EPA-R4-73-028d
August 1973
Environmental Monitoring Series
V
\
UJ
a
-------
EPA-R4-73-028d
GUIDELINES FOR DEVELOPMENT
OF A QUALITY
ASSURANCE PROGRAM
Reference Method for the Determination
of Sulfur Dioxide in the Atmosphere
by
Franklin Smith and A. Carl Nelson, Jr.
Research Triangle Institute
Research Triangle Park, North Carolina 27709
Contract No. 68-02-0598
RTI Project 43U-763
EPA Project Officer: Dr. Joseph F. Walling
Quality Assurance and Environmental Monitoring Laboratory
National Environmental Research Center
Research Triangle Park, North Carolina 27711
Prepared for
OFFICE OF RESEARCH AND MONITORING
U.S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D.C. 20460
August 1973
-------
This report has been reviewed by the Environmental Protection Agency and
approved for publication. Approval does not signify that the contents
necessarily reflect the views and policies of the Agency, nor does
mention of trade names or commercial products constitute endorsement
or recommendation for use.
11
-------
PREFACE
Quality control is an integral part of any viable
environmental monitoring activity. The primary goals of
EPA's quality control program are to improve and document
the credibility of environmental measurements. To
achieve these goals, quality control is needed in nearly
all segments of monitoring activities and should cover
personnel, methods selection, equipment, and data
handling procedures. The quality control program will
consist of four major activities:
• Development and issuance of procedures
• Intra-laboratory quality control
• Inter-laboratory quality control
• Monitoring program evaluation and
certification
All these activities are essential to a successful quality
control program and will be planned and carried out
simultaneously.
Accordingly, this fourth manual of a series of five has
been prepared for the quality control of ambient air
measurements. These guidelines for the quality control
m
-------
of sulfur dioxide measurements in the atmosphere have been
produced under the direction of the Quality Control Branch
of the Quality Assurance and Environmental Monitoring
Laboratory of NERC/RTP. The purpose of this document is
to provide uniform guidance to all EPA monitoring activities
in the collection, analysis., interpretation, presentation,
and validation of quantitative data. In accordance with
administrative directives to implement an agency-wide
quality control program, all EPA monitoring activities are
requested to use these guidelines to establish intralaboratory
quality assurance programs in the conduct of all ambient air
measurements for sulfur dioxide. Your comments on the
utility of these guidelines, along with documented requests
for revision(s), are welcomed.
All questions concerning the use of this manual and
other matters related to quality control of air pollution
measurements should be directed to:
Mr. Seymour Hochheiser, Chief
Quality Control Branch
Quality Assurance and Environmental
Monitoring Laboratory
National Environmental Research Center
Research Triangle Park, North Carolina 27711
IV
-------
Information on the quality control of other
environmental media and categorical measurements can be
obtained by contacting the following person(s):
Water
Mr. Dwight Ballinger, Director
Analytical Quality Control Laboratory
National Environmental Research Center
Cincinnati, Ohio 45268
Pesticides
Dr. Henry Enos, Chief
Chemistry Branch
Primate and Pesticide Effects Laboratory
Environmental Protection Agency
Perrine, Florida 33157
Radiation
Mr. Arthur Jarvis, Chief
Office of Quality Assurance-Radiation
National Environmental Research Center
Las Vegas, Nevada 89114
During the months ahead, a series of manuals will
be issued which describe guidelines to be followed during
the course of sampling, analysis, and data handling. The
use of these prescribed guidelines will provide a uniform
approach in the various monitoring programs which allows
the evaluation of the validity of data produced. The
implementation of a total and meaningful quality control
\
program cannot succeed without the full support of all
monitoring programs. Your cooperation is appreciated.
-------
TABLE OF CONTENTS
Section
I INTRODUCTION 1
II OPERATIONS MANUAL 3
- GENERAL 3
PRESAMPLING PREPARATION 5
Reagent Preparation (Step 1) 5
Flow Rate Calibration (Step 2) 7
Absorber Preparation (Step 3) 20
Sample Identification (Step 4) 21
Package For Shipment (Step 5) 21
SAMPLE COLLECTION 23
Sampling Apparatus Assembly (Step 6) 23
Operational Check of System (Step 7) 32
Sample Collection (Step 8) 34
i
Sampling Handling (Step 9) 35
SAMPLE ANALYSIS 36
Verify Documentation (Step 10) 36
Recalibrate the Critical Orifice (Step 11) 36
Reagent Preparation For Analysis (Step 12) 37
Develop a Calibration Curve (Step 13) 41
Colorimetric Analysis (Step 14) 45
DATA PROCESSING 46
Perform Calculations (Step 15) 46
Document and Forward Data (Step 16) 47
vii
-------
TABLE OF CONTENTS (Cont'd)
Section Page
SPECIAL CHECKS FOR AUDITING 47
Flow Rate/ Volume Check 48
Measurement of Reference Samples 50
Data Processing Check 52
SPECIAL CHECKS TO DETECT AND IDENTIFY TROUBLE 52
Average Temperature Environment of Sample
Between Collection and Analysis 53
Measurement of pH of Final Solution 54
Interference Checks 54
Volume of Absorbing Reagent 55
Check of Sample Timer 55
FACILITY AND APPARATUS REQUIREMENTS 56
Facility 56
Apparatus 56
III SUPERVISION MNUAL 58
GENERAL 58
ASSESSMENT OF DATA QUALITY 60
Required Information 62
Collection of Required Information 62
Treatment of Collected Information 64
SUGGESTED STANDARDS FOR JUDGING PERFORMANCE
USING AUDIT DATA 66
-------
TABLE OF CONTENTS (Cont'd)
Section Page
COLLECTION OF INFORMATION TO DETECT AND/OR IDENTIFY TROUBLE 66
Identification of Important Parameters 66
How to Monitor Important Parameters 71
Suggested Control Limits 71
PROCEDURES FOR IMPROVING DATA QUALITY , 74
PROCEDURES FOR CHANGING THE AUDITING LEVEL TO GIVE THE
DESIRED LEVEL OF CONFIDENCE IN THE REPORTED DATA 77
Decision Rule - Accept the Lot as Good If No Defects
Are Found (i.e., d = 0) 77
Decision Rule - Accept the Lot as Good If No More
Than One (1) Defect Is Found (i.e., d <_ 1) 77
MONITORING STRATEGIES AND COST 78
Reference Method (AO) 78
Modified Reference Method (AS) 79
Modified Reference Method Plus Action A2
(A7 = Al + A2 + A4) 79
IV WNAGBENTIWJUAL 81
GENERAL 81
AUDITING SCHEMES 82
Statistics of Various Auditing Schemes 83
Selecting the Auditing Level 87
Cost Relationships 90
Cost Versus Audit Level 92
-------
TABLE OF CONTENTS (Concl'd)
Section Page
DATA QUALITY ASSESSMENT 96
Assessment of Individual Measurements 96
Overall Assessment of Data Quality 99
DATA QUALITY VERSUS COST OF IMPLEMENTING ACTIONS 103
DATA PRESENTATION 107
PERSONNEL REQUIREMENTS 108
Training and Experience 108
OPERATOR PROFICIENCY EVALUATION PROCEDURES 109
REFERENCES 111
APPENDIX - REFERENCE METHOD FOR THE DETERMINATION OF
SULFUR DIOXIDE IN THE ATMOSPHERE (Pararo-
saniline Method) 112
-------
LIST OF FIGURES
Figure No. Page
1 Operational Flow Chart of the Measuring Process 4
2 Schematic of Setup for Calibrating a Rotameter
Against a Wet Test Meter 11
3 Rotameter Calibration Data Sheet 12
4 Typical Rotameter Calibration Curve 13
5 Schematic of Setup for Calibrating a Critical
Orifice Against a Wet Test Meter 16
6 An Absorber (24-Hr Sample) Filled and Assembled
for Shipment 22
7 Sampling Apparatus Assembly 24
8 Sampling Train With Sample Air Filter Installed 25
9 Installation of Pinch Clamp on Vacuum Pump Inlet Line 26
10 Close-up View of Absorber Installation 28
11 Removal of Dummy Absorber From Sampling Train 29
12 Installation of Critical Orifice in the Sampling Train 30
13 Connecting Critical Orifice to Exhaust Manifold 31
14 Sample Record Sheet 33
15 Flow Chart of Quality Control Checks in the
Auditing Program 61
16 Data Qualification Form 65
17 Data Flow Diagram for Auditing Scheme 84
18A Probability of d Defectives in the Sample If the
Lot (N=100) Contains D% Defectives 85
18B Probability of d Defectives in the Sample If the
Lot (N=50) Contains D% Defectives 86
-------
LIST OF FIGURES (Concl'd)
Figure No. Page
19A Percentage of Good Measurements Vs. Sample Size
for No Defectives and Indicated Confidence Level 88
19B Percentage of Good Measurements Vs. Sample Size
for 1 Defective Observed and Indicated Confidence
Level 89
20 Average Cost Vs. Audit Level (Lot Size N-100) 95
21 Critical Values of Ratio s /o Vs. n 100
22 Added Cost ($) Vs. MSE (%) for Alternative Strategies 105
23 Sample QC Chart for Evaluating Operator Efficiency 110
xii
-------
LIST OF TABLES
Table No. Page
1 Apparatus Used in the Manual Pararosaniline Method 57
2 Suggested Performance Standards 67
3 Methods of Monitoring Variables 72
4 Suggested Control Limits for Parameters and/or Variables 73
5 Quality Control Procedures or Actions 75
6 P(d defectives) 83
7 Required Auditing Levels n For Lot Size N=100
Assuming Zero Defectives 87
8 Costs Vs. Data Quality 90
9A Costs if 0 Defectives Are Observed and the Lot is
Rejected 91
9B Costs if 0 Defectives Are Observed and the Lot is
Accepted 91
10 Costs in Dollars 92
11 Overall Average Costs for One Acceptance-Rejection Scheme 93
12 Critical Values of s /a 99
13 Assumed Standard Deviations for Alternative Strategies 106
-------
ABSTRACT
Guidelines for the quality control of Federal reference
method for sulfur dioxide are presented. These
include:
1. Good operating practices
2. Directions on how to assess data and qualify
data
3. Direction on how to identify trouble and
improve data quality
4. Directions to permit design of auditing
activities
5. Procedures which can be used to select action
options and relate them to costs.
The document is not a research report. It is designed
for use by operating personnel.
This work was submitted in partial fulfillment of
Contract Durham 68-02-0598 by Research Triangle Institute
under the sponsorship of the Environmental Protection
Agency. Work was completed as of July 1973.
XIV
-------
SECTION I IliTRODUCTION
This document presents guidelines for implementing a quality assurance
program for the manual (non-continuous) measurement of atmospheric sulfur
dioxide by the pararcsaniline method.
The objectives of this quality assurance program for the pararosaniline
method of measuring atmospheric sulfur dioxide are to:
(1) provide routine indications for operating purposes
of unsatisfactory performance of personnel and/or
equipment,
(2) provide for prompt detection and correction of
conditions which contribute to the collection of
poor quality data, and
(3) collect and supply information necessary to describe
the quality of the data.
To accomplish the above objectives, a quality assurance program must contain
the following components:
(1) routine training and/or evaluation of operators,
(2) routine monitoring of the variables and/or
parameters which may have a significant effect on
data quality,
(3) development of statements and evidence to qualify
data and detect defects, and
(4) action strategies to increase the level of precision
in the reported data and/or to detect -equipment
defects or degradation and to correct same.
'.»
Implementation of a quality assurance program will result in data that are
more uniform in terms of precision and accuracy. It will enable each
monitoring network to continuously generate data that approach the highest
level of accuracy'attainable with the pararosaniline method.
-------
This document Is divided into four sections or chapters. They are:
Section I, Introduction - The introduction lists the overall
objectives of a quality assurance program and delineates the
program components necessary to accomplish the given objectives.
Section II, Operations Manual - The Operations Manual sets forth
recommended operating procedures, instructions for performing
control checks designed to give an indication or warning that
Invalid or poor quality data are being collected, and instruc-
tions for performing certain special checks for auditing purposes.
Section III, Supervision Manual - The Supervision Manual contains
directions for 1) assessing sulfur dioxide data, 2) collecting
information to detect and/or Identify trouble, 3) applying
quality control procedures to improve data quality, and 4) varying
the auditing or checking level to achieve a desired level of
confidence in the validity of the outgoing data. Also, monitoring
strategies and costs as discussed in Section IV are summarized in
this manual.
Section IV. Management Manual - The Management Manual presents
procedures designed to assist the manager in 1) detecting when
data quality is inadequate, 2) assessing overall data quality,
3) determining the extent of independent auditing to be per-
formed, 4) relating costs of data quality assurance procedures
to a measure of data quality, and 5) selecting from the options
available the alternative(s) which will enable one to meet the
data quality goals by the most cost-effective means. Also,
discussions on data presentation and personnel requirements are
included in this manual.
The scope of this document has been purposely limited to that of a field
document. Additional background information is contained in the final
report under this contract.
-------
SECTION II OPERATIONS WNUAL
GENERAL
This operations manual sets forth recommended operating procedures for
the colorimetric measurement of sulfur dioxide in the atmosphere using the
non-continuous pararosaniline method. Quality control procedures and
checks designed to give an indication or warning that invalid or poor
quality data are being collected are written as part of the operating
procedures and are to be performed by the operator on a routine basis.
In addition, the performance of special quality control procedures and/or
checks as prescribed by the supervisor may be required of the operator
on certain occasions.
The accuracy and/or validity of data obtained from this method depends
upon equipment performance and the proficiency with which the operator
performs his various tasks. This measurement method from reagent
preparation through sample collection, analysis, and data reporting is a
complex operation. Presenting detailed instructions covering all alter-
natives has not been attempted in this document. Rather, general guide-
lines are presented with special emphasis on quality control checks and
decision.rules applicable to known problem areas. The operator should
make himself familiar with the rules and regulations concerning the
Reference Method as written in the Federal Register, Vol. 36, No. 84,
Part II, April 30, 1971 (reproduced as the Appendix of this document for
convenience of reference).
Instructions throughout this document are directed primarily toward a
24-hour sampling period with comments on 30-minute and 1-hour sampling
periods when appropriate. Also, an auditing or checking level of 1 check
out of every 14 sampling periods or once a calendar month, whichever
occurs first, is used. Sampling period durations and auditing levels are
subject to change by the supervisor and/or manager. Such changes would
not alter the basic directions for performing the operation. Also,
certain control limits as given in this manual represent best estimates
for use in the beginning of a quality assurance program and are, therefore,
subject to change as field data are collected.
It is assumed that all apparatus satisfies the reference method specifi-
cations and that the manufacturer's recommendations will be followed when
using a particular item of equipment (e.g., spectrophotometer and wet
test meter).
The sequence of operations to be performed during each sampling period is
given in Figure 1. Certain operations such as reagent preparation and
flow-rate calibration (when a rotameter is used as the flowmeter) are
performed periodically. The remaining operations are performed during each
sampling period. The operations are classified as presampling preparation,
sample collection, sample analysis, and data processing. Each operation or
step in the process is identified by a block. Quality checkpoints in the
measurement process, for which appropriate quality control limits are
•assigned, are represented by blocks enclosed by heavy lines. Other
-------
PRE SAMPLING PREPARATION
1. PREPARE OR OBTAIN OXIDANT-FREE DISTILLED WATER.
PREPARE AND CHECK ABSORBING REAGENT.
2. CALIBRATE FLOWMETER/FLOW CONTROLLER (ROTAHETER/CRITICAL ORIFICE'
AGAINST A STANDARD AS SCHEDULED.
3. SELECT PROPER ABSORBER FOR SAMPLING PERIOD OF INTEREST.
CLEAN AND ASSEMBLE THE ABSORBER.
4. RECORD IDENTIFYING INFORMATION IN THE OPERATIONAL DATA LOG BOOK
AND ON THE ABSORBER TUBE.
5. PREPARE ABSORBER FOR DELIVERY/SHIPMENT TO
THE SAMPLING SITE.
SAMPLE COLLECTION
6. ASSEMBLE SAMPLE COLLECTION SYSTEM IF NECESSARY
INSTALL ABSORBER AND CRITICAL ORIFICE INTO THE
SAMPLING TRAIN.
7. PERFORM SYSTEM OPERATIONAL CHECKS AND RECORD ON
THE SAMPLE RECORD SHEET.
8. AMBIENT AIR IS SAMPLED AT A FIXED FLOW RATE FOR THE
DURATION OF THE SAMPLING PERIOD.
9. REMOVE THE EXPOSED ABSORBER FROM THE SAMPLING TRAIN,
PREPARE THE SAMPLE FOR DELIVERY/SHIPMENT.
OBSERVE AND RECORD ANY LOCAL CONDITIONS WHICH MIGHT
AFFECT THE POLLUTION LEVEL.
SAMPLE ANALYSIS
10. CHECK ALL DOCUMENTATION UP TO THIS POINT FOR
ACCURACY AND COMPLETENESS.
11. RECALIBRATE THE CRITICAL ORIFICE, IF USED, AS IN STEP 2.
RECORD THE FINAL FLOW RATE IN THE FLOW-RATE CALIBRATION
LOG BOOK.
12. PREPARE THE REAGENTS USED IN THE ANALYSIS OF
S02 SAMPLES.
13. DEVELOP A CALIBRATION CURVE USING EITHER THE
SULFITE OR PERMEATION TUBE METHOD.
14. MEASURE THE ABSORBANCE OF THE SAMPLE SOLUTION AT 548 nm.
DATA PROCESSING
15. PERFORM NECESSARY CALCULATIONS TO ARRIVE AT AN AVERAGE
CONCENTRATION OF SO, IN M9/m3 AT 25°C AND 760 ra*g FOR
THE SAMPLING PERIDOT
16. RECORD CONCENTRATIONS AND IDENTIFYING DATA ON SAROAD FORM,
ATTACH AUDIT DATA AND FORWARD FOR ADDITIONAL INTERNAL
REVIEW OR TO USER.
RECALIBRATE
CRITICAL
ORIFICE
12
REAGENT
PREPARATION
(ANALYSIS)
13
DEVELOP
CALIBRATION
CURVE
FIGURE 1. OPERATIONAL FLOW CHART OF THE ftAsupwG PROCESS
-------
checkpoints involve go/no-go checks and/or subjective judgments by the
operator with proper guidelines for decision-making spelled out in the
procedures. T/teAe op&uitioni, and c/iecfei ate at4cuA4ed 4equen£uz££{/
one. ptogi.fc64fc4 ^tep by Atep through the. sequence o£ action* in F-tgate J.
The operator is responsible for maintaining certain records. Specifically,
the following log books are maintained:
(1) Operational Data Log Book. Sample identification data,
completed gas sample record sheets, and analysis results
are filed or recorded in the operational data log book.
(2) Flow-Rate Calibration Log Book. Critical orifice or
rotameter calibration data are recorded, dated and signed
in this log book.
(3) SC>2 Calibration Log Book. Completed calibration curves
and control sample measurement data are maintained in the
calibration log book.
PRESAMPLING PREPARATION
Operations included in this classification are (1) preparation of absorbing
reagent, flow-rate calibration, absorber preparation, sample identification,
and preparation of the absorber and flowmeter (critical orifice) for
shipment .
Reagent Preparation (Step 1)
In reagent preparation for sampling and analysis all chemicals should be
ACS analytical reagent grade or better. Unless otherwise indicated,
references to water implies distilled water free from oxidants. Only
Class A volumetric glassware should be used. All weights should be made
on an analytical balance sensitive to the nearest 0.1 mg, or better.
Distilled Water - Water must be free from oxidants. It should preferably
be double-distilled from all glass apparatus. The purity of the water
should be checked when first purchased/prepared and before preparing a
batch of reagents for sampling or analysis.
Tei-t ^OfL pusUty - Distilled water can be tested for purity from oxidants
in the following manner.
(1) Add 0.20 mi of KMnO^ solution (0.316 g/£) to a mixture
of 500 m£ of the distilled water and 1 m£ of H-SO, in a
stoppered bottle of chemically resistant glass.
(2) If the permanganate .color (blue) does not disappear
completely after standing for 1 hour at room temperature,
consider the water suitable for use.
(3) If the permanganate does disappear, the water must be
purified before using.
5
-------
pfioc.e.duAe. _ Water faliing tne purity test can be purified as
follows:
(1) Add one crystal each of potassium permanganate and
barium hydroxide for each liter of distilled water.
(2) Redistill the water in an all-glass still.
(3) Perform the test for purity as described above.
(4) Repeat the purification procedure and the test for
purity until the water checks pure-.
This distilled water, free of oxidants, is used any time water is required
in preparing reagents for sampling or analysis .
Sto/uige. 0& (LLt>tUULejd WOteA - Oxidant free water should at all times be
protected from atmospheric contamination by storing in container made of
material that has been proven to be resistant to solvation by, or reaction
with, the water. Also, any tubing used during reagent preparation should
be of high resistant material. Also, when removing water from the storage
container the replacement air should be drawn through a vent guard (e.g.,
a drying tube filled with equal parts of 8-20-mesh soda lime, oxalic acid,
and 4-8-mesh calcium chloride, each compound being separated from the other
by a glass wool plug) .
Absorbing Reagent - Prepare a batch of absorbing reagent as directed in
Subsection 6.1.2 of the Appendix, page 113. The preparation and subsequent
handling of the absorbing reagent should be carried out under a ventilation
hood.
pH c/iecfe - It is suggested that the pH of the absorbing reagent be
checked with a pH meter and a glass electrode standardized in a buffer
solution with a pH of about 4.5. If the weighings and volumetric measure-
ments have been performed accurately and the chemicals and distilled water
are of the specified purity, the absorbing reagent will have a pH of approxi-
mately 4.0. The reagent is acceptable for use if the pH is between 3 and 5.
However, -a significant deviation from a pH of 4.0 (e.g., pH < 3.5 or > 4.5)
although acceptable, would tend to indicate poor laboratory procedure or
impure chemicals and should serve as a warning to exercise more care in
preparing the next batch. In the event that the pH is outside the 3 to 5
range, discard the reage'nt and check the chemicals and procedure before
preparing another batch.
cAecfe - As a check on the collection efficiency of
a new batch of absorbing reagent, it is suggested that dual samples be
collected and analyzed using two absorbers, one filled with absorbing
reagent from the old batch and the other filled with the new reagent.
-------
Compute the difference in the two measured concentrations by
d = (yg S0,/m3) - (yg S0,/m3)
*• n L o
where d = the difference in yg SO-An ,
o
(yg S00/m ) • = SO. concentration measured with new reagent,
' n L
o
and (yg S09/m ) = SO. concentration measured with old reagent.
* o *
o o
If d is less than or equal to [44 + .063 (yg S0?/m ) ] yg S0?/m , the new
o
reagent is accepted as no different than the old reagent (at the 3o level) .
If d is greater than the above value and there is no reason to suspect the
old absorbing reagent of having deteriorated, or an analysis error, the
new batch should be discarded and additional absorbing reagent prepared.
If a calibrated S0_ permeation tube is used for developing a calibration
curve (see Subsection 8.2.2 of the Appendix), the results include the
correction for collection efficiency and the above check need not be made.
/i&agent - Absorbing reagent can be stored in a
stoppered container. It is normally stable for 6 months. It should be
checked visually before each use and discarded if a precipitate has formed.
Flow Rate Calibration (Step 2)
3 3
Flow rates in the range of about 900 to 1100 cm /min, 450 to 550 cm /min,
o
and 180-220 cm /min are used for 30-minute, 1-hour, and 24-hour sampling
periods, respectively. It is recommended that rotameters or critical orifices
used for the 30-minute and 1-hour sampling periods be calibrated against a
wet test meter or a soap-bubble meter. Flowmeters used for 24-hour sampling
should be calibrated against a soap-bubble meter for best results.
It should be noted that the above flow rate ranges are only approximate.
However, as long as these ranges are observed the flow rate will not be so
great as to (1) result in entrainment of the absorbing reagent, (2) cause
excessive evaporation of the absorbing reagent (e.g., more than 10 to 20 per-
cent of the total quantity), or (3) collect large quantities of S02 necessi-
tating the dilution of samples before analysis. The lower limit will in
general result in a sample volume of air large enough to ensure the
collection of a quantity of SO- in the analytical range.
Rotameters should be removed from the system and cleaned every six months,
or after 30 days of operation, whichever occurs first. It is recommended
that rotameters be calibrated after being cleaned, at any signs of erratic
behavior, or any time the flow rate measured by the rotameter differs by
-------
as much as 4 percent from the flow rate measured by a calibrated flowmeter.
Calibrated rotameter as used here refers to a rotameter that has been cali-
brated against a wet test meter or a soap-bubble meter.
Critical orifices are calibrated before and after each sampling period.
Rotameter Calibration - A rotameter is calibrated within an apparatus
similar to the field sampling collection device. The absorber assembly
should conform to the same specifications as the field collection device,
and contain a volume of distilled water or absorbing reagent equal to the
volume of absorbing reagent used in the field device.
CaLUotuJL&Lon o^ a. fLotam&t&L agcuLn&t a. w&t tut mete/t - Figure 2 shows a
typical setup for calibrating a rotameter against a wet test meter. The
calibration procedure is as follows:
(1) Set up the apparatus as shown in Figure 2 making the
connections as short as possible and of the same inside
diameter as is used in the field sampling train.
(2) Start the air flowing through the system and allow to
flow for 5 to 10 minutes to allow the water in the wet
test meter to reach saturation with the air.
(3) Before and after the complete calibration run, read and
record room temperature and barometric pressure. Record
average values on the calibration form in Figure 3. Use
average values for subsequent calculations.
(4) Adjust the flow rate to about 20 percent of full scale
for the rotameter with the needle valve.
(5) Take a pair of timed readings on the wet test meter (for
best results use complete revolutions of the wet test
meter), under steady flow, for each of five or more
uniformly spaced points on the rotameter scale, going
from low values to high values. Repeat, going from high
to low. Record rotameter reading, total flow, elapsed
time of run, manometer reading at the wet test meter,
manometer reading at the rotamter, and T (temperature of
the liquid in the wet test meter) for each run on the
calibration sheet of Figure 3.
(6) Convert all temperature and pressure readings to absolute
units.
°C + 273 = °K
in H20 x 1.87 = mmHg.
-------
(7) For each run correct the volume measured by the wet test
meter to the volume measured by the rotameter by
Vr " \T /\P / " m'
where V = volume of air passing through the rotameter at
r T and P ,
r r'
T = temperature of air entering rotameter (taken as
the temperature of the liquid in the absorber),
P = barometric pressure minus the reading of the
manometer just upstream of the rotameter.
V = volume measured by the wet test meter,
m
P = barometric pressure plus the reading of the
m manometer on the wet test meter,
and T = temperature of the liquid in the wet test meter.
(8) For each V calculate the corresponding flow rate by
dividing the corrected volume, V , by the elapsed time,
t, in minutes for that run as
Qr = Vr/t .
(9) For each run use a general form of the rotameter equation to
calculate
" • (<>,") ftf2
where I = rotameter reading (considered dimensionless here)
3
Q = flow rate from procedure 8 above (cm /min),
P. = pressure at rotameter (barometric pressure minus
manometer reading)(mmHg),
T = temperature of air at rotameter (°K), and
K = determinable rotameter equation constant with
appropriate units to make the equation
dimensionally consistent.
9
-------
(10) Plot on Cartesian Coordinates K.(y-axis) versus I (x-axis)
for all runs. Use regression analysis or by eye construct
a best-fit curve to the data points. The curve should
approximate a straight line with zero slope. Figure 4 is
a hypothetical rotameter calibration curve. Mark the values
of T and P on the graph.
(11) Check all plotted points and rerun any that deviates more
than + 5 percent from the best-fit curve for a fixed rota-
meter indication. Compute the percent deviation by
K -K
percent deviation = —- x 100
K2
where K- = the K value of the plotted point, and
K2 = the K value read from the best-fit curve for
the same rotameter reading as for KI.
(13) Forward calibration curve and data to supervisor for his
approval. File the approved data and curve in the cali-
bration log book.
a. /wtamvteA agcun&t a. 4oap-faubfa£e me^eA - Calibrating a
rotameter against a soap-bubble meter can be accomplished with the same type
apparatus setup as shown in Figure 2 with a soap-bubble meter replacing the
wet test meter. The calibration procedure is as follows:
(1) Connect the apparatus as shown in Figure 2 with the soap-
bubble meter replacing the wet test meter. Use apparatus
having the same specifications as the actual sampling
train.
(2) Either immediately before or after the calibration,
measure and record on the calibration form in Figure 3
the ambient temperature and barometric pressure.
(3) Adjust the flow rate to about 20 percent of full scale
for the rotameter with the needle valve.
(4) Take a pair of timed readings on the soap-bubble meter
under steady flow, for each of five or more uniformly
spaced points on the rotameter scale, going from low
values to high values. Repeat, going from high to low.
Record rotameter reading, manometer reading at rotameter,
total flow and elapsed time for each run on the
calibration sheet of Figure 3.
Note: In Figure 3 the column for manometer for the wet
test meter reading is not used when calibrating
against a soap-bubble meter; also, T becomes the
same as room temperature.
10
-------
MANOMETER || IPI
ROOM_
AIR
TO
VACUUM
PUMP
IM FINGER
TRAP
Figure 2: Schematic of Setup for Calibrating a Rotameter
Against a Wet Test Meter
-------
ROTAMETER SERIAL N0._
LOCATION
CALIBRATED WITH
AMBIENT/ROOM TEMPERATURE (Average)
ATMOSPHERIC PRESSURE (Average).
CALIBRATED BY
RELATIVE HUMIDITY_
DATE
Test
Point
1
2
3
4
5
6
7
8
9
10
Rotameter
Readi ng
Total
Flow
Elapsed
Time
Manometer Reading
Wet Test
Meter
Rotameter
» (W'Av
nOW'v»
Qr = Vr/t
K=(Qr/I)(Pr/Tr)1/2
Figure 3: Rotameter Calibration Data Sheet
-------
FLOWMETER NO..
LOCATION
u
UJ
5
4
I
TEMPERATURE AT CALIBRATION (°K)
ATMOSPHERIC PRESSURE AT CALIBRATION (mmHg)_
CALIBRATED BY
I
I
I
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
ROTAMETER READING (ARBITRARY UNITS)
Figure 4: Hypothetical Rotameter Calibration Curve
13
-------
(5) Convert all temperature and pressure readings to absolute
units.
°C + 273 = °K
in H20 x 1.87 = mmHg.
(6) For each run, correct the flow rate measured by the soap-
bubble meter to the flow rate measured by the rotameter by
where Q = flow rate through the rotameter at T and P ,
3 r r
cm /min.
T = temperature of air entering rotameter (taken
as the temperature of the liquid in the absorber) ,
P = barometric pressure minus the reading of the
manometer just upstream of the rotameter, mmHg,
Q = flow rate measured by the soap-bubble meter, cm /min,
P = barometric pressure, mmHg,
and T = room temperature, °K.
m
(7) For each run use a general form of the rotameter equation to
calculate
1/2
K- (Q/D l-
r
r
where I = rotameter reading (considered dimensionless here) ,
3
Q = flow rate from procedure 6 above (cm /min) ,
P = pressure at rotameter (mmHg) ,
T = temperature at rotameter (°K),
and K = determinable rotameter equation constant with appropri-
ate units to make the equation dimensionally consistent.
14
-------
(8) Plot on Cartesian coordinates K (y-axis) versus I (x-
axis) for all runs. Use regression analysis or by eye
construct a best-fit curve to the data points. The
curve should approximate a straight line with zero slope.
Figure A is a hypothetical rotameter calibration curve.
(9) Check all plotted points and rerun any that deviates more
than + 5 percent from the best-fit curve for a fixed
rotameter indication. Compute the percent deviation by
K - K
percent deviation = x 100
K2
where K. = the K value of the plotted point,
and K_ = the K value read from the best-fit curve
for the same rotameter reading as for K .
(10) Forward calibration curve and data to supervisor for his
approval. File the approved data and curve in the
calibration log book.1
Calibration of Critical Orifice - A critical orifice is calibrated within
an apparatus similar to the sampling train used in the field. That is, the
absorber assembly and tubing should conform to the same specifications as
the ones used when sampling. It is recommended that either a soap-bubble
meter or a calibrated rotameter be used to calibrate critical orifices for
3
24-hour sampling where the flow rate is around 200 cm /min. A wet test
meter, soap-bubble meter, or calibrated rotameter can be used for calibrating
the orifices for 30-minute and 1-hour sampling.
The calibration procedure is written in terms of a wet test meter with
special notes when the procedure would be different if a soap-bubble meter
or rotameter were used. The procedure is ,as follows:
(1) Connect the apparatus as shown in Figure 5.
(2) Turn the vacuum pump ON and check the sampling train for
leaks by slightly adjusting each connection while watching
the manometer (the one next to the critical orifice) for
any fluctuation in pressure. Experience will indicate
approximate pressure drops for the different sampling trains.
Too small a pressure drop indicates leaks in the system and
too high pressure drops indicates a restriction in the
sample lines.
Eliminate all leaks or restrictions before continuing.
15
-------
THERMOMETER
MANOMETER
ROOM_
AIR
^
MANOMETER
HYPODERMIC
NEEDLE
RUBBER
SEPTUM
TO
-VACUUM
PUMP
TRAP
Figure 5: Schematic of Setup for Calibrating a Critical Orifice
Against a Wet Test Meter
-------
(3) To assure that the orifice is critical, read the vacuum
gauge on the vacuum pump (see Figure 7; the vacuum gauge
is mounted on top of the pump and connected to the pump's
inlet port) and compare that reading to the pressure
immediately upstream of the critical orifice. This upstream
pressure is obtained by subtracting the reading of the mano-
meter directly upstream of the critical orifice from
barometric pressure. The orifice is critical if
P > 0.55 P
v — u
where P = reading of vacuum gauge (mmHg),
and P = barometric pressure minus manometer
reading (mmHg).
See Reference 1 for a more detailed discussion of critical flow.
For the 24-hour sampling train the upstream pressure is usually
on the order of 735 mmHg (-29 in Hg) therefore, the vacuum
gauge reading must be greater than 405 mmHg (> 16 in Hg) for the
orifice to be critical. Normal practice is to use a pump that
will maintain a vacuum greater than 507 mmHg (20 in Hg).
If sufficient vacuum- cannot be obtained, the system probably has
leaks or the vacuum pump needs replacing.
(4) Measure and record on the calibration record sheet; (1) ambient/
room temperature, (2) barometric pressure, and (3) relative
humidity.
(5) A. Make at least three timed runs when using a wet test meter.
Time at least one complete revolution of the wet test meter
per run.
Record the total flow (volume) measured by the wet test
meter, elapsed time of run, and both manometer readings
for each run.
B. If a soap-bubble meter is used, make at least three timed
runs. Record the elapsed time, and the manometer (for the
critical orifice) reading for each run.
C. When using a calibrated rotameter make three readings by
shutting the pump OFF momentarily between readings. Record
the rotameter reading and the manometer (for the critical
orifice) reading for each run.
17
-------
(6) Correct the indicated volume/flow rate to the temperature
and pressure at the orifice by:
A. For a wet test meter use
for each of the three runs;
where V = volume of air passing through the critical
r 3
orifice at T and P , cm ,
T = temperature of air entering the orifice (taken
as the temperature of the liquid in the absor-
ber which should be room temperature), °K,
P = barometric pressure minus the reading of the
manometer just upstream of the orifice, mmHg,
V = volume of air measured by the wet test meter, cm ,
m
P = barometric pressure plus the reading of the
manometer on the wet test meter,
and T = temperature of the liquid in the wet test meter, °K.
Note: Under this setup the difference in V and V should be
fairly small since T = T and P differs f?om P by
about 25 mmHg or less.
Compute the critical volumetric flow rate of the orifice at T
and P for each of the runs by
Qr - Vr/t
where t = the elapsed time of the run in minutes.
Compute the average, Q , from the three runs and use this value
in subsequent calculations.
18
-------
B. When using a soap-bubble meter convert the measured
volumetric flow-rate to the volumetric flow rate at
the orifice for all three runs by
where Q = flow rate as would be measured by the
soap-bubble meter at orifice conditions
of T and P , cm^/min,
r r
T = temperature of air upstream of the
orifice (taken as temperature of the
absorbing reagent which should be room
temperature, CK,
P .= barometric pressure minus reading of the
manometer upstream from orifice, mmHg,
Q = flow rate measured by soap-bubble meter,
m 3
cm /min,
/
P = barometric pressure, mmHg,
and T = room temperature, °K«
m
Note: The difference in Q and Q_ should be fairly small
r TD
when using the setup of Figure 5 since T = T
r m
and P differs from P by approximately 25 mmHg.
r m
C. When a rotameter is used as the standard, it should -itself
be calibrated against a high grade soap-bubble meter as
discussed in Step (2), starting on page 10. Proceed as
follows:
(1) Average the three rotameter readings to get I.
From the calibration curve for the rotameter
determine the value of K corresponding to a
rotameter reading of I.
(2) Compute the measured flow rate by
19
-------
(3) Convert the measured flow rate to the flow rate
at orifice conditions of T and P by
pr yI/2
K PrJ
where Q = flow rate at the orifice conditions
r 3
of T and P , cm /min,
T = temperature of air upstream of orifice
(taken of absorbing reagent or room
temperature), °K,
P = barometric pressure minus reading of
the manometer upstream of critical
orifice, mmHg,
Q_ = flow rate measured by rotameter at
3
conditions T and P , cm /min,
m m
T = ambient room temperature, °K,
m
and P = barometric pressure, mmHg.
m
(7) Once Q has been determined by either of the above three methods,
calculate
K = Qr/(Tr)1/2
(8) Record the K value and date of calibration in the calibration log
book for that critical orifice.
A critical orifice should be discarded at any time its original K value
cannot be duplicated to within + 2 percent after having been cleaned.
Absorber Preparation (Step 3)
Selection of Absorbers - Select absorbers appropriate for the sampling
period to be used according to the specifications given in Section 5 of the
Appendix, page 113.
Cleaning Absorber Assembly - Before each use it is recommended that the
absorber assembly be cleaned by washing with hot water and placed in an acid
bath consisting of 1 part nitric acid, 2 parts hydrochloric acid, and 4 parts
distilled water for 1 hour. Then rinse thoroughly with distilled water and
allow to dry.
20
-------
Dispensing of TCM (Absorbing Solution) Into Absorbers - Transfer quantities
of TCM, appropriate for the sampling period to be used, as specified in
Section 7 of the Appendix into the absorber.
Note: If the situation requires shipping apparatus from the central
laboratory to the field site and vice versa, it is recommended
that for 30-minute and 1-hour samples and all-glass midget
impinger remain on site and the TCM be shipped in test tubes
with screw-on caps using a teflon or equivalent inner liner on
the caps. The TCM would then be quantitatively transferred
to the impinger by the field operator.
For 24-hour sampling the use of a set-volume reagent dispenser, such as a
50 mil repipet, for dispensing absorbing reagent into the absorbers is fast
and reduces the chance of a volume error.
Assemble the absorber as shown in the drawing of Figure 6. Two tubing
caps are fitted on the tube ends to guard against contaminating the
absorbing reagent. Shrinkable tape can be used around the tube and closure
to add mechanical stability and discourage tampering. Alternatively, a
threaded absorber tube and screw-on cap could be used. To guard against
accidentally forcing absorbing reagent out of the absorber, it is important
to always install the tube cap on the impinger tube first. Color coding
the tube cap and tube would aid field personnel in identifying the cap and
act as a reminder to install that cap first.
On the first filling and after assembly, the absorber should be scribed or
otherwise permanently marked at the 50 mi level. Glass impingers for
30-minute and 1-hour sampling come from the factory with a permanent scale.
Sample Identification (Step 4)
Each absorber should have the following identifying information semi-
permanently affixed by a stick-on label or marked directly on the absorber
with a felt-tipped pen:
(1) date absorber was prepared,
(2) sampling site number, and
(3) date to be used.
Obtain a calibrated critical orifice, if used in the system, and attach to
the absorber (see Step 5 below). Record the absorber identifying infor-
mation, orifice number and date to be shipped in the laboratory log book.
Package for Shipment (Step 5)
In situations where shipping by mail is required, a container like or
equivalent to that used by EPA in the NASN program should be used. The
container is a block of wood with drilled holes of the proper sizes for
holding the absorbers and critical orifices. This type container remains
serviceable for a long period of time.
21
-------
Polypropylene
2-Port Tube
Closure
Glass
Impinger
Polypropylene
Tube
Tube Cops
Heat Shrink
Tope
Etched 50 ml.
Mark
Absorbing
Reagent (TCM)
Figure 6: An Absorber (24-Hr Sample) Filled and Assembled
For Shipaent
22
-------
SAMPLE COLLECTION
Sampling Apparatus Assembly (Step 6)
A typical sampling system consists of a collection unit, vacuum pump, and
a 7-day timer as shown in Figure 7. The collection unit, in this case, is
a metal box which houses the sampling train. .The sampling train is
maintained at a near constant temperature with a thermostatically controlled
heater.
Sampler Assembly - General guidelines are given for assembling a sampling
system used by EPA in the NASN program as shown in Figure 7. Assemble as
follows:
(1) Attach "the membrane sample air filter unit to the glass
inlet manifold. The connecting tubing passes through a
cutaway in the sampler box. Support for the filter is
provided by the tubing resting on the cutaway. Care
should be exercised when installing or changing the
filter so as not to break the glass manifold. Figure 8
is a picture of a sampling train with the filter
installed.
(2) Attach the sample inlet line consisting of polypropylene
tubing and funnel to the sample air filter as shown in
Figure 7. Extend the funnel and sample inlet line out-
of-doors through a window or other opening. Support and
secure the tube and funnel in position so that they will
not come loose during sampling. The funnel should be
positioned to hang down so that rain will not be drawn
into the sampler.
(3) Connect the metal exhaust manifold to the intake of the
vacuum pump with tygon tubing. Place a pinch clamp on
this section of tubing as shown in Figure 9, but do not
tighten. Be sure all connections are airtight and that
there are no constrictions in the tubing.
(4) Electrically connect the vacuum pump to the timer (see
Figure 7) and the timer to a 24-hour 110 VAC outlet. The
sampler box is also connected to a 24-hour outlet.
Installation of Absorber and Critical Orifice - Correct installation of the
reagent-filled absorber and calibrated orifice is vital to the collection
of a valid sample. Color-coding schemes for the absorbers and tubing are
frequently used to assist in locating absorbers in the right positions and
for making correct connections. One such scheme is used in the following
procedure. A close-up view of the absorber installation is given in
23
-------
Figure 7: Sampling Apparatus Assembly
-------
Figure 8: Sampling Train With Sample Air Filter Installed
25
-------
Figure 9: Installation of Pinch Clamp on Vacuum Pump Inlet Line
-------
Figure 10. In this setup the top of the absorber around the impinger tube
and the tube cap on the Impinger tube are color-coded. An accordian type
of tubing Is used to connect the glass intake manifold to the Impinger
tube. Smooth tubing Is used for all other connections.
The Installation procedure Is as follows:
(1) Remove the dummy absorber (or used absorber) from the
sampling train by gently but firmly pulling the Inlet
and outlet tubing from the absorber as shown In
Figure 11.
(2) Exchange the reagent-filled absorber for the dummy
absorber.
(3) Transfer the caps from the reagent-filled absorber to
the dummy absorber. Caution: Always install the tube
cap or tubing on the impinger tube (color-coded side of
absorber) first and remove them last, otherwise some of
the absorbing reagent/sample may be forced out the
impinger opening and be lost.
(4) Gently but firmly insert the accordian inlet tube (from
the glass manifold) onto the color-coded tube (impinger
tube) of the absorber lid. Make certain the fit is
airtight.
(5) Gently but firmly insert the smooth tubing (from the
trap) to the unpainted outlet tube of the absorber.
(6) Remove the critical orifice (calibrated hypodermic needle)
from the shipping container. Note: It is important that
the needle be used and returned with the absorber that it
was received with.
For systems using a calibrated rotameter to measure flow-
rate the rotameter remains in the sampling train until it
is removed for cleaning and calibration.
(7) Insert the new needle into the center of the rubber stopper
attached to the membrane filter unit as shown in.Figure 12.
The needle must be put in straight. If the needle is
accidently bent during installation, it should be discarded
and another absorber and needle used.
(8) Slide the base of the needle onto the metal exhaust manifold
as shown in Figure 13. Manipulate the needle and/or rotate
the trap, if necessary, to obtain a tight connection.
27
-------
S3
oe
Figure 10: Close-up View of Absorber Installation
-------
Figure 11: Removal of Dummy Absorber From Sampling Train
29
-------
Figure 12: Installation of Critical Orifice in the Sampling Train
-------
Figure 13: Connecting Critical Orifice to Exhaust Manifold
-------
(9) Recheck the arrangement, alignment, and tightness of all
connections. The following points should be verified:
(a) The accordian tubing from the glass manifold goes
to the color-coded side of the absorber lid;
(b) the needle is not bent or obstructed;
(c) the needle forms a tight fit at the exhaust manifold;
(d) the inlet filter and tubing is tightly connected, and
(e) the vacuum pump connections are tight.
Note: In situations where the sampling train is not housed in a closed box
the absorber must be shielded from sunlight during and after sampling. One
means of accomplishing this is to wrap the absorber in aluminum foil.
Operational Check of System (Step 7)
(1) If a critical orifice is used in the system completely
close the pinch clamp between the sampler and the
vacuum pump (see Figure 9).
(2) Turn the timer switch to the ON position. This should
start the vacuum' pump.
(3) For systems using a critical orifice, record the vacuum
gauge reading to the nearest whole number on the Sample
Record Sheet of Figure 14, in the space marked "Start-Clamp."
The gauge should read above 508 mmHg (20 in Hg) vacuum. If
it does not, make sure the pinch clamp is closed and the
tubing is securely connected to the pump inlet. If the
vacuum reading remains below 508 mmHg (20 in Hg) the pump
should be replaced. Continue to (4) below.
If a rotameter is used check and if necessary adjust the
system flow rate to the prescribed value (e.g., 200 cm3/min
(.2 £/min) for a 24-hour sample) and record the value on the
Sample Record Sheet under "Remarks." Continue with (5)
below.
(4) Open the pinch clamp. Record the vacuum gauge reading on the
Sample Record Sheet in the space marked "Start-Open." The
gauge should read a little less than the reading with the
clamp closed. If it reads below 508 mmHg (20 in Hg), check
for a loose connection at the pump.
32
-------
SAMPLER SERIAL NUMBER
SITE IDENTIFICATION: CITY OR TOWN
SHIPPING PACKAGE NUMBER
, STATE , SAMPLER LOCATION
WIND
DIRECTION
[ ] CALM
[ ] LIGHT
[ ] GUSTY
DATE
[ ] 0000 TO 2400
HOURS
[ ] OTHER EXPLAIN
VISIBILITY
[ ] CLEAR
[ ] HAZY
METER READING
START
END
OPEN
CLAMP
HUMIDITY
[ ] CLEAR
[ ] SCATTERED
[ ] OVERCAST
AVERAGE
TEMP (°C)
AVERAGE
PRESSURE (mmHq)
REMARKS & UNUSUAL CONDITIONS OR ACTIVITIES NEAR THE SITE
(jj
Figure 14: Sample Record Sheet
-------
(5) Record on the Sample Record Sheet the date that the sample
was collected (one day only), and the time (if other than
0000-2400 explain under "Remarks").
(6) Gently lift the sampling train halfway out of the sampler
box (do not tilt the train) as in Figure 8. Examine the
absorber to make certain that it is bubbling. If not,
check for loose connections or a plugged line in the train
and correct.
(7) Turn the timer OFF and set for sampling period.
Sample Collection (Step 8)
The sampling period is started and stopped by the timer. The absorbing
reagent should be protected from direct sunlight during and after
sampling by covering the absorber with aluminum foil if it is to remain
outside the sampler or shipping block for any period of time. At the end
of the sampling period, perform the following operations:
(1) With the timer switch ON, record the vacuum gauge reading
on the Sample Record Sheet in the space marked "End-Open"
if a critical orifice is being used and continue to (2)
below. For systems using a rotameter read and record the
final flow rate, turn the timer switch to OFF, and continue
with (4) below.
(2) Close the pinch clamp as tight as possible. Record the
vacuum gauge reading on the Sample Record Sheet in the
space marked "End-Clamp."
(3) Turn the timer switch to the OFF position. Open the
pinch clamp.
(4) Check the condition of the membrane filter. Replace if
it is discolored or cracked.
(5) If the ambient temperature is below the thermostat setting
in the sampling box, check to see that the thermostat and
heater are working.
(6) If possible, obtain an average temperature of the absorbing
reagent for the sampling period. For 30-minute and 1-hour
sampling periods a temperature reading immediately before,
after or any time during the period could be used as an
average for the sampling period.
34
-------
For 24-hour sampling periods two approaches are discussed
for arriving at an estimate of the average temperature.
(a) Place a minimum maximum thermometer in the thermostated
box used to house the sampling train. Report the
average of the minimum and maximum temperatures for
that sampling period. The thermometer has to be reset
before each sampling period.
(b) If a thermometer is not available for (a) above, the
average temperature can be reported as the normal
thermostatic setting of the box or ambient temperature,
whichever is greater.
(7) Determine and report the barometric pressure by maintaining
a barometer on-site, or in the near vicinity. Read and
report the barometric pressure each sampling period. Alter-
natively, the pressure can be obtained from the nearest"
weather station provided the altitudes of the weather station
and sampling site do not differ by more than about 61 m
(-200 ft).
The barometric pressure seldom varies more than 2 or 3 per-
cent from a mean value for a given site. Therefore, unless
a high degree of accuracy is required, an average pressure
derived from 20 to 30 days of data can be used as the
barometric pressure for that site. Also, barometric pressure
is not required when a critical orifice is used to control
the flow rate.
Sampling Handling (Step 9)
(1) Remove the exposed absorber from the sampling train. Remove
the accordian tubing (from the galss manifold) from the
absorber first. Check the absorber and if any of the reagent
has evaporated, use fresh absorbing reagent to bring it up
to the 50 ml mark (24-hour sample). Record on the Sample
Record Sheet the quantity of absorbing reagent (mX,) or state
under "Remarks" that there was no evaporation. If the absor-
ber is estimated to have less than 35 mfc the sample should be
invalidated by so stating on the record sheet. The absorber
and record sheet should be forwarded to the supervisor. If
there are signs that indicate that part of the absorbing
reagent was forced out of the absorber into the system,
absorbing reagent should not be added and the situation noted
on the Sample Record Sheet.
(2) Replace the tube caps on the exposed absorber. Be sure to
place the cap over the impinger tube opening near the painted
side of the absorber lid first. Press both caps on firmly
and place the absorber in the shipping block. Note: where
glass impingers are used replace any evaporated absorbing
reagent, shake thoroughly, then transfer the exposed reagent
to a test tube with a teflon-lined, threaded cap. Place the
test tube in the shipping block.
35
-------
(3) If a critical orifice was used, disconnect the base end
of the hypodermic needle from the metal exhaust manifold
and remove the needle from the rubber stopper and place
the needle in the shipping block with the exposed
absorber. Note: The needles must be returned with the
absorber that it was used with.
(4) Record on the Sample Record Sheet any unusual activities
or conditions near the site such as fires involving
burning coal or oil, large coal burning power plants,
smoking stacks, rain, snow, fog, inversions, or other
conditions that may affect the pollution level.
(5) Fill out the Sample Record Sheet in duplicate. Fold the
original copy and wrap around the absorber in the shipping
block. File the duplicate copy in the site log book.
(6) Deliver the sample to the laboratory or pack the absorber
(and critical orifice) in a mailing container, affix the
return mailing label, and mail promptly.
SAMPLE ANALYSIS
Verify Documentation (Step 10)
Check the Sample Record Sheet for any missing information that would
invalidate the sample. If sufficient information is not available and
cannot be obtained from the field personnel, the sample should be invali-
dated at this point. Visually check the absorber to see if it has 50 m£
of reagent. If it does not have 50 m£, check for signs of leaks and check
the Sample Record Sheet for comments by the field operator. If the
absorber is low and the field operator's entry on the Sample Record Sheet
says that there was 50 m£ when mailed, continue the analysis. If the
absorber is visually low and there are no consents by the field operator
on the Sample Record Sheet, invalidate the sample. Also, any 24-hour
sample that required more than 15 m£ of TCM to be added should be
invalidated.
Recalibrate the Critical Orifice (Step 11)
Recalibrate the critical orifice in the same manner that it was calibrated
originally in Step 2. Record the new K value in the calibration log book.
If the final K value differs from the initial K value by more than 10 per-
cent, the sample should be invalidated. Compute the percent deviation by
K - K
percent difference = - * 100
Ki
where K± = the initial K value (cm3/min)/(°K)1/'2
and Kf = the final K value (cm3/min)/(°K)1/2.
36
-------
If the final K value is within 10 percent of the initial K value, compute
the average flow rate for the sampling period by:
•, /-,
_ o
where Q = the average flow rate (cm /rain) ,
KA and Kf = initial and final calibration K values,
respectively,
and T, = average temperature reported by field
operator, °K.
Use Q in subsequent calculations.
Reagent Preparation For Analysis (Step 12)
Prepare reagents for analysis according to the directions given in
Subsection 6.2 of the Appendix. Class A volumetric glassware should be
used. The analytical balance should be checked before preparing a batch
of reagents by weighing a standard weight between 1 and 3 grams. If the
measured and actual weights agree within +0.4 mg proceed with the
preparation. Record the actual and measured weights in the laboratory
log book. If the weights differ by more than + 0.4 mg, continue the
preparation but report to the supervisor that the balance calibration needs
checking. This should be carried out by a manufacturer's representative.
Sulfamic Acid (0.6 percent) - Dissolve 0.6 g sulfamic acid in a 100 mi
volumetric flask with water, and bring to mark. Keep in a glass-stoppered
flask while not in use. Prepare fresh daily.
Formaldehyde (0.2 percent) - Dilute 5 mi formaldehyde solution (36-38 per-
cent) to 1,000 ml with distilled water. Keep in a stoppered container
while not in use. Prepare fresh daily.
Stock Iodine Solution (0.1 N) - Place 12.7 g iodine in a 250 mi beaker; add
40 g potassium iodide and 25 mi of water. Stir until all is dissolved,
then transfer to a 1,000 mi flask and dilute to the mark with distilled
water. Keep the solution in a glass-stoppered, dark, bottle and store in
a cool place.
Working Iodine Solution ^0.01 N) - Prepare approximately 0.01 N iodine
solution by diluting 50 mi of stock solution to 500 mi with distilled water.
Keep in a glass-stoppered, dark, bottle or flask. It is recommended that
this solution be prepared fresh from stock daily.
Starch Indicator Solution - Triturate 0.4 g soluble starch and 0.002 g
mercuric iodide (preservative) with a little water, and add paste slowly
to 200 mJl boiling water. Continue boiling until solution is clear; cool,
and transfer to a glass-stoppered bottle. Alternatively a solution of
stabilized starch for volumetric determinations can be purchased
commercially. 07
-------
Stock Sodium Thiosulfate Solution (0.1 N) - Prepare a stock solution by
dissolving 25 g sodium thiosulfate (Na.S.O • 5H.O) in 1,000 mA freshly
boiled, cooled, distilled water and add 0.1 g sodium carbonate to the
solution. Allow the solution to stand 1 day before standardizing. To
standardize, accurately weigh to the nearest 0.1 mg, 1.5 g primary standard
potassium iodate (KIO,) dried at 180° C for 1 hour and cooled in a dessicator.
Dilute to volume in a 500 m£ volumetric flask. To a 500 m£ iodine flask,
pipet 50 m£ of iodate solution, add 2 g potassium iodide and 10 mil of 1 N
hydrochloric acid. Stopper the flask. After 5 minute titrate with stock
thiosulfate solution to a pale yellow. Add 5 mJl starch indicator solution
and continue the titration until the blue color disappears. Calculate the
normality of the stock solution as follows:
N - x 2.80
M
where N - normality of stock thiosulfate solution,
M » volume of thiosulfate required, m£,
W - weight of potassium iodate, grams,
3
j 0 80 10 (conversion of g to mg) x Q.l (fraction iodate used)
35.67 (equivalent weight of potassium iodate)
Keep this solution stored in a glass-stoppered bottle or flask.
This solution should not be used without being restandardized if stored for
more than one month.
Sodium Thiosulfate Titrant (0.01^ N) - Accurately pipet 100 m£ of the stock
thiosulfate solution into a 1,000 mH volumetric flask. Dilute to the mark
with freshly boiled, cooled distilled water. This 0.01 N solution is not
stable, and must be prepared fresh daily from the stock thiosulfate
solution. Keep in a glass-stoppered flask or bottle when not in use.
Normality - Normality of stock solution x 0.100.
Standardized Sulfite Solution for Preparation of Working Sulfite -
TCM .Solution - Dissolve 0.3 g of sodium metabisulfite (Na.,S03) in
500 mJl of recently boiled, cooled, distilled water. (Sulfite solution
is unstable; it is therefore important to use water of the highest purity
to minimize this instability.) This solution contains the equivalent of
320 to 400 pg/mJl of SO-. The actual concentration of the solution is
38
-------
determined by adding excess iodine and backtitrating with standard sodium
thiosulfate solution. To backtitrate, pipet 50 m£ of the 0.01 iodine into
each of two 500 m& iodine flasks (A and B). To flask A (blank) add 25 ml
distilled water, and to flask B (sample) pipet 25 mSL sulfite solution.
Stopper the flasks and allow to react for 5 minutes. Prepare the working
sulfite-absorbing reagent solution (see below) at the same time iodine
solution is added to the flasks. By means of a buret (50 mil buret)
containing standardized 0.01 N thiosulfate, titrate each flask in turn to
a pale yellow. Then add 5 m£ of starch solution to the flask and shake
thoroughly. Continue the titration until the blue color disappears. Store
in a glass-stoppered bottle or flask.
Working Sulfite-TCM Solution - Pipet accurately 2 m£ of the standard
solution into a 100 mJl volumetric flask and bring to mark with 0.04 M
absorbing reagent. Calculate the concentration of sulfur dioxide in the
working solution:
yg S02/m4 = (A~B)(N)(32,000) x
where A = volume thiosulfate for blank, m&,
B = volume thiosulfate for sample, m£,
N = normality of thiosulfate titrant,
32,000 = milliequivalent weight of SO-, pg, and
0.02 = dilution factor.
This solution is stable for 30 days if kept at 5°C (refrigerator). If not
kept at 5°C prepare fresh daily.
Pararosaniline Reagent - To a 250 m£ volumetric flask add 20 mX, stock
pararosaniline solution. Add an additional 0.2 mA stock solution for each
percent the stock assays below 100 percent. Then add 25 m& of 3 M phosphoric
acid and dilute to volume with distilled water. This reagent stored in a
glass-stoppered bottle is stable for at least 9 months.
This equation as written in Subsection 6.2.9 of the Appendix is in error
by a factor of 25.
39
-------
Preparing a Purified Pararosaniline Stock Solution - In preparing a
purified pararosaniline stock solution the method referenced in
Subsection 6.2.10.2 of the Appendix for purifying the dye is as follows
(Ref. 2):
(1) Place 100 m£ each of 1-butanol and IN HCl in a large
separatory funnel (250 m£) and allow to equilibrate.
Note: Certain batches of 1-butanol contain oxidants
that create an SO. demand. Before using check by
placing 20 m£ of 1-butanol with 5 ml of 20 percent
potassium iodide (KI) in a 50 mil. separatory funnel and
shake thoroughly. If a yellow color appears in the
alcohol phases, redistill the 1-butanol from silver
oxide, and collect the middle fraction, or purchase a
new supply of 1-butanol.
(2) Weigh 100 mg of pararosaniline hydrochloride (PRA), in a
small beaker. Add 50 m£ of the equilibrated acid (drain
the acid from the bottom of the separatory funnel in (1)
above) to the beaker and let stand for several minutes.
(3) To a 125 m£ separatory funnel, add 50 m£ of the equili-
brated 1-butanol (draw the 1-butanol from the top of the
separatory funnel in (1) above). Transfer the acid
solution (from (2) above) containing the dye to the funnel,
and extract. The violet impurity will transfer to the
organic phase.
(4) Transfer the lower (aqueous phase) into another separatory
funnel and add 20 m£ of 1-butanol; extract again.
(5) Repeat the extraction procedure with three more 10 m£
portions of 1-butanol. This procedure usually removes
almost all of the violet impurity that contributes to
the blank.
(6) After the final extraction, filter the acid phase through
a cotton plug into a 50 m£ volumetric flask and bring to
volume with 1 N HCl. This stock reagent will be yellowish
red.
Assaying the PararosanilineJStock Solution - The concentration of
pararosaniline hydrochloride (PRA) need be assayed only once after prep-
aration. It is also recommended that commercial solutions of pararosaniline
be assayed when first purchased. The assay procedure is as follows (Ref. 2):
(1) Prepare a buffer stock solution with a pH of 4.69 by
dissolving 13.61 g of sodium acetate trihydrate in
distilled water in a 100 m£ volumetric flask. Add
5.7 m£ of glacial acetic acid and dilute to volume with
water.
40
-------
(2) Take 1 m& of the stock solution obtained from the
purification process or from a commercial source and
dilute to the mark in a 100 m£ volumetric flask with
distilled water.
(3) Transfer a 5 m£ aliquot to a 50 m£ volumetric flask.
Add 5 mJi, of 1M acetate-acetic acid buffer solution from
(1) above and dilute the mixture to the mark with
distilled water. Let the mixture set for 1 hour.
(4) Measure the absorbance at 540 nm with a spectrophoto-
meter. Compute the percent of nominal concentration
of PRA by
% PRA
K
W
where A = the measured absorbance of the final
mixture (absorbance units),
W = the weight in grams of the dye used in the
assay. For example, 100 mg of dye was used
to prepare 50 mX. of solution in the purifica-
tion procedure and 1 mJl of the solution was
used in the assay, then W = 0.002 g
(•—• x 100 mg). When obtained from commercial
sources use the stated concentration to compute W.
K = a constant whose value has to be determined
for a given spectrophotometer and associated
equipment. For example, when using 1-cm
optical path length cells, 0.04 mm slit width
in a Beckman DU spectrophotometer, K = 21.3.
Note: For other spectrophotometers and equipment, K can
be determined by assaying a batch of PRA of known
purity (i.e., in the above equation % PRA is known,
A and W is measured then K can be calculated).
Develop a Calibration Curve (Step 13)
Procedure With Sulfite Solution - Develop a calibration curve as directed
in Subsection 8.2.1 of the Appendix, page 114. Obtain at least 6 data
points and construct the best-fit, straight line using the method of
least squares. Check the calibration curve to determine if it satisfies
the following requirements:
41
-------
(1) The slope Is between 0.03 + 0.002 absorbance unit/
yg SO.. A slope outside these limits indicates that
an Impure dye or an improperly standardized sulfite
solution was used. The calibration should be repeated
and if the slope/remains outside the above limits, the
sulfite solution should be restandardized and the purity
of the dye checked to determine the trouble. If
necessary, prepare all new reagents and develop a new
calibration curve.
(2) The calibration curve Intercept on the absorbance axis
is between 0.163 + 0.039 with the calibration performed
at room temperature. The above numbers represent the
mean and 3cr values for the calibration curve intercept
from a collaborative test of the method (Ref. 6).
If the Intercept is outside the above limits and the room
temperature is approximately 22°C, the calibration curve
should be repeated after checking the spectrophotometer
for proper calibration. If the intercept remains outside
the above limits, with the temperature ~22°C, chemicals
and laboratory procedure should be checked before preparing
new reagents and developing a new calibration curve.
The process of checking and repeating should be continued
until a calibration curve with an intercept within the
above limits is obtained.
(3) Check each calibration point and repeat any point deviating
more than 0.8 yg SO, (20 value for measuring control
samples from Ref. 6) from the best-fit curve for a given
absorbance value. Average the two values and replot the
point. Repeat until all points are within 0.8 yg SO. of
the best-fit curve.
(4) Compute the reciprocal of the slope and record on the
calibration curve. File the calibration curve in the
calibration log book after having the supervisor accept
the results by signing and dating the calibration curve.
A new calibration curve must be developed when:
(1) a new batch of reagents is made, or
(2) a control sample cannot be measured within +1.2 yg S0«
of its actual value when the control sample is prepared as
directed in Subsection 8.2.1 of the Appendix, page 114.
42
-------
Procedure with BO 2 Permeation Tubes - Procedures for generating standard
atmospheres and preparation of a calibration curve are given in Sub-
section 8.2.2 of the Appendix. SO. permeation tubes having nominal shelf
lives (at 22°C) ranging from 4 to 24 months are commercially available
(Ref. 3). Tubes with a certified permeation rate can be obtained from the
National Bureau of Standards (NBS) . Alternatively permeation tubes may be
prepared and calibration. However, this should only be attempted by
experienced personnel in a well-equipped laboratory.
CatibMvUon orf SO* permeation tube* - A detailed discussion of the required
apparatus and procedures for preparation and gravimetric calibration of
permeation tubes is given in Reference 4. A method for volumetric ally
calibrating SO- permeation tubes is given in Reference 5. Permeation rates
determined both volumetrically and gravimetrically for 4 tubes showed
volumetric determinations averaging 3 percent higher than gravimetric
determinations (Ref. 5).
The use of either of the above techniques requires a trained/experienced
operator, therefore, step-by-step procedures are not attempted here. The
operator should refer directly to the above references for guidance.
chuck - A permeation tube, whether prepared and calibrated in
the laboratory, purchased from a manufacturer with a stated permeation
rate, or obtained from the NBS with a certified permeation rate, should
be checked periodically throughout its useful lifetime. A permeation tube
to be used continuously at a given temperature can be monitored in the
following manner.
;
(1) Equilibrate the tube at the temperature it is to be
used.
(2) Weigh the tube to the nearest 0.1 mg with an analytical
balance sensitive to 0.01 mg.
The tube should be handled with teflon-tipped forceps
with special care exercised not to expose the tube to
dust or other contaminants. Also, any static charge
should be removed from the tube.
(3) Record the weight to the nearest 0.1 mg, the date and
time of weighing to the nearest minute in the cali-
bration log book for that permeation tube.
(4) The tube should be reweighed at intervals equivalent
to 1/10 of the expected life of the tube or, once a
month, whichever is shorter.
43
-------
(5) For each new weighing compute a permeation rate using
the previous weighing and the elapsed time. Record
this value in the calibration log .book.
(6) Compute the percent deviation from the stated or
certified value by
P - P
percent deviation - -~ x 100
a
where P • actual (certified) permeation rate
(ug/min)
and P = measured permeation rate (vig/min).
With proper handling and use ? an SO. permeation tube usually maintains a
constant permeation rate (at a constant temperature) as long as any
liquid is visible in the tube. Therefore, a drop in the permeation rate
indicates that the tube has been damaged or that the liquid S0? is nearly
gone. It is suggested that the certified or stated permeation rate be
used as long as the measured rate is within +_ 5 percent of the original
value. If the measured permeation rate is less than 95 percent of the
stated or certified permeation rate, the tube should be replaced if there
is little or no liquid SO- in the tube. If the tube still has a good
supply of liquid SO-, the new measured permeation rate should be used and
the tube reweighed at time intervals equivalent to a weight loss of about
10 mg as computed by the permeation rate until 3 successive measured
permeation rates agree within + 5 percent of the average value of the three.
The average value should be used as long as future checks are within
+ 5 percent.
PeveCojatng a. caJUJbJULtion. CUAve. - Develop and construct a calibration curve
according to the directions given in Subsection 8.2.2 of the Appendix,
page 114.
Check each point on the calibration curve and redo any point that deviates
more than * (0.8 yg SO-) from the best-fit curve. Replot the average of
the old and new values and construct a new best-fit curve. Continue the
replication until the average of the two most recent measurements falls
within + (0.8 ug SO.) of the best-fit curve.
Determine the total ug S0_ in the sample by multiplying the known concen-
3
tration of the calibration apparatus output (yg S0_/m ) times the total volume
3 L
of calibration air sampled (m ).
-------
Compute the reciprocal of the calibration curve slope and record the value
on the calibration curve. File the curve in the calibration log book
after it has been dated and signed by the supervisor.
A new calibration must be performed when:
(1) a new batch of reagents are prepared, or
(2) when a control sample cannot be measured within
+ (1.2 yg SO^) of its actual value when prepared
with an SCL permeation tube.
Colorimetric Analysis (Step 14)
Prepare the sample, reagent blank, and control sample as directed in
Subsection 7.2 of the Appendix, page 114.
Set the spectrophotometer to a wavelength of 548 nm. Allow at least
5 minutes for the spectrophotometer to warm up. If necessary, adjust the
zero control to bring the meter needle to 0 on the percent transmittance
scale. Standardize the light control by inserting a cell filled with
distilled water into the sampler holder and adjusting the light control
as required, until the meter reads 100 percent transmittance. The complete
transmittance scale should be checked with a calibrated set of filters from
the National Bureau of Standards any time a control sample cannot be
measured with +1.2 yg S0_ of its known value.
It is recommended that a reagent blank and control sample be measured
before each set of determinations. Record the absorbance value of the
reagent blank and the total yg S0? measured for the control sample in the
data log book with the time to the nearest hour. If the absorbance blank
is within + 0.03 absorbance unit of the calibration curve absorbance
intercept and, if the measured value of the control sample is within
+1.2 yg SO,, of the actual value, proceed to analyze the field samples.
If the reagent blank and/or control sample falls outside the above limits
the reagent blank, control sample, and calibration curve should be checked
by replication as necessary to satisfy the above limits before analyzing
field samples.
It is recommended that when first implementing a quality assurance program
the oH of the final solution be measured for a sample from each
sampling site. If the pH is less than 1.4 or greater than 1.8, the sample
should be invalidated and reported to the supervisor. If the pH is
between 1.5 and 1.7, accept as good and if the pH is between 1.4 and 1.5,
or 1.7 and 1.8, the exact pH value should accompany the reported concen-
tration value for that sample. In any case, checks should be made in an
effort to determine why the pH was outside 1.6 + 0.1 and corrective action
taken.
45
-------
The pH should seldom, if ever, deviate from 1.6 + 0.1, but large errors
result if it does. Therefore, it is suggested that the pH of the final
solution be checked for 1 sample a month for each sampling site. Report
the measured pH to the supervisor and record the pH value with the site
identification data in the laboratory log book.
Record all measured values (i.e., reagent blanks, control samples,
samples) in sequential order in the data log book. The date and time
should be recorded with each set of determinations. The reagent blank
should only be used with the set of determinations in which it was
measured.
DATA PROCESSING
Perform Calculations (Step 15)
Rotameter - When a rotameter is used to measure the flow rate, perform the
following calculations.
(1) Obtain the rotameter reading, I and I,, average temper-
ature, T,, average pressure, Pf, and sampling period time
as reported on the Sample Record Sheet by the field
operator.
(2) From the rotameter calibration curve determine the K values
for readings of I and lf.
(3) Compute the average flow rate, Qf, at field conditions by
(4) Calculate the volume sampled, V,, at field, conditions by
V£ = Qf * t
where t is the sampling period time in minutes.
(5) Correct the sampled volume to volume, VR, at reference
conditions by
V*> = Vf(cm > X X X 10
(6) Compute the average measured SO. concentration as directed
in Section 9.2 of the Appendix, page 114.
46
-------
Critical Orifice - When a critical orifice is used to control flow rate
perform the following calculations.
(1) Obtain Q as computed in Step 11, page 36. Also, obtain
the sampling period time, t, the average temperature, I,,
and the average pressure, P.., as reported by the field
operator.
(2) Compute the volue, V,t sampled at field conditions by
Vf = Qf x t
where t is the sampling period time in minutes.
(3) Correct the sampled volume to reference conditions by
VRU) = Vf(cm3) x ^x |2t 10~3 (£/cm3).
o
(4) Compute the average measured SO. concentration in pg/m
as directed in subsection 9.2 of the Appendix, page 114.
Document and Forward Data (Step 16)
o
Record the average concentration of S0_ in pg/m with required identifying
information on the appropriate SAROAD Data Form. See Users Manual: SAROAD
(Storage and Retrieval of Aerometric Data), APTD-0663, for detailed
instructions for accomplishing this. The original calculations should be
filed in the operational data log book.
SPECIAL CHECKS FOR AUDITING
Proper implementation and conduct of an auditing program will allow one to
estimate data quality in terms of precision and accuracy at a given level
of confidence. To realize maximum benefits from an auditing program, it
should be conducted independently of the routine operation of the sampling
network. That is, checks should be made by individuals other than the
regular operator. Furthermore, the checks should be performed without any
special preparation or adjustment of the system (see page 96 for further
discussion) .
It is felt that in conjunction with the special checks given in the
operating procedures three audit checks will be sufficient to properly
assess data quality. The checks Include:
47
-------
(1) a check of the average flow rate or volume of air
sampled for sample collection,
(2) measurement of control samples to evaluate the
analysis process, and
(3) a data processing check to evaluate calculation
and recording errors.
A checking or auditing level of 7 checks (n = 7) out of 100 sampling periods
(N = 100) is used here for illustration purposes when sampling is carried
out on a daily basis. For cases where one sample is collected every sixth
day, a minimum auditing level of 1 check per month is recommended. This
would result in an auditing level of approximately 3 checks (n = 3) for a
lot size of 15 (N = 15) for data reported quarterly. The supervisor will
specify the auditing level to be used according to local monitoring needs.
Directions for performing each of the checks are given here. Proper use
of the resulting data along with desirable control limits are given in the
Supervision Manual starting on page 62.
Flow Rate/Volume Check
Volume Check - The volume of air sampled can be checked with a calibrated
wet test meter for 30-minute and 1-hour sampling periods with approximate
flow rates of 1000 cm /m (1 £/min) and 500 cm /min (.5 fc/min) respectively.
Note: A flow rate of 500 cm /min is below the normal range of wet test
meters. Hence, the meter must be calibrated at this exact flow rate
against a soap-bubble meter. If the meter accuracy varies from check to
check by more than + 2 percent, it should not be used for this low flow
rate.
The regular operator prepares the sampler for collecting a sample in the
usual manner recording usual site parameters such as temperature and
pressure. If a critical orifice is used in the sampling train the person
performing the check should provide a calibrated orifice. The sampler
is allowed to run for 30-minutes of 1-hour as appropriate.
Compute the volume of air measured by the wet test meter corrected to
reference conditions by
m
48
-------
where VR(C) = volume measured by wet test meter at reference
conditions, £,
V = volume measured by wet test meter at T and
m P £ m
m' '
P = barometric pressure plus meter pressure read
from the manometer, mmHg,
and T = temperature of liquid in wet test meter, °K.
Compute the volume of air sampled as measured by the rotameter/critical
orifice and correct to reference conditions according to the procedure in
step 15, page 46. Designate this volume as V (0).
K
Compute the percent difference in the two volumes by
V (C) - V (0)
Vc>
Report the values of Vn(C), V^CO), and d with site identification data, date,
K K
and signature of person performing the check to the supervisor.
The. Cample. cotte.cte.d da>u.ng the. c/iecfe u> not a. vatid sample. and Ahoutd be.
4.n\jatida£nd and ^ofiwaAdad to the.
If d from the above calculations is equal to or greater than 9 the sampling
train including rotameter/critical orifice should be checked and corrective
action taken before sampling is resumed.
Flow-Rate Check - For 24-hour samples a flow-rate check is recommended as
a means of auditing the sample collection phase of the measurement process.
The check is performed as follows:
(1) The regular operator prepares the sampler for sample
collection as usual. This includes filling in the
Sample Record Sheet.
(2) The individual performing the audit inserts a cali-
brated rotameter in the sample inlet line just
upstream of the outside filter (see Figure 8) .
(3) With the calibrated rotameter in place the sample is
collected in the usual manner.
49
-------
(4) The individual performing the audit reads the cali-
brated rotameter before, I* and after, I*, the sampling
period and at one point, preferably the midpoint, during
the sampling period, I*.
(5) Following the procedure given in Step 15, page 46, for
rotameters, calculate the volume at reference conditions
for the calibrated rotameter, V (C). In this case the
K.
equation for computing the average flow rate is
' + K' + Kl\ /I' + I1
i i f) _i s
3 / V 3
(6) Compute the sampled volume, VD(0), measured by the
K.
regular rotameter/critical orifice and correct to
reference conditions according to the procedure given
in Step 15, page 46.
(7) Compute the percent difference by
vc> -
vR(0
(8) Record l!, I', I', I., I,, and d and forward to the
i m t i i
supervisor.
(9) If d is equal to or greater than 9, trouble shooting
should be performed and corrective action taken before
sampling is resumed.
Measurement of Reference Samples
Measurement of a reference sample prepared independently of normal
operations can be used to evaluate the precision and accuracy of the
analysis phase of the measurement process. A reference sample as used
here refers to a sample prepared by an individual other than the regular
operator using reagents prepared independent of those used for normal
operations and used for auditing purposes. A control sample as referred
to in the operating procedures implies a solution prepared by the regular
operator from the normally used reagents and used periodically to verify
that the analysis process is under control.
50
-------
The procedure for performing the check is given below. The frequency of
performing the check will be specified by the supervisor. Further
discussion is given in the Supervision Manual on page 63»
(1) Prepare a reference sample with a known concentration in
the same manner as was used when developing a calibration
curve (see Section 8 of the Appendix). The concentration
should be varied over the range of normally measured
values from audit to audit.
(2) Have the regular operator to measure the reference sample.
The operator must not know the true concentration of the
reference sample and preferably not know that it is a
reference sample.
(3) Obtain the operator's finding (yg S0_) and the true
L m
concentration (yg SO-) of the sample and compute the
difference by c
d = (yg SO.) - (yg SO.)
L m z c
(4) Record (yg SO-) , (yg SO.) and d and forward to the
^ m ^ c
supervisor.
It is recommended that the supervisor maintain a record
of the d values in chronological order and that analysis
be stopped, checks made to determine the likely cause(s),
and corrective action taken before the analysis is
resumed anytime:
(a) a d value equal to or greater than 1.2 yg SO-
is observed,
(b) two consecutive d values (i.e., results from two
consecutive audit checks) exceed +0.8 yg SO in
the same direction, or
(c) three consecutive d values (i.e., results from three
consecutive audit checks) exceed +0.4 yg SO- in the
same direction.
(5) If the control sample measurement was outside the above
limits, the calibration curve should be checked by
measuring a reagent blank and control sample. If
necessary, a new calibration curve should be developed
and other corrective actions taken as deemed necessary.
After corrections are made, a new reference sample should
be measured and, if acceptable, analysis resumed.
51
-------
(6) If the supervisor reports that the reference sample
measurement of (A) above was within limits, continue
normal operations.
Data Processing Check
In auditing data processing procedures, it is convenient and allows for
corrections to be made immediately if checks are made soon after the
original calculations have been performed. In particular, this allows
for possible retrieval of additional explanatory data from field personnel
when necessary.
The check must be independent; that is, performed by an individual other
than the one who originally reduced the data. The check is made starting
with the raw data and continuing through recording the concentration in
o
yg/m on the SAROAD form.
If the mass concentration of SO. computed by the check differs from the
original value by as much as + 3 percent, all samples collected since the
previous audit are checked and corrected. The check value is always
reported as the correct value.
Record the check and original values in the operational data log book
and report the values to the supervisor.
SPECIAL CHECKS TO DETECT AND IDENTIFY TROUBLE
The following checks are recommended as a means of checking the adequacy
of different phases of the measurement process. The checks are:
(1) estimate the average temperature environment experienced
by a sample between collection and analysis,
(2) interference checks,
(3) measurement of the pH of the final solution at time of
analysis,
(4) visual check of the volume of absorbing reagent in the
exposed absorber before analysis, and
(5) a check on the accuracy of the sampler timer.
52
-------
Errors resulting from any one of the above areas would not be detected by
the auditing process. However, errors from all except the first source
should be small if the suggested operating procedures are being followed.
Checks 2 through 5 should be performed for each sampling site when:
(1) a quality assurance program is first initiated in order
to identify potential problem areas, and
(2) periodically thereafter by the supervisor as an overall
check on the measurement process.
Check number 1 should be made for each site when:
(1) a quality assurance program is first initiated, and
(2) when any change in sample handling or shipping occurs.
Average Temperature Environment of Sample Between Collection and Analysis
For samples that spend several days in transit from the sampling site to
the analysis laboratory, the loss of SO- at elevated temperatures can be
significant (Ref. 2). From the geographical location of the site and
laboratory, the season of the year, and the mode of transport make a rough
estimate of the average temperature that the sample will be exposed to from
the completion of the sampling period to the time of analysis. It is
recommended that the following corrections be applied to the measured
concentration at reference conditions for any sample with an unrefrigerated
time lapse greater than 1 day between collection and analysis:
(1) If the estimated average temperature is 5°C or less
no correction is necessary.
(2) If the estimated average temperature is between 5 and
22°C the corrected concentration should be calculated
by (Ref. 2)
(lag S02/m3)
(yg
where ( g S00/m ) = the corrected concentration of
2 c so2,
3
( g S0_/m ) = the measured concentration of
2 m so2,
and D = the elapsed time between collection
and analysis in days.
53
-------
(3) If the estimated average temperature is between 22 and 28°C
the corrected concentration is calculated by (Ref. 7)
(pg SO /m3)
(yg S07/mJ) = m
c 1.0 - 0.02 x
These corrections are only rough estimates. The most desirable situation
would be to have D very small or the temperature controlled at a known level.
Measurement of pH of Final Solution
The absorbance of a sample is highly sensitive to the solution pH (Ref. 2).
Therefore, it is suggested that samples from each site be checked initially
to determine if the ambient atmosphere contains pollutants in sufficient
quantities to affect the pH of the final solution and thereafter as a check
on the reagents used in the analysis when control samples cannot be
measured within specified limits. The pH should be measured }ust prior to
the analysis. Any sample having a pH outside of 1.6+0.1 should be investi-
gated and corrective action taken. The primary suspect would be an error in
the quantity of phosphoric acid added to the sample.
Interference Checks
Control samples and reference samples with known concentrations of S0» are
measured to evaluate the analysis phase of the measurement process. In
addition to those measurements, it is recommended that periodically
reference samples be spiked with one or more of the principal known inter-
ferences to evaluate the effectiveness of chemicals and procedures used to
reduce or eliminate the interference.
Spiked samples should be used when the ambient atmosphere is known to
contain higher than normal concentrations of a particular interference
or any time the reported S02 values appear larger than expected from a
set of determinations.
One means of performing the check is as follows:
(1) Prepare two control samples with identical amounts of
the working sulfite-absorbing reagent solution (see
Subsection 6.2.9 in the Appendix).
(2) Spike one of the control samples with a known quantity of
the interference of interest.
(a) Nitrates or nitrites can be used to prepare a
solution of known concentration of nitrogen oxides.
(b) Salts of the heavy metal of interest can be used to
prepare solutions of known concentrations for
spiking the control sample.
54
-------
(3) Have the operator measure a reagent blank, the unspiked
control sample, and the spiked control sample in sequence.
(4) Compare the measured concentration of the two control
samples by computing the difference
difference = (vg SO ) - (yg SO.)
2 2 2
where (ug SO-) = total measured concentration of the
1 unspiked sample,
(pg S0?) = total measured concentration of the
2 spiked sample.
(5) If the difference is equal to or less than +1.2, it is
assumed that there is no difference in the sample.
(6) If the percent difference is greater than the above
limits the appropriate chemicals and procedures should
be checked and corrective action taken to eliminate
interferences.
Volume of Absorbing Reagent
An error in the volume of sample prior to analysis is transferred directly
to the measured concentration.
The operator removing the absorber from the sampling train is responsible
for replacing any absorbing reagent that was evaporated during sample
collection for 24-hour samples that have to be mailed to the laboratory.
The operator performing the analysis is responsible for bringing the
sample volume to the specified amount for samples delivered directly to
the laboratory from the field site.
Absorbers and associated Sample Record Sheets should be checked periodi-
cally just prior to analysis to determine if the sample volumes are
accurate and to see if the operating procedures are being followed.
Check of Sample Timer
In 24-hour samples an error of + 14 minutes represents an error in the
measured concentration of only 1 percent. However, accurate timing is
essential for 30-minute and 1-hour samples. If automatic timers are used
for the shorter time periods, it should be checked at least every six
months against a calibrated time piece such as a stop watch. Alternatively,
an elapsed time indicator could be used in conjunction with an uncalibrated
timer.
55
-------
Any deviation in the sample period time as measured by the standard and
that indicated by the regular means of measuring the time greater than
+ 1 percent from the standard timer should result in replacing the timer
or employment of a new technique for measuring the sampling period time.
FACILITY AND APPARATUS REQUIREMENTS
Facility
Primary facilities required for the pararosaniline method are a central
laboratory and individual sampling stations. The laboratory should be
equipped for
(1) reagent preparation,
(2) storage of chemicals and reagents
(3) calibration of critical orifices or rotameters, and
(4) sample analysis.
The laboratory should be equipped with an automatic all-seasons air
conditioning unit capable of maintaining a pre-set temperature within
+ 3°C (5°F). It should also be equipped with a hood and exhaust fan
large enough to accomodate the acid bath used for cleaning the glassware.
The sample collection unit should include a dark box for housing the
bubbler and protecting it from direct sunlight. Also, the box should be
thermostated to prevent condensed moisture from freezing in the sampling
train lines in cold weather.
Apparatus
Items of equipment with approximate costs are listed in Table 1. Each
item is checked according to whether it is 1) required in the reference
method, 2) used to control a variable or parameter, 3) required for
auditing purposes, or 4) used to monitor a variable. The permeation tube
setup is. included as an alternate calibration method.
56
-------
Table 1. APPARATUS USED IN THE MANUAL PARAROSANILINE METHOD
Item of equipment
CALIBRATION
1. Flow-rate calibration unit
2. Wet test meter
3. Permeation tube calibration setup
4. Spectrophotometer calibration
unit
5. Set of standard weights
6. Soap bubble meter
SAMPLING
7. 24-hr sampling system
8. 30-min/l-hr sampling system
9. 24-hr timer
10. Midget impinger
11. 100 mJl threaded bubbler
12. Filter holder
ANALYSIS
13. Spectrophotometer (manual)
14. Analytical balance
15. pH meter and probe
16. Misc. glassware
17. Reagents (cost/100 samples)
Approx.
cost 1973
300
1,000
2,200
100
110
50
250
250
40
10
4
2
600-900
1,400
350
100
10
Assoc.
error
Calibration
Calibration
Calibration
Calibration
Calibration
Calibration
Ref.
method
/
/
/
/
/
/
1
/
/
^
Variable
Monitoring
/
/
Auditing
equipment
^
/
-------
SECTION III SUPERVISION WNUAL
GENERAL
Consistent with the realization of the objectives of a quality assurance
program as given In Section I, this manual provides the supervisor with
brief guidelines and directions for:
(1) the collection and analysis of information necessary for
the assessment of high volume data quality,
(2) isolating, evaluating, and monitoring major components
of system error,
(3) changing the physical system to achieve a desired level
of data quality,
(4) varying the auditing or checking level to achieve a
desired .level of confidence in the validity of the
outgoing data, and
(5) selecting monitoring strategies in terms of data quality
and cost for specific monitoring requirements.
This manual provides brief directions that cannot cover all situations.
For somewhat more background Information on quality assurance see the
Management Manual of this document. Additional information pertaining
to the Pararosaniline Method can be obtained from the final report for
this contract and from the literature referenced at the end of the
Management Manual.
Directions are written in terms of a 24-hour sampling period with
reference to 30-minute and 1-hour sampling periods when appropriate and
an auditing level of n-7 checks out of a lot size of N=100 for illustra-
tion purposes. Special instructions for auditing operations where
sampling is performed every sixth day are given also. Information on
additional auditing levels is given in the Management Manual.
Specific actions and operations required of the supervisor in implement-
ing and maintaining a quality assurance program as discussed in this
Manual are summarized in the following listing.
58
-------
(1) Data Assessment
(a) Set up and maintain an auditing schedule.
(b) Qualify audit results (i.e., insure that checks
are independent and valid).
(c) Perform necessary calculations and compare with
suggested performance standards.
(d) Make corrections or alter operations when standards
are exceeded.
(e) Forward acceptable qualified data, with audit results
attached, for additional internal review or to user.
(2) Routine Operations
(a) Obtain from the operator immediate reports of
suspicious data or malfunctions. Initiate corrective
action or, if necessary, specify special checks to
determine the trouble; then take corrective action.
(b) On a dally basis, evaluate and dispose of (i.e., accept
or reject) data that have been identified as question-
able by the operator.
(c) Examine operator's log books periodically for complete-
ness and adherence to operating procedures.
(d) Approve sample record data sheets, calibration data,
etc., for filing by operator.
(e) File auditing results.
(3) Evaluation of Operations
(a) Evaluate available alternative monitoring strategies
in light of your experience and needs.
(b) Evaluate operator training/instructional needs for
your specific operation.
59
-------
ASSESSMENT OF DATA QUALITY
Procedures for implementing and maintaining an auditing program to assess
data quality are presented in this section. Throughout this discussion
and the rest of this document, the term "lot" is used to represent a set
or collection of objects (e.g., measurements or observations), and the
"lot size" designated as N is the number of objects in the lot. The
number of objects in the lot to be tested or measured is called the
"sample size" and is designated as n. The term "auditing level," used
Interchangeably with "checking level," is fully described by giving the
sample size, n, and the lot size, N.
A valid assessment of a lot of SO. data generated with the manual
pararosaniline method can be made at a given level of confidence with
information derived from special checks of key operations in the measure-
ment method. Figure 15 summarizes the quality control checks applied
at various check points in the measuring process. Each check or opera-
tion is represented by a box. They are in the order in which they would
be performed on a particular sample as it progresses through the process.
Boxes enclosed by heavy lines represent 100 percent sampling; i.e., these
checks are made on each sample passing through the system. The two checks
enclosed by dashed lines are not scheduled at any given rate but are
carried out as specific events take place (e.g., a new batch of absorbing
reagent is prepared or at the beginning of a set of determinations). The
remaining three checks are to be performed at the prescribed auditing
level.
All but one of the checks are treated on a go/no-go basis. That is, a
standard or limit is defined and the lot or individual item is accepted
or rejected on the basis of the check results. Certain rejected lots
or samples are corrigible; i.e., they are capable of being corrected.
Specifically, samples are corrected for delay between collection and
analysis and lots rejected because of data processing errors are
accepted after the errors have been located and corrected. The one check
not treated on a go/no-go basis is the volume/flow-rate check. This
check is performed at the prescribed auditing level. Action for
correcting system deficiencies may be taken as the result of any one
check, however, there is usually no clear-cut way of correcting previous
data. Therefore, results of this check are reported and used in
assessing data quality along with the results from measuring reference
samples as described in the Management Manual, page 96. Also, in some
instances results from the data processing checks may be requested by
the manager.
60
-------
I 1
! MEASURE pH OF I
I ABSORBING REAGENT I
I I
L— T J
CALIBRATE
FLOW RATE
CHECK THE pH OF A NEW BATCH OF ABSORBING
REAGENT BEFORE USE. ACCEPT THE BATCH IF
3 < pH < 5.
CALIBRATE THE CRITICAL ORIFICE BEFORE
EACH SAMPLING PERIOD. CALIBRATE THE
ROTAMETER, IF USED, PERIODICALLY.
CHECK SYSTEM
FLOW RATE
CHECK THE VACUUM GAUGE READING TO INSURE
CRITICAL FLOW WHEN CRITICAL ORIFICES ARE
USED. ACCEPT IF VACUUM IS GREATER THAN
507 m*g. CHECK AND ADJUST SYSTEM FLOW
RATE WHEN ROTAMETER IS USED.
REPORT
'irV-"i
TO SUPERVISOR
VOLUME/FLOW-RATE
CHECK
AUDIT THE VOLUME/FLOW RATE ON 7 OUT OF 100
SAMPLING PERIODS. REPORT A DEFECT IF
d,. i 9 (SEE PAGE 71 FOR DEFINITION OF d, .)
SAMPLING
PERIOD
TIME (T)
ACCEPT SAMPLE IF 23 < T < 25 HOURS FOR
24-HOUR SAMPLES. OTHERWISE INVALIDATE
THE SAMPLE AND INFORM THE SUPERVISOR.
UNUSUAL
CONDITIONS
r
MEASURE I
REAGENT BLANK ,
AND '
J CONTROL SAMPLE
ACCEPT IF NO UNUSUAL CONDITIONS ARE EVIDENT.
E.G., ENTRAPMENT OF ABSORBING REAGENT, LOSS
OF SAMPLE DURING HANDLING, OBVIOUS EQUIPMENT
MALFUNCTIONS, ETC.
A REAGENT BLANK AND A CONTROL SAMPLE IS
MEASURED AT THE BEGINNING OF EACH SFT OF
DETERMINATIONS.
REPORT
TO SUPERVISOR
MEASURE
REFERENCE
SAMPLES
TIME BETWEEN
COLLECTION/ANALYSIS
AUDIT THE ANALYSIS PHASE BY MEASURING 7
REFERENCE SAMPLES RANDOMLY DISPERSED AMONG
100 FIELD SAMPLES. REPORT A DEFECT IF .
d2j i 1.2 u9 S02 (SEE PAGE 73 FOR
DEFINITION OF d-,).
MAKE CORRECTION TO THE MEASURED CONCENTRATION
FOR DELAYS BETWEEN COLLECTION AND ANALYSIS.
DATA
PROCESSING
CHECK
REDO CALCULATIONS ON 7 OUT OF 100 SAMPLES.
ACCEPT THE 100 SAMPLES IF: 1) ALL CHECK
CALCULATIONS ARE WITHIN + 3 PERCENT OF THE
ORIGINAL. OR 2) ALL CALCULATIONS HAVE BEEN
REDONE AND CORRECTED.
REPORT TO
SUPERVISOR
AS VALID SAMPLES
REPORT ALL VALID SAMPLES TO THE SUPERVISOR.
ASSEMBLE DATA
INTO HOMOGENEOUS
LOTS OF 100 SAMPLES
SUMMARIZE AND ATTACH AUDIT RESULTS TO
SAROAD FORM.
FORWARD
FOR ADDITIONAL
INTERNAL REVIEW
FIGURE 15: FLCH CHART OF QUALITY CONTROL CHECKS IN THE AUDITING PROGRAM
61
-------
Required Information
For an auditing program as described above the required information for
evaluating data quality, includes results from the following checks:
(1) volume/flow-rate checks,
(2) reference sample measurement, and if requested,
(3) data processing checks.
Directions for performing the above checks are given in the Operations
Manual, page 47. Directions for insuring Independence and proper randomi-
zation in the auditing process and for the evaluation of the results are
presented in this section.
Collection of Required Information
Volume/Flow-Rate Check - A volume check using a wet test meter is
recommended for samples where the flow rate is 500 cm^/min. or greater.
A flow-rate check using a calibrated rotameter is recommended for samples
where the flow-rate is less than 500 cm^/min. procedure for performing
the check. Samples from individual sites should be combined Into lots.
For sites where 50 or more samples are collected each quarter, a minimum
of 7 randomly spaced checks per quarter is recommended. A minimum of 3
checks per quarter is recommended for sites operating every sixth day,
thereby generating 15 or les& samples a quarter.
Randomly select 7 sampling periods from the coming quarter for sites
where the lot size is expected to be as large as 50. Record dates.
The operator should not be aware of when the checks are to be performed.
Remember that when volume checks are made the sample is invalid. Hence,
the check should be made before or after the regular sample is collected.
The flow-rate check can be made on the actual sample.
For sites where the lot size is 15 or less, randomly select 1 sampling
period each month. Record these dates and perform the checks as scheduled.
Directions for performing the check are given in the Operations Manual,
page 47.
0^ data. - Obtain the audit results from the individual performing
the audit (see Section on "Special Checks for Auditing," page 47).
Using the check value, Vc^ and the value determined by the operator, V0j ,
compute the percentage difference by
62
-------
where V . = the Volume measured by the wet test meter (corrected
th
to reference conditions) for the j check performed
during an auditing period,
V . = the Volume measured by the regular method (i.e., flow
°-' rate times the sampling period time) ,
and d. . - the percent difference in the volumes measured for the
f~H
j check, the subscript 1 identifies the check as
being a volume or flow-rate check.
Report d-.., d..-, d.,, d ,, d , d ,, d. _ and the auditing level on the form
in Figure 16, page 65.
Measurement of Reference Samples - Reference samples prepared independent
of the normal operations are used to evaluate the analysis phase of the
measurement method.
0$ Ae^CAence AOmptu - Reference samples should be made using a
working sulfite-TCM solution (see Subsection 6.2.9 in the Append' -.) that was
prepared independent, both personnel and chemicals, from the regularly used
solution. Alternatively, reference samples can be prepared with an SO-
permeation tube set up as shown in Figure A2 and A3 of the Appendix, page
116. Reference samples should span a range from 0.20 to 1.0 absorbance
units. The specific values should be varied from audit to audit.
fan. peJifioiming cnecfe - The operator routinely measures 1 control
sample, that he has prepared, for every set of determinations as a check on
the continued reliability of the calibration curve. The measurement of
independent reference samples evaluates the overall analysis procedure
including operator, chemicals, and technique.
Reference samples should be dispersed randomly throughout the samples
awaiting analysis at the rate of 7 reference samples per 100 field samples.
This auditing level can be followed as long as a total of 100 samples will
be analyzed per quarter. If possible the reference samples should not be
recognizable to the analyst as a reference sample.
TVieotmeitt oft daJta. - Obtain from the person that performed the audit (see
page 44) the measured, (yg SO-) , and check on known value, (yg SO.)
and compute the difference m c
d = (yg SO ) - (yg SO,)
2J 2 cj 2 mj
where j is the j time that the check has been made during a given
auditing period.
63
-------
Data Processing Check - Independent checks on data processing errors are
made as directed in the Operations Manual, page 52.
check - A data processing check should be made
on 7 out of 100 samples. The check should be made by an individual other
than the operator who performed the original calculations. The check
should begin with the raw data through the point of recording the
concentration on the SAROAD form.
Tn&Ltmwt. 0£ data. - Compute the percent difference between the check
3 3
value, (yg SO./m ) , and the original value, (yg S0,/m ) , by
' c z o
(yg S02/m3) - (yg S02/m3)
d, = - - - r - — * 100
J (yg S0,/m3)
where j is the jfc time that the check has been made during a
given auditing period.
Treatment of Collected Information
Identification of Defects - One procedure for identifying defects is to
evaluate auditing checks in pairs; i.e., d..., d.., d . d ., d . d «, ....,
d._ d-_. If one or both members of the pair are defective, it counts as
one defect. No more than one defect is declared per set. Data processing
errors should be corrected when found, and are not, therefore discussed
here.
Any set of auditing checks in which the value of d. . or A^. is greater
than +9 or + 1.2 yg S02 respectively, will be considered a defect.
These values are assumed to be the 30 values and are discussed in the
subsection on Suggested Standards for Judging Performance on page 66.
As field data become available, these limits should be reevaluated and
adjusted, if necessary.
Reporting Data Quality - Each lot or quarter of data submitted with SAROAD
forms or tapes should be accompanied by the minimum data qualifying
information as shown in Figure 16. The individual responsible for the
quality assurance program should sign and date the form. As an illustra-
tion, values from the subsection on Suggested Standards for Judging
Performance, page 66, are used to fill in the blanks in Figure 16. The
reported auditing rate is the rate in effect at the beginning of the
auditing period. An increase or decrease in auditing rate during the
auditing period will be reflected by the total number of checks reported.
The reason for change should be noted on the form.
-------
Supervisor's Signature_
Reporting Date
Auditing Rate for Data Errors: n - 7, N - 100
Definition of Defect: |dj, |> 9,
1.2
and |d3Ji 3
Number of Defects Reported_
(should be circled in the table below)
Audit
1. Volume/Flow-Rate Check
2. Measurement of Reference
Samples (d2.)
3. Data Processing Check
Cd3j>
Check Values
dll
d21
d31
d12
d22
dd32
d13
d23
d33
_
_
Q
In
d2n
d3n
Data processing errors are corrected when found at the ± 3 percent
level and are therefore, not reported as defects.
Figure 16: Data Qualification Form
65
-------
Check values (i.e., d's, d- 's and d,,'s) are calculated as directed on
pages 62 and 63 and reported in Figure 16. Values of d, need be reported
only if requested by the manager. All reported check values exceeding the
definition of a defect should be marked for easy recognition by circling
on the form. In case of a defect from the measurement of reference samples
(d. ) report the defective measurement as well as the measurement made
after the trouble is corrected. Both values would be in the same column
of Figure 16 (i.e., two values of one d,,. would be reported) with the
defect circled. The other value will be used in evaluating data quality as
described in the Management Manual, page 96.
SUGGESTED STANDARDS FOR JUDGING PERFORMANCE USING AUDIT DATA
Results from a collaborative test of the pararosaniline method (Ref. 6)
show that precision is a function of the SO- concentration. The
performance standard given below in Table 2 for measurement of reference
samples represents the 3o value from that test. This standard should be
reevaluated and adjusted as field data become available.
The suggested standard for comparing the volume/flow-rate values is no
more than a rough estimate. There is the error associated with the
initial and final calibration of the critical orifice and the accuracy
of the wet test meter. Under field conditions a standard deviation of
+ 3 percent of the actual value seems reasonable. Again this standard
should be reevaluated as field data become available.
COLLECTION OF INFORMATION TO DETECT AND/OR IDENTIFY TROUBLE
In a quality assurance program one of the most effective means of
preventing trouble is to respond immediately to reports from the operator
of suspicious data or equipment malfunctions. Application of proper
corrective actions at this point can reduce or prevent the collection of
poor quality data. Important error sources, methods for monitoring
variables, and suggested control limits for each source are discussed in
this section.
Identification of Important Parameters
Measurement of sulfur dioxide in the ambient atmosphere by the manual
pararosaniline method requires a sequence of operations and events that
yield as an end result a number that serves to represent the average mass
of sulfur dioxide per unit volume of air over the sampling period.
The measurement process can be roughly divided into three phases. They
are: 1) sample collection, 2) sample analysis, and 3) data processing.
66
-------
Table 2. SUGGESTED PERFORMANCE STANDARDS
Standards for Defining Defects
u
1. Volume/Flow-Rate Check; |d..| > 9
2. Measurement of Reference Samples; IcU-J - 1-2 Ug S02
Standard for Correcting Data Processing Errors
3. Data Processing Check; Jdj.| > 3
Standards for Audit Rates
4. Suggested minimum auditing rates for data error;
number of audits, n - 7; lot size, N - 100; allowable number of
defects per lot, d - 0.
Standards for Operation
5. If at any time d - 1 is observed (i.e., a defect is observed) for
either .d.. or d,., increase the audit rate to n - 20, N - 100 until
the cause has been determined and corrected.
6. If at any tine d - 2 is observed (i.e., two defects are observed In
the same auditing period), stop collecting data until the cause has
been determined and corrected. When data collection resumes, use
an auditing level of n - 20, N - 100 until no defects are observed
in three successive audits.
7. If at any time either one of the two conditions listed below is
observed, 1) Increase the audit rate to n - 20, N • 100 for the
remainder of the auditing period, 2) perform special checks to
identify the trouble area, and 3) take necessary corrective action
to reduce error levels. The two conditions are:
(a) two (2) d.. values exceeding ± 6, or
four (4) d, . values exceeding ± 3.
(b) two (2) d» values exceeding ± 0.8 yg SO-, or
four (4) d2. values exceeding ± 0.4 yg S02.
67
-------
Sample Collection - The sample collection phase of the measurement method
contains numerous sources of error. Some errors can be eliminated by a
conscientious operator, while others are inherent and can only be
controlled. The sources of error are:
(1) flow- rate calibration,
(2) determination of the volume of air sampled,
(3) elapsed time between sample collection and analysis,
(4) exposure of sample to direct sunlight, and
(5) entrainment or loss of sample other than by evaporation.
- Flow rates for critical orifices calibrated against
a wet test meter at different times, by different individuals, but with the
same equipment showed a standard deviation of less than 2 percent of the
mean (Ref. 1). When a large population of laboratories is considered, the
variability would undoubtedly increase significantly. Rotameter calibra-
tions would not be expected to be any more precise than critical orifice
calibrations. Large flow-rate calibrations will be detected by the volume/
flow-rate check as part of the auditing procedure. However, to maintain
the error at or near a minimum the wet test meter or calibrated rotameter
used for calibration should be checked against a high quality soap-bubble
meter at least once a quarter.
$ the. votume. 0$ CUM. &cunpte.d - The volume of air sampled is
estimated from the calibration of the critical orifice before and after
sampling (flow-rate readings before and after sampling if a rotameter is
used) and the sampling period time. Such estimate assumes that any change
in flow rate during the sampling period is linear with time. Nonlinear
changes due to such things as temporary plugging of the line or critical
orifice from condensed moisture are not detected by this method. Also,
a system leak between the absorber and the critical orifice would introduce
an error in the calculated volume unless detected and corrected by the
operator prior to sample collection.
Errors in the calculated volume are checked as part of the auditing process
by using a wet test meter on-site to measure the integrated volume. A
discrepancy in the integrated volume measured by the wet test meter and the
volume calculated in the usual manner implies that there could be:
(1) system leaks,
(2) non-linear changes in flow-rate,
(3) an error in the timer, or
(4) flow-rate calibration error.
Each item should be checked and verified.
68
-------
For the 24-hour sampling period where a calibrated rotameter is used to
read the flow rate before, during, and after the sampling period, only
items (1) and (4) above can be detected. There is only a small possibility
of detecting a temporary change in flow rate (item 2) . The tinier should be
checked independently against an elapsed time indicator at least every six
months.
An additional source of error in estimating the sample volume is due to
the inability to determine an average temperature- for the sampling period.
This is more important for 24-hour samples than it is for the shorter
periods. A method for estimating the average temperature is suggested
in the Operations Manual, page 34.
Elapsed time between sample collection and analysis - It has been shown that
exposure of the sample to temperatures above about 5°C results in SO^ losses.
A loss of about 1.8 percent per day occurred at 25°C and no losses were
observed at 5°C (Ref. 2).
It is obvious that long delays between sample collection and analysis at
unrefrigerated or uncontrolled conditions will result in sizeable errors.
Every means should be employed to minimize the time that a sample is
exposed to temperatures above 5°C.
jJ Aoumpti to ctiJie.c£ Aan^igkt - Exposure of the sample to direct
sunlight during or after collection can result in deteoriation of the
sample. Losses from 4 to 5 percent were experienced by samples exposed
to bright sunlight for 30 minutes (Ref. 2). Therefore, it is recommended
that the bubbler be wrapped in tin foil anytime it is exposed to the direct
sunlight for more than 1 or 2 minutes that might occur in the normal
transfer of 24-hour samples from the sampling train box to the shipping
block.
EwtSULLm
-------
Sample Analysis - To realize a high level of precision and accuracy from
the analysis phase of the method the analyses must be done carefully, with
close attention to temperature, pH of final solution, purity of the
chemicals and water, and standardization of the sulfite solution (Ref. 2).
Important parameters in the analysis phase include:
(1) purity of the chemicals and water,
(2) spectrophotometer calibration,
(3) pH of the solution being analyzed, and
(4) temperature at analysis compared to temperature
at calibration.
Pu/l^Cj/ OfJ dnJUMJ^]JLt> ami WcuteA - Purity of the chemicals and water is
extremely important in obtaining reproducible results because of the high
sensitivity of the method. As recommended in the Operations Manual the
purity of the pararosaniline dye is checked and, if necessary, purified and
assayed before use. The water used must be free of oxidants and should be
double-distilled when preparing and protected from the atmosphere when
stored and transferred from container to container. All other chemicals
should be ACS grade and special care exercised not to contaminate the
chemicals while in storage .or when removing portions for reagent
preparation.
ca&i.faAa£ton - The spectrophotometer should be adjusted
for 100 percent transmittance when measuring a sample cell of pure water
prior to each set of determinations. Also, the calibration of the wave-
length scale and the transmittance scale should be checked periodically or
any time reference samples cannot be measured within prescribed limits after
checks of the reagents and water have failed to identify the trouble.
The calibration of the wavelength scale can be checked by plotting the
absorption spectrum (in the visible range) of a didymium glass which has
been calibrated by the National Bureau of Standards.
The transmittance scale can be checked using a set of filters from the
National Bureau of Standards.
pH 0& the. -&o£o£uw bexjtg anatyzzd - For this method the maximum sensitivity
occurs when the pH of the final solution is in the region 1.6 + 0.1 (Ref. 2)
If care is exercised in reagent preparation and the analysis phase the pH
should always be within the above limits. Phosphoric acid provides the pH
control; hence, it should be checked first when the pH is detected outside
the above range.
70
-------
- Temperature affects the rate of color formation and fading
of the final color. Also, the reagent blank has a very high temperature
coefficient. Therefore, it is important that the temperature at analysis
be within + 2°C of the temperature at which the calibration curve was
developed. If the normal room temperature varies more than + 2°C from a
set value it is recommended that a constant-temperature bath be used if
a high degree of accuracy is desired.
Large variations in temperature would be detected as error in measuring
control samples and/or reference samples.
Data Processing - Data processing, starting with the raw data through the
act of recording the measured concentration on the SAROAD form, is subject
to many types or errors. The approach used in the Operations Manual,
page 52 means that one can be about 55 percent confident that no more
than 10 percent of the reported concentrations are in error by more than
+ 3 percent.
The magnitude of data processing errors can be estimated from, and
controlled by, the auditing program through performance of periodic checks
and making corrections when large errors are detected. A procedure for
estimating the bias and standard deviation of data processing errors is
given in the Management Manual, page 98.
How to Monitor Important Parameters
Table 3 lists sbme of the parameters that need to be monitored when using
the manual pararosaniline method for measuring atmospheric SO-. Excess
error attributable to one or more of variables 1, 2, 6, 7, 8, and 9 will
be detected by the auditing program. Variable 3, elapsed time between
sample collection and analysis is determined for each sample as part of
the normal operating procedures. Variables 3 and 4 involve an operation or
subjective decision by the field operator. They can only be monitored by
observing the operator on-site. This can be carried out in conjunction
with the performance of the volume/flow-rate audit.
Suggested Control Limits
Appropriate control limits for individual variables will depend on the
level of performance needed. Table 4 gives suggested performance
standards for measuring reference samples and for volume/flow-rate checks.
Standards given for the measurement or determination of the sampled volume
or of the average flow rate for the sampling period are merely estimates.
They are not based on actual data. The error involved in making this
determination has been divided into five components. They are: 1) errors
in calibrating the rotameter or critical orifice, 2) system leaks between
the bubbler and flowmeter, 3) intermittent plugging of lines or critical
orifice, loss of power, etc., 4) error in determining the average
temperature during the sampling site, and 5) the error in determining the
average pressure during the sampling period.
71
-------
Table 3. METHODS OF MONITORING VARIABLES
Variable
Method of Monitoring
1. Flow-rate calibration
2. Volume of air sampled
3. Elapsed time between sample
collection and analysis
4. Exposure of sample to direct
sunlight
5. Entrainment or loss of sample
other than by evaporation
6. Purity of chemicals and water
7. Spectrophotometer calibration
8. pH of final solution
9. Temperature
10. Data processing errors
Measurement of volume/flow rate as
part of the auditing process.
Same as (1) above.
Dates of collection and analysis are
taken from the Sample Record Sheet
and Operational Data Log Book,
respectively, for every sample.
Observe operator technique on-site.
Observe operator technique on-site.
Purity of pararosaniline and water
are checked as part of the oper-
ating procedure. Other chemicals
are purchased as ACS grade.
Spectrophotometer calibration is
checked with calibrated filters and/
or didymium glass when a reference
sample cannot be measured within
limits.
Measure the pH of the final
solution of 1 sample from each site
when first starting a quality
assurance program and once a month
afterwards.
Temperature should be monitored with
a wall thermometer, or if a constant
temperature bath is used, a ther-
mometer immersed in the bath.
Data processing checks are performed
as a part of the auditing program.
72
-------
Table 4: SUGGESTED CONTROL LIMITS FOR PARAMETERS AND/OR VARIABLES
Parameter /Variables
1. Determination of volume/average flow rate
a) Calibration error
b) System leaks between bubbler or flowmeter
c) Intermittent plugging of lines
d) Error in average temperature of
sampling period
e) Error in average pressure of sampling
period
Total:
i
d,=d + d, + d + d, + d
1 a D c d e
V 2 2 2 2 2
a + a, -fa + a, + a
abode
2. Measurement of reference samples
Suggest
Mean
dj_ = -0.04(X)*
da = o
dfe = -0.02(X)
d = -0.02(X)
c
dd = 0
de = o
dj = -0.04(X)
d2 = 0
ed Performance S
Standard
Deviation
o^ = 0.025(X)
a = 0.02(X)
a, = 0.007(X)
b
a = 0.007(X)
c
a. = 0.01(X)
d
a = 0.006(X)
e
a^ = 0.025(X)
a2 = 0.4 yg S02
andards
Upper Limit
(+3o)
-0.10(X), +0.06(X)
+1.2 yg S02
*-
X = mean or average value.
-------
System leaks and Intermittent plugging of lines or loss of electrical power
would result in a calculated volume of air larger than the volume actually
sampled. This in turn would negatively bias the measured concentration.
Therefore, d^ and d of Table 4 are shown as negative biases. The other
errors are assumed to be normally distributed about a mean value of zero.
More accurate estimates of these variables can be made as data from the
quality assurance program becomes available.
Combining the means and standard deviations of component errors as
d = d +d+d + d. + d
1 a T> c d e
and
•A/a2 + a2 + a2 + a2 + a2
shows that at this level of control the suggested performance standard
for measuring the volume /flow-rate is satisfied as is evidenced by
and
a " °
r
The suggested standard for the measurement of reference samples is the value
obtained in a collaborative test of the method (Ref. 6).
PROCEDURES FOR IMPROVING DATA QUALITY
Quality control procedures designed to control or adjust data quality may
involve a change in equipment or in operating procedures. Table 5 lists
some possible procedures for improving data quality. The applicability
or necessity of a procedure for a given monitoring situation will have to
be determined from results of the auditing process or special checks
performed to identify the important variables. The expected results are
given for each procedure in qualitative terms. If quantitative data are
available or reasonably good estimates can be made of the expected change
in data quality resulting from implementation of each procedure, a graph
similar to that in Figure 22, page 105 of the Management Manual can be
constructed. The values used in Table 13 and Figure 22 are assumed and
were not derived from actual data.
For making cost estimates, a reference system consisting of a sampler
equipped with a rotameter and the routine performance of those control
checks spelled out in the Operations Manual is assumed.
74
-------
Table 5. QUALITY CONTROL PROCEDURES OR ACTIONS
Procedure/Action
AO Reference Conditions
Al Correct for
Temperature
A2 Minimize sample
shipping and
storage time
A3 Temperature
control
A4 Personnel
training
Description of
Action
System using routine
procedure as given in
the Operations Manual
For 24-hour samples
obtain from weather
station or make on-
site measurement of
temper atu re . Es t imat e
an average value for
the period and correct
sampled volume to
298°K.
Minimize time between
sample collection and
analysis by mailing
sample immediately
after collection, use
special mailing, and
analyze as soon after
reaching the labora-
tory.
Use a constant-temper-
ature bath for cali-
bration and analysis.
Hold a 1-week work
shop for operators
from all phases of
the measurement
process once a year.
Expected Results
bias (T) =-0.04(X)*
OT =(41)[0.7+0.001(X)]
Reduces error in
calculated sample volume.
Reduce the error due the
loss of SC>2 with time
before analysis.
Reduce variability in the
analytical phase due to
temperature fluctuations .
Reduce operational errors
through increased aware-
ness of the pitfalls and
increased interest in
doing a good job.
Costs
Equip ."""
15
100
250
none
Personnel
none
100
10
200
Total
15
200
260
200
*_
X = true concentration of SO
2'
-------
Equipment, manpower requirements, and the continuing cost of labor and
supplies are estimated for each procedure. For these estimates technician
time was valued at $5 per hour and engineering time at $10 per hour. Equip-
ment life was taken as 5 years. All calculations were based on a sample lot
of 100 and an average sampling rate of 60 samples per year per sampling site.
A procedure for selecting the appropriate quality control procedure to
insure a desired level of data quality is given below:
(1) Specify the desired performance standard, that is,
specify the limits within which you want the deviation
between the measured and the true concentration to
fall a desired percentage of the time. For example,
to measure within + 12 percent of the true value,
95 percent of the time, the following performance
standards must be satisfied:
I* ± 2STI 1 °-12 x V8 S02/m3.
(2) Determine the system's present performance level from
the auditing process, as described on page 101 of the
Management Manual, by setting
and
a
T
If the relationship of (1) above is satisfied, no
control procedures are required.
(3) If the desired performance standard is not satisfied,
identify the major error components by performing
special checks.
(4) Select the quality control procedure(s) which will
give the desired improvement in data quality at the
lowest cost. .Figure 22 on page 105 of the Management
Manual illustrates a method for accomplishing this.
The relative position of actions on the graph in Figure 22 will differ for
different monitoring networks according to type of equipment being used,
available personnel, and local costs. Therefore, each network would need
to develop its own graph to aid in selecting the control procedure
providing the desired data quality at the lowest cost.
76
-------
PROCEDURES FOR CHANGING THE AUDITING LEVEL TO GIVE THE DESIRED LEVEL OF
CONFIDENCE IN THE REPORTED DATA
The auditing process does not in itself change the quality of the reported
data. It does provide a means of assessing the data quality. An Increased
auditing level increases the confidence in the assessment. It also increases
the overall cost of data collection.
Various auditing schemes and levels are discussed on page 82 of the
Management Manual. Numerous parameters must be known or assumed in order
to.arrive at an optimum auditing level. Therefore, only two decision
rules with two levels of auditing each will be discussed here.
For conditions as assumed on page 82 of the Management Manual, a study of
Figure 20, page 95, gives the following results. These conditions may or
may not apply to your operation. They are included here to call attention
to a methodology. Local costs must be used for conditions to apply to
your operation.
Decision Rule - Accept the Lot as Good If No Defects^ Are Found
(i.e.,_d - 0).
Most Cost Effective Auditing Level - In Figure 20 the two solid lines are
applicable to this decision rule, i.e., d = 0. The cost curve has a
minimum at n = 7 or an auditing level of 7 checks out of 100 sampling
periods. From the probability curve it is seen that at this auditing
level there is a probability of 0.47 of accepting a lot as good when the
lot (for N = 100) actually has 10 defects. The associated average cost
is 240 dollars per lot.
Auditing Level for Low Probability of Accepting Bad Data - Increasing the
auditing level to n = 20, using the same curve in Figure 20 as above,
shows a probability of 0.09 of accepting a lot as good when the lot
actually has 10 defects. The average cost associated with this level of
auditing is approximately 425 dollars per lot.
Decision Rule - Accept the Lot as Good If No More^ Than One (1) Defect
Is Found (i.e., d <. 1).
Most Cost Effective Auditing Level - From the two dashed curves in
Figure 20 it can be seen that the cost curve has a minimum at n = 14. At
this level of auditing there is a probability of 0.55 of accepting a lot
of data as good when it has 10 defects. The average cost per lot is
approximately 340 dollars.
Auditing Level for_Low Probability of Accepting Bad Data - For an auditing
level of n = 20 the probability of accepting a lot with 10 percent defects
is about 0.36 as read from the d _< 1 probability curve. The average cost
per lot is approximately 375 dollars.
It must be realized that the shape of a cost curve is determined by the
assumed costs of performing the audit and of reporting bad data. These
costs must be determined for individual monitoring situations in order
to select optimum auditing levels.
77
-------
MONITORING STRATEGIES AND COST
Selecting the optimum monitoring strategy in terms of cost and data quality
requires a knowledge of the present data quality, major error components,
cost of implementing available control procedures, and potential increase
in system precision and accuracy.
A methodology for comparing strategies to obtain the desired precision of
the data is illustrated in the Management Manual, page 103, Table 5,
page 75 lists control procedures with estimated costs of implementation
and expected results in terms of which error component (s) are affected by
the control. Numerical values of the expected results as given in Table 13,
page 106, are estimates and were not derived from actual data.
Three system configurations identified as best strategies in Figure 22,
page 105, of the Management Manual are summarized below.
Again, local costs and expected results derived from field data are
required to select optimum strategies by this method.
Reference Method (AO)
Description of Method - This refers to a typical sampling operation as
described in the Federal Register. Routine operating procedures as given
in the Operations Manual are to be followed with special checks performed
to identify problem areas when performance standards are not being met.
Corrections for delay between collection and analysis and for average
temperature are not made. An auditing level of n=7, N=100 is to be carried
out for this strategy. This method or strategy is identified as AO in
Table 13 and Figure 22 in the Management Manual.
Costs - Taken as reference or zero cost.
Data Quality - Data quality can be described by
(yg S09/m3) = (yg SO /m3) - ? + 3o -
£ rv £ m 1
where (yg S0«/m ) = true average concentration of sulfur
T dioxide, and
o
(yg S0_/m ) = measured average concentration of
m sulfur dioxide.
78
-------
Taking the hypothesized values of the bias and standard deviation from
Table 13 and using in the above relationship shows that for a true
3 o
concentration, (yg S02/m ) , of 380 yg S02/m , the measured value,
(yg S09/m ) , would fall within the following limits
* m
308 < (yg S0,/m3) < 422
L m
approximately 99.7 percent of the time.
Modified Reference Method (A5)
Description of Method^ - This strategy is identical to the reference
method above except that corrections are made for the average temperature
for the sampling period (Al) and special personnel training in the form
of a yearly workshop.
Costs - The average cost per 100 samples is estimated at 215 dollars above
the cost of the reference method (see Figure 22, page 105).
Data Quality - From Table 13, values of bias and standard deviation are
seen to be T = -0.03 percent and ST = C """ ' -' " - --- - -------
The data quality would be described by
(yg SO,/m3)T = (yg S0,/m3) + (0.03 4- 3 * 0.035)(yg S0,/m3) .
* 1 L m ~ ' T
For a true concentration 380 yg S0»/m the measured value would fall
within the limits
328 < (yg S00/m3) < 409
L m
approximately 99.7 percent of the time.
Modified Reference Method Plus Action A2 (A7 = Al + A2 + A4)
Description of Method - This method is identified as A7 in Figure 22 of
the Management Manual. This method is the same as the modified reference
method above with the addition of Action A2 which would reduce errors due
to loss of SO. by minimizing time between collection and analysis.
79
-------
Costs - Average cost per lot is estimated at 415 dollars above the cost
of the reference method.
Data Quality - From Table 13 the data quality would be described by
(yg S0,/m3) = (yg S07/m3) + (0.01 + 0.03 ) (yg S0,/m3) .
2 T l m ~~ L T
a
For a true concentration of 380 yg S0_/m the measured value would fall
within the limits
336 < (yg S07/m3) < 416 .
m
Results from these estimated values show that in going from Method AO to
Method A7, the data spread is decreased by about 24 percent and the range
is more evenly distributed about the true concentration value.
80
-------
SECTION IV MANAGEMENT MANUAL
GENERAL
The objectives of a data quality assurance program for the Pararosaniline
Method of measuring the concentration of S0_ in the ambient air were given
in Section I. In this part of the document, procedures will be given to
assist the manager in making decisions pertaining to data quality based on
the checking and auditing procedures described in Sections II and III.
These procedures can be employed to:
(1) determine the extent of independent auditing to be
performed,
(2) detect when the data quality is inadequate,
(3) assess overall data quality,
(4) relate costs of data quality assurance procedures
to a measure of data quality, and to
(5) select from the options available to the manager
the alternative(s) which will enable him to meet
the data quality coals by the most cost-effective
means.
The determination of the extent of auditing is considered in the subsection
entitled "Auditing Schemes." Objectives 2 and 3 are discussed in the
subsection entitled "Data Quality Assessment," page 96. Finally,
Objectives 4 and 5 above are described in the subsection entitled "Data
Quality Vs. Cost of Implementing Actions," page 103. The cost data are
assumed and a methodology provided. When better cost data become
available, improvements can be made in the management decisions.
If the current reference system is providing data quality consistent with
that required by the user, there will be no need to alter the physical
system or to increase the auditing level. In fact several detailed
procedures could be bypassed if continuing satisfactory data quality is
implied by the audit. However, if the data quality is not adequate, e.g.,
either a large bias and/or imprecision exists in the reported data, then
(1) increased auditing should be employed, (2) the assignable cause
determined, and (3) the system deficiency corrected. The correction can
take the form of a. change in the operating procedure, e.g., minimize
delay between collection and analysis of sample by special mailing; or
it may be a change in equipment such as the installation of an improved
temperature control system. An increase in the auditing level will
increase the confidence in the reported measure of precision/bias and aid
in identifying the assignable cause(s) of t.he large deviations. The level
of auditing will be considered in the next subsection.
81
-------
The audit procedure and the reported results can serve a two-fold purpose.
They can be used to (1) screen the data, by lots of say N = 50 or 100, to
detect when the data quality may be Inadequate, and (2) calculate the bias
and precision of the audited measurement and hence estimate the bias/
precision of the final reported concentration of SO. In the ambient air.
In order to perform (1), suggested standards are provided for use In
comparing the audited results with the reported values and a defect Is
defined In terms of the standards. This approach requires only the
reporting of the number of defects In the n auditing checks. In the second
method above, It Is required to report the measures of bias/precision in
the audits. These values are then used in assessing the overall data
quality. Approach (1) is suggested as a beginning step even though it will
not make maximum use of the data collected in the auditing program. The
simplicity of the approach and the explicit definition of a defect will
aid in its implementation. After experience has been gained in using the
auditing scheme and in reporting and calculating the results, it is
recommended that approach (2) be implemented.
It is Important that the audit procedure be independent of previously
reported results and be a true check of the system under normal operating
procedures. Independence can be achieved, for example, by providing a
control sample of unknown concentration of SO- to the operator and
requesting that he measure and report the concentration of the sample.
To Insure that the check is made under normal operating procedures, it
is required that the audit be performed without any special check of the
system prior to the audit other than that usually performed during each
sampling period.
AUDITING SCHEMES '
Auditing a measurement process costs time and money. On the other hand,
reporting poor quality data can also be very costly. For example, the
reported data might be used to determine a relationship between health
damage and concentrations of certain pollutants. If poor quality data are
reported, it is possible that invalid inferences or standards derived from
the data will cost many dollars. These implications may be unknown to the
manager until some report is provided to him referencing his data; hence,
the importance of reporting the precision and bias with the data.
Considering the cost of reporting poor quality data, it is desirable to
perform the necessary audits to assess the data quality and to invalidate
unsatisfactory data with high probability. On the other hand, if the data
quality is satisfactory, an auditing scheme will only increase the data
measurement and processing cost. An appropriate tradeoff or balance of
these costs must be sought. These costs are discussed under Cost
Relationships.
82
-------
Now consider the implication of an auditing scheme to determine or judge
the quality of the reported data in terms of an acceptance sampling
scheme. Let the data be assembled into homogeneous lots of N = 50 or
100 sampling periods. Suppose that n periods are sampled in the manner
suggested in Section III. That is, the N = 50 or 100 sampling periods are
subdivided into equal time intervals (as nearly equal as possible); then
one day is selected at random during each interval. Figure 17 gives a
diagram of the data flow, sampling, and decision making process for an
auditing level of n = 7.
Statistics of Various Auditing Schemes
Suppose that the lot size is N = 100 periods (days), that n = 7 periods
are selected at random, and that there are 5% defectives in the 100, or
5 defectives. The probability that the sample of 7 contains 0, 1, ..., 5
defectives is given by the following.
p(0 defectives) =
10
( 7 )
and for d defectives
p(d defectives) = x"'v' " , d < 5.
\ 7 )
The values are tabulated below for d = 0, 1, ..., 6 and for the two data
quality levels.
Table 6. P(d defectives)
Data Quality
D=5% Defectives D=15% Defectives
0
1
2
3
4
5
6
*(5\ (95\ 5! 95!
\0/ V 7/ 0'5' 7!88!
/100\ / 100! V
\ 7 ) \,7!93! )
0.6903
0.2715
0.0362
0.0020
0.00004
« 0
= 0
95. 94. ..89
100'99-"94
a?
0.3083
0.4098
0.2152
0.0576
0.0084
0.0007
«0
= 0.6903.
-------
Data Flow
Lot 1
H - 100
Day*
Lot 2
N » 100
Days
Sample
n - 7
Periods (days)
Observe
d • 0 defect*
Observe
d - 1 defect
Calculate Costs of
Accepting and
Rejecting the Lot
Accept Data If
1. Cost Comparison
Favors This Action
2. Data Quality Is
Acceptable
Reject Data
Otherwise
Figure 17: Data Flow Diagram for Auditing Scheme
84
-------
M
0)
4J
CO
t
u
01
-------
: d <; 1, D - 6X
-- d • 0, D - 20Z
10 15
Sample Size (n)
Figure 18B: Probability of d Defectives in the Sample If the
Lot (N • 50) Contains DZ Defectives.
This graph is for a lot sice of N « 50. Only whole numbers of defectives
are physically possible; therefore, even values of D (i.e., 6, 10, and
20 percent) are given rather than the odd values of 5 and 15 percent as
given in Figure ISA.
86
-------
Figure 18A gives the probabilities of d = 0 and d <_ 1 defectives as a
function of sample size. The probability is given for lot size N = 100,
D ™ 5 and 15% defectives, for sample sizes (auditing levels) from 1 to 25.
For example, if n = 10 measurements are audited and D = 5% defectives,
the probability of d = 0 defectives is 0.58. Figure 18B gives the proba-
bilities for lot size N = 50, for D = 6, 10, and 20% defectives, and for
d = 0 and d £ 1. These curves will be used in calculating the cost
relationships.
Selecting the Auditing Level
One consideration in determining an auditing level n used in assessing
the data quality is to calculate the value of n which for a prescribed
level of confidence will imply that the percent of defectives in the
lot is less than 10 percent, say, if zero defectives are observed in the
sample.* Figures 19A and 19B give the percentage of good measurements
in the lot sampled for several levels of confidence, 50, 60, 80, 90, and
95%. The curves in Figure 19A assume that 0 defectives are observed in
the sample, and those in Figure 19B that 1 defective is observed in the
sample. The solid curves on the figures are based on a lot size of N = 100;
two dashed curves are shown in Figure 19A for N = 50; the differences
between the corresponding curves are small for the range of sample sizes
considered.
\
For example, for zero defectives in a sample of 7 from a lot of N = 100,
one is 50% confident that there are less than 10% defective measurements
among the 100 reported values. For zero defectives in a sample of 15
from N = 100, one is 80% confident that there are less than 10% defective
'measurements. Several such values were obtained from Figure 19A and
placed in Table 7 below for convenient reference.
Table 7. REQUIRED AUDITING LEVELS n FOR LOT
SIZE N = 100 ASSUMING ZERO DEFECTIVES
Confidence Level D = 10% 15% 20%
50%
60%
80%
90%
95%
7
9
15
20
»25
<5
6
10
15
18
<5
<5
8
11
13
Obviously, the definition of defective need not always be the same and
must be learly stated each time. The definitions employed herein are
based on results given in the collaborative test report (Ref. 6).
87
-------
N - 100
N - 50
10 15
Sample Size (n)
Figure 19A: Percentage of Good Measurements Vs. Sample Size
for No Defectives and Indicated Confidence Level
88
-------
L© Sis© (n)
19B:
89
-------
Cost Relationships
The auditing scheme can be translated into costs using the costs of
auditing, rejecting good data, and accepting poor quality data. These
costs may be very different in different geographic locations. There-
fore, purely for purposes of illustrating a method, the cost of auditing
is assumed to be directly proportional to the auditing level; For n • 7
it is assumed to be $155 per lot of 100. The cost of rejecting good
quality data is assumed to be $600 for a lot of N - 100. The cost of
reporting poor quality data is taken to be $800. To repeat, these costs
given in Table 8 are assumed for the purpose of illustrating a methodology
of relating auditing costs to data quality. Meaningful results can only
be obtained- by using correct local information.
Table 8. COSTS VS. DATA QUALITY
Data Quality
Reject Lot of
Data
"Good"
"Bad"
D ^ 10%
Incorrect Decision
D > 10%
Correct Decision
Lose cost of performing Lose cost of performing
audit plus cost of reject- audit, save cost of not
ing good quality data. permitting poor quality
(-$600 - $155) data to be reported.
($400 - $155)
Accept Lot of
Data
Correct Decision
Lose cost of performing
audit.
(-$155)
Incorrect Decision
Lose cost of performing
audit plus cost of de-
claring poor quality
data valid.
(-$800 - $155)
Suppose that 50 percent of the lots have more than 10 percent defectives
and 50 percent have less than 10 percent defectives. (The percentage of
defective lots can be varied as will be described in the final report
under the contract.) For simplicity of calculation, it is further assumed
that the good lots have exactly 5 percent defectives and the poor quality
lots have 15 percent defectives.
Cost of performing audit varies with the sample size; it is assumed to
be $155 for n - 7 audits per N - 100 lot size.
90
-------
Suppose that n - 7 aeasurements out of a lot of N «• 100 have been audited
and none found to be defective. Furthermore, consider the two possible
decisions of rejecting the lot and accepting the lot and the relative
costs of each. These results are given in Tables 9A and 9B.
Table 9A. COSTS IF 0 DEFECTIVES ARE OBSERVED AND THE LOT IS REJECTED
Reject Lot
D - 5Z
D - 15%
Correct
Decision
P2 • 0.31
C2 =400 - 155
Incorrect
Decision
Px = 0.69
Cj^ = -600 - 155
Net Value ($)
p^ - -$521
P2C2 = $76
Cost
+ p2C2--$445
Table 9R. COSTS IF 0 DEFECTIVES ARE OBSERVED AND THE LOT IS ACCEPTED
Accept Lot
D - 5%
D - 15%
Correct
Decision
pj - 0.69
C3 - -155
Incorrect
Decision
P2 = 0.31
C4 - -800 - 155
Net Value ($)
PXC3 = -$107
P2C4 = -$296
Cost
- -$403
The value PjCp*) ln the above table is the probability that the lot is
5% (15Z) defective given that 0 defectives have been observed. For
example ,
91
-------
/probability that the lot is 5X defective\
\ and 0 defectives are observed /
pl " ~7lot is 5X defective and\ t/lot is 15* defective and>
p\ 0 defectives observed / \ 0 defectives observed )
0.5(0.69) + ' 0.5(0.31)
/probability that the lot is 15Z defective)
\ and 0 defectives are observed /
P2 " /lot is 5Z defective and
p\ 0 defectives observed
A + /lot is 15Z defective and\
/ P\ 0 defectives observed /
0.5(0.31)
0.5(0.31) + 0.5(0.69)
- 0.31
It was assumed that the probability that the lot is 51 defective is 0.5.
The probability of observing zero defectives, given the lot quality is
5% or 15X, can be read from the graph of Figure ISA.
A similar table can be constructed for 1, 2, ..., defectives and the net
costs determined. The net costs are tabulated in Table 10 for 1, 2, and
3 defectives. The resulting costs indicate that the decision preferred
from a purely monetary viewpoint is to accept the lot if 0 defectives are
observed and to reject it otherwise. The decision cannot be made on this
basis alone. The details' of the audit scheme also affect the confidence
which can be placed in.the data qualification; consideration must be
given to that aspect as well as.to cost.
Table 10. COSTS IN DOLLARS
Decision
Reject Lot
Accept Lot
0
-445
-403
d • number of
1
-155
-635
defectives
2
+101
-839
3
+207
-928
Cost Versus Audit Level
After the decision criteria have been selected, an average cost can be
calculated. Based on the results of Table 10, the decision criterion
is to accept the lot if d - 0 defectives are observed and to reject the
lot if d • 1 or more defectives are observed. All the assumptions of the
previous section are retained. The auditing level is later varied to obtain
the data in Figure 20, page 95.
92
-------
One example calculation is given below and summarized in Table 11. The
four cells of Table 11 consider all the possible situations which can
occur, i.e., the lots may be bad or good and the decision can be to
either accept or reject the lot based on the rule indicated by Table 10.
The costs are exactly as indicated in Tables 9A and 9B. The probabili-
ties are computed as follows.
• (prob. that the lot is 51 defective and 1 or more
defects are obtained in the sample)
- (prob. that the lot is 5% defective)(prob. 1 or
more defectives are obtained in the sample given
the lot is 5% defective)
- 0.5 (0.31) - 0.155
Similarly q_, q_, and q, in Table 11 are obtained as indicated below.
q£ - 0.5 (0.69) - 0.345
q3 = 0.5 (0.69) « 0.345
q = 0.5 (0.31) - 0.155 .
The sum of all the q's must be unity as all possibilities are considered.
The value of 0.5 in each equation is the assumed proportion of good lots
(or poor quality lots). The values of 0.31 and 0.69 are the conditional
probabilities that given the quality of the lot, either d - 0 or d - 1 or
more defectives are observed in the sample. Further details of the
computation are given in the final report of this contract.
Table 11. OVERALL AVERAGE COSTS FOR ONE
_ ACCEPTANCE - REJECTION SCHEME _ _ __
Decision
Reject any lot of
data if 1 or more
defects are found.
Accept any lot of
data if 0 defects
are found.
Good Lots
D - 5%
qx - 0.155
Cx - -$755
q3 - 0.345
C3 - -$155
Bad Lots
D - 15%
q2 - 0.345
C2 - $245
q4 - 0.155
C4 - -$955
q^ + q2C2 - -$ 32
q3C3 + q4C4 ° ~$202
Average Cost » -$234
93
-------
In order to interpret the concept of average cost, consider a large
number of data lots coming through the system; a decision will be made
on each lot in accordance with the above, and a resulting cost of the
decision will be determined. For a given lot, the cost may be any one
of the four costs, and the proportion of lots with each cost is given
by the q's. Hence the overall average cost is given by the sum of the
product of q's by the corresponding C's.
In order to relate the average cost as given in Table 11 to
the costs given in Table 10, it is necessary to weight the costs in
Table 10 by the relative frequency of occurrence of each observed
number of defectives, i.e., prob(d). This calculation is made below.
Mo. of Decision Costs ($) from
Defectives Rule Table 10 Prob(d) Cost * Prob(d)
d - 0 Accept - 403 0.50 -$201.5
1 Reject • - 155 0.34 - 52.7
2 Reject 101 0.1255 12.6
3 Reject 207 0.030 6.2
4 Reject 244 0.0042 1.0
Totals 0.9997 -$234.4
Thus the value -$234 is the average cost of Table 11 and the weighted
average of the costs of Table 10. The weights, Prob(d), are obtained
as follows:
Prob(d-O) - Prob(lot is good and d-0 defectives are observed)
+ Prob(lot is poor quality and d-0 defectives are observed)
- 0.5(0.69) + 0.5(0.31) - 0.50 .
This is the proportion of all lots which will have exactly 0 defectives
under the assumptions stated. For d - 1, 2, 3, and 4, the values of the
probabilities In parentheses above can be read from Table 6.
Based on the stated assumptions, the average cost was determined for
several auditing levels as indicated in Table 11. These costs are given
in Figure 20. One observes from this figure that n • 7 is cost effective
given that one accepts it only if zero defectives are observed. (See
curve for d-0.)
If the lots are accepted if either 0 or 1 defectives are observed, then
referring to the curve d <_ 1, the best sampling level Is n - 14. The
curve of probability of d - 0 (d £ 1) defectives in a sample of n from a
lot of N - 100 measurements, given that there are 10% defectives in the
lot, is also given on the same figure.
94
-------
Probability
if d - 0
I
Q
0)
N
0)
tH
o.
en
0>
01
o
at
<*j
ai
o
00
c
(LI
ca
o
>»
(0
•8
5 10 15
Audit Level (Sample Size)
Figure 20: Average Cost Vs. Audit Level
(Lot Size N = 100)
95
-------
Another alternative is to accept all data without performing an audit.
Assuming that one-half (50%) of the lots contain more than 10% defectives,
the average cost on a per lot basis would be 0.5(-$800) = -$400. This,
however, would preclude qualification of the data. Regardless of cost,
it would be an unacceptable alternative.
DATA QUALITY ASSESSMENT
In this section, approach 2 is considered; that is, the precisions and
biases of the individual measurements and operational procedures are
estimated. These results are then used to make an assessment of the
data quality. Although it is theoretically possible to make an overall
assessment, e.g., similar to what was done in the collaborative test
program, this is not a practical possibility due to the high cost. The
following audit scheme is one which is considered reasonable in both cost
and effort.
Assessment of Individual Measurements
Assume for convenience that an auditing period consists of N = 100 days
(or sampling periods). Subdivide the auditing period into n equal or
nearly equal periods. Make one audit during each period and compute the
deviations (differences) between the audit values and the stated values
(or previously determined values as measured by the operator) as indicated
in the Supervision Manual. For example, if seven audits (n = 7) are to be
performed over 100 sampling periods (N = 100), the 100 periods can be
subdivided into 7 intervals (6 with 14 periods and 1 with 16 periods).
Select one day at random within each interval and perform the suggested
audits. The operator should not be aware of when the checks are to be
performed.
For sites operating every sixth day, a minimum of three audits per quarter
is recommended. Samples from individual sites can be grouped into logical
lots, e.g., all sites for which a single operator is responsible, to form
data lots of at least 50 samples. This approach insures that the audit
level will exceed n = 7 for the combined sites and resulting data.
In order to assess the data quality using measures of bias/precision, the
checks are to be combined for the selected auditing"period, and the mean
difference or bias and the standard deviation of the differences are to
be computed as indicated below.
The formulas for average bias and the estimated standard deviations are
the standard ones given in statistical texts (e.g., see Ref. 9). The
level of sampling or auditing, n, will be considered as a parameter to
be selected by the manager to assess the quality of data as required.
96
-------
(1) Flow Rate/Volume Check
n
Bias = d,
'1 n
where
d.. . = percent difference in the average flow rate
using the three-point and two-point approxi-
mation, or if suitable instrumentation can be
obtained to measure the integrated volume, the
percent difference in volumes computed by the
auditing equipment and by regular means using
the initial and final flow rates (see page 46).
Standard Deviation = s^
where
-------
Standard Deviation
where
.K.i - v2
V .-i •
« ° the average bias, and
s. = the estimated standard deviation of the
differences in the measured and known
.
concentration of control samples.
(3) Data Processing Check
n
Bias = d
, ,
J n
where
d,. = percent deviation of the concentration of SO^
as calculated by the operator and by that
calculated from the audit.
Standard Deviation = s» =
>J« - V
~\ n - 1
where
d, = the average bias, and
s_ = the estimated standard deviation of the
data processing errors.
Individual checks on the standard deviations of the three audits can be
made by computing the ratio of the estimated standard deviation, s , to
the corresponding suggested performance standard, 0 , given in Table 12.
If this ratio exceeds values given in Table 12 for any one of the two
audits, it would indicate that the source of trouble may be assigned to
that particular aspect of the measurement process. Critical values of
this ratio are given in Figure 21 as a function of sample size and two
levels of confidence. Having assessed the general problem area, one then
needs to perform the appropriate quality control checks to determine the
specific causes of the large deviations.
98
-------
Table 12. CRITICAL VALUES OF s./o
Level of
Confidence Statistic
90% sjL/a1
95% si/ai
•L X
Audit Level
n=5
1.40
1.54
n-10 n-15
1.29 1.23
1.37 1.30
n-20
1.20
1.26
n-25
1.18
1.23
s. = estimated standard deviation
o. = hypothesized or suggested standard deviation.
Audit
Flow Rate (or Volume) theck
Control Sample
Data Processing Check
Suggested Performance Standard
o1 = 0.025 x yg SO
o2 = 0.4 yg S02*
a = 0.03 * yg S02/m3
*a is based on information in Ref. 6 it must be converted to a percent
for use in the following subsection.
Overall^ Assessment of Data Quality
The values of d^, d2, d^, s.^, s2 and s» above measure the bias and
variation of the reported data for the three selected audits. As stated
previously, these three audits do not completely evaluate the measurement
process. A partial or limited evaluation is used because of the high cost
of a complete evaluation and the unavailability of a cheap and reliable
method of generating test atmospheres of specified concentration with high
precision from one laboratory to another. These three measures can be
combined to estimate the bias and standard deviation of that aspect of the
process which is audited. The standard deviation of the measured concentra-
tion of S02 is calculated by using the individual standard deviations expres-
sed in percent error as
(in %)
99
-------
1.60
1.50
m
o
o
-------
Using the estimated coefficient of variation OT (in percent) and the
mean concentration of SCL one obtains
ST (yg S02/m3) = 3T (in %) • yg S02/m3 ,
an estimated standard deviation of the part of the measurement process
which is audited.
Development of a Model - The final report of this project discusses a
modeling technique for combining the errors of observation and/or the
variation of environmental effects to make an overall assessment and the
assumptions required in the use of the technique. The basis for the model
is the equation for the measured concentration of S0~, i.e.,
, (A - A ) 103 B
(1) yg S02/mJ 2_ -S. x D
R
where the individual parameters are defined in the Federal Register and
included in the Appendix for convenience. Each of the variables in
equation (1) is further modeled in terms of the errors/variations in the
method of measurement, operator, and environmental effects as described in
the final report on this project. When the model is completely formulated,
an analysis of the mean concentration (or bias) and its estimated standard
deviation can be performed. For the assumed variations in each of the
variables the bias was negative and about 5%, the standard deviation was
3
about 10%, both based on 400 yg/m as the true concentration. The bias
estimate can be obtained by substituting into the computational formula
3
for yg S02/m the averages adjusted for the measured biases and thus
estimating the bias on the reported concentration. This calculated bias
is denoted by T. The composite of these two measures into a mean square
error (MSE), i.e.,
MSE ="y(Bias effect)2 + (Std. Dev.)2
= 11.2%
yields a value which might be compared to reproducibility of results
obtained in the field. It must be emphasized that to obtain these
results, estimates of the standard deviations of results were obtained from
Ref. 6 when available and by statistical or engineering judgment otherwise.
101
-------
The estimates of T and a obtained from the model can be used in reporting
the bias and precision as suggested by the following. The true
concentration of SO. should fall in the following interval, where
3
(yg S00/m ) is the measured concentration,
* m
(yg S0,/m3) - T + 25
t- m — JL
approximately 95% of the time, or within the interval
(yg S0_/m3) - T + 3o_,
*• m 1
approximately 99.7% of the time. When computed from audit data, the
coefficients of o_ are actually dependent on the number of audits conducted,
If n is large, say about 25 or larger, the value 2 (or 3) is appropriate.
In reporting the data quality, the bias, overall standard deviation, and
auditing level should be reported in an ideal situation (see the section
entitled "Data Presentation" for further discussion). More restricted
information following approach 1 is suggested in the Supervision Manual
as a minimal reporting procedure.
If the overall reported precisions/biases of the data meet or satisfy the
requirements of the user of the data, then a reduced auditing level may be
employed; on the other hand, if the data quality is not adequate, assign-
able causes of large deviations should be determined and appropriate
action taken to correct the deficiencies. This determination may require
increased checking or auditing of the measurement process as well as the
performance of certain quality control checks, e.g., monitoring of
temperature variations over the 24-hour sampling period.
Identification of the Important Parameters - The next step in the modeling
process was to use the model to identify the critical parameters, i.e.,
those parameters which may cause the greatest variation in the measured
concentration, SO., if their variation is of the order of magnitude assumed
in the analysis. Two types of analyses were employed to determine the
critical parameters and the combined effect of all the parameters on the
3
variation in the measured concentration of yg S0_/m .
The first type was a sensitivity or ruggedness analysis which identified
and ranked the critical parameters, made certain checks on the adequate
of a linear approximation to the developed model, and estimated the
variation (as measured by the standard deviation) of SO. through the use of
a linear approximation. This latter technique was a straightforward
102
-------
application of error analysis. The second analysis procedure was a
Monte Carlo simulation in which each of the parameters was assigned a
distribution of values: for example, the slope of the calibration curve
was assumed to be normally distributed with given mean and standard
deviation. This simulation analysis provided a listing of the simulated
values of concentration in ascending order and calculated the mean and
standard deviation and other pertinent characteristics of this distri-
bution. These analyses are described in further detail in the final
report of this contract.
Results from the above analyses may not be valid for one specific
situation, but should be a reasonably good evaluation of average precision
and accuracy obtainable over a large population of measurements. The
results indicate that if the operating procedures recommended in the
Operations Manual were adhered to, the measured data would have a mean value
negatively biased from the true value of concentration (about) 5%, and a
standard deviation of approximately 10% of the mean value
o 3
(3_ = 0.10 x ug S02/m ) for a true concentration of 400 yg SOj/m .
Values derived from the above analyses were used to arrive at suggested
performance standards, and to a certain extent, for suggested control
limits given for certain checks in the Operations Manual.
The standard deviation of SCL is a measure of the precision or variation
of the reported values of S0_ as estimated by the model. It is to be
noted that this measure depends on the estimated standard deviations of
each of the variables and on the coefficients in the model, which are
dependent on the form of the model. These values can be checked using
the biases and standard deviations computed from actual field data.
v
DATA QUALITY VERSUS COST OF IMPLEMENTING ACTIONS
The discussion and methodology given in a previous section were concerned
with the auditing scheme (i.e., level of audit or sample size, costs
associated with the data quality, etc.). Increasing the level of audit
of the measurement process does not by itself change the quality of the
data, but it does increase the information about the quality of the
reported data. Hence, fewer good lots will be rejected and more poor
quality data will be rejected. If the results of the audit imply that
certain process measurement variables or operational procedures are major
contributors to the total error or variation in the reported concentration
of SO™, then alternative strategies for reducing these variations need to
be investigated. This section illustrates a methodology for comparing the
strategies to obtain the desired precision of the data. In practice it
would be necessary to experiment with one or more strategies, to determine
the potential increase in precision, and to relate the precisions to the
relative costs as indicated herein. Several strategies are considered,
but only a few of the least costly ones would be acceptable, as illustrated
in Figure 22. The assumed values of the standard deviations and biases for
each strategy and audit are not based on actual data, except for the
reference method. In this case values were taken from the results of the
collaborative test program (Ref. 6).
103
-------
Several alternative actions or strategies can be taken to increase the
precision of the reported data. For example, if'the temperature variations
are large, the measurement methods may vary and, cause variation in
o
lag SOj/m . Under these conditions additional control equipment for
temperature variation can reduce the variation of the measured responses
by calculated amounts and thus reduce the error of the reported concen-
trations. In this manner, the cost of the added controls can be related
to the data quality as measured by the estimated bias/precision of the
reported results.
In order to determine a cost efficient procedure, it is necessary to
estimate the variance for each source of error (or variation) for each
strategy and then to select the strategy or combination of strategies
which yields the desired precision with minimum cost. These calculations
are summarized in Table 13 with assumed costs of equipment and control
procedures. The various strategies are given in Table 5 of Section III.
Suppose that it is desired to make a statement that the true SO- concen-
% £.
tration is within 12% of the measured concentration (for simplicity of
discussion all calculations were made at a true concentration of
380 ug/m ) with approximately 95 percent confidence. Minimal cost control
equipment and checking procedures are to be employed to attain this
desired precision. Examining the graph in Figure 22 of cost versus
precision, one observes that A2 is the least costly strategy that meets
the required goal of 2MSE £ 12 or MSB <_ 6 percent. The mean square error
(MSB = /a2 + r2) is used in this analysis as a means of combining the
bias (T) and standard deviation (a) to obtain a single measure of the
overall dispersion of the data. The assumed values of the MSB's of the
measured concentrations of S0» for the alternative courses of action are
given in Table 13. The costs for the various alternatives are given in
Table 5 of Section III and in Table 13.
Suppose that it is desired that MSE be less than b% and that the cost of
reporting poor quality data increases rapidly for MSE greater than 4%, then
strategy A6 = (A2 + A4) appears best because it meets the goal of MSE being
less than 4%, its costs of implementation is $400/100 samples. However,
strategy A5 costs $215 to implement and results in a cost of about $40 for
reporting poor quality data; an overall cost of $255 compared to $400 for
A6. Based on the assumed values A5 would be best. This example demon-
strates the need for the manager to obtain estimates of the improvements
in data quality which can be attained through various actions. This
assumption is illustrated by the cost curve given by the solid line in
Figure 22^ For any alternative strategy, the cost of reporting poor
quality data is given by the ordinate of this curve corresponding to the
strategy.
104
-------
o
Ul
,rt 400
\jj
LJ
_J
O_
^^r
§ 300
a:
LJ
a.
"2" 200
h-
(/)
O
0
Q
LJ
g 100
^
A7 = (AI + A2 + A4 )
\ 0A6 = (A24-A4) •
\ 1
\ 1
\ 1
\ 1 COST OF REPORTING
\ / POOR QUALITY DATA
\ 1 '^^^
BEST \ /-*
STRATEGIES \ / A3
'•^ \ 1 Q
~~~-~^ ^ \ /
\ i5s(M+fiL
\ A2 I
A4 \ * /
\ /
\ /
v/
\
/ N\
/ \
/ \
/ \
/ \
/ \
/ \AI
/ , V AO
A \ 1 IX 1 1 **f 1
^>
04 5 6 7 8 9
MSE (PERCENT)
Figure 22: Added Cost ($) vs. MSE (%) for Alternative Strategies
-------
Table 13. ASSUMED STANDARD DEVIATIONS FOR ALTERNATIVE STRATEGIES-'
I/
1. Flow rate d
°1
2. Control d.
s ample
°2
3. Data d
processing
°3
CTT (%)
Negative bias = -(%)
MSE(%)-/
Added cost ($) per
100 samples
AO
-4
°1
0
°2
0
G3
5.0
4
6.5
0
Al
-4
0<6ai
0
a2
0
°3
4.7
4
6.2
15
A2
-4
°1
0
0.7o2
0
a3
5.1
1
5.2
200
A3
-4
01
0
0.8a2
0
a3
4.7
4
6.2
260
A4
-3
0.7a
0
0.8a2
0
0.7a3
3.8
3
4.8
200
A£'
-3
0.42^
0
0.8a2
0
0.7a3
3.5
3
4.6
215
*&
-1
0.7a
0
0.56a2
0
0.7a3
3.8
1
3.9
400
AT*/
-1
0.42a1
0
0.56a2
0
0.7a3
3.5
1
3.7
415
I/ 31
— a = 0.4 ug S02, for a 24-hour sample at 380 yg S02/m where 0.32 m of air
is sampled, a_ = 0.4 y SO- is equivalent to a standard deviation of 3.3%.
a and o_ are assumed to be 2.5% and 3% respectively of the average or mean
- 3
value, X = 380 yg S02/m .
d is also expressed as the bias in % of the mean concentration,
X - 380 pg S02/m3.
2/
-A5 = .Al + A4, A6 = A2 + A4, A7 = Al + A2 + A4.
3/ /~2 2 2~
— o_ = Wan + a. + a- = percent variation in measured concentration of S0_,
•L T X fc J '
-------
DATA PRESENTATION
A reported value whose precision and accuracy (bias) are unknown is of
little, if any, worth. The actual error of a reported value—that is,
the magnitude and sign of its deviation from the true value—is usually
unknown. Limits to this error, however, can usually be inferred, with
some risk of being incorrect, from the precision of the measurement
process by which the reported value was obtained and from reasonable
limits to the possible bias of the measurement process. The bias, or
systematic error, of a measurement process is the magnitude and direction
of its tendency to measure something other than what was intended; its
precision refers to the closeness or dispersion of successive independent
measurement generated by repeated applications of the process under
specified conditions, and its accuracy is determined by the closeness to
the true value characteristic of such measurements.
Precision and accuracy are inherent characteristics of the measurement
process employed and not of the particular end result obtained. From
experience with a particular measurement process and knowledge of its
sensitivity to uncontrolled factors, one can often place reasonable
bounds on its likely systematic error (bias). This has been done in the
model for the measured concentration as indicated in Table 13. It is
also necessary to know how well the particular value in hand is likely to
agree with other values that the same measurement process might have
provided in this instance or might yield on measurements of the same
magnitude on another occasion. Such information is provided by the
estimated standard deviation of the reported value, which measures (or is
an index of) the characteristic disagreement of repeated determinations
of the same quantity by the same method and thus serves to indicate the
precision (strictly, the imprecision) of the reported value.
A reported result should be qualified by a quasi-absolute type of statement
that places bounds on its systematic error and a separate statement of its
standard deviation, or of an upper bound thereto, whenever a reliable
determination of such value is available. Otherwise, a computed value of
the standard deviation should be given together with a statement of the
number of degrees of freedom on which it is based.
As an example, consider strategy AO in Table 13. Here, the assumed
standard deviation and bias for a true SO- concentration of 380 yg S02/m
are OT - 19 ug S02/m3 (5.0% of 380 pg/m3) and T - -15 vg S02/m3,
respectively. The results would be reported as the measured concentration
(yg S0,/m ) minus the bias and with the following 2a limits along with
i m
the audit level and lot size Nj e.g.,
(yg S00/m3) + 15 + 38, n - 7, N - 100.
z m
107
-------
For concentration other than 380 yg SO./m , the overall standard deviation
is obtained by
I o o o
/at\ }**.&.£• -
O_CA) " V°1 ° + °o » an°
» * O *
/
a (yg/m ) - aT(%) x yg SO./m .
PERSONNEL REQUIREMENTS
Personnel requirements as described here are in terms of the pararosaniline
method only. It is realized that these requirements may be only a minor
factor in the overall requirements from a systems point of view where
several measurement methods are of concern simultaneously.
Training and Experience
Director - The director or one of the professional-level employees, in
addition to formal training in chemistry, should have a basic understanding
of statistics as used in quality control. He should be able to perform
calculations, such as the mean and standard deviation, required to define
data quality. The importance of and requirements for performing independent
and random checks as part of the auditing process must be understood. Three
references which treat the above-mentioned topics are listed below:
Probability and {Statistics for Engineers. Irvin Miller and
John E. Freund, published by Prentice-Hall, Inc., Englewood,
N. J., 1965.
Introductory Engineering Statistics, Irwin Guttman and
S. S. Wilks, published by John Wiley and Sons, Inc., New York,
N. Y., 1965.
The Analysis of Management Decisions, William T. Morris,
published by Richard D. Irwin, Inc., Homewood, Illinois, 1964.
Operator - There are or can be two levels of operation involved in the
manual pararosaniline method.
First, an operator or chemist involved in the reagent preparation or
analysis should have formal training in chemistry. A person with a
technical or community college background in chemistry with on-the-job
experience and close supervision from an experienced chemist could
adequately prepare reagents and analyze S0_ samples.
Field operations involve sample collection and handling only and require
no high-level skills. A high school graduate with proper supervision and
on-the-Job training can become effective at this level in a very short time.
108
-------
An effective on-the-job training program could be as follows:
(1) Observe experienced operator perform the different tasks
in the measurement process.
(2) Study the operational manual of this document and use it
as a guide for performing the operations.
(3) Perform operations under the direct supervision of an
experienced operator.
(4) Perform operations independently but with a high level
of quality control checks utilizing the technique
described in the section on Operator Proficiency
Evaluation Procedures below to encourage high quality
work.
Another alternative would be to have the operator attend an appropriate
basic training course sponsored by EPA.
OPERATOR PROFICIENCY EVALUATION PROCEDURES
One technique which may be useful for early training and qualification
of operators is a system of rating the operators as indicated below.
Various types of violations (e.g., invalid sample resulting from operator
carelessness, failure to maintain records, use of improper equipment, or
calculation error) would be assigned a number of demerits depending upon
the relative consequences of the violation. These demerits could then be
summed over a fixed period of time of one week, month, etc., and a
continuous record maintained. The mean and standard deviation of the
number of demerits per week can be determined for each operator and a
quality control chart provided for maintaining a record of proficiency of
each operator and whether any changes In this level have occurred. In
comparing operators, it is necessary to assign demerits on a per unit work
load basis in order that the inferences drawn from the chart be consistent.
It 4& not wuc.u>t>
Of$ evaJLuation. The. &upeA\>e. -i/t at> a. me.an& o
when and what kind o£ ini>tsiu.c£ioni> and/'on training 4Jt> needed.
A sample QC chart is given in Figure 23 below. This chart assumes that
the mean and standard deviation of the number of demerits per week,
are 5 and 1, respectively. After several operators have been evaluated
for a few weeks, the limits can be checked to determine if they are both
reasonable and effective in helping to improve and/or maintain the quality
of the air qulaity measurement.
The limits should be based on the operators whose proficiency is average
or slightly better than average. Deviations outside the QC limits, either
above or below, should be considered in evaluating the operators. Identify-
ing those operators whose proficiency may have improved is just as important
as knowing those operators whose proficiency may have decreased.
109
-------
The above procedure may be extended to an entire monitoring network (system),
With appropriate definitions of work load, a continuous record may be
maintained of demerits assigned to the system. This procedure might serve
as an incentive for teamwork, making suggestions for improved operation
procedures, etc.
1234 5 6 7 8 9 10 11 12 13
Time Intervals (Weeks)
Figure 23: Sample QC Chart for Evaluating Operator Proficiency
110
-------
REFERENCES
1. J. P. Lodge, Jr. et al., "The Use of Hypodermic Needles as Critical
Orifices in Air Sampling," Journal of the Air^ Pollution Control
Association 16 (4), April 1966, pp. 197-200.
2. F. P. Scaringelli, B. E. Saltzman, and S. A. Frey, "Spectrophoto-
metric Determination of Atmospheric Sulfur Dioxide," Analytical
Chemistry 39, page 1709, December 1967.
3. Metronics Dynacal Permeation Tubes, Product Bulletin No. 20-70,
Metronics Associates, Inc., Palo Alto, Calif. 94304.
4. F. P. Scaringelli et al., "Preparation of Known Concentrations of
Gases and Vapors with Permeation Devices Calibrated Gravimetrically,"
Analytical Chemistry 42 (8), July 1970, pp. 871-876.
5. B. E. Saltzman et al., "Volumetric Calibration of Permeation Tubes,"
Environmental Science and Technology 3 (12), December 1969,
pp. 1275-1279.
6. H. C. McKee et al., "Collaborative Study of Reference Method for
Determination of Sulfur Dioxide in the Atmosphere (Pararosaniline
Method)," Southwest Research Institute, Contract CPA-70-40, SwRI
Project 21-2811, San Antonio, Texas, September 1971.
7. "Tentative Method of Analysis for Sulfur Dioxide Content of the
Atmosphere (Colorimetric)," in Methods of Air Sampling and Analysis,
Intersociety Committee, published by American Public Health
Association, Washington, D. C., 1972.
8. Kenneth D. Reiszner and Philip W. West, "Collection and Determination
of Sulfur Dioxide Incorporating Permeation and West-Gaeke Procedure,"
Environmental Science and Technology 7 (6), June 1973, pp. 526-531.
9. John Mandel, The Statistical Analysis of Experimental Data,
Interscience Publishers, Division of John Wiley & Sons, New York, N.Y.,
1964.
Ill
-------
APPENDIX
REFERENCE METHOD FOR THE DETERMINATION
OF SULFUR DIOXIDE IN THE ATMOSPHERE
(PARAROSANILINE METHOD)
Reproduced from Appendix A, "National Primary and Secondary
Ambient Air Quality Standards," Federal Register, Vol 36,
No. 84, Part II, Friday, April 30, 1971.
112
-------
GULES AND REGULATIONS
APPENDIX A.—REFERENCE METHOD roR THE
DETEnuiNxnoN or SULFUR DIOXIDE IN THE
ATMOSPHERE (PABAROSANIUNE METHOD)
1. Principle and Applicability. 1.1 Sulfur
dioxide Is absorbed from air In a solution of
potassium tetrachloromercurate (TCM). A
dlchlorosulfltomercurate complex, which re-
sists oxidation by the oxygen In the air, is
formed (1. Z). Once formed, this complex is
stable to strong oxldants (e.g., ozone, oxides
of nitrogen)* The complex la reacted with
pararosanlllne and formaldehyde to form In-
tensely colored pararosanlllne methyl sul-
fonlc acid (J). The absorbance of the solu-
tion Is measured spectrophotometrlcally.
1.3 The method In applicable to the meas-
urement of sulfur dloxldo In ambient air
uolng sampling periods up to 24 hours.
2. Rcnge and Sensitivity. 3.1 Concentra-
tions of sulfur dioxide in the range of 36 to
l.OSO Mg/m.° (0.01 to 0.40 p.p.m.) con be meas-
ured under the conditions given. One can
measure concentrations below 38 Mg./m.° by
sampling larger volumes of air, but only If
the absorption efficiency of the particular sys-
tem Is first determined. Higher concentra-
tions can be analyzed by using smaller goo
samples, a larger collection volume, or a suit-
able aliquot of the collected sample. Beer's
Law Is followed through the working range
from 0.03 to 1.0 absorbance units (0.8 to 37
vg. of sulftte Ion In 26 ml. final solution com-
puted as SO:).
2.2 The lower limit of detection of sulfur
dioxide In 10 ml. TCM Is 0.76 vg., (based on
twice the standard deviation) representing a
concentration of 26 (ig./m:aOj (0.01 p.p.m.)
In an air sample of 30 liters.
3. Interferences. 3.1 The effects of the
principal known Interferences have been
minimized or eliminated. Interferences by
oxides of nitrogen are eliminated by sulfamlc
acid (0, 5), ozone by time-delay (6). and
heavy metals by EDTA (ethylenedlamlne-
tetroacetlc acid, dlsodlum salt) and phos-
phoric acid (4, 6,). At least 80 ng. Fa (III),
10 vg. Mn(II). and 10 tig. Cr(III) In 10 ml.
absorbing reagent can be tolerated In the
procedure. No significant Interference was
found with 10 MB. CU (II) and 22 *g. V(V).
4. Precision, Accuracy, and Stability. 4.1
Relative standard deviation at the 96 percent
confidence level Is 4.6 percent for the ana-
lytical procedure using standard samples. (5)
4.2 After sample collection the solutions
ore relatively stable. At 22* C. losses of sulfur
dioxide occur at the rate of 1 percent per
day. When samples are stored at 6* C. for
30 days, no detectable losses of sulfur diox-
ide occur. The presence of EDTA enhances
the stability of SO, In solution, and the rate
of decay Is Independent of the concentration
of SOi. (7)
5. Apparatus.
S.I Sampling.
5.1.1 Absorber. Absorbers normally used
In air pollution sampling are acceptable for
concentrations above 25 /ig./m.« (0.01 p.p.m.).
An all-glass midget Implnger, as shown in
Figure Al, Is recommended for 30-mlnute and
1-hour samples.
For 24-hour sampling, assemble an ab-
sorber from the following parts:
Polypropylene 2-port tube closures, special
manufacture (available from Bel-Art Prod-
ucts, Pequannock, N.J.).
Glass Implngers, 6 mm. tubing, 6 Inches
lone, one end drawn to small diameter such
i'in- No. 79 jewelers will pass through, but
No. 78 jewelers will not. (Other end fire
polished.)
Polypropylene tubes, 164 by 32 mm. Nal-
gene or equal).
5.1.2 Pump. Capable of maintaining an
air pressure differential greater than 0.7 at-
mosphere at the desired flow rate.
5.1.3 Air Flowmeter or Critical Orifice.
A calibrated rotameter or critical orifice ca-
pable of measuring air flow within ±2 per-
cent. For 30-mlnute sampling, o 22-gauge
hypodermic needle 1 Inch long may be used
as a critical orifice to give a flow of about 1
liter/minute. For 1-hour oampilng. a 23-
gauge hypodermic needle five-eighths of an
inch long may be.used as a critical orifice to
give a flow of about O.B liter/minute. For
34 hour sampling, a 27-gauge hypodermic
needle three-eighths of an Inch long may be
used to give a flow of about 0.3 liter/minute.
Use a membrane filter to protect the needle
(Figure Ala).
6.2 Analysis.
5.2.1 Spectrophotometer. Suitable for
measurement of absorbance at 548 nm. with
an effective spectral band width of less than
15 nm. Reagent bland problems may occur
with opectrophotometers having greater
spectral band width. Tho OTvelongth cali-
bration of the Instrument ohould be vorlflod.
If tranamlttanco lo moasurod, this can bo
converted to aboorbanoa:
A=logM(l/T)
6. Reagents.
6.1 Sampling.
6.1.1 Distilled water. Must be free from
oxldants.
6.1.2 Absorbing Reagent (O.Od M Potaa-
sium Tetrachloromercurate (TCM) ). Dissolve
10.86 g. mercuric ehlorlde, 0.066 g. EDTA
(thylenedlomlnetetraocetlc acid, dlsodoum
salt), and 6.0 g. potassium chloride In water
and bring to mark In a 1,000-ml. volumetric
flask. (Caution: highly poisonous. If spilled
on skin, flush oS with water Immediately).
The pH of this reagent ohould be approxi-
mately 4.0, but It hon been shown that there
Is no appreciable difference In collection
efficiency over the range of pH 8 to pH 3.(7)
The absorbing reagent Is normally stable for
6 months. If a precipitate forma, discard the
reagent.
5.2 Analysis.
6.3.1 Sul/amic Acid (0.0 percent). Dis-
solve 0.6 g. sulfamlc acid In 100 ml. distilled
water. Prepare fresh daily.
6.2.2 Formaldehyde (0.2 percent). Dilute
5 ml. formaldehyde solution (36-38 percent)
to 1,000 ml. with distilled water. Prepare
dally.
6.2.3 Stock Iodine Solution (0.1 N). Place
12.7 g. Iodine In a 250-ml. beaker; add 40 g.
potassium Iodide and 36 ml. water. Stir until
all Is dissolved, then dilute to 1,000 ml. with
distilled water.
3.2.4 Iodine Solution (0.01 W). Prepare
approximately 0.01 N Iodine solution by di-
luting 60 ml. of otocts oolutlon to SOO ml.
with distilled water.
6.3.8 Starch Indicator Solution. Triturate
0.4 Q. soluble starch and 0.003 g. mercuric
Iodide (preservative) with a little water, and
add the paste slowly to 200 ml. boiling water.
Continue boiling until the oolutlon lo clear;
cool, and transfer to a glass-stoppered bottle.
6.3.6 Stock Sodium Thioyul/ate Solution
(0.1 N). Prepare a stock solution by dissolving
35 g. sodium thlooulfate (NaJSiCa-BHtO) In
1,000 ml. freshly boiled, cooled, dlctlllod water
and add 0.1 g. sodium carbonate to the solu-
tion. Allow the solution to stand 1 day before
standardizing. To standardize, accurately
weigh, to the nearest 0.1 mg., 1.8 g. primary
standard potassium lodate dried at 180° C.
and dilute to volume In a 600-ml. volumetric
flask. To a 500-ml. Iodine flosS, pipet 80 ml.
of lodate solution. Add 2 g. potassium iodide
and 10 ml. of 1 N hydrochloric acid. Stopper
the flask. After 5 minutes, titrate with otocft
thloffulfate solution to a palo yollow. Add 8
ml. starch Indicator solution and continue
the tltratlon until the blue color dlsappaoro.
Calculate the normality of the otccS
solution:
W
N=—X3.80
N=Normality of stock thlosulfate oolu-
tlon.
M=Volume of thlosulfate required, ml.
w=Weight of potassium lodate, gramo.
3.80 =
10"(conversion of g. tomg.) X0.1 (fraction lodate used)
36.67 (equivalent weight of potassium lodate)
6.2.7 Sodium Thiosulfate Titrant (0.01 N).
Dilute 100 ml. of the stock thlosulfate solu-
tion to 1,000 ml. with freshly boiled distilled
water.
Normality = Normality of stock solution
X 0.100.
6.2.8 Standardize Sulflte Solution for
Preparation of Working Sulftte-TCM Solu-
tion. Dissolve 0.3 g. sodium metablsulflte
(NajSA.) or 0.40 g. sodium sulflte (Na.SO,)
In 500 ml. of recently boiled, cooled, distilled
water. (Sulflte solution is unstable; it Is
therefore Important to use water of the high-
est purity to minimize this Instability.) This
solution contains the equivalent of 320 to 400
jtg./ml. of SO,. The actual concentration of
the solution is determined by adding excess
Iodine and back-titrating with standard
sodium thlosulfate solution. To back-titrate,
plpet 50 ml. of the 0.01 N Iodine into each of
two 500-ml. iodine flasks (A and B). To flask
A (blank) add 25 ml. distilled water, and to
flask B (sample) plpet 25 ml. sulflte solution.
Stopper the flasks and allow to react for 5
minutes. Prepare the working sulflte-TCM
Solution .(6.2.9) at the same time iodine
solution Is added to the flasks. By means of
a buret containing standardized 0.01 N thlo-
eulfate, titrate each flask In turn to a pale
yellow. Then add 5 ml. starch solution and
continue the tltratlon until the blue color
disappears.
6.2.9 Working Sulflte-TCM Solution. Plpet
accurately 2 ml. of the standard solution Into
a 100 ml volumetric flask and bring to mark
with 0.04 M TCM. Calculate the concentra-
tion of sulfur dioxide in the working solu-
tion:
-------
RULES AND REGULATIONS
properly standardized.
6.2.10.2 Preparation o/ Stock Solution. A
specially purified (M-100 percent pure) so-
lution of pararosanlllne, which meets the
above ipeclflcatlons, Is commercially avail-
able in the required O.SO percent concen-
tration (Harleco*). Alternatively, the dye
may be purified, a stock solution prepared
and then assayed according to the proce-
dure of Searlngelll. et al. (4)
6J.11 Pararotaniiine Reagent. To a 280-
ml. volumetric flaak, add 20 ml. stock par-
aroaanlllne solution. Add an additional 0.2
ml. stock solution for each percent the stock
assays below 100 percent. Then add 28 ml.
S If phosphoric add and dilute to volume
with distilled water. This reagent Is stable
for at least 9 months.
7. Procedure.
T.I Sampling. Procedures are described
for short-term (SO minutes and 1 hour) and
for long-term (34 hours) sampling. One can
select different combinations of iaitir-1'frg
rate and time to meet special needs. Sample
volumes should be adjusted, so that linearity
Is maintained between absorbenee and con-
centration over the dynamic range.
7.1.1 30-Hinute and 1-Hoiir SompUnps.
Insert a midget Implnger Into ^he sampling
system, Figure Al. Add 10 ml. TOM solution
to the Implnger. Collect sample at 1 liter/
minute for 80 minutes, or at 0.8 liter/minute
for 1 hour, using either a rotameter, as
shown in Figure Al, or a critical orifice, as
shown in Figure Ala, to control flow. Shield
the absorbing reagent from direct sunlight
during and after sampling by covering the
Implnger with aluminum foil, to prevent
deterioration. Determine the volume of air
sampled by multiplying the flow rate by the
time In minutes and record the atmos-
pheric pressure and temperature. Remove
and stopper the Implnger. If the sample
must be stored for more than a day before
analysis, keep It at 8* O. In a refrigerator
(see 4.2).
7.1.2 M-Hour Sampling. Place 60 ml.
TOM solution in a large absorber and col-
lect the sample at 03 uter/mlnute for M
hours from midnight to midnight. Make sure
no entralnment of solution results with the
Implnger. During collection and storage pro-
tect from direct sunlight. Determine the
total air volume by multiplying the air flow
rate by the time In minutes. The correction
of 24-hour measurements for temperature
and pressure Is extremely difficult and Is not
ordinarily done. However, the accuracy of
the measurement will be improved If mean-
ingful corrections can be applied. If storage
Is necessary, refrigerate at 8* O. (see 4.2).
7.2 Analytii..
73.1 Sample Preparation. After collection.
If a precipitate Is observed in the sample,
remove It by eentrlfugatlon.
7.9.1.1 SO-tlinute and 1-Wour Sample*.
Transfer the sample quantitatively to a 38-
ml. volumetric flask; use about 8 ml. distilled
water tor rinsing. Delay analyses for SO min-
utes to allow any ocone to decompose.
7.2.1.2 tt-Hour Sample. Dilute the entire
sample to 80 ml. with absorbing solution.
Plpet 8 ml. of the sample Into a SB-mi.
volumetric flask for chemical analyses. Bring
volume to 10 ml. with absorbing reagent.
Delay analyses for 20 minutes to allow any
ozone to decompose.
7.2.2 Determination. For each set of de-
terminations prepare a reagent blank by add-
ing 10 ml. unexposed TCM solution to a 28-
mL volumetric flask. Prepare a control solu-
tion by adding 2 ml. of working aulflte-TOIf
solution and 8 ml. TCM solution to a 28-ml.
volumetric flask. To each flask containing el-
•Hartmen-Leddon, 60th and
Avenue, Philadelphia, PA 19143.
Woodland
ther sample, control solution, or reagent-
blank, add 1 ml. 0.8 percent sulfamle
add and allow to react 10 minutes to de-
stroy the nitrite from oxides of nitrogen.
Accurately plpet In 2 ml. 0.2 percent
formaldehyde solution, then 8 ml. par-
ameantllne solution. Start a laboratory
timer that has been set for SO minutes. Bring
al) flasks to volume with freshly boiled and
cooled distilled water and mix thoroughly.
After 80 minutes and before 60 minutes, de-
termine the absorbanoes of the sample (de-
note as A), reagent blank (denote as A.) and
the control solution at 648 nm. using l-em.
optical path length cells. Use distilled water,
not the reagent blank, as the reference.
(Nonl This Is Important because of the color
sensitivity of the reagent blank to tempera-
ture changes which can be Induced In the
cell compartment of a spectrophotometer.)
Do not allow the colored solution to stand
In the absorbanee cells, because a film of dye
may be deposited. Clean cells with alcohol
after use. If the temperature of the determi-
nations does not differ by more than 2* O.
from the calibration temperature (8.2). the
reagent blank should be within 0.03 absorb-
anee unit of the y-intercept of the calibra-
tion curve (8.2). If the reagent blank differs
by more than 0.03 absorbanee unit from that
found In the calibration curve, prepare a new
curve.
7.2.8 Aotorbonce Range. If the absorbanoe
of the sample solution ranges between 1.0
and 2.0. the sample can be diluted 1:1 with
a portion of the reagent blank and read
within a few minutes. Solutions with higher
absorbanee can»be diluted up to sixfold with
the reagent blank in order to obtain onscale
readings within 10 percent of the true ab-
sorbanoe value.
8. Calibration and tglcimciet.
6.1 riovrtnetert and* Hypodermic Needle.
Calibrate flowmetera and hypodermic nee-
dle (•) against a calibrated wet test meter.
8.2 Calibration Curves.
8,3.1 Procedure vtth Sulflte Solution. Ac-
curately plpet graduated amount* of the
working sulflte-TCM solution (8.3.9) (such
as 0, OJ, 1, 2, 3, and 4 ml.) Into a series of
28-ml. volumetric flasks. Add sufficient TOM
solution to each flask to bring the volume to
approximately 10 ml. Then add the retraining
reagents as described In 7.3.3. For maximum
precision use a constant-temperature bath.
The temperature of calibration must be
maintained within ±1* C. and in the range
of 20* to SO* O. The temperature of calibra-
tion and the temperature of analysis must be
within 2 degrees. Plot the absorbanee against
the total concentration in »g. BO. for the
corresponding solution. The total «g. SOi In
solution' equals the concentration of the
standard (Section 8.2.9) In t%. BOi/ml. times
the ml. sulflte solution added (ng. BOi=
*g-/ml. BOiXml. added). A linear relation-
ship should be obtained, and the y-intercept
should be within 0.03 absorbanee unit of the
•ero standard absorban.ee. For maximum pre-
cision determine the line of best fit using
regression analysis by the method of least
squares. Determine the slope of the line of
beet flt, calculate Its reciprocal and denote
as Bi. Bi Is the calibration factor. (See Sec-
tion 6.2.10.1 for specifications on the slope of
the calibration curve). This calibration fac-
tor can be used for calculating results pro-
vided there are no radical changes in
temperature or pR. At least one control
sample """^'"'"g a known concentration of
8O» for each series of determinations, Is
recommended to Insure the reliability of this
factor.
933 Procedure wttv SO> Permeation
rubes.
8.2.2.1 General Contiaerationt. Atmos-
pheres containing accurately known amounts
of sulfur dioxide at levels of Interest can be
prepared using permeation tubes. In the
systems for generating these atmospheres,
the permeation tube emits BO, gas at a
known, low, constant rate, provided the tem-
perature of the tube Is held constant (±0.1*
C.) and provided the tube has been accu-
rately calibrated at the temperature of use.
The BO. gas permeating from the tube is
carried by a low flow of inert gas to a mix-
Ing chamber where It is accurately diluted
with 80,-free air to the level of interest and
the sample taken. These systems are shown
schematically In Figures A3 and AS and have
been described In detail by OKeeffe and
Ortman («), Scarlngelll, Frej, and Saltaman
(10), and Bcartngelll, O"Keeff«, Rosenberg,
and Bell (11).
»333 Preparation of Standard Atmot-
pheret. Permeation tubes may be prepared
or purchased. Scarlngelll, O'KeeRe, Rosen-
berg, and Bell (11) give detailed, explicit
directions for permeation tube calibration.
Tubes with a certified permeation rate are
available from the National Bureau of Stand-
ards. Tube permeation rates from 0.2 to 0.4
«g./mlnute Inert gas flows of about 80 ml./
minute and dilution air flow rates from 1.1
to 16 liters/minutes conveniently give stand-
ard atmospheres containing desired levels
of SO, (28 to 890 fg./m.>; 0.01 to 0.18 p.p.m.
SO.). The concentration of SO, In any stand-
ard atmosphere can be calculated as follows:
P*10«
C=—
R4+R,
Where:
O = Concentration of SOi, *g./m.3 at ref-
erence conditions.
p =Tube permeation rate, *g-/Klnute.
R<=Flow rate of dilution air, Uter/mlnute
at reference conditions.
Ri=Ftow rate of Inert gas, liter/minute at
reference conditions.
8.3.3.8 Sampling and Preparation o/ Cali-
bration Curve. Prepare a series (usually six)
of standard atmospheres containing 8Oi
levels from 28 to 890 M. 8O,/m.'. Sample each
atmosphere using similar apparatus and tak-
ing exactly the same air volume as will be
done in atmospheric sampling. Determine
absorbanoes as directed In 7.2. Plot the con-
centration of SOi In «g'/m.* (x-axlsy against
A—A, values (y-ajdi). draw the straight line
of best flt and determine the slope. Alter-
natively, regression analysis by the method
of least squares may be used to calculate the
slope. Calculate the reciprocal of the slope
and denote as B,.
8.3 Sampling tfflclency. Collection effi-
ciency la above 98 percent: efficiency may
fall off, however, at concentrations below 25
«./m.«. at. 13)
9. Calculations.
9.1 Conversion of Volume. Convert the
volume of air sampled to the volume at ref-
erence conditions of 26* C. and 760 mm. Rg.
(On 24-hour samples, this may not be
possible.) p m
Vs=VX—X
760 t+273
Vt=Volume of air at 28* C. and 760 mm.
Kg. liters.
V = Volume of air sampled, liters.
p = Barometric pressure, mm. Bg.
t =Temperature of air sample, *c.
9 3 Sulfur Dioxide Concentration.
9.2.1 When sulflte solutions are used to
prepare calibration curves, compute the con-
centration of sulfur dioxide In the sample:
(A-A.) (10>) (B.)
«g. BO,/m.«= Y. D
Va
A = Sample absorbanee.
Ag^Reagent blank absorbanee.
10>=Conversion of Utere to cubic meters.
Va =The sample corrected to 28* C. and
760 mm. Bg, liters.
HDUA1 ItOISTU, VOL M, NO. 64—MIDAY, APIIL 30, 1971
114
-------
RULES AND REGULATIONS
B. =Calibration factor, ^g./absorbance
unit.
D = Dilution factor.
For 80-mlnute and 1-bour samples,
D=l.
For 24-hour samples, D= 10.
9.2 J When 8O> gas standard atmospheres
an used to prepare calibration curves, com-
pute the sulfur dioxide In the sample, by the
following formula:
80».«[.An.«=(A-A.)XB,
A = Sample absorbanoe.
Ao=Reagent blank absorbance.
B»= (See 8.3.3.3).
9.3.3 Conversion of n>./m.' to p.p.m. = lf
desired, the concentration of sulfur dioxide
may be calculated as p.p.m. SO_. at reference
conditions as follows:
p p.m. SO,=«g. 8O,/m."x3.82 x 10-«
10. «e/erence».
(2) West. P. W., and Oaeke. O. C.. "Fixa-
tion of Sulfur Dioxide as Sulntomer-
curate m and Subsequent Colon-
metric Determination", Anal. Chem.
28, 1816 (1956).
(2) Ephralms. P., "Inorganic Chemistry,"
p. 682, Edited by P.C.L. Thorne and
E. R. Roberts, 6th Edition, Inter-
science. (1948).
(3) Lyles, O. R., Dowllng, P. B.. and Blanch-
ard, V. J.. "Quantitative Determina-
tion of Formaldehyde In Parts Per
Hundred Million Concentration Lev-
el", J. Air Poll. Cont. Asioc. IS, 106
(1986).
(4) Scarlngelll, F. P., Saltzman, B. E., and
Prey. s. A., "Spectrophotometric De-
termination of Atmospheric Sulfur
Dioxide", Anal. Chem. 39,1709 (1987).
(5) Pate, J. B., Ammons, B. E., Swanson,
O. A., Lodge, J. P., Jr., "Nitrite In-
terference In Spectrophotometric De-
termination of Atmospheric Sulfur
Dioxide". Anal. Chem. 37.943 (1966).
<«) Zurlo, N. and Orlfflnl, A. M.. '•Measure-
ment of the SO, Content of Air in the
Presence of Oxides of Nitrogen and
Heavy Metals", Met. Lavoro, S3, 330
(1983).
(7) Scarlngelll, F. P.. Enters, L.. Norrls, D.,
and Hochhelser, 8., "Enhanced Sta-
bility of Sulfur Dioxide In Solution",
Anal. Chem. 42, 1818 (1970).
(8) Lodge, J. P. Jr., Pate, J. B.. Ammons,
B. E. and Swanson, a. A., "Use of
Hypodermic Needles as Critical Ori-
fices In Air Sampling." J. Air Poll.
Cont. Assoc. IS. 197 (1966).
(9) O'Keeffe, A. -E.. and Ortman, O. C.,
"Primary Standards for Trace Oas
Analysis". Anal. Chem. 38, 760 (1966).
(10) Scarlngelll, F. P., Prey, S. A., and Saltz-
man, B. E., "Evaluation of Teflon
Permeation Tubes for Use with Sulfur
Dioxide". Amer. fnd. Hygiene Asaoc.
J. 28, 260 (1967).
(11) Scarlngelll. F. P., O'Keeffe. A. E., Rosen-
berg, E., and Bell, J. P., "Preparation
of Known Concentrations of Oases
and Vapors with Permeation Devices
Calibrated Qravlmetrlcally", Anal.
Chem. 42, 871 (1970).
(12) Urone, P., Evans, J. B., and Noyes, C. M.,
"Tracer Techniques in Sulfur Di-
oxide Colorlmetrlc and Conductlo-
metrlc Methods", Anal Chem. 37,1104
(1968).
(13) Bostrom, C. B., "The Absorption of Sul-
fur Dioxide at Low Concentrations
(p.p.m.) Studied by an Isotoplc
Tracer Method", Intern. J. Air Water
Poll. 9, 33 (1965).
HYPODERMIC
NEEDLE
Figure A1a. Critical orilice How control.
WRINGER
Figure A1. Sampling train.
FEDEtAl REGISTER, VOL 36, NO. 84—ttlCAY, APRIL 30, 1971
115
-------
RULES AND REGULATIONS
TOHOOO
J
THERMOKETHl
FLOWUETER
OR CRITICAL
OfUFICB
DMBI
ItTIRRER
MWIATION
TUMMBaUR
WATER BATH
CYLINDER
> MR OH
MTROOEN
Rgm AS. Amnlw lor «MMHo nlMbi ltd fW* •*.
MT1
HOISTB. VOL M, NO. M—WIOAY, Aftll M, 1*71
116
-------
BIBLIOGRAPHIC DATA
SHEET
1. Report No.
3. Recipient's Accession No.
4. Title and Subtitle
GUIDELINES FOR DEVELOPMENT OF A QUALITY ASSURANCE PROGRAM -
Reference Method for Determination of Sulfur Dioxide in the
Atmosphere
5> Report Date
July 1973
6.
7. Author(s)
Franklin Smith and A. Carl Nelson, Jr.
8. Performing Organization Kept.
No.
9. Performing Organization Name,and Address
Research Triangle Institute
Research Triangle Park, N. C. 27709
10. Project/Task/Work Unit No.
11. Contract/Grant No.
EPA DURHAM 68-02-0598
12. Sponsoring Organization Name and Address
Environmental Protection Agency
National Environmental Research Center
Research Triangle Park, N. C. 27711
13. Type of Report & Period
Covered
Interim Contractor's r<
port
14.
15. Supplementary Notes
16. Abstracts
Guidelines for the quality control of the Federal reference method for sulfur
dioxide are presented. These include:
1. Good operating practices
2. Directions on how to assess and qualify data
3. Directions on how to identify trouble and improve data quality.
4. Directions to permit design of auditing activities
5. Procedures for selecting action options and relating them to costs.
This document is not a research report. It is for use by operating personnel.
17. Key Words and Document Analysis. 17a. Descriptors
Quality Assurance
Quality Control
Air Pollution
Quantiative Analysis
Gas Analysis
Sulfur Dioxide
17b. Identifiers/Open-Ended Terms
17c. COSATI Field/Group 13H, 14D, 13B, 07B, 07B, 14B
18. Availability Statement
19.. Security Class (This
Report)
UNCLASSIFIED
20. Security Class (This
Pa,
ge
Ul
NCLASS1FIED
21. No. of Pages
22. Price
FORM NTtS-35 (REV. 3-721
USCOMM-DC 149S2-P72
-------
INSTRUCTIONS FOR COMPLETING FORM NTIS-35 (10-70) (Bibliographic Data Sheet based on COSATI
Guidelines to Format Standards for Scientific and Technical Reports Prepared by or for* the Federal Government,
PB-180 600).
1. Report Number. Each individually bound report shall carry a unique alphanumeric designation selected by the performing
organization or provided by the sponsoring organization. Use uppercase letters and Arabic numerals only. Examples
FASEB-NS-87 and FAA-RD-^8-09.
2. Leave blank.
3> Recipient's Accession Number. . Reserved for use by each report recipient.
4* Title and Subtitle. Title should indicate clearly and briefly the subject coverage of the report, and be displayed promi-
nently. Set subtitle, if used, in smaller type or otherwise subordinate it to main title. When a report is prepared in more
than one volume, repeat the primary title, add volume number and include subtitle for the specific volume.
5- Report Date. I'.ach report shall carry a date indicating at least month and year. Indicate the basis on which it was selected
(e.g., date of issue, date of approval, date of preparation.
6. Performing Organization Code. Leave blank.
7. Authors). Give name(s) in conventional order (e.g., John R. Doe, or J.Robert Doe). List author's affiliation if it differs
from the performing organization.
8. Performing Organization Report Number. Insert if performing organization wishes to assign this number.
9- Performing Organization Name and Address. Give name, street, city, state, and zip code. List no more than two levels of
an organizational hierarchy. Display the name of the organization exactly as it should appear in Government indexes such
as USGRDR-I.
10. Projecf/Tosk/Work Unit Number, list- the project, task and work unit numbers under which the report was prepared.
11. Contract/Grant Number. Insert contract or grant number under which report was prepared.
12* Sponsoring Agency Nome and Address. Include zip code.
IX Type of Report and Period Covered. Indicate interim, final, etc., and, if applicable, dates covered.
14* Sponsoring Agency Code. Leave blank.
15. Supplementary Notes. Enter information not included elsewhere but useful, such a.*- : Prepared in cooperation with . . .
Translation of ... Presented at conference of . . . To be published in ... Supersedes . . . Supplements . . .
16. Abstract. Include a brief (200 words or less) factual summary of the most significant information contained in the report.
If the report contains a significant bibliography or literature survey, mention it here.
17. Key Words and Document Analysis, (a). Descriptors. Select from the Thesaurus of Knginecring and Scientific Terms the
proper authorized terms that identify the major concept of the research and are sufficiently specific and precise to be used
as index entries for cataloging.
(b). Identifiers and Open-Ended Terms. Use identifiers for project names, code names, equipment designators, etc. Use
open-ended terms written in descriptor form for those subjects for which no descriptor exists.
(c). COSATI Field/Group. Field and Group assignments are to be taken from the 1965 COSATI Subject Category List.
Since the majority of documents are multidisciplinary in nature, the primary Field/Group assignment(s) will be the specific
discipline, area of human endeavor, or type of physical object. The applications ) will be cross-referenced with secondary
Field/Group assignments that will follow the primary posting(s).
18. Distribution Statement. Denote rcleasability to the public or limitation for reasons other than security for example "Re-
lease unlimited". Cite any availability to the public, v-ith address and price,
19 & 20. Security Classification. Do not submit classified reports to the National Technical
21. Number of Pages. Insert the total number of pages, including this one and unnumbered pages, but excluding distribution
list, if any.
22. Price. Insert the price set by the National Technical Information Service or the Government Printing Office, if known.
FORM NTIS-35
------- |