10-3
OI
E
o
K
U
ATOMIC NUMBER
Figure 4: Detection limits for XRF
analyzer using a variety of secondary
fluorescers. Also shown are typical
ranges of trace element concentrations
in urban air (Ref. 12).
-------
10*
*T)
era
C
H
CO
103
o
o
102
101
100
100
'\.K
nr||
CHANNEL
-------
METHODS OF ANALYSIS OF
HALOCARBON AIR POLLUTANTS
BY
NORMAN E. HESTER, Ph.D.
AIR QUALITY CHEMIST
AIR MONITORING BRANCH
ENVIRONMENTAL PROTECTION AGENCY
NATIONAL ENVIRONMENTAL RESEARCH CENTER
LAS VEGAS, NEVADA
89114
-------
Presently, three major techniques are used to measure the concentrations
of fluorocarbon air pollutants. The three methods are:
1. Gas chromatography employing an electron-capture detection system,
2. Gas chromatography followed by a mass spectrometer with a single ion
detector,
3. Long-path infra-red spectroscopy.
The advantages and disadvantages of each system are discussed below and a
summary is given in Table I.
Gas Chromatography Employing Electron Capture Detection System:
Gas chromatography with an electron capture detector system (GCECD) is by :ar
the most widely used method for analysis of fluorocarbon air pollutants. Most
of the published investigations of ambient air concentrations have employed
1-9
this technique.
GCECD offers several advantages. Such systems are readily available
commercially, and they are the least expensive of the instruments used for
fluorocarbon measurement ($4,000 to $15,000 per unit). The method is very
-12
sensitive for detection for fluorocarbons: concentrations as low as 1 x 10 v/v"
for fluorocarbon 11 and 10 x 10 ^v/v for fluorocarbon 12 can be measured?
These concentrations are at least an order of magnitude below the lowest
ambient "background" levels. These systems are portable, and ambient measure-
3 6 5
ments have been taken from aboard ships ' and aircraft. GCECD systems also
offer the advantage of being able to analyze for a whole series of halocarbons
in each air sample. Fluorocarbons 11 and 12, carbontet'rachloride, trichloro-
g
ethylene, and tetrachloroethylene can all be detected in a single chromatogram.
* v/v = volume of pollutant per volume of air
73
-------
GCECD systems have some disadvantages also. The electron capture
detector's sensitivity is a function of the compound being measured, and the
compounds which the instrument can detect are limited to those molecules
containing several electronegative atoms. For example, carbontetrachloride
(CCI4) is easily detected, while the compound methyl chloride (CH3CI) is several
orders of magnitude less detectable. GCECD,as does all forms of gas chromatography,
has the limitation that real-time, continuous measurements cannot be made.
Gas Chromatography Followed by a Mass Spectrometer With a Single Ion Detector:
This system of analysis is a relatively new technique for fluorocarbon and
halocarbon analysis. It is still in the developmental stage, however, -this
method seems to be a very promising, versatile technique.^ The gns chroma-
tography - mass spectroscopy (GCMS) technique is almost as sensitive as the
electron capture technique; this method allows the detection of about
— 1 ?
5 x 10" v/v for both fluorocarbon 11 and fluorocarbon 12. This technique
does not show the compound dependence that GCECD does. This means that com-
pounds such as CH3CI, CH2CI2, etc. which cannot be easily detected by electron
capture, can now be measured in ambient air in the 10" H - 10~^^v/v range.^
A mass spectrograph of a compound is a "fingerprint" of that compound, and
this system can be used to provide a positive identification of each component
analyzed.
While the advantages of GCMS include very important features, this system
also has several disadvantages. One disadvantage is the cost which may run
from $60,000 to $150,000 per unit. The present systems used for fluorocarbon
analysis are large and cannot be used in a field situation. Being another gas
chromatograph system, this instrument cannot make real-time continuous measure
inents. The biggest disadvantage of the system is that the single ion detection
74
-------
technique, as it is now used, will allow the researcher to analyze only one
compound per sample. This means that the instrument must be "retuned" for
each component of an air sample, and that a separate sample of air is required
for every pollutant analyzed.
Long-path infra-red spectroscopy (LPIRS): The LPIRS is a versatile
technique which can be used to make ambient measurements of a broad spectrum
11 12
of air pollutants, including fluorocarbons, and other halocarbons. '
The main advantage of a long-path infra-red system is that it allows
real-time continuous measurements of halocarbons, and depending on the
absorption characteristics of the various compounds, several compounds can
be monitored at the same time. The ability to measure several compounds at the
same time is extremely important when the chemistry of fluorocarbons and
other halocarbons is examined, in that, the appearance of reaction products
can be followed. Infra-red analysis of a compound, like mass spectrometry
provides a "fingerprint" of a compound for positive identification.
Disadvantages associated with LPIRS include the rather high cost
($20,000 - $100,000) and the fact that the instruments cannot be used in
the field. The sensitivity of the instrument, which varies with the length
of the path, is at best two orders of magnitude lower than that of GCECD
and GCMS.
Other Techniques: The three methods mentioned comprise the techniques
that are presently being used, and which can measure ambient air concentrations
without preconcentration (The LPIRS can only be used for samples taken from
high concentration areas).
75
-------
If one preconcentrates a sample, it is possible to measure fluorocarbon
concentrations by more common, but less sensitive techniques (e.g. gas chrom-
atography using other detectors, shorter path infra-red systems, mass spectro-
meters, etc.). For example, cryogenic concentrating techniques have been used to
12
improve the capabilities of LPIRS , and to enable one group of workers to use
13
a more common mass spectrometer for fluorocarbon analysis. Preconcentration
techniques followed by analysis with alternate systems undoubtedly have advantages
and disadvantages, however, the amount of work in this area that has been reported
thus far is insufficient for making an adequate evaluation.
A method has been reported for measuring fluorocarbons in the stratosphere
using an infra-red spectrophotometer to measure the solar ir absorption. By
examining spectra taken on balloon flights in 1968, Murcray was able to assign
14
certain observed peaks to fluorocarbon absorbances. The volume mixing ratios
on the order of lO--^ could be detected, however the estimated error in these
measurements was + 50%. This method of measurement of fluorocarbons holds a
great deal of promise. Experiments dedicated to halocarbon measurements should
provide a great deal of information on stratospheric concentrations.
Conclusions: Complementary methods of analysis for fluorocarbons and other
halocarbons presently exist which will allow the monitoring of these compounds in
ambient air samples. The present methods are more than sensitive enough to measure
the levels of both fluorocarbons 11 and 12 in tropospheric and lower stratospheric
samples, and should be sensitive enough to measure fluorocarbon concentrations in
samples taken from the higher stratosphere. This assumes that fluorocarbon concen-
— 1 7
trations in the stratosphere are greater than 10 v/v.
76
-------
The GCECD system, because of its low cost and high sensitivity, is the most
widely used method of analysis for fluorocarbons, and will probably continue to
remain the method of choice. However, there are gaps in GCECD system capabilities
which need to be filled by the other instrumentation discussed.
The GCMS system offers more sensitivity to the whole spectrum of halocarbons
than does GCECD, and should find wide use in the analysis of halogenated hydro-
carbons which are not easily measured by electron capture.
The LPIRS system offers the capability of real-time, continuous measurements,
and will be most useful in the examination of the chemistry of fluorocarbons
and other halocarbons.
The solar infra-red system has a great deal of promise for stratospheric
measurements.
Other systems require further development and evaluation.
77
-------
TABLE I
SUMMARY OF ADVANTAGES AND DISADVANTAGES
OF METHODS OF MEASURING FLUOROCARBONS
METHOD
FLUOROCARBON
SENSITIVITY
ADVANTAGES
DISADVANTAGES
Electron Capture Fll
Gas Chromatography FI2
1 x 10"12v/v
10 x 10"12v/v
low cost, portable, can analyze
several components at the same
time
sensitivity is a function
of the compound, cannot make
continuous measurements
Gas Chromatography
plus Mass Spectro-
meter with Single
Ion Detector
Fll
F12
5 x lCT^v/v
5 x 10" v/v
sensitivity not as compound
dependent, and can measure
coumpounds not measurable by
electron capture,provides
positive identification of
compounds
high cost, not portable, cannot
make continuous measurements,
can only analyze one pollutant
per sample of air
oo
Long Path Infra- Fll 1 x
Red Spectrometer F12 1 x
10~9v/v
10~^v/v
allows continuous monitoring,
allows simultaneous monitoring
of several components, provides
posicive Identification
high cost, not portable,
lower sensitivity
-------
REFERENCES
1. Lovelock, J.E., Nature 2|30» 379 (1971).
2. Lovelock, J.E., Atmospheric Environment 6, 971 (1972).
¦ ¦ — * - 1 1 1 CW
3. Lovelock, J.E., Nature 241, 194 (1973).
.
4. Murray, A.J., and Riley J.P., Nature 242, 37 (1973).
J - I
5. Simmonds, P.E., et al., Atmospheric Enviornment 8, 204 (1974).
6. Wllkniss, P.E., et al., Nature 245, 45 (1973).
7. Su, C. and Goldberg, E., Nature 245, 27 (1973).
/V* A
8. Hester,N.E., Stephens, E.R., and Taylor, O.C., «J. Air. Pollut. Control
Assoc., 24, 591 (1974).
9. Hester, N.E.,Stephens, E.R., and Taylor, O.C., Atmospheric Environment,
(in press).
10. Rasmussen, R. and Grimsrud, E., Washington State University, private
communication.
11. McAfee, J., University of California, Riverside, private communication.
12. Hanst, P., Environmental Protection Agency, National Environmental Research
Center, Research Triangle Park, North Carolina, private communication.
13. Steadman, D. and Cicerone, R., University of Michigan, Ann Arbor,
(private communication).
14. Murcray, D.G., University of Denver, private communication.
79
-------
INSTRUMENTATION TO MEASURE GASEOUS AND PARTICULATE
POLLUTANTS FROM AIRBORNE PLATFORMS - A REVIEW
by
Robert K. Stevens
Chief, Field Methods Development Section
Chemistry and Physics laboratory
National Environmental Research Center
Environmental Protection Agency
Research Triangle Park, N.C. 27711
INTRODUCTION
Measurements of atmospheric contaminates from airborne platforms
have generally been performed with conventional hardware adapted for
this task. Problems that result from adapting state-of-the-art instru-
mentation to airborne studies can be identified in the following areas:
(1) space and power requirements; (2) temperature and pressure sensi-
tivity and (3) response time of the sensor. Commercial instrumentation
available to monitor SO2, N02> CO, 0^ and hydrocarbons normally weighs
from 30-70 pounds, requires on the average of 500 watts power and has
response time constant from 20-60 seconds. Recently, some manufacturers
have made an effort to reduce the size and power jrequirements of certain
gaseous monitoring equipment.
The text that follows summarizes the adequacy of current ambient
air measurement reference methods and areas where research is ongoing
to develop more reliable air pollution monitoring instrumentation.
DISCUSSION
Below is a table listing the air quality standards for carbon
monoxide, nitrogen dioxide, hydrocarbons, sulfur dioxide, particulate
matter and photochemical oxidants. In this table is a list of the
analytical techniques cited by EPA in the Federal Register (36(84)
8186 (1971)) as the reference methods for measuring these pollutants.
We have also classified the reference methods as adequate or in-
adequate with regard to ground level monitoring. The adequacy of each
reference method by pollutant will be discussed in detail later in the
text.
80
-------
TABLE I
STATUS OF MONITORING AT GROUND LEVEL CARBON MONOXIDE, NITROGEN DIOXIDE,
HYDROCARBONS, SULFUR DIOXIDE, PARTICULATE MATTER AND PHOTOCHEMICAL OXIDANTS
Primary
Standard
Secondary-
Standard
Substance
Carbon monoxide
Nitrogen dioxide
Hydrocarbons—
f
Sulfur dioxide —
Particulate matter
Photochemical oxidants—
mg/mJ
10—
0.10—
0.16s-
0.0S0-
0.075-
0.160^
EE
mv
mg/m:
9—
0 .05-
0. 24-
0.05-
O.OS^
40-
0. 10-
0.16-
0.565^
0. 26^-
0.160^
££mv
_ .a
Analyti cal
Reference Method
Nondispersive IR
b c
0.05— Chemiluminescent —
0.24— Flame ionization
g
0.14— Colorimetric
(pararosaniline)
High-volume sampler
0.08— Chemiluminescent
— Not to be exceeded more than once a year.
— Annual arithmetic mean.
Q
— Proposed reference method.
— Sum of all hydrocarbons other than methane.
6
— Maximum 5-hour concentration (6 to 9 A.M.) not to be exceeded more than once per year.
— Measured as sulfur dioxide.
a
—Maximum 24-hour concentration, not to be exceeded more than once per year.
— Annual geometric mean.
—Corrected for interferences due to nitrogen oxides and sulfur dioxide.
—Maximum 1-hour concentration, not to be exceeded more than once per year.
k
— Sensitivity needs improvement.
— Reliability of method is marginal at air quality standard.
— Method too complicated.
— Collection medium needs modification. Method to determine fine particles needs development.
Status
k
Adequate-
Adequate
I nadequate
Adequate
Adequate
m
n
reaction with ozone Adequate
—Calibration method needs improvement.
-------
The analytical reference methods cited in Table I are not always
the methods used for routine monitoring or even the methods used to
gather data for preparation of implementation plans.
Because of the variety of methods available to monitor these six
air pollutants, EPA has promulgated a document published in the Federal
Register which describes tests which compare a proposed method with the
reference method. If the proposed method passes the t^st to the
satisfaction of the administrator, EPA then will consider the method
equivalent to the reference procedure and the new method can be used
by state and local agencies in the preparation of implementation plans.
However, not all of the reference methods are totally adequate
for routine monitoring and the following is a status report, by pollutant,
on the adequacy of the reference methods:
Carbon monoxide: The non-dispersive IR methods is adequate for ambient
air monitoring, but more precision is desirable at the level of the
air quality standard. New instrumentation for carbon monoxide has
been developed and is currently under evaluation. The goal for instru-
mental performance characteristics includes a factor of ten increase
in sensitivity and a simplified procedure to correct for water vapor
interference.
Nitrogen dioxide: It has been proposed that the colorimetric method
originally cited as the reference method for nitrogen dioxide be re--
placed by the chemiluminescence procedure.
Hydrocarbons: The reference method for non-methane hydrocarbons is based
on chromatographic separation and measurement of methane simultaneously
with the measurement of the total hydrocarbons. The measurements are
made with a flame ionization detector. This method ?t present is con-
sidered to be unreliable when measuring concentrations at levels below
0 5 ppm. Work is ongoing tc delineate those problems and/or to improve
the reliability of the analytical method.
Sulfur dioxide: The method is based on the collection of S0o for a
24-hour period, followed by colorimetric analysis '..'hen the sjr.pie is
returned to the laboratory, ^hi? method is considered adeq-'ite bun
complicated, due to the large number of ultrapure reagents needed in
the measurement. For the gaseous pollutants, this is the only non-
instrunenta 1 reference method. Work is ongoing to develop a low cost,
reliable, simple method to serve as an equivalent procedure. In most
areas of the United States, flame photometric and coulometric instruments
are used to measure S02-
Particulate matter: Tbe HiVolume sampler method of collecting parti-
culates is marginally adequate for routine monitoring. Recent studies
82
-------
have shown the glass fiber collection surface may convert gaseous SO^
and NCL to sulfates and nitrates respectively, thus adding weight to
the filter which could provide an over-estimate of the atmospheric
particulate loading. The method provides no means of separating the
rcspirable from the non-respirable particles. Work is irt progress to
replace the HiVolume collection system in order to develop a sampler
which will separate the fine from the coarse particles. In addition
efforts ore underway to develop a collection surface where artifact
formation will be minimized.
Photochemical oxidants : The chemiluminescent ozone-ethylene reaction
is the reference method and is considered adequate. Recently, however,
EPA has been challenged with regard to the validity of the recommended
potassium iodide colorimetric method for calibrating ozone generators
(used to calibrate ozone monitors). As a result, EPA is engaged in
an intensive in-house, grant and contractural program to investigate
the ozone calibration method to determine the adequacy of the procedure.
PA will continue to maintain a quality assurance program to check
on the efficiencies of the current reference and equivalent methods.
In addition, EPA will foster development of quality assurance procedures
to assure that the data being reported on the condition of the atmos-
phere will be accurate and reliable.
AIRBORNE INSTRUMENTATION
Tabic II is a summary of the status of instrumentation for measuring
the gaseous pollutants from aircraft. The two major problems with the
methods currently employed in airborne monitoring are (1) their slow res-
ponse times and (2) their sensitivity to changes in altitude. Work is
ongoing to develop techniques to correct these operational deficiencies
but at present, a totally reliable instrument for airborne gaseous pol-
lutant monitoring studies does not exist.
The techniques employed to circumvent these problems are to calibrate
the monitors in flight at the altitude where measurements are to be made
and to correct the response time by correlating time and location and
knowing the response time of the instrument.
83
-------
TABLE II
STATUS OF MONITORING HYDROCARBONS, OZONE,
OXIDES OF NITROGEN.CARBON MONOXIDE AND SULFUR DIOXIDE FROM AIRCRAFT
CONCENTRATION
REQUIREMHNTS
PPM
ANALYTICAL
METHOD
OPERATIONAL
CHARACTERISTICS
HYDROCARBONS
0.001 - 10.0
ozone
0.01 ¦
1.0
OXIDES OI; NITROGEN
0.0J - 1.0
CARBON tMONOXIDE
0.05 -'lOO
SULFUR DIOXIDE
0.01 - 10
Flame Ionization
FID-G.C.
Catalytic Combustion
FID
Chemiluminescent
Chemiluminescent
NDIR
Conductivity
Gas Chromatography
Conductivity
Flame photometric
Electrochemical
a. Measures total H.C.
b. Requires hydrogen
a. Slow response time
b. Marginal sensitivity
c. Requires hydrogen
d. Altitude sensitive
a. Requires hydrogen
b. Altitude sensitive
c. Non-specific
d. Marginal sensitivity
a. Requires ethylene
a. Marginal sensitivity
Vibration sensitive
a.
b.
c.
a.
b.
c.
a.
b.
c.
a.
b.
c.
Non-specific
Slow response time
Inadequate sensitivity
Slow response time
Altitude sensitive
Requires hydrogen
Non-specific
Lacks sensitivity
Slow response time
Sensitivity marginal
Requires hydrogen
Altitude sensitive
a. Slow response time
b. Lacks sensitivity
84
-------
DEVELOPMENT OF STANDARD REFERENCE MATERIALS FOR
AIR QUALITY MEASUREMENT
Ernest E. Hughes
Research Chemist
Air Pollution Analysis. Section
Analytical Chemistry Division
National Bureau of Standards
Washington, D. C. 20234
ABSTRACT
The National Bureau of Standards is engaged
in a continuing program involving gaseous
Standard Reference Materials for air pollu-
tion measurements. Preparation of such
materials requires definition of the stabi-
lity, homogeneity and accuracy of the sam-
ples. This information is obtained by long
term studies of the gas systems, by develop-
ment of absolute methods of analysis and by
analysis of large numbers of samples pre-
pared in bulk. The results of studies,
extending over several years, of low con-
centration of carbon monoxide in nitrogen
and nitric oxide in nitrogen are re-
ported. Over one thousand samples of these
materials have been analyzed and the stabi-
lity with tine and the within-batch homo-
geneity have been characterized. Accuracy
is achieved by use of gravimetric standards
and with dynamic dilution systems. Accuracy
attainable by either method is described.
The use of permeation tubes of sulfur diox-
ide and nitrogen dioxide is necessary in
Some situations because of the reactivity
of the gases. Data covering the stability
and accuracy of these devices has been
Collected over a period of several years.
INTRODUCTION
The legal implications of measurements of
air pollutants whether at ambient or source
levels is such that the accuracy of the
measurement must be clearly defined. A
substance can be measured accurately in one
of two ways. First, an absolute technique
Of known accuracy may be used in which case
the measured value has accuracy equivalent
to that of the method. Alternatively, an
analytical standard may be employed together
with a precise method of comparison. In
this case, if the imprecision of the method
of comparison is considerably less than the
Uncertainty of the standard then the accu-
racy of the measurement is described by -the
accuracy of the standard.
This paper is concerned with accuracy in
the measurement of gaseous pollutants and
Reprinted by Permission. Presented at
Instrument"Society of America, Instru-
mentation — Automation Conference,
New York, N. Y.
October 28-31. 1974.
it should be recognized that few absolute
methods are available to the gas analyst.
Those that are available, volumetric
adsorption or gravimetric methods for ex-
ample, are generally useful only at rela-
tively high concentrations. A very few
electrochemical methods are applicable at
low concentrations. Consequently, accu-
rate measurements of air pollutants will
in general require the existence of a
standard of known accuracy. The Analytical
Chemistry Division of the National Bureau
of Standards has been involved in recent
years in the preparation of Standard
Reference Materials for use in gas analysis
and particularly for use in analyses asso-
ciated with air pollution measurements.
There is now available a limited number
of gas standards consisting of various
concentrations of carbon dioxide in nitro-
gen, propane in air, methane in air,
oxygen in nitrogen and a series of sulfur
dioxide permeation tubes. There are stand-
ards of nitric oxide in nitrogen and nitro-
gen dioxide permeation tubes in preparation.
Standards for vinyl chloride in air, high
concentration of sulfur dioxide in nitrogen
and a methane-nonmethane hydrocarbon in air
mixture are being investigated as potential
Standard Reference Materials.
The preparation of gas mixtures is not par-
ticularly difficult and the routine prepara-
tion of "standards" is commonplace. How-
ever, experience has shown that the accuracy
of such standards is often woefully inade-
quate for the measurements for which the
standards are intended. The major reasons
for the inadequacy are the uncertainty in
assigning an accurate value to the concen-
tration and the uncertainty in the stability
of the gas mixture during storage and use.
The solution to either of these problems is
often very time consuming and expensive and
the two uncertainties are the major reason
for the small number of Standard Reference
Materials available and for their rela-
tively high cost.
It Is obvious that accuracy and stability
85
-------
are primary requircnents for a Standard
Reference Material but in addition, avail-
ability in sufficient number and size con-
stitute further requirements. Standards
should be produced insufficient quantity
and at a range of concentrations to satisfy
the needs of diverse laboratories both
those engaged in measurement, enforcement
and research. This imposes certain prob-
lems in that the production of gas mixtures
in quantity and with homogeneous composi-
tions at trace levels has not been fully
evaluated. Commercial preparation of gas
mixtures to rigid specifications appears to
be a feasible approach to preparation in
quantity and homogeneity is currently as-
sured by individual analysis of each sample.
Thus, rapid and precise methods of inter-
comparison between the primary standards
and the potential Standard Reference Mate-
rials must often be developed and used
until the homogeneity of a batch can be
predicted from experience and from the re-
sults of a limited number of measurements.
In some cases it is not possible to prepare
mixtures of satisfactory stability in cyl-
inders and alternate standards may be feasi-
ble. Permeation tubes offer such an alter-
nate and while permeation tubes cannot be
considered primary gas standards in the
pure sense, they can be used to generate gas
mixtures of predictable concentration. The
effort necessary to evaluate the accuracy
of such devices as Standard Reference Mate-
ials is different from that involved with
gas mixtures but it is equally difficult
and time consuming.
The total development of any of these Stand-
ard Reference Materials is best illustrated
by example. A description follows of the
general procedure involved in preparation
of the carbon monoxide and nitric oxide
gaseous Standard Reference Materials and
the sulfur dioxide and nitrogen dioxide
?ermeation tubes. These were chosen as
ypical examples because of the range of
problems associated with each.
DEVELOPMENT OF A STANDARD REFERENCE MATERIAL
The development of a gaseous Standard Ref-
erence Material involves the preparation of
a series of primary standards and a deduc-
tion of the accuracy of the standards while
simultaneously investigating tlie stability
of the primary standards and of the parti-
cular gas mixture in the container in which
it will eventually be distributed as a
Standard Reference Material.
An accurate gas mixture can be prepared in
a similar manner to an accurate solution.
A measured amount of ;in analysed reagent is
mixed with a measured amount of analyzed
diluent. The concentration of the reagent
species in the resulting mixture can then
be calculated with an error dependent only
on the accuracy with which the purity of
the reagent and diluent are known and the
accuracy of measurement of the quantity of
reagent and diluent. Gas mixtures are pre-
pared most often by measurement of the pres-
sure of the reagent added to a cylinder and
of the combined pressure of reagent and
diluent but the use of gravimetric proce-
dures, in which the weight of each compo-
nent added to a cylinder is measured, has
become quite common in recent years. The
latter method could be considered more near-
ly absolute than the former but in either
case it is possible to prepare mixtures with
an error of less than one percent.
The analysis of the reagent gases is, in
some cases, a greater source of error than
the mixing procedure. It is now possible
to obtain the more common diluent gases, air
and nitrogen, which contain no significant
amounts, of the reagent gas as an impurity
but in all cases it is essential to verify
the absence or to determine the concentra-
tion of the reagent if it is present. The
assay of the pure reagent is an essential
and often quite difficult step in the pre-
paration of accurate gas standards. How-
ever, the many techniques and the variety
of instrumentation involved in this phase
of the analysis is too extensive to include
in this paper. Suffice it to say that a
great deal of effort is expended in defining
the composition of the reagent gases and
that in general the resulting uncertainties
do not contribute significantly to the error
assigned to the standard.
Primary standards are prepared by a gravi-
metric method generally to a concentration
of one percent of the starting material.
That is, starting with pure reagent and pure
diluent a one percent by weight mixture is
prepared. The combination of cylinder
weight, balance capacity and sensitivity,
and the sample weight are such that an uncer-
tainty of ± 0.1 percent could theoretically
be attained. In reality, the uncertainty
lies between 0.2 and 1 percent for most of
the mixtures investigated thus far. The
one percent mixture may now be used as start-
ing material for a second mixture with a
resulting concentration of 0.01 weight per-
cent [100 ppm). A third mixture may be
prepared from the second which will have a
concentration of 1 ppm. A series of repli-
cate samples are prepared in this manner
having concentration very near the concen-
tration desired for the series of SRM's. It
is, of course, essential that the stability
of these standards be assured. This cannot
be done directly unless an accurate method
of analysis is available and unless it is
applicable at all concentrations through
which the preparative steps are performed.
Lacking an absolute analysis it is necessary
to deduce the stability by observing the
relationship between samples using 3 relative
method of analysis of adequate precision.
Gas mixtures stored at pressure in cylinders
86
-------
tend to change composition in three ways.
First, there may occur a rapid decrease in
concentration due to adsorption on the walls
of the cylinders or in the case of certain
reactive gases by reaction with the con-
tainer. Second, slow long term decreases
in concentration may result from reaction
within the cylinder. Third, an increase in
concentration may occur as cylinder pres-
sure drops with use and previously adsorbed
material desorbs in keeping with changing
equilibria inside the cylinder. The second
effect, slow long term decrease, can be
recognized by comparison of freshly pre-
pared standards with standards that have
been stored within the container for a long
period of time. Short term changes can be
recognized by transferring an analyzed
sample to an empty container and comparing
the composition of the two. The possibi-
lity of increase of concentration due to
desorption is observed by deliberately blow-
ing down a cylinder and observing the change
or lack of change in concentration. If no
significant changes are observed due to the
above causes then one can feel some assur-
ance for the stability of the mixture.
In addition to gravimetric data it is desir-
able to measure the pressure of the various
components as they are added to a cylinder.
Analytical blunders are easily recognized
when the pressure data and gravimetric data
do not agree to within one percent of each
other.
The use of dynamic dilution methods to gen-
erate standards at the time of use can pro-
Vide an independent confirmation of the
accuracy of primary standards prepared by
gravimetric or pressure techniques. The
dynamic dilution method involves the pre-
paration of a mixture by blending two
streams of gas of individually known compo-
sition. It is necessary to measure the
volume flow of each gas with an accuracy of
better than 99 percent if the resulting
mixture is to have an uncertainty of compo-
sition similar to the gravimetric or pres-
sure standards.
The primary standards may or may not be pre-
pared in the same type of container in which
the Standard Reference Material will be
contained and it is necessary to independ-
ently assess the stability of the mixture
in both containers. This stability can
only be verified with confidence by observ-
ing the change, or lack of change, with
time over the period of time during which
stability is required. Some prediction of
the long term stability of a sample may be
ba sed on observed behavior of a sample
immediately after preparation but.only long
term experience yields the degree of con-
fidence required of a Standard Reference
Material.
PREPARATION OF CARBON MONOXIDE STANDARD
REFERENCE llATERIAL
The major difficulty associated with carbon
monoxide standards is the tendency of the
carbon monoxide concentration to decline
with time. Carbon monoxide mixtures pre-
pared and stored in stainless steel cylin-
ders remain constant for very long periods
of time but similar mixtures prepared in
mild steel cylinders, either DOT specifica-
tion 3A or 3AA, show unpredictable and
often disastrous reduction in carbon monox-
ide concentration. Figure 1 represents
data obtained on analysis of four samples
of carbon monoxide in nitrogen. The samples
were prepared from the same materials and
at the same time and under essentially iden-
tical conditions at concentrations of S, 10,
25 and 50 ppm. The samples were analyzed
periodically over a period of one year. The
behaviour shown is typical of what has been
observed with carbon monoxide mixtures
stored in cylinders. The reason for the de-
cline has not been definitely established
but is probably due to either oxidation to
carbon dioxide or carbonyl formation. In any
case, the individual container is critical
relative to the stability. Stainless steel
containers are not an economically practical
solution to the problem but a simple treat-
ment of mild steel' cylinders has resulted in
adequate long term stability. The treatment
consists of applying a thin coat of ceresin
wax to the interior surface of the cylinder.
The wax is applied hot to a heated cylinder,
which is subsequently allowed to drain thor-
oughly. Table 1 shows the results of analy-
sis of several samples of carbon monoxide
in nitrogen stored in these wax-lined cylin-
ders. It is obvious that over a period of
one year the composition has not changed.
The primary standards were prepared gravi-
metrically in stainless steel cylinders. It
was determined by comparison of numerous
samples prepared over a period of years that
no decrease in concentration of carbon mon-
oxide occurs in stainless steel. The accu-
racy with which a typical set of gravi-
metric standards of carbon monoxide can be
prepared is indicated in Table 2. This
table shows the results of intercompar ison
of two samples each at six concentrations
between 10 and 1300 ppm. The comparison
was made by a method in which the carbon
monoxide was catalytically reduced with
hydrogen to methane with subsequent meas-
urement of the methane using a flame ion-
ization detector. The precision of inter-
coinparison was between 0.1 and 0.2 percent
relative. Ttie agreement between pairs
of samples> is indicated by the agreement
in the values "Division per unit of con-
centration" which is simply the total signal
due to carbon monoxide divided by the con-
centration calculated from the gravimetric
data. It is obvious that the accuracy
of the standards may approach the pre-
cision with which they can be intercompared
87
-------
but in general does not equal it. It should
be noted that a direct comparison between
different pairs of samples is not possible
in this case because the separate pairs
were analyzed at different times under
slightly different instrumental conditions.
Analysis of large numbers of samples would
require quantities of primary standards
that would be inconvenient to produce.
Consequently, a series of secondary stand-
ards were prepared in large stainless steel
cylinders. The concentration of carbon
monoxide in these samples was determined by
numerous comparisons with primary standards.
At least three primary standards were pre-
pared at concentrations bracketing the con-
centration of the secondary standards. New
primary standards were prepared periodi-
cally to confirm the stability of the sec-
ondary standards.
Two hundred and eighty cylinders consisting
of five lots of five different concentra-
tions of carbon monoxide in nitrogen were
obtained from a commercial supplier. The
cylinders were wax-coated as previously
described prior to filling. All samples in
each lot were analyzed by comparison with
the secondary standards in a sequence of
the standard followed by five samples fol-
lowed by a repetition of the standard. Two
separate measurements were made of each
sample and standard. Some instruments drift
may occur between the analysis of the stand-
ards and if this occurs it is necessary to
interpolate the value of the instrument
calibration between the standards. This
resulted in some additional imprecision in
the intercomparison. This is illustrated
by Figure 2 where the measured value for 55
samples at the 950 ppm level is shown in
the sequence of analysis. In addition, the
manner in which the instrument calibration
changed, as evidenced by the change in the
absolute value of the signal generated by
the standard, is shown. It is obvious that
the calibration was changing rapidly early
in the analysis and the exact value was
less certain than during the period of
stable operation. A second analysis of at
least ten samples from each lot was per-
formed one months after the first analysis.
All samples that appeared to deviate from
the average as well as a number at the aver-
age were included in the second analysis
except that all samples at the 10 ppm level
were reanalyzed. The total results are
summarised in Figure 3 which is a frequency
distribution plot of the observed concentra-
tions.
All cylinders in each lot were filled from
a single batch of material and it is rea-
sonable to assume that the variation of the
individual values tor a lot represent impre-
cision in the measurement rather thair actual
differences in samples. The average for a
lot, therefore, represents the actual con-
centration of any sample in the lot and the
uncertainty of the average represents the
imprecision of the measurement.
The uncertainty assigned to the final value
of the Standard Reference Material is a
combination of four errors which are as
follows:
1. The inaccuracy of the primary standard
which is estimated to be * 0.4 percent
of the calculated value.
2. The imprecision of intercomparison of
the primary and secondary standard
which is ± 0.2 percent relative.
3. The imprecision of intercomparison of
the secondary standard and the samples
which is ± 0.2 percent relative.
4. The heterogeneity of the analytical
values which is equal to the one half
the range of the values.
The estimated upper limit for the total
error of the average is given simply by the
square root of the sum of the squares of
the individual errors. Table 3 is a sum-
mary of the estimated upper limit for the
total error for the five mixtures. The
range of values for each lot is also shown.
The somewhat higher error for the second
lot, 482 ppm, is due to the fact that the
overall precision of intercomparison for
this lot was 0.6 percent relative. This
resulted from an attempt to use a second
instrument of lesser precision for some of
the analyses. Twenty-two samples analyzed
by the more precise instrument with which
all other lots were analyzed gave an average
of 482 ppm with a total estimated error of
3 ppm. The average for the lot, 482, is con-
sidered accurate to within the ± 1 percent
limit placed on the final certified value.
NITRIC OXIDE IN NITROGEN
Nitric oxide is not a particularly reactive
gas in the chemical sense but it is diffi-
cult to prepare and store as a gas mixture
in cylinders at pressure. Two major sources
of error are recognized. These involve an
apparent reaction with resulting decrease
in concentration. This can be minimized by
vigorous pretreatment of the container with
nitric oxide mixtures of higher concentra-
tion than will ultimately be stored in the
cylinder. This treatment, however, may
result in a second source of error, which is
the desorption of nitric oxide from the
walls of the container after a period of use
in which the cylinder pressure is reduced
and which results in an increase in concen-
tration.
Figure 4 is an illustration of this latter
effect and was encountered in a study of
nitric oxide stored in small steel cylinders.
The cylinders had all been previously used
88
-------
JO
43
33
JO
•»
• ¦¦
:
— ^
¦
-a*
•
-m
•
¦
i
( .
WO 200 300
THE' N DAYS
400
FIGURE L STABUTY OF CARSON MONOXIDE MIXTURES IN CYLINDERS U30T-3AA1
fif
t 5 'S3
3 I ,sl
§3 S3?
3 © ,4S
§* •O.Jt
.
h
•0.#t
• ••
CO —i-CH.
Hi *
SEOUENCE OF ANALYSES
FIGURE I ANALYSIS OF 55 SAMPLES OF CAR0ON MONOXIDE LLUSTRATNG
THE EFFECT OF INSTRUMENT RESPONSE ON PRECISION
MO ft*
rzdZL
a
f*WR£ 3. DISTREUTKIN OF CARBON MONOXIM IN NITROGEN CONCENTRATIONS
89
-------
Table 1. Comparison of Carbon Monoxide Concentration
at One fear Interval-
Average Concentration
Number of Samples in ppm
mi
6/74
958.»
9S4.
482.
481.
94.4
94.4
47.1
47.0
9.74
9.74
; percent or
less at
all concentrations.
Table 2. Intercomparison of Carbon Monoxide Standards
of Concentration
Sample
Number
Calculated Concentration
in ppm
Signal in
Divisions
1
9.33
1128
2
10.30
1241
3
47.1
5260
4
51.5
5750
5
94.5
10915
6
103.6
11945
7
478.
54958
8
518.
59431
9
1007.
116913
10
1008.
116524
11
1275.
147900
12
1368.
159646
121.0
120.5
111.7
112.1
115.5
115.3
115.0
114.8
116.1
115.6
116.0
116.7
Table 3. Estimated Upper Limit for Total Error of
Five Lots of Carbon Monoxide in Nitrogen
Measured Average
in ppm
Range in ppm
Error in ppm
958.
5.
*5.
482.
9.
±5.5
94.4
1.5
±0.9
47.1
0.5
±0.3
9.74
0.09
±0.7
90
-------
CONCENTRATION OF NITRIC OXIDE IN PPM
IS E
M
ESLm
&
jlbi rerflnmfff7.fr
.tnn.m mi
EiuBimBimmm I;
FIGURE 6. DISTRIBUTION OF NITRIC OWE IN NITROGEN CONCENTRATIONS
994PPM EtPPM
477PPM
94.0PPM
• O « • A
s5 i «< >* v
nm n
46.0PPH
o
t
o>
3
r
1
FIGURE 7. CONCENTRATION OF NITRIC OXIDE MIXTURES IN ALUMINUM CYUNDERS
15 SAMPLES AT EACH CONCENTRATION)
-------
to store mixtures of nitric oxide or had
been specially conditioned by "soaking" with
a mixture of a high concentration of nitric
oxide prior to use. The soaking appears
to reduce or eliminate the decrease in con-
centration that had been observed to occur
when new or untreated cylinders were used
but it is obvious that increases in con-
centration can occur with time. Unfortu-
nately, no exact record was'kept of the
cylinder pressure during this series of
experiments but the frequency of analysis
and the pressure observed at approximately
seven months indicate that a steady decline
in pressure occurred from about 500 psi to
essentially zero pressure. The single
points on Figure 4 are the results obtained
on analysis of a single sample at each con-
centration while the broken line is the
estimated average for a total of four sam-
ples analyzed at each concentration.
Neither wax lining of cylinders nor the use
of stainless steel cylinders appeared to1
offer any advantage over steel cylinders
of DOT 3AA specification. Aluminum cyl-
inders which have recently become commer-
cially available and which meet Department
of Transportation requirements for inter-
state shipment may constitute a more sat-
isfactory container for mixtures of this
sort. However, considerable work has yet
to be done to prove the feasibility of
aluminum cylinders. This work is currently
in progress simultaneous with the prepara-
tion of nitric oxide mixtures in nitrogen
as Standard Reference Materials.
Primary standards for nitric oxide mixtures
have been prepared by gravimetric, pressure
and dynamic dilution techniques. Because
of the potential instability of these mix-
tures it is necessary to prepare primary
standards more frequently than with most
other gaseous systems investigated.
Gravimetric mixtures c
an error of about ± 0.
This is shown in Table
obtained on comparison
mixtures at five diffe
are shown. The sample
an- instrument utilizin
cent reaction between
ozone.
an be prepared with
2 percent relative.
4 where the results
of IS gravimetric
rent concentrations
s were compared with
g the chemilumines-
nitric oxide and
In this case the difference between samples
approximates the imprecision in the measure-
ment of the sample and the error in the
gravimetric samples is considered to be
this imprecision. Secondary standards were
prepared for analysis of the Standard
Reference Materials for the sane reason
mentioned for the carbon monoxide mixtures.
Intercomparison of these mixtures with the
primary standard has been done with maximum
orror of i 0.6 percent relative.
The pressures of the components added to
the gravimetric mixtures were measured dur-
ing preparation of the gravimetric stand-
ards. In addition, the gravimetric stand-
ards were compared to standards generated
with a dynamic dilution system. The sys-
tem is shown in Figure 5. The major source
of error in the composition of mixtures
generated by gas blending systems of this
type lies in the uncertainty in calibration
of the flow measuring devices. Mass flow-
meters of the thermal conductivity type
were used in the system described in Figure
5 and each was calibrated. The high flow-
meter (0-10 1/min) was calibrated at sev-
eral points by passing a measured weight
of nitrogen through the meter. A weighed
cylinder of nitrogen was attached to the
system and nitrogen was allowed to flow
at a constant rate for a measured time.
The weight loss of the cylinder gives the
value of nitrogen directly and after appro-
priate minor corrections for response time
of the flowmeter the relationship between
indicated flow and actual flow was deter-
mined. The effect of very small leaks in
the calibration of the high flow system
are negligible. This is not so in the
case of the low flow system and a somewhat
less direct technique was employed. A one
liter per revolution wet test meter was
first calibrated by passing a known weight
of nitrogen through the meter. The meter
was then attached to the dilution system at
the outlet of the mass flowmeter and a
constant.flow was established. The volume
flow per upit time was then determined from
the time required for passage of one liter
of nitrogen.
The estimated error in the calculated con-
centration of a mixture prepared by blende
ing two streams of gas of known composition
is about 0.8 percent relative. This is
based on the uncertainty in the composi-
tion of the stream containing the nitric
oxide (0.2 percent), the error in calibra-
tion of flowmeter number 1 (0.5 percent)
and the error in calibration of flowmeter
number 2 (0.6 percent). The error in
calibration of flowmeter number 1 is based
on the agreement between six separate
gravimetric determinations of the flow.
The error in calibration of flowmeter
number 2 is the combined error in the cali-
bration of tfrfc wet test meter and the flow-
meter.
Table 5 is a summary of results obtained
by the three methods. The calculated value
is based on both the measured pressure and
the measured weights. The analyzed value
was obtained by measurement of the sample
using a chemiluminescent analyzer cali-
brated with mixtures of nitric oxide gene-
rated by the dynamic dilution system.
There is generally good agreement between
all samples with the exception of the
value calculated from the pressure data
for sample #133. The agreement between
the gravimetric result and the analyzer
values suggest an error in reading or
92
-------
Table 4. Intercomparison of Primary Standards
of Nitric Oxide in Nitrogen
Sample
No.
Cal. Cone,
in ppm
Signal
in mv.
Div. Per Unit
of conc.
A *
140
949.
929.
0.979
0.0
139
951.
929.
0.977
-0.2
133
960.
943.
0.982
+0.3
1?2
449
438
0.976
-0.1
143
446
436
0.976
-0.1
142
451
441
0.979
~ 0.2
145
226
215
0.951
-0.5
131
225
215
0.956
0.0
144
227
218
0.960
>0.4
149
95.2
96.2
1.011
0.0
148
94.5
95.4
1.010
-0.1
147
94.4
95.5
1.012
~0.1
150
45.4
45.0
.991
-0.3
151
46.2
45.9
.994
0.0
152
45.2
45.1
.998
~ 0.4
Average « 0.2
•Relative difference of the value from the average for
the set.
Table 5. Summary of Results Obtained by Three Methods for
Concentration of Nitric Oxide Standards
Calculated Concentration
Mixture No. Weight Pressure Observed Concentration
133 960. (942) 962.
139 951. 951. 948.
140 949. 9S0. 948.
132 449. 445. 443.
142 451. 449. 447.
143 446. 446. 447.
131 225. 225. 2^3.
144 227. 229. 225.
145 226. 227. 222.
147 94.4 93.1 94.3
148 94.5 93.5 93.9
149 95.2 94.3 95.1
150 45.4 45.0 45.3
151 46.2 46.0 45.9
152 45.2 45.2 45.3
(All concentrations are in parts per million)
93
-------
recording the pressure data for this par-
ticular sample.
"he value assigned to the primary stand-
ards is the value calculated from the
gravimetric data. The internal self con-
sistency of the set of gravimetrics shown
in Table 5 is used to describe the accu-
racy of these standards and the pressure
data and analysis against the dynamic
dilution standards is considered confirma-
tory.
A total of 26S cylinders of five different
concentrations of nitric oxide in nitrogen
were prepared commercially and have been
analyzed by comparison with the secondary
standards. The results are shown in Figure
6. The wide range of values around the
average is not entirely due to imprecision
of intercomparison but rather, may repre-
sent real differences between samples. It
has already been established that the sev-
eral very low values in each lot represent
true differences. It has not yet been
determined exactly how wide is the impre-
cision around the average but previous
experience with the method suggests a maxi-
mum value of + 0.6 percent relative but
the exact value is probably lower. For
instance, a single sample in the 46 ppra
lot was analyzed on several different days
during the analysis of the entire lot.
The value for this sample was found to be
46.0 ppm with a standard deviation of the
average of twelve measurements of 0.1 ppm
or 0.2 percent relative.
The final upper limit of the error on the
certified value will be determined by the
reason for the wide range of values. If
the range represents imprecision in the
measurement then the total uncertainty
will be somewhat larger than if the differ-
ences are real. Table 6 is an estimate
of the error derived for the two cases.
The first value of the error is based on
the same considerations as were applied to
the carbon monoxide samples except that
the obviously low values have been omitted.
The second value assumes real differences
in the samples and the error is the
expected imprecision of analysis based on
prior experience in the intercomparison of
the primary and secondary standards. Both
values are conservative and work in pro-
gress is intended to more clearly define
the error. First the nature of the wide
variation of results will be determined
by reanalysis of a representative group of
samples from each lot. However, this
analysis will not be performed until a
time interval has passed sufficient to
recogniie any further deterioration of
samples. The analyses described here were
perfo rmed approximately one month after
the cylinders were filled. If the dis-
parate results arc due to progressive
reaction of nitric oxide in the cylinder
then those samples originally found to be
low will on reanalysis'be even lower. If
the differences in concentration are due
to a fast reaction dependent on the indi-
vidual cylinder then the relationship
between samples will be approximately as
shown by the first analysis.
The cylinders in which the Standard Refer-
ence Materials were packaged were first
treated by soaking with a mixture of nitric
oxide in nitrogen at about the concentra-
tion with which they were to be finally
filled. All cylinders were new and unused
and were carefully prepared by drying,
evacuating, and flushing with oxygen-free
nitrogen. The long term stability is
expected to be adequate and will be con-
firmed by further analyses as described
above.
The possibility of an increase in concen-
tration due to desorption of nitric oxide
is being investigated by analysis of a
number of cylinders from each lot in which
the pressure is decreased stepwise from
2000 psi to about 500 psi. Thus far,
analyses do not indicate that any serious
desorption occurs but further analyses
over a longer period of time will be nec-
essary to confirm this.
Several samples in each lot were examined
foT other oxides in nitrogen. No evidence
of any other oxides within the limits of
precision of the measurement were observed.
The limits of measurement for this analysis
is considered to be about ±0.2 percent
relative.
Each lot of fifty-three samples included
five samples packaged in aluminum cylinders.
The results for these cylinders is shown
in Figure 7. Compare this to Figure 6
where the results for all samples are
shown and it is obvious that the aluminum
cylinders offer some advantage. While the
spread of values among the aluminum samples
is approximately that of the majority of
samples, there are no low values of con-
centration among the 2S aluminum cylinders.
Statistically, it is unlikely that a dis-
tribution of this sort would have occurred
if the differences between samples repre-
sented imprecisian rather than selective
reaction in specific cylinders.
In summary concerning the nitric oxide mix-
tures, it appears that a stable series of
standards at concentrations between SO ppm
and 1000 ppm will result. A maximum error
of between ± 1 or 2 percent appears rea-
sonable and the shelf life and quantity
available appear adequate.
PERMEATION TUBES
Permeation tubes consist essentially of a
tube of porous material, usually Teflon,
containing a liquid such as sulfur dioxide
or nitrogen dioxide. The material in the-
94
-------
tube permeates through the walls of the
tube at a rate determined by the vapor
pressure of the substance, and at constant
temperature the permeation rate is constant.
The dev ices are used to generate atmospheres
containing a known quantity of the substance
by placing them at fixed temperatures in a
Stream of air or other gas flowing at a
known rate.
In order to serve as a satisfactory stand-
ard a permeation tube must permeate at a
predictable and unchanging rate throughout
its useful lifetime. The rates are deter-
mined by measurement of the weight loss
per unit time at a number of fixed tempera-
tures. The rate (R) is described by the
equation, log R => mT - b where T is the
temperature in degrees absolute and m and
b are empirically determined constants.
The tubes are calibrated by periodic meas-
urement of the weight while maintaining the
tube at a constant temperature. The value
of the constants, a and b, are determined
by calibration at several different tem-
peratures. Permeation tubes must maintain
a predictable rate even after subjection
to varying temperature excursions both of
use and storage.
Sulfur Dioxide Permeation Tubes
Sulfur dioxide permeation tubes have been
studied in detail and no significant prob-
lems have been recognized in either use,
storage, or calibration. Sulfur dioxide
permeation tubes are available as Standard
Reference Materials in three lengths, 2, S
and 10 cm. The permeation rate of these
SRM's is about 0.25 ug/minute per centi-
meter of tube length. The relative accu-
racy with which the rate is known at 25 °C
is ± 0.5 percent for 5 and 10 cm tube and
one percent for 2 cm tubes. The relative
accuracy at the 20 0 and 30 0 calibration
points is one percent for 5 and 10 cm and
two percent for 2 cm tubes. These values
are considered conservative and are based
on at least nine measurements of the rate
of each tube at three different tempera-
tures.
A long term investigation of the behavior
of sulfur dioxide permeation tubes is con-
tinuing. The normal lifetime of a tube
stored at room temperature is about nine
months but the lifetime can be extended by
low temperature storage between periods of
use. A number of tubes fabricated as long
as three years ago and retained from cali-
brated batches of Standard Reference Mate-
rial have recently been recalibrated. The
results of the recalibratioii and the rates
measured before storage at low temperatures
are shown in Table 7.
In addition to these tubes a number of tubes
manufactured in 1S69 are being studied.
Early rate data for most of these tubes is
not available but tho rates measured recent-
ly are consistent with rates measured ear-
lier for other tubes in the same batch.
One tube that was calibrated was found to
have a rate of 2.58 ± 0.2 microgram per
minute in 1970 and in June of 1974 the rate
was measured at 2.61 t 0.2 micrograms per
minute at the same temperature.
The results obtained thus far are all
reassuring and it appears that no problems
arise with sulfur dioxide permeation tubes
frpm the effects of storage. Further,
there is no evidence of changing rate as
the quantity of sulfur dioxide decreases
in the tube with use. It has been
observed that the rate for a sulfur di-
oxide permeation tube remains constant
as long as any liquid remains in the tube
no matter how small the quantity may be.
Tube 7-36 shown in Table 7 is a good
example of this. The value at 25 °C meas-
ured during 6/74 represents the last use-
able measurement because the liquid dis-
appeared shortly after the weighing used to
determine this rate. The rate at 25 °C
remained constant over a period of two
years, which time included both use at 25 °
and storage at low temperatures.
Nitrogen Dioxida'-Permeation Tubes
Permeation tubes appear at present to be
the most practical method for conveniently
preparing mixtures of nitrogen dioxide in
air of accurately known concentration.
Nitrogen dioxide permeation tubes, however,
are not as predictable in use as sulfur
dioxide tubes have proven to be. There are
two primary difficulties associated with
nitrogen dioxide permeation tubes. First,
a tube constructed similarily to a sulfur
dioxide tube has a permeation rate several
times faster with a resulting shorter life-
time. Second, the rate has been found to
decrease significantly with time. The solu-
tion to the problem of a high rate has been
solved through the design shown in Figure 8.
The device consists of a glass reservoir
connected to what is in effect a short per-
meation tube. This design has the advantage
over other designs in that the thickness of
the permeation wall is such that equilibrium
is established rapidly and that the total
weight of the tube is small enough to allow
rapid calibration through measurement of
small changes in weight.
The solution to the problem of declining
rate appears to lie in drying of the nitro-
gen dioxide with which the tube is filled
and protecting the tube from subsequent
exposure to high concentration of water
vapor.
Nitrogen dioxide permeation tubes can be
calibrated with an accuracy equal to sulfur
dioxide tubes of similar output. However,
greater caution must be exercised when
weighing tubes because of the hygroscopic
nature of the external surface of the
95
-------
Table 6. Estimated Upper Limit of the Error of
the Nitric Oxide Standard Reference Material
Estimated Error
Average For The Lot Range Imprecision Real Differences
994. 20. ±13. *11.
477. 14. *8. ±5.
253. 8. * 4. l 3.
94.2 2.4 ± 1.5 * 1.0
45.0 0.8 t 0.6 * 0.5
(All values are in parts per million.)
Table 7. Effect of Low Temperature Storage on the
Observed Rate of Sulfur Dioxide Permeation
Tubes
Tube
Storage Time
Rate in
Micrograms Per
Minute
at Indicated
Temperature
Number
in Months
Original Rate
Rate
6/74
25 °C
30 UC
25 °C
30 °C
2-45
39
2.90
2.96
3-9
35
3.10
3.13
3-10
35
3.12
3.13
7-36
25
0.58
0.85
0.59
0.85
8-11
24
1.50
2.20
1.49
2.21
fi-50
24
1.96
1.95
14-49
7
3.29
4.82
3.32
4.78
14-50
7
3.17
4.59
3.19
4.64
Table 8. Calibration of Fifteen Nitrogen Dioxide
Permeation Devices
Tube Number
Average Rate
s
37-20
1.15
.014
37-21
1.23
.016
37-23
1.33
.015
37-24
1.41
.013
38-2
0.980
.012
38-3
1.15
.012
38-4
1.18
.017
38-6
1.21
.011
38-7
1.05
.008
38-8
1.39
.014
38-10
1.03
.010
38-12
1.21
.012
38-13
1.19
.008
38-14
1.06
.016
38-1S
1.24
.013
(The rate and
standard deviation are
in units of
micrograms per minute at 25 °C.)
.96
-------
permeating area. Typical results are sum-
marized in the calibration of fifteen tubes
at 25 °C as shown in TabH' 8. The stand-
ard deviation shown is based on fifteen
measurements extending over a period of six
months.
Nitrogen dioxide permeation tubes of the
design described here arc intended for use
in the range between 20 ° and 30 °C. Expo-
sure to temperatures above 35 °C may
permanently increase the rate at lower
temperatures by several percent depending
on the length of exposure at higher tem-
perature. Storage at low temperatures
has no permanent effect on the rate in the
useable temperature range but low tempera-
ture storage is not advised because of the
long lifetime of the tube. These tubes
when initially filled to 90 percent of
capacity have a useful lifetime of about
two years depending on the actual rate for
an individual tube.
CONCLUSION FIGURE 8. N3S NITROGEN DIOXIDE PERMEATION DEVICE
The development of a Standard Reference
Material for air pollution measurements
involves a detailed study of the accuracy
and stability of such standards. The devel-
opment has been illustrated by reference to
several specific Standard Reference Mate-
rials which probably represent the median
in terms of the total developmental effort.
Other systems will be studied but the time
required for full development and the
necessity for constant surveilance precludes
any extensive inventory for some time to
come.
ACKNOWLEDGMENT
The author has reviewed what constitutes a
major part of the program of the Air Pollu-
tion Analysis Section of the Analytical
Chemistry Division and wishes to give due
credit to all those in the section who
were involved. The work described has
been supported by the Analytical Chemistry
Division and the Measures for Air Quality
Office of the National Bureau of Standards
and by the Environmental Protection Agency.
tWii Km
91
-------
MEASUREMENT OF OZONE AND OXIDES OF NITROGEN
IN THE LOWER ATMOSPHERE FROM AN AIRBORNE PLATFORM
J. B. Tommerdahl, J. H. White, R. B. Strong
Research Triangle Institute
INTRODUCTION
A program^ was conducted during the summer of 1974 in the Ohio
River Valley; the objective of the program was to investigate high
rural ozone concentrations and their possible relationship to urban
hydrocarbon. Data were obtained by sampling for relevant gases at
five ground stations and from an aircraft instrumented to measure
ozone (0_), oxides of nitrogen (NO -NO-NO2), air temperature and obtain
air samples for subsequent hydrocarbon (HC) analysis.
Two basic flight patterns were to be flown during the field
measurement program. These were a box pattern to be flown during
stagnation and a square wave pattern aligned with the predicted wind
field. Examples of each are shown in a later section of this paper.
These patterns were to be flown, in general, at altitudes between 2000 ft
to 4000 ft above msl. In addition to these two basic patterns, vertical
ascents were to be made to 12,000 ft on days with certain meteorological
conditions.
A C-45H aircraft, two pilots, and some instrumentation were provided
by the National Environmental Research Center at Las Vegas, Nevada, for
the duration of the program. Installation of equipment and test flights
were carried out from the Raleigh-Durham airport. During the field phase
of the sampling program, mid-July thru September, the aircraft was based
at the deactivated Clinton County Air Force Base in Wilmington, Ohio.
This paper describes the instrumenation, the various steps taken to
assure valid data, gives some examples of data format and concludes with
a brief discussion of recommendations for upgrading of system performance.
AIRBORNE PLATFORM AND INSTRUMENTATION
Ai rcraft
The aircraft used in the program, a C-45H, is shown in Figure 1. This
aircraft has an instrument weight allowance of approximately 750 lbs when
operating with a load configuration of two pilots, two instrument person-
nel, and a full load of fuel. Under these conditions this aircraft has an
average cruising speed of 130-150 knots, a climb rate of 500 ft/min, a turning
— Work performed under contract to the Environmental Protection Agency
Office of Air Quality Planning and Standards
98
-------
Figure 1. C-45H Aircraft
99
-------
radius of lk miles and a flight range of approximately 5 hrs. Naviga-
tion equipment included two VOR units with DME, an ADF, and flux-gate
compass. Under normal operating conditions, 10 amps at 28 vdc from the
primary power system were available for the instrumentation.
Given an airborne platform and program requirements, it is necessary
to consider system parameters, such as: power required by instruments
and duty cycle of this power demand; total weight and space requirements
of instrumentation and distribution in aircraft; temperature control of
equipment in flight and on the ground; electrical noise of instruments and
possible adverse effects on communication and navigation equipment; and
shock mounting of instruments to reduce vibration problems.
Air Sample Probe
The air intake system is illustrated in Figure 2; it consisted of a
2-in ID aluminum probe permanently mounted on the underside of the air-
craft nose approximately 3 ft forward of any point on the plane. The probe
was lined with 1-in ID Teflon tubing to prevent or minimize reaction of air
pollutants with the manifold wall.
The other end of the Teflon tube emptied inside the aircraft cabin.
Two ^-in Teflon tubes and a temperature-sensing thermister were inserted
well into the larger 1-in Teflon tube. One of the k-in Teflon lines was
connected to a glass sample manifold from which the 0^ and NO analyzers
sampled. Ram pressure created by air rushing through the 1-in tube forced
a steady flow of sample air into the manifold. The pressure in the manifold
was essentially that of the unpressurized aircraft cabin, which prevented
the pressurization of the instrument sample inlets. The other ^-inch Teflon
line was connected to a stainless steel diaphragm pump used to fill sample
bags. The air sample inlet system was designed for a minimum of sample con-
tamination by emissions from the aircraft itself. Ram pressure caused an
airflow through the Teflon tube at a velocity sufficient to keep gases found
within the aircraft cabin from diffusing into the tube and contaminating
the air sample stream. Tests were conducted to insure that contamination from
the interior of the aircraft was not getting into the sample manifold.
Instrumentation
A block diagram of the instrumentation installed in the C-45H is shown
in Figure 3 and included an RTI solid phase ozone analyzer, a Bendix
N0-N02~N0x analyzer, a Yellow Springs temperature sensor, air sample pump,
five strip chart recorders, and a 2KW Nova inverter. All instrumentation
was shock-mounted with Areoflex twisted steel rope shock mounts. A photo-
graph of the instrumentation is shown in Figure 4.
100
-------
To instruments
Temperature
instrument
To bag
sample
pump
(For hydrocarbon analysis)
0.635 cm i.d.
Teflon .
tube '
Aircraft body
Aluminum
tube
i
2.5 cm i.d.
Teflon
tubing
Class
tubing
insert
^ ' >v
i li J-n 1 lprrl\
T
Inlet
Spacers
(aluminum &
plexiglass)
h
Am
Figure 2. Air-Sampling Probe Used on C-45 Aircraft
-------
Figure 3. Block Diagram of Aircraft Instrumentation
-------
Figure 4. Instrumentation
103
-------
The ozone analyzer used in the aircraft was a solid phase chemilumi-
nescent instrument built by RTI. It is a self-contained instrument
requiring no external pressurized gas supplies of any kind. The analyzer
operates in a cyclic mode with a calibration signal response and a measure
cycle response output once during each 2-min cycle.
The Bendix 8101-B chemiluminescent analyzer was used in the aircraft
to monitor oxides of nitrogen. Support equipment on board for this
instrument included a compressed oxygen cylinder and a two-stage diaphragm
vacuum pump. The sample inlet for the instrument was from the same mani-
fold as the ozone analyzer.
The system used for collecting grab samples consists of a stainless
steel metal bellows pump with a manganese dioxide catalytic converter
on the input to convert any ozone to oxygen. The output of the pump is
connected to a 20 1 Tedlar bag with Teflon lines using stainless steel,
quick disconnect fittings.
Primary power was supplied by the aircraft 28 vdc system which was
converted to 115 v, 60 Hz, l, by the Nova 2KW inverter. Provisions were
incorporated in the system to facilitate rapid switchover from aircraft
primary power source to ground power.
DATA VALIDATION PROCEDURES
An airborne platform presents some particularly troublesome problems
with respect to operation of ambient air gas analyzers. These include the
necessity for essentially continuous supply of power to the instruments;
the determination of the pressure or altitude effects on the detection
capabilities of each instrument; contamination due to aircraft fuel and
power plant of the air sample; and the problem of performing a dynamic cali-
bration of the total system.
Various steps were taken to insure the validity of the data. These in-
cluded: (1) low altitude passes above the base station with all instrumenta-
tion in operation, (2) comparison of readings in the aircraft while on ground
to those in the base station located nearby, (3) comparison of readings in
the aircraft with readings in another aircraft independently instrumented
and operated, and (4) hydrocarbon air sample contamination tests. Data
were also obtained during the program to demonstrate the comparability of
the RTI solid phase and Bendix gas phase chemiluminescent ozone analyzers.
Operational Procedures
Procedures for the operation of the aircraft instruments were established
and documented early in the program so data could be taken under uniform
conditions in the event of a change of operators. Also, operating from pre-
determined procedures insured more consistent operation with the same
operator.
104
-------
Preflight checks were made in order to insure that all instruments
were working and set to the proper range and mode of operation. Also,
the strip chart recorders were checked for operation and turned on. The
manifold was checked to insure that the instruments were connected proper-
ly. These procedures were set to insure that valid data collection began
as soon after takeoff as possible and that minimum data were missed due to
instruments being improperly set. Routine calibrations were conducted on
each of the gas analyzers in the aircraft each day for which a flight was
planned. Calibrations were conducted against the instruments in the van
which were considered "standards." To accomplish this, two steps were
performed: (1) calibration of a portable calibration unit and (2) cali-
bration of the analyzers using the portable calibration units.
During flight, the instrument operator annotated the strip charts with
time and observed the instruments for symptoms of normal or abnormal opera-
tion. The second observer operated the bag sample apparatus and made
pertinent observations of weather conditions and aircraft location. Period-
ically, analyzer flow rates at given altitudes were measured with a bubble
flowmeter.
The instruments aboard the aircraft were operated continuously during
the study. Immediately after landing and taxiing to the aircraft tie-
down position, a 117 V ac source was connected to the instrumentation be-
fore both engines were shut down. Prior to takeoff, at least one engine was
started, bringing a generator into operation before ground power was dis-
connected from the aircraft instrumentation system. Since the aircraft
sampling probe system was designed only for operating during flight, an al-
ternate means was needed for supplying sample air to the instruments while
on the ground. This was accomplished by connecting Jj-in Teflon lines to the
instrument inlet. Each Teflon line extended through the cockpit window and
was supported approximately 2 ft above the aircraft. Concentration values
of NO and ozone could then be read from the instruments in the aircraft and com-
pared to those values being measured in the RTI base station.
Calibration Methods
Standard operating procedure called for the calibration of all gas
analyzer instruments immediately before each flight. The gas analyzers in the
mobile van of the same type as those in the aircraft (ozone and NO) were con-
sidered standards for the aircraft calibrations. Specifically, the output
concentration of a portable calibration unit capable of generating 0^> NO, and
NO2 was determined by connecting it to the analyzers in the base station and
observing their response. The portable unit wa<; th^n carried to the aircraft
and operated under the same conditions (flow rates, pressures, line voltage,
etc.) as in the base station. The aircraft instruments were connected to the
portable calibration unit, and either the span was adjusted to its proper
setting (in the case of the Bendix N0-N02~N0X analyzer) or the internal
calibration level was determined (in the case of the RTI ozone monitor). The
calibration unit operating in this manner is merely a transfer standard im-
posing the requirement of only short-term stability on it to insure its accuracy.
105
-------
Zero determinations for the NO instrument were carried out by connecting
a clean-up system to the sample input of the analyzer. No zero adjustment
was necessary on the ozone analyzer by virtue of the way the output is com-
puted. Multipoint calibrations on the 0-j and N0-N0X-N02 analyzers in the
aircraft were performed every two weeks. The temperature-measuring instru-
ment aboard the aircraft did not require as frequent calibration due to
its inherent stability. It was calibrated at the beginning and end of the
program and twice between. The calibration consisted of a two-point cali-
bration, one at 0°C determined by an ice bath and one at ambient temperature
as measured by a lab-type mercury thermometer.
Contamination Tests
On an initial flight, the gas chromatographic analysis of the grab sam-
ples indicated a high concentration of gasoline evaporatives. The logical
deduction was that air from the interior of the aircraft was being drawn in-
to the sample along with the ambient air. The sample inlet was modified to
produce the system described and illustrated earlier. Then during an actual
data flight some contents of a lecture bottle of a known hydrocarbon were re-
leased continuously in the cabin during the collection of a bag sample. This
bag was analyzed and the concentration of the known hydrocarbon observed and
compared to levels in normal ambient levels made earlier during the flight.
The test indicated that no detectable amount of cabin air containing the high
concentration of the released hydrocarbon was evident in the bag sample.
Since the sample inlet for the hydrocarbon bag sample was approximately 6 in
downstream from the ozone-NO samplet inlet, it was assumed that the ozone-NO
sample was uncontaminated as well.
A test for bag permeation was carried out by storing bags of known con-
centration in the aircraft for long periods of time and reanalyzing them
to see if any hydrocarbons permeated the bags.
Instrument Response vs Altitude
Before the air quality analyzers could be used in the aircraft, varia-
tions in their response due to pressure changes had to be determined. The
environmental test chamber facility at NERC, Las Vegas, was used to determine
the effects of change in altitude on instrument response. This facility con-
sisted of a sealable chamber approximately 28' x 32'd x 60'h. Temperature,
dew point, and pressure could be set by control units external to the chamber
and varied during the test.
A series of tests was run on the RTI ozone monitor and a Bendix N0-N02-N0X
analyzer. A functional diagram of the test set-up is shown in Figure 5.
Each instrument was placed in the chamber and calibrated. Then a constant
concentration of the gas being measured by the instrument was input to the
instrument through a port in the side of the chamber. The calibration gas
was input to the instrument through a glass manifold open to the inside of
the chamber to insure that the pressure on the sample inlet was the same as
106
-------
SPAN
CONCENTRATION
-~
FROM CAL.
SYSTEM
INST.
SIGNAL
OUTPUTS
THERMISTOR
TEMPERATURE
PROBE
C2H4
CHAMBER
TEFLON FILTER
CHAMBER
CONTROL
110 VAC
TO CHAMBER
VACUUM SYSTEM
Figure 5. Test Set-up in Pressure Chamber
107
-------
the chamber pressure. The chamber then was depressurized in increments
of pressure up to the pressure equivalent of 12,000 ft (and beyond in some
tests) while the response of the instrument was being monitored on two
strip chart recorders. Results of multiple runs of this test were processed
to produce a curve for each instrument. From this curve the percentage
decline in instrument response at a given altitude may be determined and
the data compensated for altitude. Curves for the two instruments used
in the aircraft are included in Figures 6 and 7.
In addition to the tests run in the chamber, tests of instrument be-
havior were performed on the ozone instrument while airborne. A standard
calibration system using an ultraviolet ozone source and rotameter was used
aboard the aircraft with a modification on the output. A restriction was
placed on the output, along with a vent controlled by a valve (see Figure 8).
By closing the valve, the calibration system could be pressurized. An
altimeter was connected to the system so the pressure could be adjusted to
the same level as it was on the ground. The flow out the vent was monitored
and the length of the 1/8 in Teflon restriction adjusted so that the flow out
the vent did not exceed 1 1/min (at standard conditions). This insured that
sufficient calibration gas flowed through the manifold to keep it flushed and
not contaminated with cabin air.
Flow rates were monitored on the instruments during the tests on the
aircraft at the various altitudes. Then during routine flights, the flow
rates were checked periodically to insure that the operating conditions were
unchanged. This insured that the results obtained during the altitude tests
were still valid (including the correction curves).
The results of these in-flight calibration and operational verification
tests were in agreement with the tests run in the fixed chamber. Repeatedly
it was shown that the response of the instrument declined linearly with en-
vironmental pressure.
Ground Comparison of Aircraft and Base Station Measurements
Ground comparison tests consisted of running the aircraft instrumentation
and sampling ambient air while the aircraft was parked in close proximity to
the base station. The sampling ports' in this configuration were located ap-
proximately 70 ft apart. The port for the aircraft was approximately 12 ft
above the ground while the base station sampled air from about 24 ft above
the surface. The comparison of these data was done not so much for quantita-
tive checks of instrument calibration as for a check of the instrument per-
formance. The shape of the ozone concentration curve taken in the aircraft
was compared to shape of the one taken in the Mobile Monitoring Laboratory
every morning prior to a calibration. The calibration then should correct any
discrepancies in instrument readings. A graph of the outputs from both the
108
-------
.8
.6
.4
• Test Run //I
O Test Run if2
km
I
8
i
10
I
12
1000 ft.
Altitude
Figure 6. Response of RTI 0^ Instrument vs Altitude
-------
1.0 -
.9
.5
.4
*n
^1
«
¦ W
S>
km
3
I
I
4
8 10
1000 ft.
12
I
14
—r
16
i
18
ALTITUDE
Fiqure 7. Correction Factor for Bendix NO Analyzer
to Compensate for Changes in Altitude
110
-------
Figure 8. System Used to Calibrate Analyzers While Airborne
the aircraft O3 instrument and the instrument in the base station plotted
against time is shown in Figure 9. Discrepancies in the two as already
mentioned may be due to a number of causes, including differences in sample
manifold height.
Low-Pass Flights
Fly-bys were conducted by flying the aircraft over Clinton County Air
Force Base at an altitude between 50-100 ft above the terrain. Generally,
the aircraft approached the field from the end opposite the base station
(approach from the northeast end) and flew over the taxiway parallel to the
runway. Flying in this manner, the aircraft could safely achieve a lower alti-
tude and hold it longer than flying from another direction over grass or wooded
terrain. The time for the aircraft to fly from the end of the runway to the
base station took approximately 30 s, which gave the instrument some time to
stabilize before a reading was taken.
A plot of the results is included in Figure 10. Here the readings taken
from the RTI instrument are plotted on the ordinate against the fixed base
station 5-min instantaneous sample readings taken closest in time which were
plotted on the abscissa. Perfect agreement between the two instruments would
111
-------
260
240
220
200
180
160
140
120
100
80
60
40
0
o RTI OZONE MONITOR IN AIRCRAFT ON GROUND
° BENDIX OZONE INSTRUMENT IN BASE STATION
4
16
18
20
22
13 August
24
10
14 August
Figure 9. Ground Comparison - Ozone Instruments
-------
Figure 10. Aircraft Fly-By Data
-------
be indicated by a plot in which all points fell on a line passing through
the origin and inclined at an angle of 45 . The departure of the instru-
ment behavior from this ideal is not unexpected. During the early hours
of the day at a time when many of these passes were made, a homogeneous
concentration of ozone due to mixing had not been achieved. This was borne
out by observing the strip chart trace in the aircraft during the measure
cycle. It was frequently quite erratic, indicating variations in the con-
centrations of ozone at different points surrounding the base station.
Points taken before 1000 were omitted from the data plot; a rough linear
fit is indicated by the solid line in Figure 10. Since the base station
sample inlet was at 30 ft above ground level, it is reasonable that the
values should differ to some extent.
Comparison Flight
Another means of insurance of valid data lies in comparison of results
of two independent similar systems operating in the same environment with
no interaction. This type of test was conducted in conjunction with
Battelle Memorial Research Institute. They were operating a Cessna 173
aircraft outfitted with a REM gas phase chemiluminescent ozone analyzer
and ram pressure bag filling system. The objectives of the test were
to fly the C-45H aircraft and the Battelle Cessna through the same air
at the same time in such a way as to avoid sampling each other's exhaust.
Due to the differences in cruising speed of the two aircraft, a formation
flight could not be flown. The pattern shown in Figure 11 was agreed upon.
Both aircraft were to fly the pattern twice with the C-45H aircraft starting
the flight in front of the Battelle aircraft.
The flight was conducted on 9 August with success. A partial plot of
both the RTI-measured ozone levels and those determined by Battelle is in-
cluded as Figure 12. The readings are plotted versus location along the
flight path; consequently, there was a time difference between the times
that data at the same point was taken, and the amount of time difference
varied with position of the flight plan.
The plot shows good agreement between the RTI instrument and the REM in-
strument over the duration of the flight path, the difference being typically
less than 10% •
Gas Phase Chemiluminescent vs
Solid Phase Chemiluminescent Ozone Measurements
The ozone meter used in the aircraft was one built by RTI, under a
previous contract with EPA, using a Rhodamine B disk which chemiluminesces
with ozone. This instrument requires no ethylene bottle aboard the aircraft.
In addition, it incorporates an internal calibration source which is measured
once each cycle.
114
-------
Richmond
OMNI
Dayton OMNI
-V.v'Kt-.-
Dayton
Ohio
t' .\VJe»-j'/i
•• J
Montgomery
County
OMNI
Wilmington
(Base station)
16.1 km
Figure 11. Ozone Instrument Comparison Flight Pattern
115
-------
TIME
Figure 12. Comparison Flight - Data - 9 August 1974
-------
Tests were run using the RTI instrument alongside the Bendix gas
phase chemiluminescent ozone monitor to demonstrate the comparability
of the two methods. These tests were run in the RTI Mobile Monitoring
Laboratory with the two instruments located side by side and connected
to the sampling manifold through which ambient air was being aspirated.
The output of each was sampled once every 5 min, 24 hr per day and re-
corded on a magnetic tape. The RTI ozone monitor used for this test had
a set of three sample and hold modules and electrical analog subtraction
and division circuitry which carried out the function of comparing the mea-
sure signal level with the known internal calibrate. The magnetic tape
data was processed by computer and the data output in concentration units.
Tests were run on these instruments on two occasions: once in the
spring of 1974 (April 17, 18, 19 and May 2, 3) and again in late summer
1974 (August 9-September 1), yielding ample data for comparative analysis.
A statistical analysis was performed on the data (spring and summer) to
numerically evaluate the similarity of the readings. Results of this
analysis are shown in Table 1. For the April and May data, 15 min
averages consisting of three 5-min data points were used for the analysis;
in the August data, individual 5-min data points were used. An example of
a three-day period is shown in Figure 13. These data substantiate that the
two instruments are comparable.
Data Acquisition and Reduction
The data acquisition system used consisted of six electric writing
strip chart recorders rack-mounted in a single unit. Also included in the
unit was switching circuitry which allowed the input of the recorders to be
shorted without shorting each instrument output. This facilitated the zero
adjustment check of the recorders by allowing this operation to be carried
out with a single switch. A digital voltmeter (DVM) was included for check-
ing the instrument levels (and therefore checking the operation of the strip
chart recorders by checking the correctness of the deflection). The DVM
could be switched to any instrument output with no movement of test leads.
Since there were up to five individual records represented — concentrations
of O3, NO, N02~N0x and temperature — some means had to be utilized to corre-
late the five with each other and with the corresponding location on the
ground. The zero adjust switch was used for this purpose. Each time a sig-
nificant point was crossed in the flight pattern, the switch was activated
causing all recorders to simultaneously trace to zero and return to normal
leaving a distinguishable mark. The time of the event was marked on the ozone
recorder to the nearest minute. The operator maintained a separate log of the
times and the event which occurred.
After the data were converted to micrograms per cubic meter, it was
placed on a map illustrating the flight path at the point corresponding to
its origin. The flight path was drawn immediately after the flight consider-
ing recollections of the pilot and co-pilot and notes made by the second in-
strument operator. The data was then put on the map by observing where the
117
-------
Table 1.
STATISTICAL COMPARISON OF THE
RTI SOLID PHASE INSTRUMENT VERSUS BENDIX GAS PHASE INSTRUMENT
Time Period
Day
N*
Correlation
Means
RTI Ben
Ratio of Means
RTI/Ben
April 17
20
.9644
159.1
150.7
1:06
18
20
.9958
136.9
133.6
1.02
19
20
.9567
158.1
151.7
1.04
May 2
30
.9981
74.1
66.4
1.12
3
64
.9987
65.3
59.4
1.10
Aug. 7-9
547
.99
85.7
81.1
1.06
14-16
632
.99
94.1
95.4
.99
22-24
849
.98
121.8
120.2
1.01
28-30
617
.97
63.7
61.6
1.03
~
Number of data points included in comparison.
118
-------
240
220
200
180
160
140
120
100
80"
60
40
20
0
• — RTI Measurement
o — Bendix Measurement
• — Coincident Points
J I I ' ' L
J ' '
8 12 16 20 24
16 August
4 8 12 16
TIME (HOURS)
17 August
20 24 4 8 12 16 20
18 August
Figure 13. Gas Phase and Solid Phase Ozone Instrument Comparison
-------
particular data point fell on the strip chart in relation to noted sig-
nificant events in the flight path (turns, prominant landmarks, etc.)
and placing it on the pertinent portion of the flight path at proportionant
distance between the two nearest significant features. The times of all
turns were placed on the maps as an aid to those who analyze the data. NO
and NO2 data were omitted from the flight plan due to the fact that they
were nearly always below the detectable limits of the instrument, but
were put in tabular form along with the temperature data.
Examples of the aircraft's data may be found in Figures 14, 15, and
16. The latter example was taken from flights flown subsequent to the
Ohio River Basin study and are given to show a third type of flight pattern.
SUMMARY AND RECOMMENDATIONS
There are several improvements one would consider in a system of the
type described, most of which are practical to implement, assuming the
necessary time, facilities, and funds are available. These improvements are
as follows:
(1) Although continuous measurement, with compatible response times, of
each of the pollutants is practical at this time with state-of-the-art
analyzers, an improvement is needed in the minimum detection level of.
N0x-N0-N02 analyzers. At least an order of magnitude improvement in
detection sensitivity is needed for detection and measurement of NO
more than 500 ft above ground. Also, it would be desirable to have
improved flow-rate control and/or monitoring; especially for flights
involving altitudes varying over a range of several thousand feet.
(2) A small, light weight, simple, and reliable calibration system for in-
flight measurements would be very useful in routinely verifying the instru-
mentation system operation and calibration. Such a system would have to
be usable over the range of altitudes covered in data flight, a require-
ment not imposed on typical calibration systems currently available.
(3) Testing of the sample inlet system should be tested for dynamic
characteristics in a wind tunnel initially, if possible. In addition,
in-flight tests should be conducted to compare the aircraft measurements
with those made at a ground-based station. These could be conducted simi-
lar to those described in the Low-Pass Flight Section, with more flights
being made from several directions, at a time of day when good mixing has
occurred and with the ground sample inlet at a height compatible with the
permissible aircraft minimum flight altitude. By making several passes
from several directions a better statistical sample could be obtained
for comparison with the ground-based data system. If possible, a location
could be chosen whereby the gircraft could fly at the test altitude for
at least a minute before the comparison data point would be taken.
120
-------
Figure 14. Double-Box Flight - 13 July 1974
121
-------
|
-------
,1)1? ¦
to
CO
LAKE
MATHEWS
RIVERSIDE
V
1219
78
975
76 76 60
BANNING
RIVERSIDE TO INDIO-TRAWSPORT STCPY
Flight Date/Time Oct. 7, 1974 1044-1219 PDT
Ozone Concentration in ug/o
Time XX3"-^ In PDT
Altitude In
Dashed lines Indicate changing altitudes
1151 /
PALM SPRINGS
Figure 16. Los Angeles Basin and Vicinity
-------
(4) Incorporate automatic recording of altitude and position, along
with the primary data and any desirable status information, onto
a data acquisition system which is directly compatible with com-
puter processing equipment.
(5) Redesign of the instruments for aircraft operation for reduction
in size, weight and power consumption, and reduced susceptibility
to shock and vibration would be desirable. Rapid switchover to a
minimum power consumption mode would provide for the maintenance of
power to the critical subsystems during times of refueling, etc. with
a minimum battery back-up system.
124
-------
INSTRUMENT TIME RESPONSE AND ITS IMPLICATIONS
David T. Mage
RAPS Project Officer
Air Monitoring Branch
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
NERC-LV
-------
Introduction
It is well known that a linear air monitoring instrument with a finite
response time, defined as the time required to reach a percentage of a unit
step input signal, will not produce an exact record of the atmospheric con-
centration being sampled. Data taken by ground based monitoring stations
are recorded, reported and analyzed without correction for the response
time characteristics of the instruments. Neglect of this effect is justi-
fied in most cases, by the fact that air quality standards and health effects
are related to averages of one-hour or greater which are relatively unchanged
by this effect. However, for aircraft sampling, where the purpose of the
monitoring is to obtain spatial variability, or to measure the manner in which
a plume is dispersing these effects can and do become significant. This
paper discusses a technique by which the concentration input to the instru-
ment can be estimated from the recorded output using linear systems theory
and the instrumental transfer function.
Theoretical Development
A previous publication by the author, included as appendix A, developed
the theory and the use of the transfer function to provide a rigorous relation
of input to output. The details will not be discussed in this presentation
since they are available in the literature. A simplified model may be set up
for an arbitrary detector placed within a chamber of volume V liters with a
flow of pollutant in and out at a rate of v liters/minute. If we assume
the chamber is perfectly mixed, the outflow will have an identical concen-
tration C, to that in the chamber, and we can write a mass balance
125
-------
vCi - vC = (1)
where Ci is the input concentration which we desire to measure. This
equation is a first order linear differential equation and can be written
as Ci = C + (2)
where t = V/v is the time constant for the system. For the case of zero
lag time, this equation means that to the output C we must add t times
the slope of the curve to get the input concentration Ci at the time the
recorder is reading C. For example, let us assume the instrument with time
constant measures a sinusoidal signal of period I as C = Co + sin t.
2TT
The input to the system which produced this output is Ci = Co +• sin ~y~ t
2TT t 2TT ?TT 7-
' cos —— t. At t = o,T,2T, . . . when C = Co, Ci = Co -fr- ¦ —
" rp w *»wh. w ^ M ^ ) • • * u wv 5 I ip
For the simple case where T = t and Co = 2, the input and output curves
are shown on Figure 1.
These two curves have the same period only because of the choice of T =
2TCt . In general the input and output curves will be out of phase. The
most obvious effect is the decrease in amplitude of the fluctuations about the
mean from 1.414 to 1, a difference of about 307». The following sections
discuss these corrections in relation to aircraft measurements.
Response Time Effect on Aircraft Monitoring
For an aircraft moving at 150 knots ground speed, one second of time
corresponds to 77 meters (253 ft) of travel. To demonstrate the effect of
the response time, assume the plane is carrying an instrument which has a
first order linear response with a time constantt= 2 seconds and a lag of
1 second. If the plane were to pass through a plume of pure pollutant
(106 ppm) .077 111111 wide, this would correspond to a delta function input of
area 1 ppm-second. The response of the instrument is shown as Figure 2. The
126
-------
25T
.50T
time
.75T
FIGURE 1
127
-------
aircraft would observe the exponential curve rather than the delta
function. The instrument would not respond until the aircraft had traveled
77 meters beyond the plume and the reading would be spread out over approxi-
mately 1/2 kilometer of travel. The difference between input and output
shown here is the maximum possible and points out the need for making the
time-space correction when sampling from high speed aircraft.
In the following section reference is made to the figures in the appendix.
Let us assume the instrument is being flown at 150 knots and it responds 60
times faster than the instrument analyzed in the appendix. The figures A4,
A5 and AG could then be changed simply by changing the time scale from minutes
to seconds, as shown in Figures ? and 4. The instrument would have a 2 second
lagtime and it would take an additional 5 seconds to reach 90% of a change
in signal. The technique of using the convolution integral in finite
difference form is equivalent to treating the output as a general autoregressive-
moving average process of order m, where the transfer function is divided
into m increments. The transfer function of this particular instrument is
shown in Figures 3 and A-4. It is of interest to note that this function is
no longer the ideal exponential response and no apparent analytical expression
can be used to represent it. In Figure A-6 and 4, the instrument is assumed
to have recorded the output marked as I, the response R(t) curve caused by
the disturbance D(t) curve, marked as II. The disturbance is damped out over
1007„ and is spread over almost twice the distance. If this were a horizontal
pass through a plume, the spread of the recording would change a computed
from the order of 50 meters to about 100 meters. Another area where these
y
corrections might be necessary is that of vertical measurements. The recorded
concentrations can be used to obtain the flux in horizontal crosswind flight
128
-------
Distance (meters)
0 2 4 6 8 10
Time (seconds)
FIGURE 2
-------
FIGURE 3
TRANSFER
FUNCT ION
~L 6 8
T IME, s
-------
distance; m
-------
since the wind velocity will not change noticeably in the horizontal over
vertical flux requires use of the wind velocity U as a function of height
data directly. Turning the previous figure 90° we have what may correspond
to a vertical sounding as the plane is descending. The plume at the higher
altitude would give a greater flux than the recorded plume at the lower
altitude where the winds are diminished. This effect can be minimized by
utilizing a slow rate of descent such as 500 ft/min used by the RAPS
helicopters in the St. Louis area.
Summary and Conclusions
In using high speed aircraft and instruments with response times on the
order of seconds, high frequency disturbances may be spread over relatively
long distances. For those situations where gradients of properties, rather
than the averages of the properties are rejuired, a correction for the
response of the instrument system may be necessary. This paper reviews
the theory of linear systems and presents a technique by which any linear
instrument may be modeled and corrections made.
However the computation of
cannot be used with the recorded
132
-------
True Atmospheric Pollutant Levels by Use of
Transfer Function for an Analyzer System
David T. Mage and Jamshid Noghrey
Department of Chemical Engineering
San Jose State College
Air pollution analyzers, because of inherent lags and response times, will not pro-
duce an exact record of pollutant concentrations. A procedure is described for
determining the transfer function for an analyzer system. This transfer function is
applied to the recorded output to produce a record of the actual atmospheric
pollutant concentration at the inlet to the system. This procedure is applied to the
carbon monoxide analyzer of the Bay Area Air Pollution Control District monitoring
station in San Jose. Results show that variations in peak heights of more than 50%
can exist between actual and recorded values.
It is ;i well knouu fact that mo-it air
pollution monitoring instruments do
not respond in-taiitaneoii-1y to the
actual levels of pollutants entering their
intakes. In ventilation typo instru-
nients Ij.e.. instrument- in whic h air is
pumped thiough a detector), it taker, a
finite time for a ga- .-ample to reach the
detector. In thi-, time the sample may
become ini\ed with other ua- -ample-
wliuli pre\ iou.-ly have entered the sys-
tem in n similar manner. As a result of
this system behavior, a sudden, abiupt
change in an atmospheric pollutant level
will not be recorded instantly, and the
instrument output will show a gradual
increase to the new level after a delay
which ^characteristic of the system.
Such an abrupt change of the input of
a system is commonly referred to as a
"step" input, while the result in? output
is called the "step response'' of the sys-
tem. Figure 1 for example, shows the
step response of a sulfur dioxide (SOj)
measurement instrument to a three
minute input step of 2 ppm SO;.1
There are four important factors to
observe:
1. It took GO sec before the iu-trument
fir-t began to re.-pond.
2. It took another 1H-ec for the output
to reach I ..">1 ppm.
3. The maximum recorded value wa-
1.9 ppm not the true value of 2.0
ppm.
4. During a period of 1.S5 min the re-
corded value is above 1.5 ppm.
The inability of the monitoring instru-
ment to record the actual pollutant level
in the atmosphere can be a serious ob-
stacle for those responsible for air pollu-
tion monitoring. In the San Francisco
Hay Area, this occurrence of 2 ppm for a
3-min period would constitute a viola-
tion of the local air quality standards.
Regulation 2 of the San Francisco Hay
Area Air Pollution Control District
11' \ \ I'('I) ] -late-, in pal t "No pel -nn
-hall I all-i'. let. pel mit. -liltei .11 a'lnw
,in\ emi—1011 of -ullur dioxide ulm h 1 • ¦-
-nil- 111 ground level coin entiat mil- ot
-ull'iir dioNule at any tin en point 111 e\-
n — of I.") ppm I vol) I'm tliMT ron-ei-u-
ti\r minute-...." Tim-, fur the pur-
po-e of monitoring a -u-peeted \ 10I; 11
before initially re-poudiim and .1 total
elap-etl timi of 30 nun beloie uarhtii-
90% of the maximum value 111 re-pon-e
to a step input. Tlui.-. tin- type of
instrumentation would be inadequate
for measurement of level- as part of
an emergency alert network both be-
cause of the large delay and the re-
duction of maximum values which
would lie produced.
The method described in this paper
allows one to take into account the
instrument delay and response and thus
allows the more accurate analysis of
true time variations of air quality.
February 1972 Volume 22, No. 2
133
-------
Time above 1.5 ppm:
Unit A 1.85 min,
3 4
Time, minutes
Figure 1. Response check.
To.
Unit
impulse
»(t)
1.0
Unit
step
u(t)
0 0
Time— Time-
(a) (b)
Figure 2. Idealized inputs tor system analysis.
Theory of Transfer Functions for
Linear Systems
A linear -y-tcm i« one which exhibit*
vnin>ii- linear properties. urn* of tlio
infi-t important of which i> the principle
of -iipc'i-poMtion illustrated hv Table I.
Tin- functions ;/iU) ami y.(t) are the out-
put lespouse* for each of two different
input exc itation-;. .n(.0 and iz(t). When
the input function = Xi(t) + jrs(t)
is introduced into tin* .¦system, the out-
put then mu»t be equal to the sum of the
previous two outputs, j/3U) = iji(l) +
A more detailed description of
thi' characteri-tics of linear systems may
be found in the literature.5-3
From our investigations with air
monitoiing instrumentation, it appears
that many such instruments can be
treated, to a jiood approximation at
lea-.t. as linear systems. Each linear
system has its own unique transfer
function, and, from this transfer func-
tion. the system's response to an arbi-
trary input can be determined. The
development of such a transfer function
usually involves the use of Laplace
transforms. If f(t) is a function of t de-
fined for values of t > 0, its Laplace
transform i- obtained as follows:
= fo f(t)c--'dl (1)
Two inputs which are common in the
di>cu*sion of linear -y-tem.- are the unit
impulse o\t) and the unit step »(/).
The unit impulse is defined to be zero
for nil values of / except at i = 0. at
which point it has the value of infinity;
it is defined to have the area of unity.
The unit step is a function which is 0
for all values of t < 0 and 1 for ( > 0.
These two functions are depicted in
Figure 2. The Laplace transform of
the unit impulse is simply unity.
The response of a system to a unit
impulse excitation is called the transfer
function for the system, C(t), with
Laplace transform Cis).
The transform of the output R(s) will
be the product of the system transform
C(s) and the input transform X(s).
L[R(0] - R(s) - A'(«)• response to arrive at the true im-
pulse respon-c. From this the overall
system transform C(s) may be derived,
and then it will be ivo-^ible to calculate
the output to any input function.
Since the resulting transform of a
complex instrument can sometimes be
unwieldy to write and handle, one can
choose a more simple approach which
makes use of the convolution integral.
Here, the unit impulse response is u-cd
to estimate the output to any arbitrary
input without the neces-ity for writing
an analytical expre—ion of the tran-fer
function of the *\Mein. A rigorous
solution to the system the time response
/?(/), with zero initial condition, can be
obtained by u-
-------
From Figure 3 it is observed that such
a gas sample may take 1 ruin to reach
the detector and 3 min to pass through
it completely.
In practice, the solution to the con-
volution integral of equation (4) is quite
involved and a finite-difference method
may be used to approximate the solu-
tion. The following describe* the com-
putational method for obtaining the
output response, Ii(t), to an arbitrary
input function X(t) bv such a finite
difference approach, in which J\T repre-
sents an arbitrarily small increment
-------
Time, hours
Figure 5. Ppm CO vs time, BAAPCD, San
Jose.
Figure E. I. R(») curve. II. X(t) curve.
broken into equal increments of 0.5
mm each, corresponding to the same
increment used in determining the
transfer Mux tion.
For nurpo?c of computation, the at-
ir.o-phei if caibon monoxide was as-
sumed to 1m- constant at 1.1 ppm for the
proceeding 10 niin, a- .-lirmn On Figures
5 and 0. The viuve /?t() van then he
broken up into int-icMiiciit 1 1 ppm
/?< 4; = 1.1 ppm
ft>r>) = 1 3 ppm
ftiti) = 2 2 ppm
etc
By utilization of equation Hi), the
A" (I) cuivec.in be solved fur a-> follow.*:
Ritl = Cil)-A'« - A7>f.5) +
CV21 ¦ A" (< - 2sT)-{.5) +
Ci3)-A'U - foT) i.o) +
C'i4) • A'U - 4±T) -(.5) +
Ci• j) -.V(/ — osT) ¦ (.5) +
C(G) X(t - C,lT)-(.5) 4- ..
noting
C(l).. ,C(4) = 0 and
R{t) = 1.1 ppm, t < 0
/?(4) = ('2.0 min-')-(.5 inin) •
(l.l ppm) =>1.1 ppm
In this analy-U, A"(<) i~ the unknown,
and can l>c solved for by u stepwise pro-
cedure a> follows:
R(t) = C(5) X(t - 5ST)-(.5) +
C(6)-A'U - 6lT)-(.5) + ...
R( 5) = C(5)-A'(5 — 5 ±T)-(.o) +
C(6)-A'(5 — 6a7)-(.5) + ...
notina/?(5) = 1.3 ppm and A'(5 — 6AT)
A"(5 — 20ST) are equal to 1.1 ppm
one can apply the equations above using
the experimental values for C(5)...
C(20) to solve for the remaining un-
known, A'(5 — 5 AT).
1.3 ppm = 0.1G2 min-l-A~(o — 5AT)-
21)
t.5 niin) + 51 C(«) A'(5 — nX
n ¦ G
ST) ¦ (1.5 miu)
1.3 ppm = .OSl -Alo — oST) +
1.83S mill-l• (l.l ppm)-(.5 min)
in>r for A"(5 — 5A7\), it is found to
be equal to 3.G ppm.
Continuing the procedure, advancing
time by ST. one again solve- for
A" <5 — 5A7) but imvv .Y("> — OAT) i- no
loimerequal to 1.1 ppm. but the picviou«
value determined 3.ti ppm now become-
the value for ,V(5 — GsT). "I'llpro-
cedure is repeated, continuously index-
ing the determined value* and solving
for the single unknown eoncentration.
Tlic results of this process are shown in
Figure 6 as curve II. The maximum
recorded value of 14.3 ppm was quite
different from the true maximum input
value of approximately 34 ppm.
In conclusion, where air quality stan-
dards are based upon a timed exposure
to certain air pollution levels which are
believed to be injurious to health, it may
be necessary to calibrate the instrument
as described herein to determine the
true atmospheric values. Once the
transfer function is obtained, one can
write a computer program which will
give the output of the instrument for
any arbitrary input, as was done here.
13y doing so, it is then possible to obtain
a feel for the true nature of the input by
examination of the recorded output of
the instrument. When the data ac-
quisition system is part .of an air pollu-
tion alert system for a community, a
computerized data transmission system
can make these corrections which will
aid the calling of alerts at the earliest
possible moment.
Acknowledgment
We wi>h to thank the Hay Area Air
Pollution Control District for its co-
operation in this re-cardi and the use of
the CO analyzer at the San Jo*e Air
Monitoring .Station in undiTt;iking the
analysis and calibration technique de-
scribed in this paper. We al*o wi*h to
thank Mr. Wayne Ott of the Office of
Air Programs, U. S. Environmental
Protection Agency lor many valuable
anil helpful discussions and suggestion-*
concerning this work.
References
1. 1'iiUet. L. B.. "Coinpaii-iili of Sulphur
Dioxide Analv/ei*." l'aper pie-en led
ai lllih t'nuiri 1'ine "ii .\ieilmds in Air
Pullutiim and Industrial Ilvgiene .Stud-
ies, Fianiiscn. Call!., February
11). l'.HV).
J. Lcvcii-picl. O., Chcini'til Ucirtion Engi-
nccrmtj, Wiley. New Yuik, l!Ki'2.
3. John J. and Con-tantine II.
llmipU, Fcnlhact: Control. Analy-
sis uiul Si/nlhais, McGraw-Hill, Inc.,
New Yurk l'.KJti.
4. Murrill, 1'. \V„ Pike, 1!. W., nnd
Smith, C. 1 j., "Dynamic mathematical
model-," Ohem. Knit.. 76 (7), 151
(I OUt) )-
Di. Mage is A
------- |