United States Environmental Monitoring Systems
Environmental Protection Laboratory
Agency Research Triangle Park NC 27711
Research and Development EPA/600/4-90/003 August 1989
Quality Assurance
Handbook for
Air Pollution
Measurement
Systems:
Volume IV.
Meteorological
Measurements
Revised August 1989
Printed on Recycled Paper
-------
-------
QUALITY ASSURANCE HANDBOOK
FOR
AIR POLLUTION MEASUREMENT SYSTEMS
Volume IV — METEOROLOGICAL MEASUREMENTS
as revised August, 1989
U. S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Atmospheric Research and Exposure Assessment Laboratory
Research Triangle Park, North Carolina 277II
-------
Section No. 4.1
Revision No. 0
Date: 17 Sep 89
Page: 1 of 1
ACKNOWLEDGEMENTS
This volume of the Quality Assurance Handbook has been prepared by Thomas
J. Lockhart, CCM, of the Meteorological Standards Institute on Fox Island,
Washington. Sections 4.0.4.4, 4.1.8, 4.4.0, 4.5.0 and 4.6.0 were used from
EPA-600/4-82-060 with only slight modification. The contributions of the
authors of this work, Peter L. Finkelstein, Daniel A. Mazzarella, Thomas J.
Lockhart, William J. King, and Joseph H. White are hereby gratefully
acknow1edged.
The first draft of this volume received the contributions of the Critical
Review Panel who are:
Dr. C. Bruce Baker
Mr. Jack A. Bowen
Dr. Harold L. Crutcher, CCM
Prof. Gerald C. Gill, and
Mr. Alvin L. Morris.
Work on this volume and the workshops which followed itspublication have
been immeasurably helped by the generous cooperation of the following companies
who made their products and services available to the author:
AeroVironment Inc. Campbell Scientific, Inc.
Climatronics Corporation Hollis International Limited
Odessa Engineering Rotronic Instrument Corp.
Qualimetrics, Inc. Teledyne Geotech
Waters Manufacturing, Inc. R. M. Young Company
This revised Volume IV was produced under contract 68-02-4553, Mr. Jack A.
Bowen, Project Officer.
DISCLAIMER
Mention of trade names or commercial products does not constitute EPA
endorsement or recommendation for use.
-------
Section No. 4.ii
Revision No. 0
Date: 17 Sep 89
Page: 1 of 2
VOLUME IV
TABLE OF CONTENTS
Section
4.1 ACKNOWLEDGEMENTS
4.11 TABLE OF CONTENTS
4.0
4.0
4.0
4.0
4.0
4.0
4.1
0
1
2
3
4
4.0.5
GENERAL PROGRAM REQUIREMENTS
FOR METEOROLOGICAL MEASUREMENTS
OUTLINE, PURPOSE AND OVERVIEW
GLOSSARY
STATE OF THE ART
DATA REQUIREMENTS
MEASUREMENT REQUIREMENTS
REFERENCES
GENERAL ASPECTS OF QUALITY ASSURANCE
FOR METEOROLOGICAL MEASUREMENTS
4.1.0 OUTLINE AND SUMMARY
4.1.1 PLANNING FOR A QUALITY ASSURANCE PROGRAM
4.1.2 ORGANIZATION OF AUTHORITY AND
RESPONSIBILITY
4.1.3 QUALITY CONTROL GUIDELINES
4.1.4 TRACEABILITY PROTOCOL
4.1.5 ESTIMATING PRECISION AND ACCURACY
4.1.6 SYSTEM AUDITS
4.1.7 PERFORMANCE AUDITS
4.1.8 DATA VALIDATION PROTOCOL
4.1.9 QA REPORTS AND CORRECTIVE ACTION
4.1.10 REFERENCES
4.2 QA FOR WIND SPEED,
TURBULENCE
WIND DIRECTION AND
.0
1
.2
4.2.
4.2.
4.2.
4.2.3
4.2.4
4.2.5
4.2.6
4.2.7
4.2.8
4.2.9
OUTLINE AND SUMMARY
TYPES OF INSTRUMENTS
SPECIFICATIONS
ACCEPTANCE TESTING
INSTALLATION
CALIBRATION
OPERATIONS, MAINTENANCE AND QC
PERFORMANCE AUDIT METHODS
ESTIMATING ACCURACY AND PRECISION
REFERENCES
4.3 QA FOR TEMPERATURE AND TEMPERATURE
GRADIENTS (AT)
4.3.0 OUTLINE AND SUMMARY
4.3.1 TYPES OF INSTRUMENTS50
4.3.2 SPECIFICATIONS
4.3.3 ACCEPTANCE TESTING
4.3.4 INSTALLATION
4.3.5 CALIBRATION
4.3.6 OPERATIONS, MAINTENANCE AND QC
Pages Rev. Date
1
2
25
5
2
6
3
-7
2
33
3
3
1
2
4
6
2
1
7
2
2
83
4
4
32
5
8
4
2
20
1
3
36
3
4
4
2
1
9
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
9/17/89
-------
Section No. 4.ii
Revision No. 0
Date: 17 Sep 89
Page: 2 of 2
VOLUME IV
TABLE OF CONTENTS
Section Pages Rev. Date
4.3.7 PERFORMANCE AUDIT METHODS 8 0 9/17/89
4.3.8 ESTIMATING ACCURACY AND PRECISION 1 0 9/17/89
4.3.9 REFERENCES 1 0 9/17/89
4.4.0 QA FOR PRECIPITATION MEASUREMENTS 7 0 9/17/89
4.5.0 QA FOR RELATIVE HUMIDITY OR DEW POINT 10 0 9/17/89
TEMPERATURE
4.6.0 QA FOR SOLAR RADIATION MEASUREMENTS 8 0 9/17/89
4.7.0 QA FOR ATMOSPHERIC PRESSURE 2 0 9/17/89
-------
Section No. 4.0.0
Revision No. 0
Date: 17 Sep 89
Page: 1 of 5
Section 4.0
GENERAL PROGRAM REQUIREMENTS FOR
METEOROLOGICAL MEASUREMENTS
OUTLINE
Section Pages Rev. Date
4.0.0 OUTLINE, PURPOSE AND OVERVIEW
OF THE QUALITY ASSURANCE HANDBOOK 5 0 9/89
4.0.1 GLOSSARY 2 0 9/89
4.0.2 STATE OF THE ART 6 0 9/89
4.0.2.1 AUDITOR SURVEY
4.0.2.2 INTERVIEW SUMMARY
4.0.3 DATA REQUIREMENTS 3 0 9/89
4.0.3.1 REGULATORY PROGRAMS
4.0.3.2 RESEARCH PROGRAMS
4.0.3.3 CONTINGENCY PROGRAMS
4.0.4 MEASUREMENT REQUIREMENTS 7 0 9/89
4.0.4.1 MEASUREMENT SYSTEM
4.0.4.2 DOCUMENTATION
4.0.4.3 MOUNTING
4.0.4.4 SITING
4.0.5 REFERENCES 2 0 9/89
-------
Section No. 4.0.0
Revision No. 0
Date: 17 Sep 89
Page: 2 of 5
PURPOSE AND OVERVIEW OF THE QUALITY
ASSURANCE HANDBOOK
The purpose of this volume of the QA Handbook is to provide information and
guidance for both the meteorologist and the non-meteorologist who must make
judgments about the validity of data and accuracy of measurement systems. Care
has been taken to provide definitions to help those making these judgments to
communicate without ambiguity. Methods are described in the handbook which
will objectively define the quality of measurements so the non- meteorologist
can communicate with the meteorologist or environmental scientist or engineer
with precision of meaning.
The first section of the handbook contains a special glossary of terms
necessary to meteorology and quality assurance. Following that is an analysis
of the state of the art from information and interviews of those practicing QA
in the air quality field. The final parts of the first section define some of
the requirements for gathering data which a QA effort can compare to the
practice of acquiring data.
The second section is devoted to quality assurance and quality control as
it is applied to meteorological problems. This section is somewhat independent
of the variable being measured. Where the variable is important it is treated
individually.
The final six sections are variable-specific. The most important wind
measurement is covered in considerable detail. The temperature measurement
section concentrates on the temperature difference measurement used for
stability determination. The final four sections cover to an adequate depth
the measurement of humidity, radiation, precipitation and surface air pressure.
Examples are given where possible to help explain the methods and problems to
be found in programs of collecting meteorological data and assessing data
validity.
The need for common understanding is critical for the practice of quality
control (QC) and quality assurance (QA). This is achieved in part by the
definitions of the language used within the discipline. From that vocabulary,
the details of the systems and procedures are defined in terms of the necessary
goals.
There are a variety of QA/QC definitions in the literature and in common
usage. Volume I, Section No. 1.3 and Appendix A provide some general
definitions. Section 1.4 shows how the elements of QA are distributed and
where in the section they are described. The well known "quality assurance
wheel" is shown in Figure 1.4.1. The following discussion of definitions is
broader based to include meteorological requirements and explicit between QA
and QC.
The structure shown in Figure 4.0.0.1 below is from ANSI/ASQC Q90-1987;
American National Standard, Quality Management and Quality Assurance Standards
Guidelines for Selection and Use. The definitions in the glossary (4.0.1) and
the following descriptions are structured to fit Figure 4.0.0.1 and the
practices of meteorological measurement.
-------
Section No. 4.0.0
Revision No. 0
Date: 17 Sep 89
Page: 3 of 5
[Organizational]
\ Structure /
I Confidence \
I To The I
^Management/
QUALITY MANAGEMENT ASPECTS*
^"QUALITY SYSTEM^XN
QUALITY
CONTROL
ASPECTS•
I
I
-\ ' INTERNAL
\ * QUALITY
\
(Policy)
\
/ Operational \
I Techniques )
yAnd Activities/
(When Required
By A Contract)
EXTERNAL QUALITY
ASSURANCE ASPECTS
CONFIDENCE TO
PURCHASER
Figure 4.0.0.1 The Structure Of Quality (reprinted with permission
from the American Society for Quality Control)
This figure will be described as it applies to a meteorological company
which manufacturers instruments and provides a variety of services. It can
also apply to government organizations and temporary project-oriented
cooperative organizations. At the top of the figure and the top of any
organized effort toward quality are the QUALITY MANAGEMENT ASPECTS. These
aspects comprise the policy statement for the organization expressed by
management. The statement is in writing as a company policy and signed by the
president of the company so that there will be no misunderstanding or confusion
about the quality goals of instruments so that they do produce valid data. A
performance audit, then, is a challenge both to the instrument and to the
operator to independently verify that the measurement system is "in control."
Just as with system audits, the auditor is primarily a teacher and trainer.
The audit method should be the best possible method. The operator should be
encouraged to upgrade the calibration methods to do a better job.
Often the operator has no experience with meteorological instruments.
Often they are well qualified instrument technicians, but the auditor is an
expert, or should be. A mutually valuable goal is for the operator to learn
what methods are necessary and most complete and adopt those for the
calibration procedure. When the instruments are all working perfectly it is
because they are getting the experienced attention it takes for "in control"
operation. At this point, the audit becomes a spot checking operation
producing documentation from an independent individual verifying this "in
control" operation.
If some basic ground rules are followed, the audit is maximized as a
learning exercise. One rule is that the operator does all the instrument
handling. There is a general reluctance to handle unfamiliar instruments.
They might get broken or changed in some mysterious way. The way to become
-------
Section No. 4.0.0
Revision No. 0
Date: 17 Sep 89
Page: 4 of 5
familiar with them is to work with them. The safest environment in which to
gain this experience is in the presence of someone who is familiar with the
instruments.
Good audit methods are as close to laboratory tests as a field site and the
ingenuity of the auditor will allow. This inevitably requires a tower mounted
instrument to be taken down, complete with cable or substitute cable, so that
tests can be made in a physically convenient environment. For wind sensors,
the bearing condition is of vital importance. This can be measured only when
the sensor can be connected to the torque instrument with complete control and
freedom to move. It is not a proper test to try on a tower or mast.
Performance audits, in exactly the same way as calibrations, challenge parts of
the system. Ideally, one wants to challenge all of the system, but that is
often impossible. Known free atmospheres are not available from NBS.
Controlled atmospheres like a wind tunnel or a thermal chamber or a "sun" lamp
only challenge a part of the system. They leave out or drastically change the
important coupling function. Even controlled atmosphere devices, such as a
wind tunnel, are not available to the auditor in the field. All one can do is
impose a known condition such as the rate of rotation for an anemometer, and
measure the system response. This type of audit assumes that the
manufacturer's generic transfer function applies to this sensor, or assume that
earlier wind tunnel tests of this sensor still apply (a good assumption if the
sensor is not damaged).
Another method is the ASTM collocated transfer standard method. This is
the most complete method from the standpoint of total system error sources but
it has two drawbacks. First, it is limited to the conditions that prevail
during the audit. Secondly, it is very sensitive to exposure or siting bias.
It requires careful guidelines pointing out potential bias sources and ways to
watch for them in the data. These are covered in the variable-specific
sections.
A performance audit program using experienced independent auditors, whether
internal or external to the organization, is the first step toward establishing
a quality plan if one does not already exist. The goal of the measurement
program is to have documented data. The performance audit will point out areas
required to get the system "in control." The auditor can help implement the
establishment of a quality system, or its key elements, in order to achieve the
necessary on-going activities to keep the measurement system "in control"
continuously.
The survey which led off the work of revising this handbook exposed some
confusion in the community of meteorological auditors with regard to the
difference between performance audits and calibrations. A form letter was
composed to discuss these differences and to ask for new numbers of audits
conducted. The letter used the definitions found in the glossary (4.0.1) and
expanded on them with examples. The principal difference is independence of
responsibility. Some organizations perceived the documentation of the
condition of the system "as found" as a performance audit and the adjustment of
the system to acceptable operating conditions, documented "as left," as a
calibration. Thus, a single individual could both audit and calibrate during
the same visit. By any accepted standard of quality systems definition, this
whole process of testing and adjustment is a calibration. This properly
documented calibration is the basis for claims of data validity. All the
-------
Section No. 4.0.0
Revision No. 0
Date: 17 Sep 89
Page: 5 of 5
performance audit adds is an independent assurance that the calibrations were
thoroughly done and that the documents are complete and accurate. Such
assurance must be entirely free of potential influence.
The letter described situations where a single company can be structured to
provide both calibration and auditing services, but cautioned that the
independence of the auditor requires a management structure insulating the
auditor from the budgetary concerns of the operating organization. Responses
to this letter were few and in no case was the distinction challenged. All
agreed with the concept of independence for QA audits. Regulators should
acknowledge the distinction and require true independence.
Situations arise where the fundamental principle of independence between
calibration and audit services is difficult to follow. Small agencies may not
be able to contract for independent audits. In the interest' of documented
validity of data for all parties, innovative arrangements among different
agencies should be promoted. The individual who operates and calibrates
instruments for Agency A might be asked to audit the Agency B instruments in
exchange for the operator at Agency B auditing the Agency A instruments. This
practice would have the further benefit of stimulating communication about and
standardization of good audit methods.
-------
Section No. 4.0. 1
Revision No. 0
Date: 17 Sep 89
Page: 1 of 2
4.0.1 GLOSSARY FOR METEOROLOGY AND QA/QC
ACCURACY - is the degree of agreement of a measurement (or an average of
measurements of the same thing), X.with an accepted reference or true value, T,
usually expressed as the difference between the two values, X - T, or the
difference as a percentage of the reference or true value, 100(X-T)/T, or
sometimes expressed as a ratio, X/T.
CALIBRATION - is a MEASURE of conformance to or discrepancy from a
specification or set of criteria for an instrument or system if necessary and
an ADJUSTMENT of the instrument or system to conform to the specification or
criteria. A calibration may be performed by a person or agency within the
operating organization.
DAMPING RATIO (T?) - The damping ratio is calculated from the overshoot ratio
(Q). C2]
In
H-HI
DELAY DISTANCE (D) - The distance the air flows past a wind vane during the
time it takes the vane to return to 50 percent of the initial displacement. 12]
EXTERNAL QUALITY ASSURANCE - is the activity designed to provide the purchaser
with confidence in the quality of what is being purchased.
INTERNAL QUALITY ASSURANCE - is the activity designed to provide management
with confidence that the quality system is operating and the management policy
is being carried out.
INVERSION (+AT) - is the inverted lapse rate or an increase of air
temperature with height. There is no general limit for inversion strength.
LAPSE RATE (-AT) - is the normal decrease of air temperature with height
limited by the auto convection rate of 3.4°C/100 m.
OVERSHOOT (fl) - The ratio of the amplitude of two successive deflections of a
wind vane as it oscillates about the equilibrium position after release from an
offset position of ten degrees, as expressed by the equation
0
0.
0
n
where 9 and 9 are the amplitudes of the n and n+1 deflections,
n (n+1)
respectively.
-------
Section No. 4.0. 1
Revision No. 0
Date: 17 Sep 89
Page: 2 of 2
PERFORMANCE AUDIT - is a report of conformance to or discrepancy from a
specification or set of criteria determined by a person or agency separate
from and independent of the operating organization.
PRECISION - is the standard deviation of a series of measured values, X.,
about the mean measured value, X. (see 4.1.5.1)
QUALITY ASSURANCE - All those planned and systematic actions necessary to
provide adequate confidence that a product or service will satisfy given
requirements for quality, [i]
QUALITY CONTROL - The operational techniques and activities that are used to
fulfill requirements for quality. Ci]
QUALITY MANAGEMENT - That aspect of the overall management function that
determines and implements the quality policy, [i]
QUALITY POLICY - The overall quality intentions and direction of an
organization as regards quality, as formally expressed by top management, m
QUALITY SYSTEM - The organizational structure, responsibilities, procedures,
processes, and resources for implementing quality management, [i]
REPRESENTATIVENESS - is the extent to which a set of measurements taken in a
space-time domain reflects the actual conditions in the same or different
space-time domain taken on a scale appropriate for a specific application. [4]
STARTING THRESHOLD (S , m/s) - The lowest speed at which a vane will turn to
within 5° of 9 (the true direction) from an initial displacement of 10°. [2]
B
STARTING THRESHOLD (U , m/s) - The lowest speed at which a rotating anemometer
starts and continues to turn and produce a measurable signal when mounted in
its normal position. [3]
[i] ANSI/ASQC, 1987a: Quality Management and Quality Assurance
Standards - Guidelines for Selection and Use. ANSI/ASQC Q90-1987. American
Society for Quality Control, Milwaukee, WI 53203.
[2] ASTM, 1985b: Standard Test Method for DETERMINING THE DYNAMIC PERFORMANCE
OF A WIND VANE. (Draft 8 of D22.ll) Amer. Soc. for Testing and Materials,
Philadelphia, PA 19103.
(3) ASTM, 1985a: Standard Test Method for DETERMINING THE PERFORMANCE OF A CUP
ANEMOMETER OR PROPELLER ANEMOMETER. (Draft 6 of D22.ll) Amer. Soc. for
Testing and Materials, Philadelphia, PA 19103.
-------
Section No. 4.0.2
Revision No. 0
Date: 17 Sep 89
Page: 1 of 6
4.0.2 STATE OF THE ART
The achievement of predicted quality for a product or service can be the
delegated responsibility of an identifiable part of an organization. The
practice of elevating quality to a manstgement staff level is relatively new.
The value of quality control and the umbrella management structure of quality
assurance became clear when products, purchased against a specification, were
rejected by the purchaser. When the cost of rework or scrap absorbs the
profit, an alternative will be found. The alternative is to do it right the
first time and the path to that goal involves training, in-process inspection,
final inspection and all of the other QA functions designed to minimize scrap
and rework.
The QA profession grew during World War II and thereafter as the U.S.
Government became a significant purchaser using comprehensivespecifications,
like the well known Mil-Specs. In the '60sand '70s the practice of planned
obsolescence and using the customer as the final inspector set up our
industries for failure against foreign competition with higher quality
standards. The successful foreign producers, using the quality principles
developed in the United States, caused a resurgence of quality awareness.
The Environmental Protection Agency (EPA) recognized the need to set
standards, develop standard methods and materials, and produce a system of
quality assurance to support validity claims for the data being collected in
response to the Clean Air Act. In 1976 a Quality Assurance Handbook for Air
Pollution Measurement Systems: Volume I. Principles was published (EPA, 1976).
In 1977, Volume II. Ambient Air Specific Methods (EPA, 1977a) and Volume III.
Stationary Source Specific Methods (EPA, 1977b) were published. This program
addressed the Criteria Pollutants which were covered by federal law.
Meteorological measurements were recognized as supportive to the Criteria
Pollutant measurement program but they were secondary.
When the Clean Air Act was amended, permission for growth of source
strength (and thereby growth of industry) was granted as a con- sequence of
diffusion model predictions based on input meteorological data. Now, the law
recognized the requirement for valid and represent- ative meteorological data
and the need for a structure to provide documented assurance of validity. In
1983, Volume IV. Meteorological Measurements (EPA, 1983) was added to 'the
Handbook family.
Almost all of the QA work provided by the private sector was geared to air
and source chemistry. When meteorology was added to the technical requirements
list, a variety of solutions were applied by a variety of individuals with a
variety of technical backgrounds. The original Volume IV. was like a
background guidebook for taking meteorological measurements and general
suggestions for how QA and QC might be applied to the requirement for valid
data. This revision of Volume IV. is intended to be more specific and more
informative and more in the spirit of the other three volumes. It has not, and
cannot as yet, specify standard methods. A greater success with predictive
models is necessary before knowledge will exist which can dictate the standard
methods to assure valid input data. The premise of this Volume IV. is that
measurements worth taking are worth taking right.
-------
Section No. 4,0.2
Revision No. 0
Date: 17 Sep 89
Page: 2 of 6
4.0.2.1 Auditor Survey
This handbook is intended to document the methods currently in use in
meteorological QA/QC and to point to methods which are optimum for meeting the
requirements suggested or defined in various EPA publications. A starting
point toward this goal is a survey of all those active in performance and
system auditing of meteorological measurement programs. Figure 4.0.2.1 is a
copy of the survey form sent to as many people with experience in auditing as
could be found. The initial list, shown by company and location in Table
4.0.2.1, grew considerably with help from all the EPA Regions and many state
and local agencies. The number of survey forms returned from each company is
also shown.
Table 4.0.2.1 - Original Survey List
COMPANY
* AeroVironment, Inc.
Dames & Moore
Desert Research Institute
* Enviro. Monitoring & Services Inc.
Environmental Research & Tech.
* Environmental Research & Tech.
Galson Technical Services, Inc.
Meteorological Standards Inst.
Research and Evaluation Assoc.
* Research Triangle Institute
Roy F. Weston, Inc.
RTP Associates
Technical Environmental Enter.
Tennessee Valley Authority
* TRC Environmental Consultants
CITY/STATE
Monrovia, CA
Atlanta, GA
Reno, NV
Thousand Oaks, CA
Fort Collins, CO
Concord, MA
E. Syracuse, NY
Fox Island, WA
Chapel Hill, NC
RTP, NC
West Chester, PA
Denver, CO
Aurora, CO
Muscle Shoals, AL
E. Hartford, CT
NUMBER
4
2
2
1
1
4
1
1
1
1
1
1
1
* indicates companies chosen for in-depth interview
Of the 70 or so forms sent originally or copied and distributed within
an organization, 49 forms were returned. The summary of these responses is
shown numerically on Figure 4.0.2.1. The number of audits represented by the
survey is 12,195, where the definition of an audit is the challenge of one
instrument measuring a meteorological variable. Each respondent was asked to
qualify himself by specialty, using three or more if necessary but indicating a
priority of 1,2 or 3. Some managers reported for their organization of
auditors. The responses to the questions were not weighted by numbers of
audits. As with most surveys, a few points are useful but action should not be
based on the survey results. Of the 49 survey forms returned with data, 21
came from the original list, 9 came from local, state or federal agencies and
19 came from others. Of this 19, 5 came from utilities in the Northeast U.S.
(3 from Pennsylvania Power & Light) showing a close relationship to Regional
Meteorologists and interest in QA/QC in the area.
-------
Section No. 4.0.2
Revision No. 0
Date: 17 Sep 89
Page: 3 of 6
Survey of Meteorological Measurement QC/QA People
(please print or type your responses)
NAME COMPANY
ADDRESS ADDRESS
PHONE < ) < )
I am a _ meteorologist. _ chemist, _ environmental scientist, QA/QC
professional, _ instrument technician, electronics tech., enaineer.
met.tech., modeler, manager. _ da£a analyst, field hand.'(use
T?2,3 if you are more than one) This in-formation will be summarized
without the use o-f names or companies or agencies so please be candid.
Consider an audit or challenge as a QA observation o-f the response
o-f an instrument to a known input and consider a calibration as an
OPERATIONAL testing and adjustment, as necessary, o-f an instrument.
1. I-f one meteorological audit is de-fined as a challenge to one
variable or one variable o-f a system, about how many audits have you
per-formed in 19QO-1984 , 1985 , 1986 , 1987 ?
2. Did you 34 usual 1 y J~£_sometimes ^fcnever use a written procedure-?
3. If a calibration is de-fined as the testing and adjustment o-f one
variable or one variable in a system, how many calibrations have you
per-formed in 198O-1984 , 1985 , 1986 , 1987 ?
4. Did vou^ 9 usual 1 v £5. sometimes _/._ never use the manufacturer's
calibration procedure?
5. When you perform an audit, do you require the operator to remove the
sensors from their mounted position? ZZLY«S ^(?No /.^.Sometimes
6. Do you require the operator to re—connect the sensor to the system
when it is presented for audit? 2jQ.Yes (»«No
7. If 6. is yes, is the re—connection made with 25»tne operational
cable, _Q_a substitute cable or ° either?
S. Do you /"|QusuaI 1 v J_/_sometimes JJJnever measure the starting torque
of each anemometer bearing assembly and transducer?
9. Do you /_5_usually ,/,£| some times £j_.never measure the starting torque
of each wind vane bearing assembly and transducer?
1O. Do you ^usually /£ somet i mes j^*|^never use the collocated transfer
standard method for auditing a wind instrument?
11. Do you 3JLusual J V ./^sometimes ^_never find the audited instrument
meets the required specification?
^f ^K .
12. Do you challenge anemometers with known rates of rotation? Jjf'es
/^ No - If yes, how many speeds ? Synchronous or measured?
13. Do you challenge direction vanes with a dividing wheel? 2l£tVes
2J_No - If yes, how many angles ? CW, CCW, both.
14. Will you fill out a more detailed questionaire as a contribution to
the quality of this project? g*oves _L_No
Figure 4.0.2.1 Survey Form and Response Summary
-------
Section No. 4.0.2
Revision No. 0
Date: 17 Sep 89
Page: 4 of 6
The number of audits, sorted by the technical background rated by the
respondent as ttl, is shown in Table 4.0.2.2. It is comforting to note that the
largest number of auditors consider themselves meteorologists first. While the
largest number of audits were reported by persons considering themselves
managers first, it is likely that those they managed were distributed like the
rest of the group. One organization reported by a manager listed 3,600 audits.
It is likely that the discipline of the person doing most of the performance
audits is meteorologist.
The information from each question gives some feeling about how the
audits reported were conducted. Question 2 shows that 76% of the auditors back
their work with a written procedure. Question 4 shows that 64% of the auditors
usually use the manufacturer's calibration procedure. This answer pertains to
the calibration function which most auditors perform as a separate part of
their job.
Table 4.0.2.2 - Survey Summary
Meteorological Performance Audits
#1 Specialty Number
Meteorologist 11
Engineer 8
QA/QC 7
Instrument Tech. 5
Manager 4
Environ. Sci. 4
Electronics Tech. 4
Chemist 3
Data Analyst 1
No indication 2
Total 49
Average number per year
Percent change
1980-84
1,291
253
387
129-
2,115
224
510
352
0
0
5,261
1,052
1985
647
194
193
702
543
70
165
220
0
0
2,102
2,102
100
1986
473
188
237
102
551
130
181
256
0
30
2,256
2,256
7
1987
623
206
318
60
644
91
212
192
0
30
2,576
3,435
52
Total
3,034
841
1,135
669
3,853
515
1,068
1,020
0
60
12,195
In question 5, 43% of the auditors either do not physi'cally inspect the
sensor or do so by performing the operator's function of climbing the tower and
removing the sensor. Volume IV. should reduce that percentage to zero.
Question 6 suggests that most auditors (677.) do both a physical and an
operational challenge of the sensor when it is down from the tower. The
conditional question 7 shows a preference for the operational cable (71%) over
a substitute cable.
Questions 8 and 9 show only 43% of the auditors usually measure the
starting torque of the anemometer and only 41% usually measure the direction
vane starting torque. It looks like when an auditor decides to make this
measurement, both sensors are included. Several respondents answered "never"
but indicated that they were getting equipment to make the measurement in the
future. Other auditors inspect the bearing assemblies with educated fingers
which tell the auditor whether or not they are "all right" but fail to provide
numerical or objective documentation.
-------
Section No. 4.0.2
Revision No. 0
Date: 17 Sep 89
Page: 5 of 6
Question 10 shows that only 19% rely on the collocated transfer standard
(CIS) method for auditing wind instruments. Volume IV. should help increase
that number. Question 11 may mean that 70% of the instruments audited are
working within specification, or it may mean that the audit methods used are
not rigorous enough to find, the discrepancies. The fact that half the audits
do not include a torque measurement, the only method short of a wind tunnel to
challenge starting threshold, points to the latter possibility. Volume IV.
should help to improve audit methods toward a standard practice so that this
question, asked in the future, will provide an unambiguous answer.
Questions 12 and 13 show a difference in challenging speed and
direction. There were 73% of the auditors who indicated the use of a simulated
speed to challenge an anemometer. The most common number of -speeds was two
(52%). followed by three (21%), then one (15%), and finally four or more (12%).
There were 86% who indicated a synchronous motor was used. This near unanimity
is probably because of the availability of synchronous motors and the lack of
availability of simple measurement systems. The measured method is the only
choice where good commercial power is not available.
The direction challenges were not as uniform. Of the 23 who indicated
the number of angles used, seven said 4, five said 6, four said 8, three said
5, and one each said 1, 12, 16 and 18. There is no consensus there. All but
two said they used both clockwise and counterclockwise rotation. The two used
clockwise.
If the survey did one thing, it demonstrated the need for guidance
toward an acceptable standard of performance auditing. It also demonstrated a
recognition of need to move toward that goal and a willingness to help in the
process. Only one of 47 said no to question 14.
4.0.2.2 Interview Summary
After the survey results were in, a series of visits was planned to talk
to private sector organizations which had a recognized role in quality
assurance of meteorological measurements. The first organization visited was
AeroVironment, Inc. of Monrovia, California. The half-day discussion with four
AV auditors was a frank exchange of methods currently in use, shortcomings of
Volume IV and suggestions for the content of the revised Volume IV. The
principle of starting torque measurements of anemometer and wind vane shafts as
a field substitute for starting threshold wind speed determination in a wind
tunnel was accepted. The principle of operators doing all the climbing or
handling of sensors was currently practiced.
The second interview was at Environmental Monitoring and Systems, Inc.
in Thousand Oaks, California. Half-day discussions with two meteorological
auditors reinforced the belief that some organizations were advanced in the
practice of meteorological QA. Comprehensively written audit procedures were
followed. Questions of the difference between an audit and a calibration were
correctly answered with authority. The need for uniform expectations or
requirements was expressed in the context of competitive bidding for providing
audit services. •It was felt that the new Volume IV could help buyers of
-------
Section No. 4.0.2
Revision No. 0
Date: 17 Sep 89
Page: 6 of 6
services to specify a scope of work in enough detail to both assure a
comprehensive service and provide a fair bidding competition.
The third interview was with Environmental Research and Technology (now
ENSR) in Concord, Massachusetts, perhaps the largest of the five organizations
in terms of meteorological services and auditing. Some different concepts and
practices were found, particularly in the area of starting threshold
determination. Seven meteorologists, field auditors and QA specialists were
present during the half-day discussion. This organization was a leader in the
field of providing meteorological monitoring services to industry. As a result
of the history of providing all services including design, installation,
operation, data summarization and QA auditing, an interesting discussion was
held on the subject of independence between operators/calibrators and auditors.
The fourth interview was with the head of the field operations
department of TRC Environmental Consultants in E. Hartford, Connecticut. This
organization was also a leader in providing full meteorological monitoring
services. Their procedures developed in a different way. They calibrated
their sensors by wind tunnel testing in their calibration facility and employed
a regular replacement of sensors in the field. All of the performance auditing
related to sensors was done by QA personnel in the calibration facility. This
method requires a spare set of sensors be available for each client. The
methods described in Volume IV for calibrating or auditing in the field are not
necessary if you have a wind tunnel and employ the interchangeable sensor
method.
The final interview'was with a meteorologist/auditor from the Research
Triangle Institute of Research Triangle Park, North Carolina. Since two of the
original Volume IV authors were at RTI when the work was written, it was not
surprising to find the methods employed to be acceptable standard methods. The
level of quality of the field standards used in auditing was the highest, as it
was with most of these organizations.
This series of interviews provided valuable insights and confirmations
about the best methods to use for meteorological quality assurance practices.
It showed the field to be well practiced at the level of the largest and best
consulting organizations. The task for Volume IV is to provide a basis for a
standard practice in this field at all levels, and to provide a measure by
which those practicing in the field can be judged by those with the final
authority to accept or reject data on the basis of documented validity.
-------
Section No. 4.0.3
Revision No. 0
Date: 17 Sep 89
Page: 1 of 3
4.0.3 DATA REQUIREMENTS
There are a variety of reasons why meteorological data are collected. Some
reasons relate to regulatory requirements or national monitoring programs.
Some data are collected for the purpose of research. Some data are collected
against the contingency that they may be needed at some future time. Sometimes
data are collected for one reason and then used for other reasons.
The philosophy upon which this volume rests is the belief that data need to
have an estimation of uncertainty before the numbers can be dignified by the
title "data." The estimation might be a simple declaration such as "The
meteorological measurement program was operated in conformance with PSD
guidelines." This cites the accuracy requirements for PSD as the uncertainty
level for the data and promises that the documentation required for validity
claims for a PSD application will be available to back these data. When such
an estimation exists and rests on documentation of performance, decisions can
be made as to whether or not these data are appropriate for the application.
4.0.3.1 Regulatory Programs
4.0.3.1.1 PSD
The regulatory program used in this document, and to some extent in
the On-Site Meteorological Program Guidance for Regulatory Modeling
Applications (EPA, 1987b), is the Ambient Monitoring Guidelines for Prevention
of Significant Deterioration (PSD) (EPA, 1987a). This is the most explicit
guideline and one requiring a quality of instrument performance available only
from sensitive instruments. Recently it seems to be used for other programs as
a "standard" of specification.
Most air quality dispersion models described in the Guideline on Air
Quality Models (EPA, 1986) used for PSD applications are Gaussian models
requiring input data which represent the conditions at the site of interest and
which follow a prescribed data content and format. The models require five
meteorological inputs. They are:
1) Wind speed - representing the average wind speed at 10 m above
the ground (and additional heights for elevated sources) during each hour
calculated by a scalar average or mean of samples taken during the hour,
usually in 15 minute increments. The samples may be the integrated wind run
during the sample period (one or two seconds is often used) or instantaneous
samples of speed. A resultant vector magnitude does not represent the initial
dilution for which the Gaussian model uses wind speed.
2) Wind direction - representing the average wind direction at 10 m
above the ground (and additional heights for elevated sources) during each hour
calculated by carefully averaging samples of wind direction or by calculating
the resultant vector direction using unity as the wind speed for each sample.
A resultant vector direction does not represent the distribution of direction
samples which occurred during the hour.
3) Temperature - representing the air temperature at the standard 2
m height above ground (and additional heights for elevated sources).
-------
Section No. 4.0.3
Revision No. 0
Date: 17 Sep 89
Page: 2 of 3
4) Stability class -, representing the site of interest can be
estimated by a variety of schemes. Turner (1970) describes a method based on
the observation of cloud cover, ceiling height and wind speed along with a
known solar elevation angle. Estimations may also be based on the distribution
of wind direction (sigma theta) or on the vertical temperature gradient (delta
T). Current research is investigating whether or not the Turner method
stability class can be estimated with measurements of solar radiation (daytime
sky cover substitute) and 2 to 10 m delta temperature (nighttime sky cover
substitute) along with wind speed, latitude and date. The method which will be
acceptable for the site of interest is determined by the regulatory authority.
5) Mixing height - may be estimated by a method described by
Holzworth (1972).
The PSD guideline provides accuracy and performance requirements
for wind speed, wind direction, temperature, vertical temperature difference,
and solar radiation along with humidity, precipitation and visibility.
Measurements for PSD permitting may, in some cases, be continued
after the new source begins to verify the estimations made by modeling.
Continued monitoring requires the same QA/QC efforts as the permit phase
required.
4.0.3.1.2 Other Programs
Meteorological measurements may be made to augment air quality
measurements made to demonstrate compliance with the National Ambient Air
Quality Standards (NAAQS) or to monitor trends.
4.0.3.2 Research Programs
Meteorological data networks may be installed for special model
validation studies. The same kind of QA/QC efforts are necessary for these
programs but they are usually applied on a shorter time scale since the
programs are relatively short in duration and the need for documented accuracy
could not be greater.
Data representativeness is a critical question as the terrain increases
in complexity. Research looks into the number and location of measurement
sites and the applicability of certain types of instruments to characterize the
complex (turbulent or stratified) air flow systems. Different instruments,
such as Doppler sodars for remote vertical sounding, sonic anemometers for
small eddy size sensitivity and low threshold speeds, laser anemometers for
long path length integration, and even the old standby bivane, are examined to
try and optimize the detection of important aspects of flow measurement for
model inputs or verification.
Meteorological data are used to find correlations with aerometric
measurements in a continuing search for better forecasting capability.
4.0.3.3 Contingency Programs
*
Industry may choose to monitor meteorological variables at
representative sites on their property to document the local air flow
-------
Section No. 4.0.3
Revision No. 0
Date: 17 Sep 89
Page: 3 of 3
conditions in case excessive concentrations are observed which might
erroneously be attributed to their source. While such programs may not fall
under any regulatory requirements, the use of the data for its contingency
purpose requires documentation sufficient to verify the data accuracy.
Other programs may be exploratory to see how on-site data compare with
public data from other sites (airports or state or local agency stations).
Such questions of representativeness cannot be convincingly answered if the
on-site data does not come from suitably sensitive instruments, properly
calibrated and maintained and subject to QA/QC effort designed to document data
validity.
It is possible to select, install, operate and document on-site
measurement systems to meet PSD requirements. Public data from airports may
differ from valid on-site data for three reasons. Representativeness deals
with different meso-scale structures in the surface layer flow driven, in part,
by the larger synoptic flow. It is common to find that airport measurements do
not represent other sites just a few miles away because the flow is different.
It is also common to find airport data to be different from on-site data
because the airport data is essentially an instantaneous sample (a one minute
average) taken within ten minutes of the end of the hour while the on-site data
for the same hour includes samples from the entire hour. Finally, the airport
instruments are selected to serve aviation where low wind speeds are of no
importance. Airport instruments do not meet PSD requirements.
-------
Section No. 4.0.4
Revision No. 0
Date: 17 Sep 89
Page: 1 of 7
4.0.4 MEASUREMENT REQUIREMENTS
4.0.4.1 Measurement System
4.0.4.1.1 Sampling
The usual period of time assigned a data value is an hour. It
is getting common to use a shorter intermediate period in the process of
generating the hour value. Fifteen minutes is the recommended intermediate
period. The fifteen minute values are usually calculated from samples taken
during the period. The number of samples is related to the accuracy with which
the samples represent the true value for the period. It has been found that
when the mean is estimated by 60 samples, the sampling error is 5 to 10
percent. Also, when the standard deviation is estimated by 360 samples, the
sampling error is also 5 to 10 percent. For this reason, the required number
of samples for sigma theta, the standard deviation of the wind direction is
equal or greater than 360 (EPA, 1987b).
If a fifteen minute period is used and if 360 samples are
required within that period, a simple calculation shows the maximum time
between samples is 2.5 seconds. How a sample is taken and what it represents
is also a consideration. If a continuous output voltage is available, as with
wind direction, a sample of the voltage can be taken at any time. If the wind
speed is calculated by measuring rate of rotation by counting pulses during a
fixed time, as is common for systems with the sensor directly connected to a
data logger (without a dedicated signal conditioner), the "sample" is really
the average for the fixed time. If samples are taken once a second and the
anemometer provides three pulses per revolution and the anemometer turns one
revolution for every 0.3 meters of air that goes through it, each pulse will
represent 0.9 m/s. If samples are taken every 2 seconds, the resolution of the
wind speed sample becomes 0.45 m/s. A 15 minute period at 2 second sampling
will have 450 samples. The average wind speed will be accurate with a
resolution of better than 0.1 m/s. The variance of the wind speed samples may
be influenced by the 0.45 m/sresolution of the sample.
Quality assurance considerations should include the
determination and documentation of the sampling procedures used in generating
the reported hourly data values.
4.0.4.1.2 On-Line Processing
There are two on-line processing programs commonly used in air
quality meteorology. One is the program used to combine wind speed and
direction samples for an hour. The other is the program used to calculate or
estimate sigma theta.
The QA role is to determine what these programs do and judge
the suitability of the programs for the measurement application. The field of
software QA for meteorology is in it infancy and methods are not standardized
as yet.
4.0.4.1.3 Data Handling
There is a need to provide data in certain formats for some
applications. If the data are machine processable in the final measurement
-------
Section No. 4.0.4
Revision No. • 0
Date: 17 Sep 89
Page: 2 of 7
step, any reconfiguration required will be handled by a program which can be
subject to software QA.
If any hand entry work is required, a data handling QC step is
required to be sure that errors of transcription do not enter the data base.
4.0.4.2 Documentation
If there is a requirement to show evidence of data validity, the process
of documenting the various QA, QC, and operational activities is important.
The added time such documentation takes is usually proportional to the degree
of preparation and training which has been applied.
4.0.4.2.1 Station Log
The station log is the journal of all happenings at the
measurement site. These include visits where no problems are found, scheduled
calibration visits and findings, unscheduled maintenance tests and repairs, and
audits. It is a truism that there are never enough field notes to reconstruct
with certainty what happened in the past, Planning for the day when such a
reconstruction may be necessary can save a long period of data from being
discarded because of inadequate documentation.
4.0.4.2.2 Reports
Any activity effecting the measurement system should be
reported. This procedure allows responsible individuals to follow these
activities without visiting the measurement site or witnessing calibrations and
audits. It also provides input to a file of activities related to the system.
Reports should include calibrations, audits, discrepancies found and corrected,
modifications or upgrades and the like. Reports do not need to be exhaustive
or glossy but they do need to be as factual and succinct.
4.0.4.3 Siting and Mounting
4.0.4.3.1 Introduction
Although good instrumentation is a necessity, proper site
selection is critical to obtain good meteorological data. It is, from an
absolute error point of view, much more important than proper placement of any
other kind of air monitoring equipment. Poor placement can and has caused
errors of 180 in wind direction, and can cause major errors in any other
meteorological variable, including wind speed, temperature, humidity, and solar
radiation.
The purpose of this section is to offer guidance in assessing
the suitability of meteorological monitoring sites. The guidance given is
based principally on standards set by the World Meteorological Organization
(WHO, 1971), the Federal Meteorological Handbook No. 1 (NWS, 1979) and the
Tennessee Valley Authority (TVA, 1977). For an understanding of flow around
obstacles and their potential bias to wind data, see Hosker (1984).
Proper siting is part of the total quality control program. Of
course, as in many other monitoring activities, the ideal may not be attainable
and, in many urban areas where air quality studies are traditionally done, it
will be impossible to find sites that meet all of the siting criteria. In
-------
Section No. 4.0.4
Revision No. 0
Date: 17 Sep 89
Page: 3 of 7
those cases, compromises must be made. The important thing to realize is that
the data will be compromised, but not necessarily in a random way. It is
incumbent upon the agency gathering the data to describe carefully the
deficiencies in the site and, if possible, quantify or at least evaluate the
probable consequences to the data.
4.0.4.3.2 Instrument Siting
The primary objective of instrument siting is to place the
instrument in a location where it can make precise measurements that are
representative of the general state of the atmosphere in that area, consistent
with the objectives of the data collection program. Because most atmospheric
properties change dramatically with height and surroundings, certain somewhat
arbitrary conventions must be observed so that measurements can be compared.
In this section, conventions published by the World Meteorological Organization
(WHO, 1971) have been adopted wherever possible. Secondary considerations such
as accessibility and security must be taken into account, but should not be
allowed to compromise data quality.
4.0.4.3.2.1 Wind Speed and Direction
"The standard exposure of wind instruments over level,
open terrain is 10 m above the ground" (WHO, 1971), however optimum measurement
height may vary according to data needs. Open terrain is defined as an area
where the horizontal distance between the instrument and any obstruction is at
least ten (10) times the height of that obstruction. An obstruction may be
man-made (such as a building) or natural (such as a treeHFigure 4.0.3.1).
TOT
Figure 4.0.4.1
Siting wind instruments; a 10 m tower located at 'east
10 times the height of obstructions away from those
obstructions (not to scale).
The wind instrument should be securely mounted on a mast
that will not twist, rotate, or sway. If it is necessary to mount the wind
instrument on a roof of a building, it should be mounted high enough to be out
of the area in which the air flow is disturbed by the building. This is
usually 1.5 times the height of the building above the roof so that it is out
-------
Section No. 4.0.4
Revision No. 0
Date: 17 Sep 89
Page: 4 of 7
of the wake of the obstruction. This is not a good practice, however, and
should only be resorted to when absolutely necessary. Sensor height and its
height above the obstructions, as well as the character of nearby obstructions,
should be documented.
4.0.4.3.2.2 Temperature and Humidity
Temperature and humidity sensors should be mounted over a
plot of open level ground at least 9 meters in diameter. The ground surface
should be covered with non-irrigated or unwatered short grass or, in areas
where grass does not grow, natural earth. The surface must not be concrete or
asphalt or oil soaked. The standard height for climatological purposes is 1.25
to 2 m, but different heights may frequently be required in air quality
studies.
The sensors should not be closer to obstructions such as
trees and/or buildings than a distance equal to four times their height. They
should be at least 30 m from large paved areas and not close to steep slopes,
ridges, or hollows. Areas of standing water should also be avoided. Louvered
instrument shelters should be oriented with the door opening toward true north,
in the northern hemisphere.
4.0.4.3.2.3 Radiation
Solar and whole sky radiation measurements should be taken
in a location free from any obstruction to the measurements. This means there
should be nothing above the horizontal plane of the sensing element that would
cast a shadow on it. Neither should the instrument be near light colored walls
or artificial sources of radiation. Usually a tall platform or roof of a
building is the most suitable location.
4.0.4.3.2.4 Precipitation
A rain gage should be mounted on level ground so that the
mouth or opening is horizontal. The gage should be shielded from the wind but
not placed in an area where there will be excessive turbulence caused by the
shield. For example, a good location would be an opening in an orchard or
grove of trees where the wind speed near the ground is reduced due to the
canopy effect, but a location that is mostly open except for one or two trees
would not be good because of the strong eddies that could be set up by the
trees. This admittedly requires a good deal of subjective judgment but it
cannot be avoided. Obstructions to the wind should not be closer than two to
four times the obstruction height from the instrument. In open areas, a wind
shield such as that used by the U.S. National Weather Service should be used.
The ground surface around the rain gage may be natural vegetation or gravel.
It should not be paved, as this may cause splashing into the gage. The gage
should be mounted a minimum of 30 cm above the ground and should be high enough
so that it will not be covered by snow.
4.0.4.3.2.5 Meteorological Towers
It is frequently necessary to measure some meteorological
variables at more than one height. For continuous measurements or where the
height requirement is not too restrictive, towers may offer the most
advantageous measurement platform.
-------
Section No. 4.0.4
Revision No. 0
Date: 17 Sep 89
Page: 5 of 7
Towers should be located in an open level area (see Table
4.0.4.1) representative of the area under study. In terrain with significant
topographic features, different levels of the tower may be under the influence
of different meteorological regimes at the same time. Such conditions should
be well documented.
Table 4.0.4.1 Limits on Terrain and Obstacles Near Towers
Distance
from tower
(m)
0- 15
15- 30
30-100
100-300
Source: TVA,
Slope
(between)
(7.)
± 2
± 3
± 7
± 11
1977
Max. obstruction or
vegitation height
0.3
0.5-1.0
3.0
10 x ht.
(m)
(most veg.<0.3)
must be less
than distance
to obstruction
Towers should be of the open grid type of construction,
such as is typical of most television and radio broadcast towers. Enclosed
towers, stacks, water storage tanks, grain elevators, cooling towers, and
similar structures should not be used (Mollo-Christensen, 1979). Towers must
be rugged enough so that they may be safely climbed to install and service the
instruments. Folding or collapsible towers that make the instruments available
to be serviced or calibrated at the ground are desirable provided they are
sufficiently rigid to hold the instruments in the proper orientation and
attitude during normal weather conditions.
Wind instruments should be mounted above the top of the
tower or on booms projecting horizontally out from the tower. If a boom is
used, it should support the sensor at a distance equal to twice the maximum
diameter or diagonal of the tower away from the nearest point on the tower.
The boom should project into the direction which provides the least distortion
for the most important wind direction. For example, a boom mounted to the east
of the tower will provide least distortion for north or south winds. One may
wish to consider having two sets of instruments at each level, located on
opposite sides of the tower. A simple automatic switch can choose which set of
data to use (NASA, 1968). Documentation of the tower should include the
orientation of the booms.
Temperature sensors must be mounted on booms to hold them
away from the tower, but a boom length equal to the diameter of the tower is
sufficient. Temperature sensors should have downward facing aspirated shields.
The booms must be strong enough so that they will not sway or vibrate
excessively in strong winds. The best vertical location on the tower for the
sensors is at a point with a minimum number of diagonal cross members, and away
from major horizontal cross members. Even with these precautions, data
obtained while the wind blows from the sector transected by the tower may not
be free from error.
These instrument siting suggestions may seem to preclude
the use of many air monitoring sites that otherwise would be desirable, but
-------
Section No. 4.0.4
Revision No. 0
Date: 17 Sep 89
Page: 6 of 7
this need not be the case. In siting air quality monitors that are to be used
for long-term trend analysis or large geographic area coverage, it may be
perfectly acceptable to have some or all of the meteorological equipment at a
different location that better meets the large-scale requirements of the study.
As long as both sites are in the same area of interest and meet their
respective siting criteria, this should present no proolems. When the air
quality data are to be used for short-term diffusion model validation or
studies of short-term levels from specific sources, however, a meteorological
station should be located in the vicinity of the air quality sensor.
4.0.4.3.3 Station Siting
Besides care in selecting the local environment of a
meteorological sensor, it is important that care be taken in selecting station
location with respect to major man-made and topographic features such as
cities, mountains, large bodies of water, etc. Meteorological variables are
obviously affected by the large-scale surrounding features. The effect of
cities has been studied extensively (Ito, 1972; Vukovich, 1971; U.S.PHS, 1961).
Documented effects include a decrease in an average wind speed, decrease in
atmospheric stability, increase in turbulence, increase in temperature, and
changes in precipitation patterns. These changes will obviously have an effect
on the evaluation and interpretation of meteorological and air quality data
taken in an urban area.
Even more pronounced are the effects of large natural features
(Slade, 1963). Besides their obvious effect on humidity, oceans and large
bodies of water are usually at a different temperature than the nearby land.
This generates the well known land- and sea-breezes which, in many coastal
areas, dominate the wind patterns. There are also often simultaneous
differences in cloud cover between the oceans and nearby land surfaces. This
difference in thermal lag, insolation, and changes in surface roughness and
vertical temperature structure can have a profound effect on atmospheric
stability (SethuRaman, 1974).
The effects of mountains and valleys on meteorological
variables and atmospheric dispersion continue to be studied. Two of the more
interesting recent papers are by Kao (1974) and Hunt (1978). Well-known
effects include the channeling of flow up or down a valley, the creation of
drainage flow, the establishment of lee-waves, and an increase in mechanically
generated turbulence. All of these effects and others can play a major role in
the transport and dispersion of pollutants.
The important point is that almost any physical object has an
effect upon atmospheric motion. In fact, it is probably impossible to find a
site that is completely free from obstruction. This being the case, it is the
responsibility of the person choosing a monitoring site to have in mind the
various forces at work and to choose a site that will be most representative of
the area of interest. If the area is a valley or a sea coast, then the
meteorological instruments should be in that valley or near the coast; not on a
nearby hilltop or inland 30 km at a more convenient airport site. Of course
one must also keep in mind the vertical structure of the atmosphere. Winds
measured at the bottom of a 100 m valley will not represent the winds at the
top of a 200 m stack whose base happens to be in that valley.
-------
Section No. 4.0.4
Revision No. 0
Date: 17 Sep 89
Page: 7 of 7
The choice of a station for meteorological data collection must
be made with a complete understanding of the large-scale geographic area, the
sources being investigated, and the potential uses of the data. Then rational,
informed choices can be made.
Once they are made, the site should be completely documented.
This should include both small- and large-scale site descriptions, local and
topographic maps (1:24,000 scale), photographs of the site (if possible), and a
written description of the area that is adequately represented by this site.
This last point is most important for it will allow a more rational
interpretation of the data. It might state, for example, that a site
adequately represents a certain section of a particular valley, the suburban
part of a given city, or several rural counties. Whatever it is, the nature of
the site should be clearly described in a way that will clear to those who will
be using the data in the future.
-------
Section No. 4.0.5
Revision No. 0
Date: 17 Sep 89
Page: 1 of 2
4.0.5 REFERENCES
EPA, 1976: Quality Assurance Handbook for Air Pollution Measurement
Systems, Vol. I, Principles, EPA-600/9-76-005. Office of Research and
Development, Res. Triangle Park, NC 27711.
EPA, 1977a: Quality Assurance Handbook for Air Pollution Measurement
Systems, Vol. II, Ambient Air Specific Methods, EPA-600/4-77-027a. Office
of Research and Development, Res. Triangle Park, NC 27711.
EPA, 1977b: Quality Assurance Handbook for Air Pollution Measurement
Systems, Vol. Ill, Stationary Source Specific Methods,
EPA-600/4-77-027b. Office of Research and Development,
Res. Triangle Park, NC 27711.
EPA, 1982: APTI Course SI: 409, Basic Air Pollution Meteorology, Student
Guidebook EPA 450/2-82-009. Air Pollution Training Institute,
Environmental Research Center, Research Triangle Park, NC 27711.
EPA, 1983: Quality Assurance Handbook for Air Pollution Measurement
Systems, Vol. IV, Meteorological Measurements, EPA-600/4-82-060. Office
of Research and Development, Res. Triangle Park, NC 27711.
EPA, 1986: Guideline on Air Quality Models (Revised). OAQPS, U.S.
Environmental Protection Agency, Research Triangle Park, NC.,
EPA-450/2-78-027R (NTIS PB 288-783).
EPA, 1987a: Ambient Monitoring Guidelines for Prevention of
Significant Deterioration (PSD), EPA-450/4-87-007, Office of Air
Air Quality Planning and Standards, Res. Triangle Park, NC 27711
EPA, 1987b: On-Site Meteorological Program Guidance for Regulatory
Modeling Applications, EPA-450/4-87-013, Office of Air
Quality Planning and Standards, Res. Triangle Park, NC 27711.
Holzworth, G. C., 1972: Mixing Heights, Wind Speeds, and Potential for
Urban Air Pollution Throughout the Contiguous United States. Office of
Air Programs, U.S. Department of Health, Education and Welfare, Research
Triangle Park, NC. Publication Nol. AP-101.
Hosker, Jr.,R. P.,1984: Flow and Diffusion Near Obstacles, Chapter 7.
Atmospheric Science and Power Production, Darryl Randerson,
Ed. DOE/TIC-27601, pp 241-326.
Hunt, J. C. R., W. H. Snyder and R. E. Lawson, Jr., 1978: Flow structure
and turbulent diffusion around a three dimensional hill: part 1.
EPA-600/4-78-041. U.S. Environmental Protection Agency.
Ito, Masashi et al., 1972: An examination of local wind measurements in
cities. Annual Report of the Tokyo Metropolitan District Public Damage
Research Institute,3(3), pp. 27-32 (APTIC 41974 TR 186-73).
-------
Section No. 4.0.5
Revision No. 0
Date: 17 Sep 89
Page: 2 of 2
Kao, S. K., H. N. Lee and K. I. Smidy, 1974: A preliminary analysis of the
effect of mountain-valley terrains on turbulence and diffusion.
Symposium on Atmospheric Diffusion and Air Pollution, American
Meteorological Society, Santa Barbara, CA, pp. 59-63.
Mollo-Christensen, E., 1979: Upwind distortion due to probe support in
boundary layer observations. Journal of Applied Meteorology, 18(3),
pp. 367-370.
NASA, 1968: Meteorological Measuring and Recording Equipment
Description, Calibration and Maintenance Procedures for NASA's 150
Meter Meteorological Tower Facility, Kennedy Space Center, Florida,
National Aeronautics and Space Administration, GP-465.
NWS, 1977: Federal Meteorological Handbook No. 1, Surface Observations.
U.S. Department of Commerce, Washington, D.C.
SethuRaman, S., R. M. Brown, and J. Tichler, 1974: Spectra of
atmospheric turbulence over the sea during stably stratified
conditions. Symposium on Atmospheric Diffusion and Air Pollution,
American Meteorological Society, Santa Barbara, CA, pp. 71-76.
Slade, D. H., 1968: Meteorology and Atomic Energy. U.S. Atomic Energy
Commission, TID-24190.
Turner, D. B., 1970: Workbook of Atmospheric Dispersion Estimates,
Revised. Office of Air Programs, U.S. Department of Health, Education
and Welfare, Research Triangle Park, NC. Publication No. AP-26.
TVA, 1977: Criteria for meteorological measurement site acceptance
and/or preparation (unpublished). Tennessee Valley Authority.
U.S.Public Health Service, 1961: Air Over Cities. A symposium,
Cincinnati, Ohio.
Vukovich, F. M., 1971: Theoretical analysis of the effect of mean wind
and stability on a heat island circulation characteristic of an urban
complex. Monthly Weather Review, 99(12), pp. 919-926.
WMO, 1971: Guide to meteorological instrument and observing practices.
World Meteorological Organization No. 8TP3, 4th edition, Geneva,
Switzerland.
-------
-------
Section No. 4.1.0
Revision No. 0
Date: 17 Sep 89
Page: 1 of 3
Section 4.1
GENERAL ASPECTS OF QUALITY ASSURANCE FOR
METEOROLOGICAL MEASUREMENTS
OUTLINE
Section Pages Rev. Date
4.1.0 OUTLINE AND SUMMARY 3 0 9/89
4.1.1 PLANNING FOR A QUALITY ASSURANCE PROGRAM 3 0 9/89
4.1.1.1 PROJECT DESCRIPTION
4.1.1.2 PROJECT ORGANIZATION
4.1.1.3 QA OBJECTIVE
4.1.1.4 CALIBRATION METHOD AND FREQUENCY
4.1.1.5 DATA FLOW ANALYSIS
4.1.1.6 VALIDATION AND REPORTING METHODS
4.1.1.7 AUDITS - PERFORMANCE AND SYSTEM TYPES
4.1.1.8 PREVENTIVE MAINTENANCE
4.1.1.9 QA PROCEDURES
4.1.1.10 CORRECTIVE ACTION AND REPORTS
4.1.2 ORGANIZATION OF AUTHORITY AND RESPONSIBILITY 1 0 9/89
4.1.3 QALITY CONTROL GUIDELINES 2 0 9/89
4.1.4 TRACEABILITY PROTOCOL 4 0 9/89
4.1.4.1 VOLTAGE
4.1.4.2 WIND SPEED
4.1.4.3 WIND DIRECTION
4.1.4.4 TEMPERATURE AND TEMPERATURE GRADIENTS
4.1.4.5 SOLAR RADIATION
4.1.4.6 ATMOSPHERIC WATER VAPOR
4.1.4.7 PRECIPITATION
4.1.4.8 ATMOSPHERIC PRESSURE
4.1.5 ESTIMATING PRECISION AND ACCURACY 6 0 9/89
4.1.5.1 DEFINITIONS
4.1.5.2 COLLOCATED TRANSFER STANDARDS
4.1.5.3 OTHER CONSIDERATIONS
4.1.6 SYSTEM AUDITS 2 0 9/89
4.1.7 PERFORMANCE AUDITS 1 0 9/89
4.1.8 DATA VALIDATION PROTOCOL 7 0 9/89
4.1.8.1 STRIP CHARTS
4.1.8.2 METHODS
4.1.8.3 *THE EMSL/RTP SYSTEM
-------
Section No. 4.1.0
Revision No. 0
Date: 17 Sep 89
Page: 2 of 3
4.1.9 QA REPORTS AND CORRECTIVE ACTION 2 0 9/89
4.1.9.1 OPERATIONS LOG AND MAINTENANCE REPORTS
4.1.9.2 CALIBRATION REPORTS
4.1.9.3 AUDIT REPORTS
4.1.9.4 REPORTS TO MANAGEMENT
4.1.9.5 DISCREPANCY REPORTS
4.1.10 REFERENCES 2 0 9/89
-------
Section No. 4.1.0
Revision No. 0
Date: 17 Sep 89
Page: 3 of 3
GENERAL ASPECTS OF QUALITY ASSURANCE
FOR METEOROLOGICAL MEASUREMENTS
SUMMARY
Quality assurance (QA) for meteorological measurements is a relatively new
field. There are generally two reasons for recording meteorological data. One
is to learn what the atmosphere is doing, particularly the lower part of the
boundary layer. The other is to document what the atmosphere is doing. It is
necessary to find the relevant facts and understand them to learn something .
To document what has been learned may require considerably more data and will
require some assurance the data are correct. The organizations that need valid
data are the ones which collect it, and they will write their own procedures.
This is how the National Weather Service has handled data collection for
synoptic and climatological applications. When third party requirements with
the force of law began to need meteorological data for transport and diffusion
modeling and safety analysis, the need for QA was established.
As with most specialties, HA in general has acquired its own language and
infrastructure. In this handbook, the goal is to avoid structure which has no
specific value to meteorological data validity. On the other hand, the goal is
to provide clear definitions, methods and examples which will help produce and
verify valid meteorological data. Some of the popular sayings or phrases make
valid points. The book "QUALITY IS FREE" (Crabby, 1979) promotes the idea that
it is really cheaper to do it right the first time. This concept is easy to
demonstrate with manufactured products where bad products will either cost more
through warranty repairs or lost sales and bad will. If the "product" is data
or services producing data, an awareness of the ultimate cost of the loss of
data is important. If no one ever looks at or uses the data, it is a waste
of money to buy, install and operate instruments and recording systems. Even a
facade of data is not worth the money. If there is a reason for meteorological
measurement, that reason will provide the basis for estimating the economic
down-side for producing unacceptable data.
Of course data judged "unacceptable" must have been rejected by someone for
some reason. The reason for needing the data in the first place will provide
the basis for the economic price which must be paid for either not "doing it
right the first time" or for fighting the rejection because it was done right.
Usually the reason for measuring meteorological variables is a government
regulation requirement or a need to combat potential claims of injury. It can
be argued that the fagade of data coming from the instruments on the tower will
satisfy the government regulation requirement. It has been argued that an
extra nickel spent measuring meteorological variables (or air quality, for that
matter) is a nickel lost to the bottom line profit of a manufacturing plant.
The assumption upon which this handbook rests is that the pertinent
government regulations and guidance documents will clearly define what valid
data are and how validity is proven. And further, that those people
responsible for accepting or using the data will require that the data be valid
or rejected. The expertise necessary for this determination will be found in
this handbook and its references. This should equally help those who must
collect valid data and those who must accept the claim of validity. This
assumption clearly moves the purpose for collecting data away from learning and
into the category of documenting. Documentation needs QA.
-------
Section No. 4.1.1
Revision No. 0
Date: 17 Sep 89
Page: 1 of 3
4.1.1 PLANNING FOR A QUALITY ASSURANCE PROGRAM
A formal quality assurance program should be designed into the monitoring
program so that provisions can be made in the measurement system design for
necessary quality control checks and for better monitoring of system
operations. If these activities are planned and provided for by incorporation
of special readouts, calibration equipment, spares, and procedures for their
use, then the system is more likely to perform in a satisfactory manner and
deliver valid data.
The formal plans for quality assurance are presented in a document called a
QA Plan (Lockhart, 1985c and EPA, 1987b). This plan lists all the
quality-related procedures and the frequency of their use to document the
operation of the instrument system. The QA Plan contains information under
different headings to organize all the various activities in a logical sequence
and to avoid overlooking an important step. The specifics of each plan must
relate to the needs of the program, but the general content elements are the
same.
4.1.1.1 Project Description
This introduction establishes why the documentation of meteorological
data monitoring is needed and why it is important to the organization that
valid data are collected. It also describes how the data will be used which
establishes the criteria for judging the representativeness of the data.
4.1.1.2 Project Organization
The literature of QA abounds with examples of the importance of well
defined organizational structure starting with the organization policy on
quality, endorsed in writing by top management. This provides the authority to
"do it right the first time." If the organization has no policy on quality and
if someone at the operations level is given the QA responsibility without
sufficient authority (often the case), the effort may become just another
secondary task which must be done. This is an invitation to a fagade system.
An organization will seldom build a new plant without the expertise of
architects and engineers. Meteorological data systems are often assembled from
parts picked from catalogs by experts in other fields who do not understand the
routine operating requirements for collecting valid data. A valid QA Plan is a
structure to encourage and guide organizations toward a successful collection
of needed data.
4.1.1.3 QA Objective
This section is the real QA plan. The first two sections described the
project for which the data will be used and the organization of those who will
participate in the data collection. This section contains the details of how
the QA program will monitor the collection process with the purpose of
documenting and defending claims of data validity.
4.1.1.4 Calibration Method and Frequency
Calibrations are tests, and adjustments if necessary, to relate the
instrument system to truth or validity. The evidence of this activity, the
documentation, is the foundation upon which the judgment of data validity must
-------
Section No. 4.1.1
Revision No. 0
Date: 17 Sep 89
Page: 2 of 3
rest. This section defines in advance how the calibrations will be done, how
often and by whom.
4.1.1.5 Data Flow Analysis
This section starts with samples of atmospheric conditions, a rate of
rotation of an anemometer representing wind speed for example, and describes
how the samples are combined into reported values. The section describes how
these values, perhaps hourly averages, are inspected and judged to be
acceptable or not. Finally, after validity has been established, the data are
archived in some way to become available for use as the project requires. An
experienced meteorologist reading this section will know what the data mean and
what data quality control has been applied.
4.1.1.6 Validation and Reporting Methods
Section 4.1.1.4 provides points in time at which the instrument
performance is known. This section describes what criteria are used for any
automatic data inspection programs applied between calibrations and how the
results of such programs are implemented and reported. If comparisons are made
to other similar measurements, such as a wind speed at a different location or
a different time, this section will document the methodology applied.
4.1.1.7 Audits - Performance and System Types
Audits may be required or chosen to add to the documentation some
independent evidence of the performance of the meteorological instruments
and/or the performance of those who are responsible for implementation of the
QA Plan. This section defines how often performance audits are used to
challenge the measurement instrument system and how often system audits are
used to challenge the implementation of the QA Plan or program. Also defined
is the type of auditor to be used. Internal auditors are members of the
organization who are independent from those responsible for collecting and
handling the data. External auditors are usually outside contractors. In
either case, the auditor must be experienced in the field of meteorology and
must be provided support from the operating organization. Auditing should be
the most positive learning experience for operators and a contributor to data
validity.
4.1.1.8 Preventive Maintenance
Some instruments require routine service to assure valid data. Solar
radiation sensors have glass covers which need to be properly cleaned on some
schedule (daily or weekly), depending on its location. Tipping bucket
raingages need to be checked periodically for spiders or other insects which
might take up residence in the bucket mechanism. Anemometers and wind vanes
have bearings which will need service (usually replacement) on some time scale
(quarterly or annually), depending on the environment. Some dew point sensors
require coating periodically. All such predictable service points should be
recognized and a preventive maintenance plan described for each of them in this
section.
-------
Section No. 4.1.1
Revision No. 0
Date: 17 Sep 89
Page: 3 of 3
4.1.1.9 QA Procedures
Section 4.1.1.2 describes the QA objectives. This section contains the
details of how these objectives will be met. A written procedure will both
document how the QA task will be addressed and guide a QA person through the
process. Procedures are a mechanism for establishing.technically correct
methodology which can be followed by people without the technical background or
experience to write the procedure. While it is not practical to use experts to
perform routine tasks, it is necessary to have the expert guidance to follow.
Procedures fill this need. Procedures should be controlled to the degree that
they cannot be changed without written approval of an expert. A system audit
checks to see if procedures are being followed as they are written.
4.1.1.10 Corrective Action and Reports
Documentation is the main goal of a QA Plan. General procedures will
require noting in the Site Log any activity relating to the meteorological
system. Top management, having established the policy and granted the
authority to "do it right the first time," needs to be aware of the QA
activities required by the QA Plan. If a procedure or inspection uncovers a
discrepancy with respect to the clearly written system specifications, a
discrepancy report becomes the message to initiate corrective action. Top
management needs to see these reports along with the corrective action
statements (usually a part of the discrepancy report form) to know that the
system is in control. Too often, problems must be visible to top management if
corrective action to the system is going to be initiated. Audit reports and
other performance reports are circulated and followed up by procedures
described in this section of the QA Plan.
-------
Section No. 4.1.2
Revision No. 0
Date: 17 Sep 89
Page: 1 of 1
4.1.2 ORGANIZATION OF AUTHORITY AND RESPONSIBILITY
Quality does not mean the best money can buy. It means that the customer
gets what he needs and expects, no more and no less. The way to assure a
quality product or service is to first set a top management policy in writing
stating that goal or commitment. The implementation of such a policy requires
a person who can avoid the conflict of interest of providing the product or
service and judging its quality before it is delivered. This usually means a
Quality Assurance Manager. Once the policy is set and the QA Manager is chosen
the procedures by which quality is assured can be written, usually a QA manual.
The manual might proclaim that every project with a deliverable product or
service will have a QA Plan.
The authority to establish this kind of organizational structure must be
top management. During the establishment period, top management must
participate and approve the quality organizational structure and procedures.
Once established, top management can delegate authority to the QA Manager for
operating the department. A routine feedback from the QA Manager to top
management is necessary to preserve the control of delegated authority, see
that it is being used effectively and demonstrate to the rest of the
organization a level of importance placed on the quality policy.
Once the organization is in place, and QA Plans are required, the job of
collecting wind data for a PSD (Prevention of Significant Deterioration)
application starts with a QA Plan. What are the measurement requirements? How
will the data be handled? How will the instruments be calibrated and serviced?
What procedures will be used? What outside authority will assure management
that the QA Plan is adequate? Once all the steps required to gather a year of
valid wind data have been defined, the QA people will monitor the process, help
train inexperienced operators, and build a documentation base to support the
claim of validity for the data once the year is over. This should include some
outside auditing to add to the documentation an impartial expert opinion of
satisfactory performance, or bring to the attention of the QA people and their
management any problems which might have been overlooked. The best time for an
independent audit is at the beginning of the data-taking period when the loss
of valuable time can be minimized.
The principles of this approach to quality are sound and irrefutable. The
cost is less than any other approach. All that is required is to figure out,
in sufficient detail, what and how the job is to be done before it begins and
to specify how the job will be monitored to assure satisfactory completion.
Anything less is a gamble which may or may not pay off. Most organizations do
not like to gamble, but many do.
-------
Section No. 4.1.3
Revision No. 0
Date: 17 Sep 89
Page: 1 of 2
4.1.3 QUALITY CONTROL GUIDELINES
Quality control is a process which operates in parallel with the production
of a product or a service. There is a gigantic body of literature on the
subject. Some examples of books are Juran (1979), Feigenbaum (1961) and Grant
and Leavenworth (1974). A technical organization, the American Society For
Quality Control (ASQC) exists for the promotion of quality systems. ASQC has
committees of volunteers to establish guidelines and standards for the quality
profession. One such committee is the American National Standards Institute
Accredited Standards Committee (ANSI ASC) Z-l Committee on Quality Assurance.
A product of this committee is a series of standards (ANSI/ASQC 1987a-e) in
which is stated the need to qualify quality control with an adjective
describing what is to be controlled. This need is nowhere greater than in
meteorological monitoring.
In the meteorological literature there are recent papers (Wade, 1987 and
Lockhart, 1988) in which the use of quality control is discussed. There is a
difficulty with the language used to describe a task and the perception of
control as related to those working the task. Assume the task is the
accumulation of a one year data set of wind and temperature measurements. The
QA Plan describes the goals and the specifications to which the instruments
must conform. Purchasing has bought the instruments with a purchase order
requiring conformance to performance specifications and describing how the
conformance will be tested. A suitable site has been found and a consensus has
been reached as to the representativeness of the site. Receiving inspectors
accepted the instruments and operators have installed them. The QA Plan called
for an independent performance audit at the beginning of the data year which
the system passed. The QA Plan calls for an inspection of the data on a weekly
basis by the meteorologist or environmental scientist who will be working with
the data. The QA Plan provides a procedure with which the data inspector can
communicate in writing with the operators to report questionable data and
receive an answer of special instrument checks. The QA Plan requires operators
to calibrate the instruments on a six month frequency or when problems are
found. The QA Plan specifies how the calibrations are to be done and to whom
the reports will be routed. At the end of the year the data are summarized and
made ready for use in diffusion models. Where is control and where is quality
control?
The whole program is controlled by the top management through the QA Plan.
How well the various parts of the organization carry out their responsibilities
is checked on by the QA people or person. If the receiving inspectors balk at
performing their service because they are too busy, top management participates
in the decision to either modify the QA Plan and Policy on Quality or find a
way to accomplish the receiving inspection. Management may choose to gamble
that the instruments are all right and any problems will be uncovered during
the installation and audit. This is management's prerogative and this gamble
is a pretty safe one. Without the QA Plan the manager of receiving may make
the decision without the benefit of knowing what the stakes of the gamble might
be. While the whole project is controlled by the QA Plan, there is one data
quality control function performed by the meteorologist/environmental
scientist; on a weekly basis the data are examined and accepted or rejected.
If problems or questions arise, discrepancy reports will be initiated which
operate in accordance with the QA Plan. The inspector finds that the wind
direction looks too steady and writes it up. The operator goes to the site
-------
Section No. 4.1.3
Revision No. 0
Date: 17 Sep 89
Page: 2 of 2
and, finding that the vane has been removed by a predatory bird, installs a
spare, notes the action in the site log and on the discrepancy report,
recalibrates if necessary, and reports back to the inspector closing the loop.
This is a true quality control activity. The quality of the data is being
controlled by the experienced judgment of the data inspector who looks back in
the data to find when the vane disappeared, and deletes or flags the data as
missing.
Another quality control function is the periodic calibration. This is an
instrument quality control process where adjustments are made as necessary to
keep the instruments "in control." If more frequent calibrations are used or
if the program goes on for several years, a standard control chart may be used
to visually track the "in control" status of the instruments. The data quality
control inspector must see the calibration reports and contribute to the
decision about what if anything to do to the data as a result of calibration
findings. Doing anything to the data requires very careful consideration and
thorough documentation.
When the organization policy is to achieve the level of quality the
"customer" expects, the whole organization effectively becomes a part of the QA
department. Various techniques, such a quality circles, may be used to
maintain a high level of quality through broad participation and training.
These techniques also underscore the management's dedication to the quality
policy. It is only when other criteria, such as departmental profit goals,
enjoy a higher priority than does quality that an independent "watch dog"
organization is required to achieve a published quality policy. The price for
quality is inversely proportional to its place on the priority scale. When
quality is first on the list, the most efficient and least expensive process
can be found for its achievement.
-------
Section No. 4.1.4
Revision No. 0
Date: 17 Sep 89
Page: 1 of 4
4.1.4 TRACEABILITY PROTOCOL
There is a general practice in QA/QC to use a hierarchy of standards
resting on international standards or those maintained by the National Bureau
of Standards (NBS). This results in claims of calibrations that are "traceable
to NBS." While it is difficult to define traceability in quantitative terms,
there is value in using an authority against which other instruments can be
compared. This section will discuss how this hierarchy relates to
meteorological measurements.
Most meteorological measurements contain some sensing element which reacts
to the variable of interest and the usual transducer outputs of voltage,
current or frequency.. In terms of accuracy, the response of the sensing
element is the most important and the most difficult to define (see sub-section
2, Specifications, for each variable). The measurement of the various
components of the electrical output, including digital code, is straight
forward and subject to normal methods of calibration and certification.
Protocols for "traceability to NBS" for voltage will be discussed first.
4.1.4.1 Voltage
Regular calibration labs maintain transfer standards which are sent to
NBS for calibration. These in turn are used to calibrate the lab's voltage
sources which in turn are used to calibrate a subject volt meter. This process
has gone on for years and is called "traceable to NBS." Calibration labs check
a volt meter, adjust it if necessary, and affix a calibration sticker
certifying the meter to be in calibration at the date tested and recommending
re-certification at a future date (six months to a year). This traditional
process is entirely acceptable.
Manufacturers of volt meters also have transfer standards which they use
to calibrate and certify their products. Modern digital volt meters or
volt-ohm meters (DVOM) are very stable in calibration, particularly those of
high quality. Another method for achieving "traceability to NBS" involves the
comparison of DVOMs. If one DVOM is certified as accurate, either by a
calibration lab or by the manufacturer with a transfer standard, and another
DVOM is placed in parallel across a voltage source, the uncertified one can be
certified by that comparison. This process is valuable to use in performance
auditing in order to fix the accuracy with which the operator calibrates the
signal conditioners. The process yields a relative accuracy if neither DVOM is
certified, but it becomes absolute as comparisons are made with certified
DVOMs.
What is an acceptable error in voltage measurement for meteorological
purposes? Assume a measurement system with a full scale output of 1 volt for
wind speed, wind direction and temperature difference. Assume the DVOMs are on
a range displaying millivolts where the 1 volt full scale looks like 1.000.
How important is it if the two DVOMs disagree by as much as 0.002 volts? If
the range for wind speed is 0.2 to 50 m/s, and the accuracy requirement is 0.2
m/s, what is that accuracy requirement expressed in volts?
50.0 m/s = 1.000 V
0.2 m/s = 0.2/50 x 1.000 = 0.004 V
-------
Section No. 4.1.4
Revision No. 0
Date: 17 Sep 89
Page: 2 of 4
The disagreement between the DVOMs is equivalent to half the accuracy
requirement. For all practical purposes a disagreement of this size is not
important, but an auditor would like more information. Is it a bias or a
random difference?
A good DVOM, like the Fluke 8060A, specifies its accuracy on the 2.0000
V range as ± (0.04% of reading + 2 digits). On the 20.000 V range the accuracy
is specified as ± (0.05% of reading + 2 digits). If the two DVOMs were on the
same output of 0.1000 V (5 m/s for the wind speed example), and if they were on
a range equivalent to the 2 V range stated above, they should each read 0.1000
±(0.00004 + 0.0002) or between 0.09976 V and 0.10024 V. Truncation of the
measurement to fit the display would cause the meters to read between 0.0997
and 0.1002 which would be interpreted as 4..98 and 5.01 m/s. If the difference
were as much as 2 mV (0.1 m/s), it would indicate a bias (calibration) error in
one or both of the DVOMs. If the DVOMs were each on a 20 V range, they should
read 0.100 ± (0.00005 + 0.002) or between 0.09795 V and 0.10205 V. Truncation
would force the meters to read between 0.097 and 0.102 which would be
interpreted as 4.85 and 5.10 m/s. A difference of 2 mV (0.1 m/s) in the meter
readings could be either a bias error or a random error from the 2-digit
uncertainty. Switching both DVOMs to the 2 V range would resolve the question.
Table 4.1.4.1 summarizes the accuracy of the conversion of the
transducer output to voltage output in units of voltage and units of
meteorology for the 1 volt range example for wind speed,direction and
temperature difference.
Table 4.1.4.1 - Voltage vs. Met. Unit Accuracy
Variable
Range
Volts
Wind speed
Direction
A T
0
0
0
.000 -
.000 -
.000 -
1
1
1
.000
.000
.000
Met. Units
0.0 -
360 -
-5.0 -
50.0
360
15.0
m/s
deg.
°C
Accuracy (0. 1% FS)
Volts
0.
0.
0.
001
001
001
Met. Units
0.05
0.36
0.02
m/s
deg.
C
4.1.4.2 Wind Speed
Traceability to NBS has some meaning in the measurement of wind speed.
The National Bureau of Standards Fluid Mechanics Section operates a pair of
calibration wind tunnels at their facility in Gaithersburg, MD (Washington,
D.C.). One can arrange to send an anemometer to NBS for calibration. A report
will result which describes the output of the sensor or system (rate of
rotation or volts) at a series of wind speeds. NBS states the accuracy of the
wind speed they use to be 0.1 mph. How the user implements the test report is
a different story (see 4.2.12). If the user is a manufacturer, the test report
will probably be smoothed by some least square method which predicts speed from
rate of rotation. The speed predicted by the rate of rotation of the
anemometer calibrated by NBS will then be transferred to another anemometer by
collocating them in another wind tunnel or by calibrating the wind tunnel as an
intermediate standard. If new anemometers agree with this transfer of the
performance of .the "standard" anemometer to within some margin of error, the
calibration of the new anemometer is said to be "traceable to NBS."
-------
Section No. 4.1.4
Revision No. 0
Date: 17 Sep 89
Page: 3 of 4
There is no standard wind. NBS uses a structure designed to smoothly
control the air being driven by a propeller-motor assembly. How smoothly and
uniformly the air flows through the test section is determined by testing.
The wind speed at some point in the test section is calculated from the
measured pressure difference between the pitot tube and the static pressure,
correcting for air density. The pressure difference is measured with a
manometer. Anyone can build a wind tunnel and measure its performance as
accurately as can NBS. "Traceable to NBS" provides a relative standard of
comparison with absolute errors which are small compared to the needs of the
scientific and industrial users.
4.1.4.3 Wind Direction
"Traceable to NBS" has no meaning as it relates to wind direction
(see 4.2.2.2).
4.1.4.4 Temperatureand Temperature Gradients
There is a hierarchy for temperature much the same as voltage.
Calibration labs and manufacturers maintain sensors with calibrations run by
NBS. A user can send an electrical transducer, which has a unique relationship
between resistance and temperature, to a calibration lab and get a report on
that relationship as determined by the lab's transfer standard. This
calibration is called "traceable to NBS" because the transfer standard was
calibrated there. Some concern about how the subject transducer and the
transfer standard are exposed to the "same" temperature is warranted. The test
method and test facilities are not usually certified by NBS and so the
calibration may not deserve the inferred NBS authority.
Differential temperature is nothing more than two or more temperature
measurements taken at different points. The important calibration for this
variable is one which compares one instrument to the other, a
relativecalibration. Traceability is not relevant to relative calibrations.
4.1.4.5 Solar Radiation
Traceability to an absolute measurement of solar radiation is achieved
by collocated comparisons with secondary standards at organizations such as the
Desert Research Laboratory in Arizona or at a scheduled World Meteorological
Organization (WHO) inter comparison. An absolute measurement of the intensity
of the direct beam from the sun is made with an Active Cavity Radiometer. This
instrument is based on fundamental principles. The cavity "sees" only the
direct radiation from the sun. The optically black surface in the cavity is
heated by radiation of all wave lengths. The cavity temperature is accurately
measured and the instrument yields the absolute flux of direct radiation (D) at
the measurement location. A global pyranometer with a disc located to shield
the direct beams from the disc of the sun measures the diffuse radiation (d).
With knowledge of the angle of the sun from the zenith (9), the total global
radiation (G) can be calculated by the following formula.
G » D cos 0 + d
Secondary standards, traceable to such an inter comparison, may be used t
calibrate operational pyranometers. Pyranometers which have been calibrated in
this way may be used as collocated transfer standards in the field.
-------
Section No. 4.1.4
Revision No. 0
Date: 17 Sep 89
Page: 4 of 4
4.1.4.6 Atmospheric Water Vapor
It is possible to create an absolute humidity and NBS has a facility for
doing just that. Atmospheric water vapor instruments, expressing the
conditions as relative humidity or dew point temperature can be calibrated by
NBS in a fundamental procedure requiring only the measurement of length
(volume), mass and temperature. There are standard methods for creating
relative humidity environments useful for some kinds of instruments. ASTM
(1985c) describes such a method. For air quality applications, a collocated
comparison provides adequate accuracy. If the collocated instrument is a
psychrometer and proper methods are used (ASTM, 1984) for measuring the wet-
and dry-bulb temperatures, traceability to NBS might be claimed for the
thermometers used in the psychrometer.
4.1.4.7 Precipitation
There are some measurements where traceability to NBS is possible but
not required. Precipitation measurement is essentially a measure of the volume
of liquid water (including the liquid water equivalent volume of snow) which is
collected by an area bounded by a cylinder. Calibration may require volumes of
water or equivalent weights. The accuracy required and expected from
precipitation gages will be well served by the accuracy of commercial measuring
equipment. Ordinary chemical dispensers such as graduated cylinders and burets
are accurate enough without calibration traceable to NBS. The measurement of
the area of the cylinder may be made with a commercial ruler or tape, always
keeping in mind the need for quality commercial products for the best accuracy
without extraordinary effort.
4.1.4.8 Atmospheric Pressure
Calibration labs create pressures with devices using weights. It is
possible to use weights with calibration traceable to NBS, but the accuracy
with which the atmospheric pressure is needed for air quality applications does
not require such an effort.
-------
Section No. 4.1.5
Revision No. 0
Date: 17 Sep 89
Page: 1 of 6
4.1.5 ESTIMATING PRECISION AND ACCURACY
4.1.5.1 Definitions
There are about as many definitions of precision and accuracy as there
are bodies devoted to carefully defining these terms. The definition used here
is found in EPA (1976) on page A17 as follows:
"Accuracy - The degree of agreement of a measurement (or an average of
measurements of the same thing), X, with an accepted reference or true value,
T, usually expressed as the difference between the two values, X - T, or the
difference as a percentage of the reference or true value, 100(X-T)/T, and
sometimes expressed as the ratio, X/T. "
For meteorological purposes, this concept of accuracy is acceptable.
The problem comes from knowing the "accepted reference or true value." Section
4.1.4 Traceability Protocol discusses this problem for all the meteorological
variables of interest. All data that are
used are averages or means. The formula for accuracy is
1 n
E = - V(X. - T ) - X - T (1)
n £_ i
where
E is the average error (accuracy)
Xi is the ith sample of X
T is the non-varying true value of X
n is the number of samples
i is a sample, l,2,3...n
Accuracy, the average error, or really the uncertainty in the value, X. ,
has two or three components. They are bias, conditional bias, and random
error, a statistical expression of a series of which is called precision.
Since, in some cases, bias and conditional bias can be separated, both will be
discussed.
Precision is defined in EPA (1976) as "A measure of mutual agreement
among individual measurements of the same property, usually under prescribed
similar conditions. Precision is most desirably expressed in terms of the
standard deviation but can be expressed in terms of variance, range, or other
statistic. Various measures of precision exist depending upon the 'prescribed
similar conditions.'" This is more difficult to fit to meteorological
measurements made in the atmosphere because "prescribed similar conditions" are
hard to find. Parts of the instrument system can be challenged by controlled
environments such as wind tunnels and temperature, humidity or pressure
chambers. The precision of the measurement can be found, providing it is
larger than the variability of the controlled environment. Usually it is not
-------
Section No. 4.1.5
Revision No. 0
Date: 17 Sep 89
Page: 2 of 6
larger and what is measured is the variability of the controlled environment.
The definition to be used in this handbook is as follows:
Precision is the standard deviation of a series or measured values,
X., about the mean measured value, X. The for
of the precision or standard deviation, s, is
X., about the mean measured value, X. The formula for an estimation
2
- X)
(2;
n - 1
An equivalent formula which can be used for real-time calculations in modern
data loggers or computers (Juran, 1979, p 22-7) is
(3)
n (n - 1)
Equation (3) is preferred because it introduces only two rounding errors rather
than the four introduced in equation (2). Suppose
X'. = X. + C where C is some constant value. Then
Xi = X^ - C and X = X' - C and X - X = X^ - X' .
Therefore, the standard deviation, s, of a series of values, as calculated by
(2) or (3) is identical to the standard deviation, s, of the same series of
numbers, each plus a constant. If the true value, T, is constant, the
precision of the accuracy estimate, E, is identical to the standard deviation
of the samples, X. .
Bias is defined in EPA (1976) as "A systematic (consistent) error in the
test results." It goes on with qualifiers which apply to chemical labs. For
meteorology, the definition for this handbook is simply the average difference
between the measured value X. and the true value T , (X. - T ). Using the
symbols of ASTM (1984), the.systematic difference (d) or bias is found from the
following equation.
-------
Section No. 4.1.5
Revision No. 0
Date: 17 Sep 89
Page: 3 of 6
, (X. - T. ) = X - T (4)
n
where
T. is the ith sample of the true value of T
It can be seen in these definitions that when the true value, T, does not
change, the accuracy is the same as the bias. What is the precision of formula
(D? For the non-varying true value, the accuracy with which any sample may be
determined is
EI « E ± s (s:
and for a variable T
E = d ± s (6)
When T varies, X may vary in a systematic way. For example, the case for wind
direction found in 4.2.2.2.2.3 shows a systematic difference, d, of -3.4 deg.
(orientation error) and a conditional bias (potentiometer linearity error) of
about ±0.5 deg. on some wave shape with an amplitude of about ± 2 deg. The
accuracy of the vane might be stated as
E. = -3.4 ±2.5 deg. or, removing the
orientation error, E, = 0.0 ± 2.5 deg. or, correcting for linearity
at each angle, E. = 0.0 ± 0.5 deg. There is little point in
making this final correction in operational programs. This is just an example
of how a conditional bias can be treated to decrease an error or improve the
accuracy.
4.1.5.2 Collocated Transfer Standards
If a collocated transfer standard (CTS) is used to estimate accuracy,
any correction which makes the CTS provide values closer to the true value is
worth making. The statistical method of operational comparability (ASTM, 1984)
is used when a CTS is employed to estimate the accuracy of an instrument
operating in the field. The analysis requires differences between the
instrument being challenged and the CTS, properly sited. The differences are
best if the same data logger can sample the challenged sensor, X , at the same
time as the CTS, X, , so that the difference, (X .-X,,), can be stored, squared
and accumulated. If this is not possible, the CTS data logger should form
averages over the same time period as the averages formed in the challenged
system. Then these averages are used for X and X, . The first calculation
to make is comparability from the following formula:
-------
Section No. 4.1.5
Revision No. 0
Date: 17 Sep 89
Page: 4 of 6
(7)
where
X is the ith measurement of the subject output
SL1
X is the ith simultaneous sample from the CIS
The systematic difference, d, is calculated from (4) substituting
X . and X^.for X. and T., respectively. The estimated standard deviation of
ai bi i i
the difference, s, is calculated from (8).
C2 - d2 (8)
The minimum sample size, n , required for the calculation of C is given by
equation (9). Most data loggers sample sequentially. The time between members
of a data pair to satisfy the requirement of simultaneous measurements must not
exceed one tenth the response time of the instruments. The time between pairs
of measurements must be greater than four times the response time of the
instruments to assure sample independence.
(9)
where
r is one increment of resolution reported by X
For example, a CTS wind vane operating in a speed range of 2 to 7 m/s
with a delay distance of 2 m would have a response time between 1 and 0.3 s.
If a data logger had an analog to digital conversion cycling
time faster than
0.03 s and if samples were taken no faster than every 4 s, and if the
resolution (measurement and display) of the measuring system were 1 deg., and
assuming a 5 deg. standard deviation of the difference, s, the minimum sample
size would be
n
o
3 x 5)2
= 225 samples, requiring 15 minutes @ 1 per 4 s.
There cannot be too many samples. The minimum is specified for a confidence of
99.7% or greater that the estimated mean difference, d, is within one element
of resolution (1 deg. in the example). At that confidence level, the accuracy
of the estimate 'increases (error decreases) as the square root of the sample
size. If the sample size were increased by a factor of 4 to 900, the accuracy
of the estimate would by 0.5 deg. The values of C and d found from a series of
-------
Section No. 4.1.5
Revision No. 0
Date: 17 Sep 89
Page: 5 of 6
differences are only valid for the range of conditions used for the test. The
shorter the period of time that is sampled the smaller the range of conditions
will be. A reasonable goal for a minimum CIS time period is 24 hours from the
standpoint of dynamic range variation.
Assume the CTS will provide the true value in a test in the free
atmosphere within the limits of the calibration of the CTS to some other
standard, typically 0.1 m/s with respect to some wind tunnel for speed and 1
deg. with respect to TRUE NORTH for average direction. The accuracy of the
challenged instrument is the comparability, C. The bias, d, provides the
calibration and orientation error. The standard deviation, s, provides the
irreducible random error or minimum functional precision with which two
instruments measuring the same quantity report when operated using the ASTM
D4430-84, determines the operational comparability of meteorological
measurements. Lockhart (1989) found the following values of "s for wind speed
and wind direction:
Wind speed s a 0.2 m/s
Wind direction s a 2 deg.
When s is found to be larger than these values, the assumption must be made
that either site bias or a malfunctioning sensor is at fault. Under those
conditions the calibration error, d, is also suspect.
4.1.5.3 Other Considerations
The average error, E, calculated over a uniform distribution of X, is
the same as the average difference or bias. The contribution of the standard
deviation of the difference between X and T (or the precision error) goes to
zero and the average accuracy is the bias. The measurement of meteorological
variables in the atmosphere is never really that simple.
Occasionally there is confusion between the word precision and the
resolution of the measurement system. Resolution is the fineness of the
measurement system, the output of the measurement system or the display of the
output of the measurement system. A wire-wound potentiometer in a wind
direction sensor may have a resolution of 0.355 degrees (1,000 windings of a
wire over 355 degrees). The circuit that converts the resistance of the
potentiometer to voltage may have a resolution of 0.1 degree. The display or
recorded value may show whole degrees. The whole degree may be truncated from
the output or rounded. For example, if the potentiometer wiper is at the 312th
wire (312 x 0.355 = 110.76 degrees), the voltage output (312/1000 x 1 volt =
0.3120 volts) has no resolution;, the resolution of its measurement is limited
only by the resolution of the volt meter. If the system has a digital output
with a resolution of one degree,, the output will be 110 degrees (truncated) or
111 degrees (rounded). The resolution of the sensor in this example is 0.36
degrees and the resolution of the system is 1 degree.
A wind direction system with a resolution of 10 degrees might have a
precision of ± 0.5 deg. (it would take a lot of samples to prove that
precision). On the other hand, a wind direction system with a resolution of
0.1 deg. might b,ave a precision of ± 3 deg. (because of hysteresis in the
coupling of the potentiometer to the vane in the sensor). Resolution should be
-------
Section No. 4.1.5
Revision No. 0
Date: 17 Sep 89
Page: 6 of 6
specified to match the needs of the data application and to provide sufficient
information for data QC.
Most of the discussion in this handbook includes the sensor and signal
conditioner providing an output. Good digital data systems degrade the output
so little that they contribute only a small error to the total. The resolution
and accuracy example 4.1.4.1 applies to any digital system. There is a two
digit uncertainty in the digit which represents the resolution of the
measurement. If the analog to digital converter resolves wind direction to 10
deg., the accuracy cannot be better than ± 10 deg. If the converter resolves
to 1 deg., rounds to the nearest 10 deg. and displays 10 deg., the accuracy of
the average direction may be 3 deg. while the accuracy of a single observation
is ± 5 deg.
When analog recorders are used, their error must be added to the error
of the measurement. For some recorders this error can get quite large. Seldom
considered or specified, for analog recorders which use rolls of chart paper,
is the error caused by expansion and contraction of the paper as a function of
temperature and humidity. This error added to the resolution uncertainty of
narrow paper rolls marked by a tapping bar (when such recorders are used) can
dominate the error of the system when the analog recorder data are used as
measurement data.
Random errors identified for each component of a system can be combined
to estimate the total system error by the RSS (root sura square) method. Biases
or systematic errors cannot be combined in this way. They must be added
arithmetically. See Fritschen and Gay (1979) for further discussion of error
analysis.
-------
Section No. 4.1.6
Revision No. 0
Date: 17 Sep 89
Page: 1 of 2
4.1.6 SYSTEM AUDITS
A system audit, as defined in EPA (1976, p A10) is "A systematic on-site
qualitative review of facilities, equipment, training, procedures,
record-keeping, vaiidation, and reporting aspects of a total (quality
assurance) system, to arrive at a measure of the capability and ability of the
system. Even though each element of the system audit is qualitative in nature,
the evaluation of each element and the total may be quantified and scored on
some subjective basis." In short, it is an evaluation of the suitability and
effectiveness of a QA Plan or QA Manual.
Any audit is most useful when considered as a learning or training
exercise. Given the newness of the implementation of quality systems to air
quality programs, particularly meteorology, a mechanism for "on the job
training" is useful. Given the two facts that everyone really wants to do a
good job and almost everyone is a stranger to the concepts of structured
quality systems, an audit is a valuable tool. There is really little
difference between air quality and meteorology when it comes to system audit
principles. The short section on system audits in EPA (1976) is slanted toward
air chemistry projects. In a very general way, a system audit should include
the following elements for any technical discipline.
1. Declared Agenda - The audit should not be a surprise or contain
surprises. The serious audit is well planned in advance in writing. The items
to be covered are spelled out. The agenda is structured with the help of those
to be audited, recognizing the areas where they may need special help.
2. Entry and Exit Interview with Top Management - A short introduction
meeting with the authority being audited sets the stage for the cooperation
necessary for success. Success is defined as improving the audited program
through training and education. A short exit interview will announce the
findings already discussed with the QA people being audited. The exit
interview and the audit report should contain no surprises.
3. Checklist Structure - The audit should flow along a prepared
checklist of questions, but if time is limited as it usually is, flexibility is
valuable. Special problems, either found or volunteered should be resolved
even at the expense of failure to finish the checklist.
4. Audit Report - The report should'be delivered in a timely manner,
certainly no more than 30 days after the audit, preferably within a week of the
audit. The report is the important documentation verifying the QA program is
"in control." It must contain the structure for corrective action with plans
and schedules committed in writing. An open-loop pledge or a general plan is
likely to get a low priority. The value of the audit and corrective action
must be clear to the audited organization if the system audit is to be
something other than a paperwork exercise.
5. Follow-up - The QA Plan of the organization being audited should
require some form of documentation of the completion of the tasks defined in
the corrective action plan. Completion of this task closes the loop for the
system audit.
-------
Section No. 4.1.6
Revision No. 0
Date: 17 Sep 89
Page: 2 of 2
A short visit to understand the organization, QA Plans and Procedures, and
to meet the QA person is a good first step. It is difficult to do an effective
system audit without access to all three.
Since there is a large spectrum of quality programs which might be audited,
it is important to keep in mind the original reason for the audit and the
objectives and regulations defining the quality program. One objective must be
to provide the necessary level of quality at the minimum cost as well as to
make the audit process useful and effective. The goal is to ensure valid data
with documentation backing up the claim for validity. This can happen in many
different ways, some of which may not conform to the auditor's concept of a
system. It is not the role of the auditor to redesign the system, but to
determine if the-system meets the objectives and requirements.
-------
Section No. 4.1.7
Revision No. 0
Date: 11 Aug 89
Page: 1 of 1
4.1.7 PERFORMANCE AUDITS
This handbook will concentrate on performance audits. The audit methods
for each variable will be described in each of the variable-specific sections.
The purpose of a performance audit is to determine as completely as possible
whether or not the instruments are producing valid data. It is the
responsibility of the operators to calibrate and operate the instruments so that
they do produce valid data. As mentioned in Section 4.0.0, performance audit
methods may be identical to those used in calibration. If they are different,
it is expected that the audit method is most comprehensive challenging the
greatest part of the total system. A complete calibration of an anemometer
requires a wind tunnel. Most operators do not have access to a wind tunnel and
elect to use the manufacturer's wind tunnel experience as authority for the
anemometer transfer function. This practice is generally acceptable where
manufacturers can provide test reports confirming their results. Statistics on
the distribution of error of samples of production run anemometers with respect
to the generic transfer function are necessary if the manufacturer does not
calibrate each unit during manufacture. If 100% calibration is a part of
manufacturing process, the method of calibration should be available from the
manufacturer.
A performance audit may include a challenge by a collocated transfer
standard. Such data serve to check the transfer function at a few points. The
uncertainty of the challenge is usually greater than wind tunnel challenges
where the conditions can be carefully controlled.
A performance audit on variables such as relative humidity (or dew point
temperature), solar radiation and atmospheric pressure will usually include a
collocated transfer standard comparison. This is may be the only way a
challenge can be made to the whole measurement system.
-------
Section No. 4.1.8
Revision No. 0
Date: 17 Sep 89-
Page: 1 of 7
4.1.8 DATA VALIDATION PROTOCOL
4.1.8.1 Strip Charts
Some years ago, meteorological data were recorded primarily on strip
charts. Average values were estimated by "reading" the charts. The most
common and the most useful strip chart was one where the output of the
measuring system was continuously recorded by means of a galvanometer movement
or a servomotor. Data QC meteorologists became expert in determining the
validity of the data by examining the strip chart trace, a process necessary
while checking the digitizing of the strip chart data.
When the digital data logger and computer first appeared, it was the
recommended practice to have strip chart recorders in parallel to the digital
systems. The strip chart data could be "read" and used to fill the gaps when
the digital system failed, a common occurance in early designs. Many
meteorologists found that the strip chart data contained information which was
not present in the digital listing. The most important information was the
character of the output during the period of time that the digital system was
sampling and averaging. One example is the presence and frequency of
potentiometer noise in the wind direction output. This information is an early
indicator of potentiometer failure. Whether or not the digital average was
influenced by this noise could be seen by the comparison of the two outputs,
the strip chart being used as truth.
Another example shows threshold degradation by the character of the
anemometer trace. Of course the effect of ice or freezing temperatures on
anemometers and wind vanes could often be seen on the strip chart. The digital
average value would simply be a number which met the plausibility test but was
erroneous.
Digital systems have become more reliable, accurate and capable of
on-line processing and at the same time less and less expensive. The economy
of the digital system pushed the analog recorder to a "back up" role and toward
extinction. Digital systems show promise of both large memory sufficient to
save one second data samples for time history plots and on-line diagnostic
programs to monitor output patterns for unrealistic variability or lack
thereof.
Until the technology stabalizes enough to allow a consensus to be
developed and regulatory positions to be taken, requirements for strip chart
data will be a reagon or agency specific requirement. The one clear fact is
that strip chart data, used as a data quality control tool, will result in a
better data validity protocol.
4.1.8.2 Methods
Once data are collected, they should be reviewed to screen out possible
incorrect data points before they are put into accessible storage or passed on
to the user. While the purpose of a QA program is to avoid the generation of
bad data, it is impossible to do so completely. Even in the best planned and
best conducted programs, undetected errors can be generated by faulty
equipment, noisy data transmission lines, faulty key punching, and a myriad of
-------
Section No. 4.1.8
Revision No. 0
Date: 17 Sep 89
Page: 2 of 7
other sources. Filippov (1968) offers a detailed and thorough discussion of
the various possible sources of error.
In both automatic (ADP) and manual data screening the most obvious
checks should be performed first. These include such things as being sure that
the data exist and are properly identified, the forms or files are filled out
properly, that numbers are in the blocks where they should be, letters are
where they should be, and blanks exist where nothing should be. This sort of
data editing is a subject unto itself and will not be pursued here.
Methods of editing or screening meteorological data usually involve
comparison of the measured value with some expected value or range of values.
Techniques for checking the measured value usually fall into one or more of the
following categories:
1. Comparison with upper and/or lower limit on the allowed
range of the data;
2. Comparison with known statistical distribution of
the data;
3. Comparison with spatial and/or temporal data fields; and
4. Comparison based upon known physical relationships.
A choice must also be made of what to do with the datum that does not
pass a validation procedure. Basically there are two choices, eliminate the
questionable data from the file, or flag it for further examination.
Automatically discarding data may be a viable, cost-effective option if the
screening procedure is carefully designed and each datum is not of high value.
Records must be kept of discarded data so the reason for the fault can be found
and corrected. Flagged data are examined and a decision made on their
acceptability. If unacceptable, it may be possible to correct them or
substitute a more reasonable value (Reynolds, 1979). Corrected or substituted
values should be so indicated in the data file, with an explanation of the
substitution available to the user. Alternatively, data of questionable value
may be kept in the data file under a flagged status, with a notation of why
they are questionable, so that the user can make a decision as to their
usefulness. This procedure is of questionable value to most users because the
collecting agency is frequently in the best position to make a decision on the
data.
The range test is the most common and simplest test. Data are checked
to see if they fall within specified limits. The limits are set ahead of time
based usually upon historical data or physically impossible values. Some
examples of reasonable range tests are rainfall rate greater than 10 in./hour
or wind direction not between 1 and 360 . In setting the limits, one must
take into consideration whether or not the system will select only outrageous,
extreme (i.e., impossible) values usually caused by data handling errors (such
as wind speeds greater than 100 m/s or less than zero) or just unusually high
(i.e., possible ) values, which should be examined further. This may require a
further decision on just how extreme a value should be flagged. This decision
should be based on the real impact of using extreme values should they be in
error. Considerations of the cost of incorrect data, the possibility of
correction or substitution, or replacement by obtaining new data should be
made. Unfortunately, the decision may also frequently be made on the available
resources of those who examine the flagged data.
-------
Section No. 4.1.8
Revision No. 0
Date: 17 Sep 89
Page: 3 of 7
4.1.8.2.1 Comparison with known distributions
Comparison with known statistical distributions may involve
comparison of means, standard deviations, means of extremes, or higher order
statistics. For example, Lee and Stokes (1978) report that their data base
usually had kurtosis of approximately 3 with zero skewness. Any of their
instruments, then, that showed a marked departure from these values were
considered to be in need of further verification. (Additional research is
needed to determine whether these or similar criteria could be used in other
areas..)
Lockhart (1979) suggests compressing data into a densely packed
graph where long-term (week, month, or seasonal) patterns can be easily seen.
Major departures from these subjectively seen patterns can be noted and the
data checked. Although this method of data verification is usually used to
check a particular data set against a longer term climatology, it can also be
used to check individual values. For example, one might compare a temperature
reading with the monthly average maximum or minimum plus or minus
(respectively) two or three standard deviations. This technique obviously
depends on a reliable history or representative measurements being available
from the site and is ineffective for noting significant long-term changes in
the instrument.
4.1.8.2.2 Comparison with other data fields
Screening data by comparison with fields of similar or related
data is commonly done when large amounts of data are taken and when assumptions
of spatial continuity of the meteorological variable are physically reasonable.
The most easily visualized example of this is a field of atmospheric pressure
measurements. Any value can be compared with those in a large area around it,
either visually, or by numerical interpolation. Major deviations from the
dominant pattern (a low pressure reading in the middle of a high pressure area)
are not to be expected. Of course, allowance must be made for meso- and micro-
scale phenomena such as a shortwave or pressure jump area ahead of a convective
storm.
Not all meteorological fields can be expected to have the
needed continuity. Rainfall is a notorious example of discontinuity or
microscale variations. Wind speed and direction can exhibit continuity on some
spatial scales, but care must be taken to account for the many effects, such as
topography, that can confuse the issue (See Section 4.0.4.3.2.4).
Interrelated fields can also be used to screen data. Rainfall,
for example, is unusual without clouds and high humidity while wind direction
and speed, especially above the surface layer, are related to pressure
gradients.
Fields of data in time, rather than space, are also used to
check datum points. These checks are usually made on rates of change of the
data. Checks are made both on rates of change that are too high and not high
enough. For example, atmospheric stability is not expected to change by
several classes within an hour. A wind direction reading, however, that does
not change at all for several hours may indicate that the vane is stuck
-------
Section No'. 4.1.8
Revision No. 0
Date: 17 Sep 89
Page: 4 of 7
(assuming the wind speed is not zero]
the system.
or that there is some other problem with
4.1.8.2.3 Comparison based on physical relationships
Screening checks can also be made to assure that physically
improbable situations are not reported in the data. This kind of check is not
commonly used because of the wide variety of conditions that can occur in the
atmosphere under extreme conditions. These unusual events would frequently be
noted first by some of the statistical or range checks noted above.
Table 4.1.8.1 Examples of Data Editing Criteria
Wind Speed:
Wind Direction:
Delta Temperature:
Stability:
Temperature:
Dew Point:
Pressure:
>25 m/s (NRC)
>50 kts (NCC)
>20 kts and doubles at 3-hour observation (NCC)
First 5 hourly values <±0.2 mph of next 4 (TVA)
Any recorded calm wind speed (NCC)
Same sector for more than 18 hours (NRC)
.First 5 hourly values <± 2° of the next 4 (TVA)
AT/Az >l°C/100m between 10 a.m. and 5 p.m. (TVA)
AT/Az <-l°C/100m between 6 p.m. and 5 a.m. (TVA)
AT/Az >15°C/100m (TVA)
AT/Az <-3.4°C/100m (autoconvective)(TVA)(NRC)
AT/Az changes sign twice in 3 hours (TVA)
A.B.F, or G stability during precip. (NRC)
F or G stability during the day (NRC)
A,B, or C stability during the night (NRC)
Change in stability of more than 3 classes
between 2 consecutive hours (NRC)
Same stability class for >12 hours (NRC)
9°F > mean daily maximum for the month (TVA)
9°F < mean daily minimum for the month (TVA)
> 10°F change in 1 hour at a site (TVA)
First 5 hours within ±0.5°F of next 4 (TVA)
>125°F (NCC)
<-60°F (NCC)
> 10°F change 1 hour or 20°F in 3 hours (NCC)
Dew point > temperature (TVA)(NRC)
Dew point change >7°F in 1 hour (TVA)
First 5 hours within ±0.5°F of next 4 (TVA)
>90°F (NCC)
<-60°F (NCC)
Temp. - dew point >5°F during precip. (NRC)
Temp. = dew point >12 consecutive hours (NRC)
>1060 mb (sea level) (NCC)
<940 mb (sea level) (NCC)
Change of 6 mb or 0.2 inch Hg in 3 hours (NCC)
-------
Section No. 4.1.8
Revision No. 0
Date: 17 Sep 89
Page: 5 of 7
Some screening points of this type that are used include
assuring that the dew point is not greater than the temperature, and that the
lapse rate is not greater than the autoconvective lapse rate. Checks on
stability class versus time (not allowing "strongly unstable" at night or
"stable" during the day) may also be considered in this category.
Table 4.1.8.1 gives examples of some of the data editing
criteria used by three Federal agencies: the National Climatic Center (NCC,
now NCDC), Klint (1979); the Nuclear Regulatory Commission (NRC), Fairobent
(1979); and the Tennessee Valley Authority (TVA), Reynolds (1978). Examination
of the table shows some interesting differences that can be ascribed to the
differing missions of the agencies. Because of their global concerns, the NCC
must allow a far wider range of limits on fields such as temperature and
humidity than does an agency with only local interest, such as TVA. On the
other hand, the NCC has the data available to do spatial checks over a wider
area than would be possible for many local study situations. Differences can
also be noted depending on the type of data collection (spot readings once per
hour or three hours versus continuous recordings) and major interests (synoptic
weather patterns versus stability). Filippov (1968) gives an exhaustive review
of checks used by weather services of many other countries. The criteria
listed in the table are used to identify data to be edited or challenged for
further review.
4.1.8.3 The AREAL System
On the following page is a data validation system recommended for AREAL
to replace the present system for screening meteorological data. It could be
used to screen data gathered by AREAL, contractors, or state and local
agencies. The system takes into account the variable nature of AREAL's field
activities. It does not depend on, or have the advantages of, long-term
multistation network design, nor is it labor intensive. The basic goal is that
of rapid identification of field problems, with low value assigned to
individual data points, thus allowing the discard of questionable values.
Flexibility is available, however, if an individual project's meteorological
data are judged to warrant a more critical approach.
The flow of the system is shown in Figure 4.1.8.1. All data will go
first through a hard copy auditing procedure designed to find data entry and
keypunch errors. In the hard copy audit, a percentage of data points will be
randomly selected for audit. A second, independent file of these values, as
well as the hour just before and after the hour, will be created from the
original hard copy. This file will be compared with the master file and
discrepancies noted. If there are only a few random discrepancies, these
points will be eliminated from the system. If there are several, or there
seems to be a systematic pattern of errors, the project office (the office
responsible for gathering and reducing the data) will be notified so that they
can correct and re-enter the data and correct the data entry system. The data
are next passed through a screening program, which is designed to note and flag
questionable values. Flagged data will go to the laboratory meteorological
office for review. There they will either be accepted, discarded, or returned
to the project officer if there is a large amount of questionable data. That
officer may accept, discard, or correct the data. The screening values are
given in Table 4.1.8.1. They offer a combination of range, rate of change, and
physical impossibility checks that are chosen to be reasonably restrictive. It
-------
Section No. 4.1.8
Revision No. 0
Date: 17 Sep 89
Page: 6 of 7
is anticipated that some good data will be flagged, but that most data handling
and gross instrument failure problems will be caught.
Stan
f/Vew Dai
/^-^
(Data
I Base
V
1
\^
ft T"
—^ Edit/Merge
Program
} I
(Unvalidateo
File
" S
Compansion of Q
Selected Data —
vith Data Obtained
ndependently from
Original Source
„
1
Comparison of
Subset of Data
with N.W.S. Data
I
Edit
1
/ Validated 1
Data File
_y
Correct Data
Acquii
Proce
CProce
)
jspeci/Ars Ba
ata / Indica
\ fleco
\ Prob
'ition/
ssing
dures
\Yes
\ Project Offijce
dData\No — — — •
tive of \__ Determine if Suspect
rring / Data are to be:
lem? / Approved.
/ Corrected.
/ or Deleted.
/ i 1
Flayed Meteorologica,
°ata Office
[_^ Determine if Data Unresolved
Approved. Qata
| Corrected.
\ or Deleted. t
Questionable
Data
Data Corrections.
Approvals, or
Deletions.
Figure 4.1.8.1
Schematic flow of decisions in the
AREA! data validation scheme.
Data that pass the screening program will go through a comparison
program. This program will randomly select certain values for manual
comparison with information collected by the National Weather Service. In the
selection process, one day and one hour will be chosen on which data from all
stations in the network will be audited. One day in every 20 will be randomly
chosen, and on that day, one hour between 5 and 9 a.m. (EST), will be randomly
chosen. Data from that hour and day for all stations will be printed out by
the audit program for the manual checks. The program will also make compressed
time scale plot (20 days of hourly values on one line) for each parameter for
the use of the validators.
The data "generated by the audit programs will first be compared with
National Weather Service data to see if they fit reasonably well with synoptic
-------
Section No. 4.1.8
Revision No. 0
Date: 17 Sep 89
Page: 7 of 7
conditions prevalent in that area. The meteorologist will choose the stations
to be used in the verification, and train the data clerks in the subjective
comparison procedure. All questionable data will be given to the meteorologist
for review as above. The variables to be checked in this way will include wind
speed and direction, temperature, dew point, pressure, and occurrence of
precipitation.
Naturally, if the audit checks show a problem with one or more
instruments, an attempt will be made to identify the time range of that problem
so that all questionable data can be found. Logs of bad data will also be kept
and used to identify troublesome instruments and other problems.
This system is suggested principally for AREAL, but may prove a useful
starting place for state and local air pollution agencies wishing to develop a
meteorological data validation procedure. The suggested system is very
complete and will be evaluated over a period of time. Changes to the system
may have to be made, depending upon the needs and resources of the users.
-------
Section No. 4.1.9
Revision No. 0
Date: 17 Sep 89
Page: 1 of 2
4.1.9 QA REPORTS AND CORRECTIVE ACTION
4.1.9.1 Operations Log and Maintenance Reports
In all of section 4.1 it has been stressed that the real purpose for a
QA or Quality program is to document data validity and the steps taken to make
that determination. Any activity which has the potential of affecting the
validity of the data must be reported. A report usually includes a note in the
station log indicating the time of the operator visit and visual status of the
sensors. The log is signed by the operator. If the operator found a problem
which he was authorized to fix, the log would contain the entry to that effect.
If the operator is not authorized to make the repair or does not have the
necessary parts, a maintenance report can initiate the work and the purchase
of the parts to fix the instrument. When parts are changed, serial numbers or
property numbers should be noted in the log. If a part does not have a number
(some cup assemblies and propellers do not have numbers), a number of some sort
should be assigned and marked on the part with permanent ink. The QA Plan
should provide some communication route and method by which the person
responsible for the project and the person responsible for data quality control
(if they are different people) are notified of work done on the system.
4.1.9.2 Calibration Reports
Calibration reports are the most crucial documents of a data collection
project. They are the foundation blocks which uphold the the validity claim.
Quality Control and the routine inspection of the data spans the time between
calibrations. The calibration reports will show whether or not the system is
"in control." If the system is always "in control" or operating within the
required tolerance limits stated in the QA Plan and generated by the
application, and the data QC does not have any unsolved mysteries, the data are
valid. If the calibration shows problems, the report will also show the
corrective action taken or initiated. The "as-found" and "as-left" readings
are a vital part of the calibration report. If any data "correction"
(quotation marks used because this is a very delicate subject) is applied, the
justification must rest on calibration reports on either side of the data
period and the data in between. This report requires distribution to the
project leader and the QC inspector, or at least a sign-off routing. If
corrective action is initiated but not completed, a report of completion is
required and has the same routing.
4.1.9.3 Audit Reports
Audit reports should confirm the calibration reports. If they do
not, the assumption is that the audit report is correct. Whenever a
measurement discrepancy exists, the cause of the difference must be found and
resolved. If the audit measurements are wrong, the auditor will be smarter
next time and all parties will have more confidence in the calibration reports.
It is the responsibility of the auditor to include a report of the discrepancy
between calibration data and audit data along with the explanation and solution
of the discrepancy in the audit report. It is the responsibility of the
operator to be sure that it is in the report. The documentation must be
suitable for use in a court of law.
-------
Section No. 4.1.9
Revision No. 0
Date: 17 Sep 89
Page: 2 of 2
4.1.9.4 Reports to Management
Reports to management are of value to maintain the close communication
necessary between the source of authority, top management, and the exerciser
delegated authority, the QA organization. Whatever the structure of the
organization, directed effort must be paid for and planned for through
budgeting. Top management must know how the quality program is performing its
intended, money saving role. There are about as many types of meteorological
monitoring programs as there are applications. One fairly standard one is the
2-level, 60 m tower used at most nuclear power plants. Crutcher (1984)
provides an insight into costs of a minimum system and an acknowledgement of
the annual costs involved in operations and quality assurance.
"Costs are controlled by the design and reliability of the system, as well as
the marketplace. Costs given here are approximate 1977 prices for presently
available equipment sufficient to meet the minimum requirements of the Nuclear
Regulatory Commission's Regulatory Guide 1.23 (formerly the USAEC Safety Guide
23). For the first two years these costs approximate one-third of a million
dollars. These minimum costs do not include either office or storage space.
One tower, installation and equipment $100,000
Annual maintenance cost 25,000
Annual cost of surveillance and quality assurance
(including personnel and supplies,
magnetic tape, paper etc.) 50,000
Annual cost of data listings, etc. based on 15-min.
integrating intervals and automatic logging in
digital form on magnetic tape, 13 parameters
(channels) to a page, daily summaries 60,000
The cost of a mobile tower and equipment installation is essentially 50% of the
cost of a permanent-type tower installation. Other costs remain essentially
the same."
Of particular interest in this reasonably accurate estimate (ten years ago)
is the ratio of annual costs to one-time costs, 1.35:1. Of course, the
requirement of RG 1.23 is for valid data with documentation and quality
assurance. Smaller systems cost less, but the often neglected provision for
annual operating and QA costs are still necessary if valid data are required.
4.1.9.5 Discrepancy Reports
Some systems report discrepancies as a section of another report and
some use a discrepancy report as a stand-alone vehicle to initiate corrections
and report completed corrections. If it is a stand-alone report, some system
of control is necessary to keep track of open reports and monitor progress
toward completion (called follow-up or needling).
-------
Section No. 4.1.10
Revision No. 0
Date: 17 Sep 89
Page: 1 of 2
4.1.10. REFERENCES
ANSI/ASQC, 1987a: Q90, Quality Management and Quality Assurance Standards -
Guidelines for Selection and Use. Araer. Soc. For Quality Control,
Milwaukee, WI 53203
ANSI/ASQC, 1987b: Q91, Quality Systems - Model for Quality Assurance in
Design/Development, Production, Installation, and Servicing. Amer. Soc. For
Quality Control, Milwaukee, WI 53203
ANSI/ASQC, 1987c: Q92, Quality Systems - Model for Quality Assurance in
Production and Installation. Amer. Soc. For Quality Control, Milwaukee, WI
53203.
ANSI/ASQC, 1987d: Q93, Quality Systems - Model for Quality Assurance in Final
Inspection and Test. Amer. Soc. For Quality Control, Milwaukee, WI 53203,
ANSI/ASQC, 1987e: Q94, Quality Management and Quality System Elements
Guidelines. Amer. Soc. For Quality Control, Milwaukee, WI 53203.
ASTM, 1984a: Standard Test Method for MEASURING HUMIDITY WITH A
PSYCHROMETER (THE MEASUREMENT OF WET- AND DRY-BULB TEMPERATURES,
E. 337-84. Amer. Soc. for Testing and Materials, Philadelphia, PA 19103.
ASTM, 1984b: Standard Practice for DETERMINING THE OPERATIONAL COMPARABILITY
OF METEOROLOGICAL MEASUREMENTS, D4430-84. Amer. Soc. for Testing and
Materials, Philadelphia, PA 19103.
ASTM, 1985c: Standard Practice for MAINTAINING CONSTANT RELATIVE HUMIDITY BY
MEANS OF AQUEOUS SOLUTIONS, E 104-85. Amer. Soc. for Testing and Materials,
Philadelphia, PA 19103.
Crosby, P. B., 1979: Quality is Free. McGraw-Hill Book Co.,
ISBN 0-07-014512-1.
Crutcher, H. L., 1984: Monitoring, Sampling, and Managing Meteorological
Data, Chapter 4. Atmospheric Science and Power Production, Darryl
Randerson, Ed. DOE/TIC-27601, pp. 136-146.
EPA, 1976: Quality Assurance Handbook for Air Pollution Measurement Systems,
Vol. I, Principles, EPA-600/9-76-005. Office of Research and Development,
Res. Triangle Park, NC 27711.
EPA, 1987b: On-Site Meteorological Program Guidance for Regulatory Modeling
Applications, EPA-450/4-87-013, Off. of Air Quality Planning and
Standards, Res. Triangle Park, NC 27711.
Fairobent, J. E., 1979: (personal communication) Nuclear Regulatory Commission.
Feigenbaum, A. V., 1961: Total Quality Control Engineering and Management.
McGraw-Hill, LCCCN 61-11051
-------
Section No. 4.1.10
Revision No. 0
Date: 17 Sep 89
Page: 2 of 2
Filippov, V. V., 1968: Quality control procedures for meteorological data.
World Weather Watch Planning Report, No.26, World Meteorological
Organization.
Fritschen, L. J. and L. W.Gay, 1979: Environmental instrumentation.
Springer-Verlag, NY, ISBN 0-387-90411-5.
Grant, E. L . and R. S. Leavenworth, 1974: Statistical Quality Control, Fifth
Edition. McGraw-Hill, ISBNO-07-024114-7
Juran, J. M., 1979: Quality Control Handbook, Third Edition. McGraw-Hill,
ISBN 0-07-033175-8
Klint, W. E. , 1979: Screening checks used by the National Climatic Center for
meteorological data. NCC Asheville, NC (unpublished)
Lee, J. T. and J. Stokes, 1978: Tall tower and aircraft instrumentation quality
control procedures - development and application. Proceedings 4th
Symposium on Meteorological Observations and Instrumentation. Amer.
Meteor. Soc., pp 3-6.3-6.
Lockhart, T. J., 1978b: Data graphics for assessment of measurement quality.
Proceedings of the National Conference on Quality of Environmental
Measurements, Denver, Nov. 27-29, pp 123-132.
Lockhart, T. J., 1985c: Quality assurance of meteorological measurements. In:
"Quality Assurance for Environmental Measurements," ASTM STP 867.
J. K. Taylor and T. W. Stanley (eds). Amer. Soc. for Testing and Materials,
Philadelphia, PA. 253-259.
Lockhart, T. J., 1989a: Comments on "A Quality Control Program for Surface
Mesonetwork Data". J. Atmos. and Oceanic Technol., 6, 525-526.
Lockhart, T. J., 1989b: Accuracy of the collocated transfer standard method
for wind instrument auditing. J. Atmos. Oceanic Technol., 6, 715-723.
Reynolds, G. W. and D. E. Pittman, 1978: The TVA meteorological data acceptance
analysis program. Proceedings 4th Symposium on Meteorological Observations
and Instrumentation. Amer. Meteor. Soc., pp 3-6.3-6.
Reynolds, G. W., 1979: Final acceptance review for TVA meteorological data.
Presented at Quality Assurance in Air Pollution Measurement Conference,
Air Pollution Control Association, New Orleans.
Wade, C. G., 1987: A quality control program for surface mesometeorological
data. J. Atmos. and Oceanic Technol., 4, 435-453.
-------
-------
Section No. 4.2.0
Revision No. 0
Date: Sep 17 S9
Page: 1 of 4
Section 4.2
QA FOR WIND SPEED, WIND DIRECTION AND TURBULENCE
OUTLINE
Section Pages Rev. Date
4.2.0 OUTLINE AND SUMMARY 4 0 9/89
4.2.1 TYPES OF INSTRUMENTS 4 0 9/89
4.2.1.1 COMMON MECHANICAL SENSORS
4.2.1.1.1 Wind Speed Sensors
4.2.1.1.2 Wind Direction Sensors
4.2.1.1.3 Fixed Component Sensors
4.2.1.2 SECONDARY EFFECT SENSORS
4.2.1.2.1 In-situ Sensors
4.2.1.2.2 Remote Sensing Devices
4.2.2 SPECIFICATIONS 32 0 09/89
4.2.2.1 WIND SPEED
4.2.2.1.1 Threshold
4.2.2.1.1.1 Definition
4.2.2.1.1.2 Threshold Measurement
4.2.2.1.1.3 Starting Torque Measurement
4.2.2.1.2 Accuracy
4.2.2.1.2.1 Definition
4.2.2.1.2.2 Measurement of Accuracy
4.2.2.1.2.3 Application of Accuracy
4.2.2.1.2.4 Precision
4.2.2.1.3 Distance Constant
4.2.2.1.3.1 Definition
4.2.2.1.3.2 Measurement of Distance Constant
4.2.2.1.4 Off-Axis Response
4.2.2.2 WIND DIRECTION
4.2.2.2.1 Threshold
4.2.2.2.1.1 Definition
4.2.2.2.1.2 Threshold Measurement
4.2.2.2.1.3 Torque Measurement
4.2.2.2.2 Accuracy
4.2.2.2.2.1 Definition
4.2.2.2.2.2 Measurement of Sensor Accuracy
4.2.2.2.2.3 Measurement of Orientation Accuracy
4.2.2.2.2.4 Expression of Accuracy
4.2.2.2.2.5 Precision
4.2.2.2.3 Delay Distance (Distance Constant)
4.2.2.2.3.1 Definition
4.2.2.2.3.2 Measurement of Delay Distance
4.2.2.2.4 Overshoot or Damping Ratio
4.2.2.2.4.1 Definition
4.2.2.2.4.2 Measurement of Overshoot
,4.2.2.2.5 Dynamic Vane Bias
-------
Section No. 4.2.0
Revision No. 0
Date: Sep 17 89
Page: 2 of 4
Section Pages Rev. Date
4.2.2.3 TURBULENCE
4.2.2.3.1 Definition
4.2.2.3.2 Direct Measurement of Sigma Theta
4.2.2.3.3 Statistical Summaries
4.2.3 ACCEPTANCE TESTING 5 0 9/89
4.2.3.1 WIND SPEED
4.2.3.1.1 Threshold
4.2.3.1.2 Accuracy
4.2.3.1.3 Distance Constant
4.2.3.2 WIND DIRECTION
4.2.3.2.1 Threshold
4.2.3.2.2 Accuracy
4.2.3.2.3 Delay Distance and Overshoot
4.2.3.3 MEASUREMENT SYSTEM
4.2.4 INSTALLATION 8 0 9/89
4.2.4.1 GENERAL CONSIDERATIONS
4.2.4.2 WIND SPEED
4.2.4.3 WIND DIRECTION
4.2.4.3.1 Exposure
4.2.4.3.2 Orientation
4.2.4.3.2.1 True Solar Noon Method
4.2.4.3.2.2 Solar Azimuth Method
4.2.5 CALIBRATION 4 0 9/89
4.2.5.1 WIND SPEED
4.2.5.1.1 System Accuracy
4.2.5.1.2 Component Accuracy
4.2.5.2 WIND DIRECTION
4.2.5.2.1 System Accuracy
4.2.5.2.2 Component Accuracy
4.2.6 OPERATIONS, MAINTENANCE AND QUALITY CONTROL 2 0 9/89
4.2.6.1 OPERATIONS
4.2.6.2 MAINTENANCE
4.2.6.2.1 Routine and Preventive Maintenance
4.2.6.2.2 Corrective Maintenance
4.2.6.3 QUALITY CONTROL
4.2.7 PERFORMANCE AUDIT METHODS 20 0 9/89
4.2.7.1 GENERAL CONSIDERATIONS
4.2.7.1.1 Who
4.2.7.1.2 What
4.2.7.1.3 Where
4.2.7.1.4 When
4.2.7.1.5 How
-------
Section No. 4.2.0
Revision No. 0
Date: Sep 17 89
Page: 3 of 4
Section Pages Rev. Date
4.2.7.2 WIND SPEED
4.2.7.2.1 Sensor Control Method
4.2.7.2.2 CIS Method
4.2.7.2.3 "W" Anemometers
4.2.7.3 WIND DIRECTION
4.2.7.3.1 Sensor Control
4.2.7.3.2 CIS Method
4.2.7.3.3 Vertical Wind Direction,
4.2.7.4 TURBULENCE OF
-------
Section No. 4.2.0
Revision No. 0
Date: Sep 17 89
Page: 4 of 4
QUALITY ASSURANCE FOR WIND SPEED, WIND DIRECTION
AND TURBULENCE
SUMMARY
This section discusses all aspects of the task of monitoring the wind at a
particular site with an emphasis on quality assurance. A background chapter
describes the nature of wind and the kinds of instruments commonly used to
monitor its speed and direction. This section describes in detail the
important aspects of the operation of conventional anemometers and wind vanes.
Some discussion of secondary effect sensors is provided but the handbook is not
intended to cover these instruments. The background information and the
detailed information found in the following chapters are necessary for two
kinds of tasks. One is to execute the responsiblity for the collection of
valid data. The other is to audit or judge how well the first task was
performed within the goals or regulations which caused the measurements to be
made in the first place.
Specifications is the longest and in some ways the most detailed section.
The premise is that an understanding in depth of the way the common sensors
work is necessary before purchasing, installing and operating the instruments.
Specifications set the performance parameters for the instrument or system.
Careful definitions are given along with test methods which will equip the user
to verify or to judge the work of others who verify conformance to
specification.
Once the specifications are clearly understood, the process of purchasing
and acceptance testing can be considered. The contention is that quality
assurance is a vital aspect of defining that which is to be purchased and
verifying the performance of the delivered system. When the valid system is in
hand, the installation can be planned and implemented. The important process
of orientation of the wind vane to TRUE NORTH is described in detail.
Calibration is a foundation on which claims of data validity are built.
This important function may be practiced in a number of phases of the
monitoring program. This chapter stresses documentation of the calibration
findings and methodology. The use of the most inclusive methods practical in
field conditions is advocated. Once the calibrated system has been installed,
the routine performance of operational checks, preventive and corrective
maintenance and quality control operations begin. The documentation of these
operations provide -the framework, resting on the foundation of calibration, to
support the claim of data validity.
Performance audits add confidence to the documentation that the system is
in control. Performance audit methods must be the most comprehensive methods
possible for field challenges. This chapter describes some recommended audit
methods and audit forms to support the methods. The performance audits and
calibrations provide the data for the estimation of accuracy and precision
described in the following chapter.
There is not room to include all the details or background information
which might be needed or desired. A list of the references used for this
section is found at the end of the section. If the reader needs additional
information or is curious about peripheral subjects, the references will
provide the ansWers or a start in search of the answer.
-------
Section No. 4.2. 1
Revision No. 0
Date: 17 Sep 89
Page: 1 of 4
4.2.1 TYPES OF INSTRUMENTS
There are many ways to detect wind as it passes a point on Earth. Only
those ways which reference a fixed point (or volume) will be considered in this
handbook. This class of measurements is expressed in Eulerian coordinates
where properties of the air are assigned to points in space at each given time
(Huschke, 1970). The other class of measurement is expressed in Lagrangian
coordinates. It is good to keep in mind that Eulerian measurements are
frequently used in Lagrangian models. Or, in other words, monitoring data
measuring wind on a tower are used to estimate where parcels of air move and
how the concentrations of constituents of the parcel change in the process.
It is necessary to understand just how the measurement is made to
adequately do the following:
° write procurement specifications,
o adopt and apply acceptance testing methods,
o site the sensors in the representative flow of interest,
° perform calibration and maintenance services,
o establish an effective quality control (QC) operation, and to
o evaluate audits used to estimate precision and accuracy of the data.
This section will describe how various kinds of instruments work. The
thoroughness of the description in this handbook will be proportional to the
frequency of use of the instrument in air quality applications.
Another background point deals with the nature of wind. There can be no
question about the wind requiring a vector to describe fully a single
measurement. The vector has direction in spherical coordinates (azimuth with
respect to TRUE NORTH and elevation with respect to a horizontal plane) and
length (speed) along that direction. It is common for many air quality
applications to deal only with the presumably horizontal flow as measured by
vanes and anemometers. In this case, the horizontal components of the vector
are expressed as an azimuth angle from which the wind is blowing and the speed
at which the air is passing the point of measurement. While each sample of
wind requires both speed and direction, it is common to measure them
separately. A series of samples may be summarized in different ways depending
upon the application. The arithmetic mean of the samples is recommended for
many applications (EPA, 1987b). The standard deviation of the samples is used
to describe the level of turbulence in the air.
4.2.1.1 Common Mechanical Sensors
4.2.1.1.1 Wind speed sensors
Common anemometers are either cup assemblies turning on a
vertical axis or propellers turning on a vane-oriented horizontal axis. The
cup anemometer is an empirical sensor in that the relationship between the rate
of rotation and the wind speed is determined by testing rather than defined by
theory. It is a linear relationship, for all practical purposes, above its
threshold non-linearity and through the range of important application. It is
an aerodynamic shape which converts the wind pressure force to torque (hence
rotation) because of asymmetrical lift and drag. Its dynamic performance
characteristics (starting threshold and distance constant) are density
-------
Section No. 4.2.1
Revision No. 0
Date: 17 Sep 89
Page: 2 of 4
dependent but its transfer function (rate of rotation vs. wind speed) is
independent of density. The cup is not very efficient and creates turbulence
as the air flows through and around it. The linear speed of the center of a
cup is only a half to a third the linear speed of the air turning it (Mason and
Moses, 1984). The cup anemometer is omnidirectional to horizontal flow but
exhibits a complicated reaction to vertical components. It may indicate speed
slightly greater than the total speed when the flow is non-horizontal
(MacCready, 1966).
The propeller anemometer is a more efficient shape. The
helicoid propeller is so efficient that its transfer function can be specified
from theory (Gill, 1973). It creates little turbulence as the air flows mostly
through it, turning like a nut on invisible threads. The propeller measures
wind speed when it is oriented into the wind by a vane. Its -errors from
imperfect alignment with some mean vector are small, being nearly proportional
to the cosine of the angle of misalignment.
In either of these types of anemometers, the rate of rotation
is sensed by some transducer. Tachometer generators, a.c. frequency
generators, light choppers and shaft revolution counters have all been used.
It is important to know how the transducer works if the performance of the
anemometer is to be challenged for a QA purpose.
4.2.1.1.2 Wind direction sensors
The wind vane is perhaps the simplest of instruments. A fin is
tied to a vertical shaft such that when force is applied to the area by the
wind, it will turn the shaft seeking a minimum force position. The
relationship of the shape, size and distance from the axis of rotation of the
fin to the bearing assembly and transducer torque requirements determines the
starting threshold. These attributes of the fin area along with its
counterweight determine the dynamic performance characteristics of overshoot
(damping ratio) and delay distance (distance constant) of the direction vane.
While its equilibrium position is insensitive to density, the dynamic response
characteristics and threshold are density dependent.
Vane design is of little importance if the average wind
direction is all that is required. If turbulence parameters are of interest,
as they usually are or should be, the design of the vane becomes important.
The vane transducer Is usually a potentiometer, but synchros, shaft encoders,
capacitors and Hall effect devices have been used. It is fairly common to find
theorange of the sensor to be "540 degrees" rather than the physically true
360 . The reason is related to the problem of a continuous range (a circle)
with a discontinuous output (0 to n volts). It is important to know how the
transducer works if the performance of the wind vane is to be challenged for a
QA purpose.
A special direction vane is the bivane which has the vertical
range of ± 45° to 60°in addition to the full azimuth circle. The additional
range brings with it the need to neutralize gravity by having a perfectly
balanced vane assembly. Bivanes can be conditionally out of balance, such as
happens when dew^forms and then evaporates from the tail fins. The effect of
this imbalance on threshold and performance is complicated. Horizontal vanes
-------
Section No. 4.2.1
Revision No. 0
Date: 17 Sep 89
Page: 3 of 4
can be designed to be stable in the horizontal even when slightly out of
balance. The effect of this design is to add the vane horizontal restoring
force to the wind force, again a complication.
4.2.1.1.3 Fixed component sensors
Propeller anemometers exhibit something like a cosine response
to a wind along some line other than the axis of the propeller. The degree
with which this response represents a cosine is a function of the design of the
propeller. If the cosine response is perfect, the fixed propeller accurately
reports the component of the wind parallel to the axis of the propeller. If
three propellers were located on fixed X, Y and Z axes, the three outputs would
define the components of a three dimensional wind vector. From a QA
perspective, the accuracy of the wind speed and direction data are related to
the determination of the component errors and the algorithms used to correct
for them. Often ill-defined are the errors from the interference of one
propeller on another and the errors when the beyond-threshold-nonlinearity
speed has not been reached. It is the speed of the component parallel to the
anemometer axis that the propeller responds to, not the total speed. A 5 m/s
wind with a 5° up angle and 5° off the Y axis will provide a 0.44 m/s wind for
the W and X propellers. A 50/4 error in the X propeller because of threshold
nonlinearity would cause an insignificant 0.014 m/s error in the wind speed and
a 2.5° error in wind direction. A 50% error in the W propeller for the same
reason would cause a 50% error in the W component (0.22 m/s reported rather
than the true 0.44 m/s).
4.2.1.2 Secondary Effect Sensors
4.2.1.2.1 In-situ sensors
Several meteorological instrument books contain information on
a variety of wind instruments. See Mason and Moses (1985) and Middleton (1953)
for greater depth and variety.
The three component sonic anemometer is considered in some
circles as the standard for wind measurement. For those applications where the
contribution of small eddys is important, it is an excellent choice. As with
many of the secondary effect sensors, it is a research tool requiring
considerable attention from the operator. It is not a good choice for routine
monitoring. It has its own set of error sources when it is used for measuring
long-term (tens of minutes) averages and standard deviations. Single component
sonic anemometers deployed as W sensors may become the standard for this
difficult measurement. The secondary effect for this instrument is the
transport of sound waves in the air.
The hot wire or hot film anemometer is also a research tool
which measures the wind speed component perpendicular to a heated cylinder.
The secondary effect for this instrument is the removal of heat by the air
measured by the current it takes to replace the heat. There are some new
designs of this .type of instrument which are intended to be monitoring
instruments.
-------
Section No. 4.2.1
Revision No. 0
Date: 17 Sep 89
Page: 4 of 4
4.2.1.2.2 Remote sensing devices
There are two Doppler shift instruments which measure wind
remotely by analyzing the return from transmitted energy pulses. The most
important for the boundary layer applications is the acoustic Doppler (SODAR)
sounder. It depends on the back scatter from small temperature differences
that tag the air for motion measurement. The other Doppler uses
electromagnetic energy to measure winds through the troposphere. There are
also some Doppler applications using lasers as the energy source. These
systems are complicated. A QA effort related to systems of this type will
require special study of the system and ingenuity to find other ways of
measuring what they are measuring. In the case of the SODAR which is being
used in monitoring applications for air quality programs, a reasonableness test
with another instrument capable of measuring winds aloft is more likely to
produce useful information than is a challenge designed to report the accuracy
of the SODAR.
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 1 of 32
4.2.2 SPECIFICATIONS
The purpose of defining specifications is to give unambiguous meaning to
the terms used by all those who are concerned that the instruments and systems
selected and operated will meet the needs of the application or project. This
starts with procurement specifications and ends with supporting claims of data
quality. These specifications provide the basis for receiving inspection and
testing. The wind is the most important variable to be measured and its
specifications are the most complicated. Specifications discussed here will
also include some aspects of the measurement system.
Project and application requirements vary. To make this handbook as
specific as possible, the examples used will be consistent with those presented
in the On-Site Meteorological Program Guidance for Regulatory Modeling
Applications (EPA, 1987b). The specifications will be discussed in order of
their importance and then summarized at the end of the sub-section.
4.2.2.1 Wind Speed
4.2.2.1.1 Threshold
4.2.2.1.1.1 Threshold definition
One of the keys to a good wind sensor is a low threshold.
The threshold is also the one performance characteristic which will certainly
change with time because of bearing degradation. There is no standard
definition of threshold so different manufactur-ers may apply different tests
to establish their threshold specifica-tion. Absence of a standard or
definition of the specification makes it difficult to specify a meaningful
value. The following definition comes from Standard Test Method for
DETERMINING THE PERFORMANCE OF A CUP ANEMOMETER OR PROPELLER ANEMOMETER (Draft
6) (ASTM, 1985):
"Starting threshold (U ,m/s)—the lowest wind speed at which a
rotating anemometer starts and continues to turn and produce
a measurable signal when mounted in its normal position."
A starting threshold specification, 0.5 m/s for example, should include a
footnote describing the meaning of the specification. In the example above, it
might say: 0.5 m/s (1)
(1) "as determined by wind tunnel tests conducted on production
samples in accordance with ASTM D22.ll test methods."
All rotating anemometers are non-linear as they go from not
turning to turning at a rate predicted by their linear transfer function.
Note that the definition does not require linear output at threshold, only
continuing turning and measurable signal. If the manufacturer provides an
accuracy specification which is independent of speed, the presumption is that
the accuracy specification is met at threshold. Consider a hypothetical cup
with a transfer function, i.e., the relationship between rate of rotation and
wind speed, as follows:
U = 0.2 + 1.5 R
where U is wind speed (m/s) and
R is rate of rotation (rps)
The transfer function would have been found by using a least squares fit
(linear regression) to wind tunnel data. The ASTM method uses the wind speeds
well above the starting threshold to avoid bias from the non-linear threshold.
In Figure 4.2.2.1 the lowest 2 m/s of the hypothetical performance curve is
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 2 of 32
shown along with the contribution of the offset to the system output. The
variable part of the transfer function (U = 1.5 R) coming from the cup rotation
is shown theoretically as the straight line from 0.2 m/s to an output of 1.8
m/s when the wind speed is 2 m/s. The triangles show the actual output from
the cup rotation. They start to turn at 0.3 m/s (threshold) and reach the
theoretical line at about 0.8 in/s. The parallel line through the origin simply
adds the constant offset to the cup rotation output. The measurement error is
the difference between the diamonds in the figure and the ideal straight line.
It starts at +0.2 m/s, goes to -0.1 m/s at 0.3 m/s, and then gets smaller as
the nonlinearity of the threshold decreases.
The offset is defined either by the linear regression or
by the arbitrary choice of the manufacturer. If it is the former, the starting
threshold will always be larger than the offset. If it is the latter, the
starting threshold may be either side of the offset. The manufacturers of the
common small three cup anemometer often set an offset voltage in their signal
conditioner as shown in Figure 4.2.2.1. For this hypothetical cup, the offset
voltage is critical to its meeting the accuracy specification discussed in
4.2.2.1.2. Sensitive propeller anemometers have a much smaller offset because
they develop more force (torque) per m/s. Some offsets are so small that there
is no advantage or need to use an offset voltage. See Baynton (1976) and
Lockhart (1977) for further discussion of the errors of rotational anemometers,
particularly at the threshold.
Cup Anemometer Performance
Threshold Analysis
Hypothetical Transfer Function
U (m/s) = 0.2 + 1.5 R (rps)
From Cup Rotation and Offset Voltage
Ideal
O Actual
From Cup Rotation
Theoretical
V Actual
i i i i i r
0.2 0.4 0.6 0.8 1.0
Wind Speed (m/s)
*
Figure 4.2.2.1 A hypothetical cup anemometer threshold analysis.
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 3 of 32
4.2.2.1.1.2 Threshold Measurement
There is only one way to measure starting threshold. It
requires a wind tunnel capable of accurate operation below 1 m/s. One standard
methodology is defined in ASTM (1985) and described in Lockhart (1987).
However, it is possible to estimate the starting threshold by matching the
torque which is required to keep a cup or propeller from turning at a known
wind speed (in a wind tunnel) with the starting torque of the anemometer
bearing assembly. Lockhart (1978) provided the torque relationship as a
function of wind speed for four anemometer shapes.
Table 4.2.2.1 contains values calculated with these
data by using the relationship
T = K u2
where T is torque (g cm /sec )
u is the square of the wind speed (m/s)
K is a constant for the aerodynamic shape (g)
The values in the table were calculated from this formula using
the K values from Lockhart (1978).
Table 4.2.2.1 - Torque Developed vs. Wind Speed
Speed
(m/s)
0.1
0.2
0.3
0.4
0.5
1.0
cup
#1
(g-cm)
0.014
0.056
0.126
0.224
0.350
1.4
cup
#2
(g-cm)
0.027
0.108
0.243
0.432
0.675
2.7
prop
#3
(g-cm)
0.049
0. 196
0.441
0.784
1.225
4.9
cup
#4
(g-cm)
0.148
0.592
1.332
2.368
3.700
14.8
#1 Teledyne Geotech 170-42 (20.3 g) K= 1 . 4
#2 MR I Model 1022 (48.3 g) K= 2.7
#3 R.M. Young Model 21180 (9.7 g) K= 4.9
#4 MRI Model 1074 (186.8 g) K=14.8
The torques listed are those acting on the sensor when the sensor is restrained
in a wind field at the speed listed. If the sensor bearing assembly has a
starting torque less than the torque provided at that speed, and the restraint
is removed, it will start turning. The torque watch used for the low speed end
of the wind tunnel tests was a Waters Model 366-3 with a range from 0.003 oz-in
(0.216 g-cm) to 0.03 oz-in (2.16 g-cm). To convert oz-in to g-cm, multiply by
72.
The method of using a starting torque measurement to find
the sensor starting threshold will become standard only with the publishing of
K constants by the manufacturers. One manufacturer (R.M. Young Co.) provides
the K value for anemometers. These values are shown in Table 4.2.2.2.
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 4 of 32
Table 4.2.2.2 - Anemometer K Values
Type
Polypropylene Cupwheel
Polypropylene Propeller
Polystyrene Propeller
Polystyrene Propeller
Model
No. 12170C-100cm
No. 08234-18x30cm
No. 21282-19x30cm
No. 21281-23x50cm
K
1.4
2.5
3.6
5.0
4.2.2.1.1.3 Starting Torque Measurement
The starting torque of an anemometer bearing assembly will
increase in time because of wear and dirt. The starting torque, with the cup
assembly or propeller removed, can be measured. Starting torque measurement is
simple in concept but sometimes difficult in application. An experienced
meteorological instrument technician can tell if a bearing assembly is in need
of service by simply feeling the shaft or rotating or spinning the shaft and
listening to its sound. The trouble with this practice is that it is not
quantitative. It works for field servicing instruments but does not provide
documentation suitable for a quality control program. Another qualitative
practice is to roll the sensor slowly over a smooth horizontal surface watching
the shaft not turn as the sensor
turns around it (see Figure
4.2.2.2). Set screws and other
asymmetries apply a torque which
•; keeps the shaft from turning while
the sensor moves around it. If the
applied torque could be measured,
this method would be quantitative.
The measurement of the
starting torque of the bearing
assembly provides only an
Figure 4.2.2.2 Climatronics F460 torque approximation of the starting
test for speed sensor threshold of the anemometer,
particularly cup anemometers. The absence of the cup weight may lower
the starting threshold of the cup bearing assembly but there is no evidence
that this is an important consideration. At this point in time there is no
better way to estimate and document in the field and in units of wind speed
this important specification, the starting threshold of the anemometer.
The direct measurement of starting torque requires some device which can
apply a known torque. The most common, perhaps, is the Waters Torque Watch. A
model 366-3 is shown in Figure 4.2.2.3 applied to a Climatronics cup anemometer
sensor. The measurement requires some degree of care and skill. The torque
watch has a square shaft which fits into a square hole in the connecting
fixture. The torque watch is turned while holding its shaft in line with the
anemometer shaft, without end loads. The indicator is watched and when the
shaft turns the maximum reading is recorded. This process needs to include at
least one full turn of the anemometer to be sure the maximum friction in the
bearing assembly is encountered. The torque watch measures either clockwise or
counterclockwise. Use only the rotation sense required by the cup assembly or
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 5 of 32
propeller. The range of the torque watch may not be as sensitive as one would
like. If the model 366-3 is used on anemometer #1 in Table 4.2.2.1, the
threshold (0.003 oz-in or 0.216 g-cm) will not measure equivalent speeds below
0.4 m/s. If the torque watch turns the shaft
without reaching the lowest scale point, all
that can be said is that the starting threshold
of the anemometer is less than 0.4 m/s. ,; ^
Another torque watch is the GM-CM Torqmeter
781 with a 0.1 to 2.1 g-cm range (0.001-0.029
oz-in), shown in Figure 4.2.2.4 mounted to a
Teledyne Geotech 1565C wind direction sensor.
A third torque measuring device is the simple
Torque Disc, model 18310 made by R.M. Young Co.,
shown in Figure 4.2.2.5. This is a fundamental
device which does not need expensive
calibration. Weights (screws) are attached at
distances from the center of rotation. The
force of gravity provides g-cm torques at the
center of rotation of the intentionally out of
balance disc. The shaft being tested must be
horizontal and symmetrical in mass. A cup
anemometer shaft which does not turn while the
sensor is slowly rolled along a flat surface
will not work with the Torque Disc. The g-cm
torque applied equals the weights and distances
when the weights are in the same horizontal
plane as the shaft. Calibration results from
weighing the weights and measuring the distances.
Figure 4.2.2.3 Waters
Torque Watch
An appropriate interface
fixture would allow the Torque Disc to be used to calibrate a torque watch.
Figure 4.2.2.4 Cm-Cm
Figure 4.2.2.5 Young Torque Disc
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 6 of 32
There are several ways to measure torque but the available
instrumentation to make the measurement is limited. Figure 4.2.2.6 is a
collection of spring-type torque watches, spring scales and circular discs
capable making torque measurements within narrow ranges and specific
orientations. It i& necessary to become familiar with these devices and how
they are correctly used.
A - Waters Torque Watch
Model 366-3
(0.2-2.0 g-cm)
B - Waters Torque Watch
Model 651X-3
(18-360 g-cm)
C - GM-CM Torqmeter
" Model 781
(0.1-2.1 g-cm)
D - Young Torque Disc
Model 18310
(0.1-15 g-cm)
E - Haldex AB
Gram Gauge
(1-10 g)
F - Young Gram Gauge
Model 18330
(0-10 g)
Figure 4.2.2.6 Various Torque Measuring Devices
4.2.2.1.2 Accuracy
4.2.2.1.2.1 Definition
The classic definition of accuracy is the comparison of a
measured value to a true value expressed as a bias term plus or minus a random
uncertainty (precision). The bias term may be conditional with respect to the
best fit straight line; it may vary with wind speed or angle of attack.
When accuracy is specified, the kind of true value to be
used to test the accuracy claim must also be specified. Usually the buyer
expects the "true value" to be the wind speed where the anemometer is sited.
The manufacturer expects the "true value" to be the near laminar flow of a
calibration wind tunnel. Some auditors expect the "true value" to be the
output predicted by the transfer function when the anemometer is rotated at a
known rate. Let us label the kind of accuracy as follows:
A(l) - accuracy with respect to the horizontal component
of wind speed at the sited location,
A(la) - instrument response
A(lb) - siting representativeness
A(2) - accuracy determined in a wind tunnel, and
A(3) - accuracy of conversion of rate of rotation to output.
A(3) is the easiest to measure and represents most of the
claims for data accuracy from audit reports. It requires, usually, a measure
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 7 of 32
of the offset voltage (or equivalent m/s) and the combination of offset and
output from the rate of rotation to output converter.
A(2) requires a calibrated wind tunnel and it is a more
difficult and more expensive accuracy to determine. It does provide a check on
the manufacturer's generic transfer function plus the variation in production
individuals represented by the individual being tested. The wind tunnel
calibration test provides a specific transfer function which can be used to
change the signal conditioning to get the smallest error of an A(2) type for
that specific system. Serial numbers should be recorded with the test results.
Some manufacturers do not identify their cup assemblies or
propellers with numbers. Assume an A(2) accuracy is found in a wind tunnel
test and compared to the generic transfer function used for A(3) operations.
Assume the test shows that the A(3) value is uniformly 10% low. Assume the
operators took a year of data with A(3) accuracy tests showing insignificant
errors. The error to the A(2) level is a 10% bias and the data can be
corrected for the year. This action requires either evidence or good reason to
believe that there was no physical change to the cup assembly over the year and
that the individual tested was the one used during the year.
There are still unknowns of the A(l) type to consider.
These are usually conditional biases and often impossible to define. They may
be recognized and their impact estimated. There are two types of these errors.
One is the consequence of the anemometer design in the flow field it is to
measure., A(la). The other is a result of assumptions of representativeness,
The discussion of A(la) errors requires an understanding
of conventions for the use of u, v and w. In traditional diffusion
applications, the statistics for wind representing a period of time refer to u
as the speed of the horizontal component along the direction of the mean wind,
v as the speed of the horizontal wind component perpendicular to the mean
direction, and w as the speed of the vertical wind component. Another
convention applies to fixed component anemometers such as the UWJ propeller
array. Here, the U is the east-west component of the wind in a Cartesian
coordinate system (a west wind is positive); V is the north-south component of
the wind (a south wind is positive); and W is the vertical component of the
wind (upward moving wind is positive) (Stull, 1988).
MacCready (1966) characterized errors in anemometers when
operating in a turbulent flow. A cup anemometer has a u-error because of a
different response "constant" to an increasing speed than to a decreasing
speed, so-called overspeeding. With modern sensors, this is usually a small
error of a percent or two, depending on sensor design and height above ground.
A cup anemometer has no v-error since it is insensitive to changes in
direction, but it does have a w-error caused by non-horizontal flow. This
error can easily be 10% and larger with some designs (Lockhart, 1987). A vane
oriented propeller will have small v-errors and w-errors from misalignment.
These will be small because the propellers respond nearly as the cosine of the
misalignment angle, 2% for a 10 degree misalignment. The u-error is too small
to measure for light weight helicoid propellers. These are all Ada) errors
and they vary as the wind varies.
A(lb) errors deal with the assumption is that the
anemometer is measuring what the true wind would be at the point of measurement
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 8 of 32
if the anemometer were not there. This is a question of representativeness and
not instrumentation but it can have a large impact on the question of data
accuracy. The influence of any supporting structure can bias the flow which
the anemometer faithfully measures. If the assumption goes further to equate
the measurement to its physical height above ground, and if the anemometer is
mounted on a 2 m pole on top of a large 8 m building, the bias with respect to
a 10 m flow over a flat field will be the fault of the building. These errors
are of the A(lb) type. If the pole is on the edge of the building, the
distortion of the building will provide non-atmospheric errors of the A(la)
type to be combined with the A(lb) type. These types of errors are very
difficult to define and virtually impossible to correct. Data from an
anemometer mounted in a questionable site, after the A(2) and A(3) errors have
been calibrated out, could be compared with data from a vane oriented propeller
anemometer mounted in a space where the subject anemometer is assumed to
represent. The difference in these collocated measurements may be used to
estimate the magnitude of A(l) errors. The A(2), A(3) and A(la) errors are the
ones to concentrate on minimizing. See 4.2.4.1 for siting guidance.
4.2.2.1.2.2 Measurement of accuracy
The accuracy of an anemometer is found by comparing its
output to the known speed in a wind tunnel. A calibrated wind tunnel has
uncertainties associated with its operation. These incldde instrumentation
errors in measuring the wind speed in the tunnel when it is empty (0.1 mph or
0.05 m/s in the NBS wind tunnel above 2 mph) and the inhomogeneity of wind
speed in the test section away from the boundary layer (a function of the
tunnel design). The turbulence level in the wind tunnel test section should be
homogeneous across the test section with most of the energy in eddy sizes which
are small compared to the size of the anemometer. When this is true, and it
usually is, turbulence does not influence the calibration. Fluctuations in the
tunnel speed can be thought of as long wave length longitudinal turbulence.
This "turbulence" can influence the calibration without careful measurement
synchronization and time averaging.
When an anemometer is placed in the test section for
calibration, consideration must be given to blockage errors, which are
dependent on the ratio of the size of the instrument to the size of the test
section. Also interference errors, which are dependent on the placement of the
anemometer with respect to the wind tunnel instrumentation need to be
considered. Small calibration wind tunnels may themselves be calibrated with
an anemometer which has been run in the NBS wind tunnel. It is also a common
practice to run two anemometers side by side, one of which has an NBS
calibration curve. It is prudent to reverse the positions from time to time to
verify test section homogeneity. It is not reasonable to expect such
calibrations, even though they are "traceable to NBS" by some definition, to
have an accuracy better than 0.2 mph (0.1 m/s).
While a calibrated wind tunnel is the recognized standard
method for calibrating an anemometer, a fundamental (but not very practical)
calibration is possible by moving an anemometer over a measured length in a
measured period of time through still air (Lockhart, 1985b and Stearns, 1985).
Most manufacturers have samples of their products calibrated by NBS to
establish for their design a generic relationship between wind speed and rate
of rotation, measured by counting pulses, frequencies or output voltages. This
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 9 of 32
relationship is then used as the transfer function to define what the signal
conditioning electronics or other output devices require to express the
measured rate of rotation in units of wind speed. Some manufacturers test and
adjust each cup wheel or propeller to fit the generic relationship within some
error band.
Baynton (1976) discusses the calibration of anemometers
and shows the results of tests of 12 different kinds of anemometers in the
National Center for Atmospheric Research (NCAR) wind tunnel. He compares his
calibration to the manufacturer's calibration or generic relationship. Except
for the Aerovane, which probably was too large for the NCAR tunnel, the
difference was within ±3%. He also discusses the difference between an
anemometer transfer function which goes through the origin, of the form
Y = bX,
and the transfer function with an offset or threshold, of the form
Y = a + bX.
Table 4.2.2.3 summarizes data from his Table 1 and Table 5.
Table 4.2.2.3 Wind Tunnel Test Results From Baynton
Type of Anemometer
Gill 4-blade helicoid propeller
Gill 3-blade helicoid propeller
Aerovane helicoid propeller
Taylor Biram' s propeller
Case 1 la Sensitive
Thornthwaite
INSTAAR t
Climet 011-1
TechEcology $
Gill 3-cup
Electric Speed
Bendix Total lizer Model 349
MRI Model 1074 ft
* a is not significantly greater than
a
(m/s)
0.073*
0.011»
0.233
0.145
0.467
0.331
0.316
0.265
0.275
0.250
0.610
0.588
0.087»
zero
t Institute for Arctic and Alpine Research,
* Analysis of NBS data provided by L.
ft Data from Lockhart (1977)
E is the ratio of the test result to
b
(m/rev. )
0.309
0.487
1.356
0.255
1.404
1.476
1.597
1.382
1.391
1.057
2.728
2.605
2.314
Boulder,
E
1.03
1.03
0.93
1.00
0.98
0.97
1.01
1.03
0.97
Colo.
Petralli
the mfg
. ' s calibration.
Baynton lists the value of "a" for the 12 anemometers he tested, which ranged
from nearly zero (0.01 m/s) to 0.6 m/s. One purpose of his paper was to
caution users of "wind run" instruments of the errors associated with ignoring
"a." Wind run is generally used to describe those anemometers which count
shaft rev-olutions over a long period of time or are geared to a counter or
recorder in such a way that the output is in units of speed. These instruments
cannot pro-vide an offset. Lockhart (1977) shows that some cup designs,
specifically the Meteorology Research, Inc. Model 1074 used on the Mechanical
Weather Station, can be nearly as accurately described without an offset (a»0)
as with one (a=0.03). Table 4.2.2.4 lists the NBS data and the linear
regressions to sup-port this fact. The residual errors from each model are
plotted in Figure 4.2.2.7.
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 10 of 32
Table 4.2.2.4 NBS Test Data for MRI Model 1074
Test Output
No. Freq.
4
7
8
9
10
11
12
13
14
15
16
|7
14
28
88
112
240
380
500
425
755
830
1005
1255
1500
1755
1995
iiOJ
18 2530
Test date - 11/18/75
Tunnel
Speed
dph)
0.9
1.3
2.8
3.7
4.7
15..1
19.9
24.9
29.8
34.7
39.6
49.5
59.3
69.6
79.6
90.2
100.7
Y K
Tunnel Output
Speed /132
(i/s) (rps)
0.4 0.11
0.21
0.49
0.67
0.35
1.82
2,88
3.79
4.73
5.72
6.67
0.6
1.3
1.7
2.1
4.3
6.8
8.9
11.1
13.3
17.7
22.1
7.61
9.51
31.1 13.30
35.6 15.11
40.3 17.16
45.0 19.17
HR1, Altadena, Calif.
1
i.i/S)
o.:s
0.53
1.19
1.59
2.02
4.29
6.77
8.90
11.12
13.43
15.64
17.86
22.30
26.64
31.17
35.42
40.21
44.91
llif.
.inear Kegn
(i/S)
-0.12
-0.05
-0.07
-0.06
-0.08
-0.00
0.02
0.01
-0.01
0.10
0.13
0.16
0.17
0.13
0.05
-0.16
-0.11
-0.10
Node! 1074
ession
K"
fi/s)
0.25
0.50
1.15
1.56
1.99
4.26
6.75
8.88
11.10
13.41
15.63
17.35
22.29
26.64
31.17
35.43
40.23
44.93
8 132/rev.
r-Y
(s/s)
-0.15
-0.08
-o.to
-0.09
-0.11
-0.03
-0.00
-0.02
-0.03
0.09
0.12
0.15
0.16
0.13
0.05
-0.15
-0.10
-0.08
Regression Cutout:
Constant
Std Err of Y Est
R Squared
Mo. of Observations
Degrees of Freedoi
I Coefficient(s)
Std Err of Coef,
r
No. 1-18
032039
,106914
0.999949
18
16
2.341656
0.004174
Regression Output:
Constant
Std Err of Y Est
R Squared
No, of Observations
Degrees of Freedoi
K Coeffident(s)
Std Err of Coef.
to.
r
1-18
0
0.106039
0.999946
18
17
2.344297
0.002764
Linear Regression of NBS Data
U.3 -
0.4 -
0.3 -
0.2 -
< 0.1 -
e
P
t -0.1 -
UJ
-0.2-
-0.3 -
-0.4 -
-n s -
/+ ' *
rv
K
4-
_^f —
s*^
/
•t-
T~-^_
^^x
\
\
\
V--
V--^
fc *
KEY
X'
+ X"
0 10 20 30
Wind Speed (m/s)
Figure 4.2.2.7 Residua 1 Errors from HRI Model 1074
40
50
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 11 of 32
A similar analysis for a propeller anemometer is shown in
Table 4.2.2.5. These data come from a test in the Atmospheric Environment
Service (AES) of Canada wind tunnel on a propeller anemometer being used in a
"round robin" experiment to estimate the accuracy of wind tunnel calibrations.
Each test was run for 100 seconds. The tunnel speed is an average of one
second samples taken every ten seconds by AES. The sensor count is a total for
100 seconds from the light chopper delivering 10 pulses per revolution. Each
test was replicated and tests 3 and 4 were also replicated in tests 19 and 20.
Two linear regressions were run. The first, and best fit, allowed the intercept
of the X axis, or zero offset in ASTM language, to be calculated. The second
forced the straight line through the origin. This latter method yields a
constant slope or pitch (meters per revolution) which when multiplied by the
rate of rotation (revolutions per second.) results in wind speed (meters per
second). The residualerror from these two regressions are plotted in Figure
4.2.2.8.
It is characteristic for helicoid propellers to show a
better correlation with wind tunnel speeds than does a cup anemometer. This is
because propellers generate torque uniformly without sensitivity of position.
Three-cup assemblies, on the other hand, produce three peaks and three valleys
in torque for each revolution (Lockhart, 1985). Either type of anemometer can
be calibrated to an accuracy sufficient for most applications.
4.2.2.1.2.3 Application of accuracy specifications
An accuracy specification should include enough
information to define the type of accuracy intended and the method by which
accuracy claims may be tested. Here are a few examples of accuracy
requirements.
In the Ambient Monitoring Guidelines for Prevention of
Significant Deterioration (PSD) (EPA, 1987a), it states that for horizontal
wind systems "Wind speed systems should be accurate above the starting
threshold to within 0.25 m/s at speeds equal to or less than 5 m/s. At higher
speeds, the error should not exceed 5 percent of the observed speed (maximum
error not to exceed 2.5 m/s)." In the On-Site guide (EPA, 1987b) in 8.1.1 it
states "Accuracy (error) (1 ) (2) s(0.2 m/s +5% of observed)
(1) as determined by wind tunnel tests conducted on production
samples in accordance with ASTM D-22.11 test methods. 21 (sic)
(2) aerodynamic shape (cup or propeller) with permanent serial
number to be accompanied by test report, traceable to NBS,
showing rate of rotation vs. wind speed at 10 speeds."
By implication, the latter specification refers to accuracy type A(2), although
the expectation is that A{1) will be included by careful siting. This
expectation must be addressed with experienced subjective judgment.
Assume a system is to be used in accordance with EPA
(1987b), and an "off the shelf" anemometer is purchased. The manufacturer
states that the sensor delivers 30 pulses per revolution (ppr) with a transfer
function from revolutions per second, R (rps), to wind speed, U (m/s), of
U ,„., , 0.224 * = 0.244
0.224 (m/s) + 1.410 (m/r) R (rps).
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 12 of 32
Table 4.2.2.5 NBS Test Data for Young 27106 at AES
Test
No.
I
i
i
T
N>
4
5
6
7
a
9
10
11
12
13
14
IS
16
17
18
19
20
Output
Count
J)
3660
3675
10839
10911
13108
13207
25418
25433
33006
33003
40163
40139
47636
47577
55102
55105
62613
62785
10930
10931
fiES
Tunnel
Speed
ii/5)
1.14
1.14
3.27
3.27
5.42
5.42
7.59
7.61
9.82
9. S3
11.96
11.95
14.13
14.14
16.35
16.30
18.53
13.56
3.29
3.23
Y
Tunnel
Speed
it/5)
1.14
1.14
3.27
3.27
5.42
5.42
7.59
7.61
9.82
9.53
11.96
11.95
14.13.
14.14
16.35
16.30
18.53
18.56
3.29
3.28
X
eqression-
Output
/1000
(rps)
3.66
3.68
10.89
10.91
18.11
18.21
25.42
25.43
33.01
33.00
40.16
40.14
47.64
47.58
55.10
55.11
62.61
62.79
10.93
10.93
r
(»/s)
1.15
1.15
3.23
3.29
5.41
5.44
7.57
7.53
9.81
9.81
11.92
11.92
14.13
14.11
16.33
16.34
18.55
18.60
7 1C
•J. i7
3.29
X'-Y
(«/S)
0.01
0.01
0.01
0.02
-0.01
0.02
-0.02
-0.03
-0.01
-0.02
-0.04
-0.03
0.00
-0.03
-0.02
0.04
0.02
0.04
0.00
0.01
r
(I/E)
1.09
1.09
3.23
3.24
5.37
5.40
7.54
7.55
9.80
9. SO
11.92
11.91
14.14
14.12
16.35
16.36
18.58
18.63
3.24
3.24
X'-Y
il/5)
-0.05
-0.05
-0.04
-0.03
-0.05
-0.02
-0.05
-0.06
-0.02
-0.03
-6.04
-0.04
0.01
-0.02
0.00
0.06
0.05
0.07
-0.05
-0.04
Regression
Constant
Std Err of
R Squared
Output:
Y Est
No. of Observations
Degrees of
Freedot
X Coefficient^)
Std Err of
Regression
Constant
Std Err of
R Squared
Caef.
Output:
Y Est
No. of Observations
Degrees of
X Coeffici
Std Err of
Freedoc
ent'.s)
Coef.
V
No. 1-20
0.067574
0.023965
0.9999S4
20.00
18.00
0.295224
0.000279
X1
No. 1-20
0.000000
0.043523
0.999944
20.00
19.00
0.296803
0.000268
U) 10 pulses oer revolution counted for 100 seconds.
Ataosonertc
Environtent
Service of Canada
April 12,
1983 with
J. Earle
Chaoaan, Young 27
106 8 10/rev.
Linear Regression of AES-RR Data
0.5 -
0.4 -
0.3 -
0.2 -
0.1 -
0 -
0.2 -
0.3 -
0.4 -
0.5 -
Residual Error Analysis
7 f
V **. — -fe
IT »
_™jj,...
7 7
_J '
>
KEY
V X"
X>
10
15
19
Wind Speed (m/s)
Figure 4.2.2.8 Residual Errors from Helicoid Propeller
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 13 of 32
Assume that an NBS test was conducted after a year of
operation and the results provided a least squares analysis of
U (m/s) = 0.301 (m/s) + 1.387 (m/r) R (rps).
What action is suggested by this finding? Throughout the year the operator had
the electronics trimmed to output 0.224 m/s when the cups were not turning and
14.325 m/s when the cups were turning at 10 rps. The A(3) error in converting
R to U is 0.00. The A(2) error can be expressed as follows:
Generic transfer function: U = 0.224 + 1.410 R
Wind tunnel (truth): U' = 0.301 + 1.387 R
The error (E, m/s) is U - U' or E » -0.077 + 0.023 R
In terms of the measured speed, U, E = -0.081 + 0.0163 U
Table 4.2.2.6 compares this error with the specification at different rates of
rotation. The 1.6 percent overestimation of speed by the generic transfer
function is not large enough to bother with data correction. The data meet the
accuracy guidelines with two thirds of the allowable error unused. At the next
calibration the system should be adjusted to the wind tunnel derived transfer
function.
Table 4.2.2.6 Wind Speed Errors
R
(rps)
0.000
1.225
3.388
6.993
14.203
U
(m/s)
0.224
1.951
5.001
10.084
20 . 250
U'
(m/s)
0.301
2.000
5.000
10.000
20.001
E
(m/s)
-0.077
-0.049
0.001
0.084
0.249
Allowed
(m/s)
±0.20
±0.30
±0.37
±0.45
±0.81
Used
(%)
38
16
0
19
31
4.2.2.1.2.4 Precision
The definition of accuracy describes a bias term and a
variable term akin to precision. Traditionally, precision describes the
uncertainty with which a measuring process or instrument realizes the measured
value when that being measured is the same thing and is repeatedly measured.
The key to finding the variability of a measuring process is to use a
non-varying subject. In meteorology, and particularly in anemometry, it is not
possible to have, with certainty, a non-varying subject. The ASTM subcommittee
D-22.11 dealt with this problem by writing the Standard Practice for
DETERMINING THE OPERATIONAL COMPARABILITY OF METEOROLOGICAL MEASUREMENTS -
D4430-84 (ASTM, 1984). This work was patterned after Hoehne (1973) in which he
defines Functional Precision as the root-mean-square of a progression of
samples of the difference between simultaneous measurements made by identical
instruments collocated in the atmosphere. Operational comparability applies to
two different kinds of instruments rather than identical ones. This method
recognizes that, from an operational perspective, the precision of a
measurement can be estimated by know-ing how well identical or similar
instruments measure the "same" flow.
An EPA project collected data in Boulder, Colorado in 1982
to add to the literature some estimates of comparability. Finkelstein et al.
(1986) published in the refereed literature the material published by NOAA in
Kaimal et al. (1984). Lockhart (1988) re-analyzed these data and concluded
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 14 of 32
that the operational precision of anemometers at 10 m is no larger than 0.2
m/s. Operational precision is the standard deviation of a series of difference
measurements which is equivalent to the operational comparability with all the
bias (mostly calibration error) removed.
An expression of accuracy for an anemometer operating on a
10 m tower in the atmosphere can be expressed as some function of speed [f(u)l,
which comes from the wind tunnel test, A(2), plus or minus 0.2 m/s. This
estimate does include the influence of turbulence on the sensor since the 0.2
m/s comes from collocated cups and vane oriented propellers operating in a
turbulent summer environment.
4.2.2.1.3 Distance Constant
4.2.2.1.3.1 Definition of distance constant
ASTM (1985a) defines distance constant as the distance the
air flows past a rotating anemometer during the time it takes the cup wheel or
propeller to reach (1-1/e) or 63 percent of the equilibrium speed after a step
change in wind speed. The step change is specified as one which increases
instantaneously from 0 to the equilibrium speed. The step change is simulated
by releasing a restrained anemometer in a wind tunnel running at the
equilibrium speed. Several authors, among them Acheson (1988), Hayahsi (1987),
Lockhart (1987), and Snow et al. (1988), have commented on the difference
between the distance constant to an increasing step function and the distance
constant to a decreasing step function. The difference is larger with larger
and heavier cup wheels, as is the size of the resulting overspeeding error.
Snow et al. (1988) point out that a system including a sensor and an analog
signal conditioner will have a combination distance and time constant.
4.2.2.1.3.2 Measurement of distance constant
Most manufacturers will provide the distance constant of
their product. These are usually derived from tests of prototype sensors
during the development phase of the product. The variation from individual to
individual in a production model is not large nor important. It is important
to use a standard test and standard definitions if distance constant
specifications are to be meaningfully compared to other designs and
requirements.
EPA (1987a) does not specify a distance constant for
anemometers. EPA (1987b) does suggest in the Instrument Procurement section
8.1 a distance constant of <5 m at 1.2 kg/m (standard sea-level density). As
with accuracy, this reference uses a footnote to specify the ASTM test method.
The reason why distance constant is included is to urge
users to buy high quality responsive sensors. Heavy sensors with long distance
constants are more likely to produce overspeeding errors, which overstate the
average wind speed. If they are used to measure turbulence, they will
underestimate sigma u because of a failure to respond properly to eddy sizes
smaller than twice the distance constant.
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 15 of 32
4.2.2.1.4 Off-Axis Response
This specification, while included in ASTM (1985) and
recognized in the literature as a source of error, is not included in EPA
requirements or suggestions. It is mentioned here for completeness and in
anticipation of future specifications when more data have been published on the
subject.
The off-axis errors from helicoid propellers are nearly cosine
errors. When a vane-oriented propeller is turned in a wind tunnel so that the
wind is at some angle to the axis of rotation of the propeller, the propeller
slows down. The indicated speed from this misorientation of the propeller is
nearly equal to the total speed times the cosine of the angle of
misorientation. That is, if the indicated speed from a propeller is 5.00 m/s
and the propeller is being held 10 degrees off the true axis of the flow by the
aligning wind vane, the true speed is the indicated speed (5) divided by the
cosine of 10 degrees (0.9848) or 5.08 m/s. In natural turbulent flow, a vane
located behind the propeller may not keep the propeller perfectly aligned with
the wind. Small misalignments result in small errors since the cosine of a
small angle is nearly one.
The off-axis errors from a cup anemometer with a properly
oriented vertical axis will depend on the design of the cup wheel and the angle
from horizontal from which the wind reaches the cups. MacCready (1966) and
Kondo et al. (1971) show that cup anemometers overstate the wind speed when the
air flow is not horizontal. Kondo shows the overestimation by the cups tested
to be 5 percent when the standard deviation of the elevation angle is 17
degrees and 10 percent at 25 degrees. Siting on ridges or building tops or
anywhere the distortion of the flow over an object produces a steady
non-horizontal flow will result in errors which will be unknown.
The figure from MacCready (1966) showing the response of
various anemometers to the elevation angle of the wind is reproduced here as
Figure 4.2.2.9.
(1) cos 8 CURVE
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 16 of 32
Consider the fact that a wind of 5 m/s with an elevation angle
of 30 degrees will have a horizontal component of 5 x cos (30) = 4.33 m/s. If
the presumption is that the cup anemometer is providing the speed of the
horizontal component of the wind, and if the cup performs like a "total speed
sensor" in the range of ±50 degrees as the figure suggests, the 5 m/s the cup
reports is a 15% overestimation of the true horizontal speed of 4.33 m/s. A
propeller anemometer will report the horizontal component because it does have
a nearly cosine response. Operating side by side in this 30 degree wind, the
cup will report 5 m/s and the propeller will report 4.33 m/s and each will be
"right."
In addition to the horizontal component dilemma, the cup
anemometer tends to overestimate even the total wind. This is particularly
noticeable when the air is rising and flows past the support column creating a
wake which interferes with the normal cup aerodynamics. The figure shows this
effect to be about 10% at +30 degrees for "standard small cups." This 30
degree rising air example suggests that the side by side anemometers mentioned
above will really be reporting 5.5 m/s (10% off-axis error for the cup) and
4.33 m/s for the true horizontal speed from the propeller, or a 27%
overestimation of the horizontal component by the cup anemometer.
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 17 of 32
4.2.2.2 Wind Direction
4.2.2.2.1 Threshold
4.2.2.2.1.1 Definition
As with wind speed measurement, a key to a good wind vane
for air pollution applications is a low threshold. The threshold is the one
performance characteristic which will certainly change with time because of
bearing degradation. Most wind vanes use potentiometers to convert position to
output voltage. Potentiometers have bearings or bushings which will wear and
add to the starting threshold. There is not a standard definition for wind
vane threshold, although the ASTM Standard Test Method for DETERMINING THE
DYNAMIC PERFORMANCE OF A WIND VANE (ASTM, 1985b) offers the following
candidate. The 9 in the definition is the equilibrium direction of the vane
B
in a wind tunnel at about 10 m/s.
Starting threshold (S , m/s) is the lowest speed at
which a vane will turn to within 5° of 9 from an
B
initial displacement of 10 .
Even this definition runs into some problems in interpretation. If the vane
must move at least from 10° to 5° at the threshold speed, is the offset
sensitivity really 5° rather than 10 ?
The requirement in EPA (1987a) for PSD applications states
"Wind direction and wind speed systems should exhibit a starting threshold of
less than 0.5 meter per second (m/s) wind speed (at 10 degrees deflection for
direction vanes)." Does this mean that a vane that moves from a 10
displacement to 9.5° at 0.5 m/s has a starting threshold of 0.5 m/s? The newer
EPA (1987b) on-site guidance says
"Threshold (1) * 0.5 m/s
(1) as determined by wind tunnel tests conducted on production
samples in accordance with ASTM D-22.11 test methods."
The reason the ASTM committee required the vane to move from 10° to 5° was to
relate the starting threshold to accuracy. With wind speed, there is a way to
correct for threshold nonlinearity; for wind direction there is not. It seemed
best to establish the range of operating speeds to correspond to the range
where accuracy requirements are met. ASTM assumed 5° for wind direction as a
reasonable accuracy.
When torque measurements began their use as a measure of
starting threshold, the question became clearer. If the vane is required to
move to 5 there should be enough torque developed by the wind speed working on
the tail area exposed at 5 from the wind tunnel centerline or the true wind
direction to turn the shaft assembly and transducer. This sounds like a 5
threshold requirement, and perhaps that is a better description. As will be
shown later& there is a big difference between the torque developed at some
speed at 10 and the torque developed at the same speed at 5°. The nature of a
standard test method is less important than the application of a standard
method everyone uses and regulatory performance requirements consistent with
that test method. This handbook will use the 10° offset moving to 5° on
release in the wind tunnel (the ASTM method) as the criteria for starting
threshold. The relevant torque for this definition is that at 5°.
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 18 of 32
4.2.2.2.1.2 Threshold measurement
The measurement of starting threshold requires a wind
tunnel capable of accurate operation below 1 m/?. One standard methodology is
defined in ASTM (1985b) and described in Finke3stein (1981). Just as it is
with wind speed, it is possible to measure the torque which results from the
force of the wind on a wind vane as the torque measurement device holds the
vane at some angle from the wind tunnel centerline, say 10 . Lockhart (1978)
describes wind tunnel test data using two very different wind vane designs, the
front-damped Meteorology Research, Inc. (MRI) Model 1074 and the more
traditional Teledyne Geotech (TG) Model 53.2.
Another body of wind vane torque data exists as a result
of tests run by the R. M. Young Company (RMY). Their tests used a DC "torque
motor" as the transducer for vanes mounted in their wind tunnel. The torque
motor current was linearly correlated to torque measured with a series of
Waters Torque Watches. The torque motor drove the vane to each of four
positions, plus and minus five degrees and plus and minus ten degrees from the
wind tunnel centerline. A measurement of current was taken at each position
and at each of 12 wind speeds varying from 0.3 to 6 m/s, depending on the vane
design. Table 4.2.2.7 lists the average constant, K, which was found by a
linear regression of the motor current (torque) to the square of the wind
speed, with the intercept forced to zero. They tested all of their products
along with some vanes from other manufacturers.
Table 4.2.2.7 - K Values for Vanes at Two Angles to the Wind
Vane Type
Wind Sentry (RMY 03301)
Wind Monitor (RMY 05103)
Wind Monitor AQ (RMY 05305)
Propvane (RMY 08003)
Microvane (RMY 12302)
Bivane-19 cm fin (RMY 17003)
Anemometer Bivane (RMY 21003)
Propeller Vane-23 cm (RMY 35003)
Long Vane (Vaisala WAV 15)
Short Vane (Vaisala WAV 15)
Black Aluminum (Met One 024A)
High Damping Ratio (Met One024A)
F460 Vane (Climatronics 100075)
where: K = T/U2, E = Std. Err.
Offset Angle
5°
1
10
16
15
25
14
17
19
3
2
13
19
16
of
K
.8
.6
.8
.9
.0
.5
.1
.0
.6
.0
.8
.9
.0
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
Coeff.
E
006
080
126
061
414
188
141
127
047
015
181
194
322
and
0
10
O
K
3.
23.
37.
38.
57.
37.
45.
46.
7.
4.
28.
29.
r =
7
3
0
8
5
6
6
5
8
3
4
8
K(
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
10°
E
017
114
260
304
760
367
457
378
049
029
394
497
)/K(
r
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
1.
5°)
1
2
2
4
3
6
7
4
2
2
1
9
The ratio of the 10 degree K value to the 5 degree K value
seems to be lower for high aspect ratio vanes. A square vane has an aspect
ratio of 1. The Propvane, Microvane, Bivane and Anemometer Bivane are examples
of designs with an aspect ratio of 1. A rectangular vane which is two times as
high as it is long (along the tail boom) would have an aspect ratio of 2. The
Wind Monitors are examples of this design. The F460 vane has an aspect ratio
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 19 of 32
of 4 and a K ratio of 1.9. The "high aspect ratio" TG Model 53.2, whose torque
data (natural log of torque vs. natural log of displacement angle) are shown in
Figure 4.2.2.10, has an aspect ratio of 6 and has a K value ratio of 1.4 at
0.45 m/s. Differences in torque between 5 degrees and 10 degrees could not be
measured at 2.2 m/s (the K value ratio therefore equals one at that speed).
The High Damping Ratio (Met One) also has a high aspect ratio and also could
not provide a stable torque reading at 10 degrees. The MRI 1074 (aspect
ratio of 2) has a K ratio of 2.6. This design is more difficult to compare to
other vanes because of its front damping vane.
Wind Vane Torque Data - Two Vanes
1975 Data - (see Lockhart 1978)
a>
3
CT
i_
O
\—
c
5 -
4 -
KEY:
MR! Model 1074
TG Model 53.2
-500
-100
-50
cu
D
CT
w
O
h-
In Displacement Angle (deg.)
Figure 4.2.2.10 Torque measurements as a function of vane angle,
If the starting torque of the shaft of a direction vane bearing
and transducer assembly is to be interpreted in terms of wind speed, an
expression of torque as a function of speed is required. Each expression is
specific to the vane design and an offset angle. Take, for example, the Wind
Monitor AQ shown in Table 4.2.2.7. The expression for a 10 degree offset is
T = 37 U?
If a starting torque were found to be 5.9 g-cm, that
measurement can be expressed as a threshold wind speed of 0.4 m/s (0.9 mph). A
0.41 m/s wind at a 10 degree angle from the vane position will produce enough
torque to move the vane closer to the wind direction. The expression for this
wind vane for a 5 degree offset is
T = 16.8 U?
The same starting torque of 5.9 g-cm will require a wind speed of 0.6 m/s (1.3
mph) to move the vane closer than 5 degrees to the true wind direction.
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 20 of 32
Table 4.2.2.8 - Wind Vane Torque vs. Wind Speed and Angle
Wind
Speed
U
(m/s)
1.3
2.2
4.5
Offset Angle 9 (deg. )
5°
A
T T!
(g-cm) (g-cm)
72 69
216 217
O
10 A
T T2
(g-cm) (g-cm)
72 76
166 177
562 557
»\
T T3
(g-cm) (g-cm)
108 97
324 278
1152 1163
«\
T T
4
(g-cm) (g-cm)
144 150
432 430
T is the measured torque holding the vane at offset angle 9
K is the linear regression coefficient when a = 0
A
T is the predicted torque using K from the following:
A 1 , A . fi
Tj- 19.55 U1-0 T2= 50.22 U
A 2 A 2
T3» 57.47 U T4= 88.81 U
The data from Lockhart (1978) for the MRI Model 1074 are
shown in Table 4.2.2.8 to demonstrate the complexity of the dynamic performance
of some vane designs. A simple expression is useful to convert a torque
measurement to a wind speed. The simple vane designs listed in Table 4.2.2.7
fit a U2 expression quite well. The 5 degree and 10 degree data for the Model
1074 define a different slope than U on the log-log plot of Figure 4.2.2.11.
An expression of U1'6 fits the data well enough to use to extrapolate the
experimental data for this vane design to other wind speeds. The physical
reason for this unusual dynamic performance is probably related to the effect
of the front damping vane and the relatively large support column. The
vortices shed by the column only effect the rear vane.
The question remains, should the 5 degree K value be used
or the 10 degree K value? For the purpose of making a conservative estimate of
starting threshold for performance accurate to 5 degrees, the 5 degree K value
is recommended. The user should not expect this torque-defined threshold to
agree with the "starting threshold" published by manufacturers. Only after a
test is specified, like the ASTM test, can a 5 degree K value be expected to
agree with the data sheet values.
4.2.2.2.1.3 Torque measurement
Starting torque measurements of a wind vane may be made in
either of two general ways. If the vane can be removed, a torque watch can be
used to measure the starting torque of the bearing assembly and transducer (see
Figure 4.2.2.3). For this method to be most accurate, an equivalent weight of
the removed vane must be placed on the shaft to simulate the end loads of the
shaft of the bearings.
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 21 of 32
Wind Vane Torque Data - MRI 1074
1975 Data - (see Lockhart, 1978)
\
CD
0)
13
CT
i_
O
-1,000
E
u
0)
3
CT
- 100
- 50
0.6 0.8 1
In Wind Speed (m/s]
Figure 4.2.2.11 Torgue measurements as a function of wind speed.
If the vane cannot be removed or the choice is to not
remove it, the starting torque can be measured by imposing a force at a
measured radial distance from the axis of rotation. A spring-type gram scale
at 10 cm from the axis of rotation will yield g-cm after dividing by 10. On
some designs it is impossible to impose the force at 1 cm. In the interest of
accuracy, it is better to use a longer distance so the length part of the
measurement can easily be just a few percent. Of course the trade-off for
accurate distance is small force, an equally troublesome source of uncertainty.
Figure 4.2.2.12 shows different gram scales used on a Young Wind Monitor AQ.
Figure 4.2.2.12 Starting torque measurements on a wind vane.
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 22 of 32
If the vane is left on, the space used for the measurement
must be devoid of any air movement. Human breath provides a force which can
bias the measurement. It is also important that the axis of rotation be
vertical to negate any imbalance in the vane assembly.
For either method, the full 360 degrees of rotation of the
vane should be challenged with the highest torque found being reported as the
starting torque (worst case).
4.2.2.2.2 Accuracy
4.2.2.2.2.1 Accuracy definition
There is no transfer function for a wind vane comparable
to that for an anemometer, unless the conversion of shaft position to output
voltage is taken to be such a function. The vane is assumed to be accurately
placed, on average, downwind from the axis of rotation, when the wind is steady
and its speed is well above the threshold. If the vane is bent in some way, a
bias will be introduced (see Dynamic Vane Bias later in this section). This is
seldom large enough to be of concern.
The accuracy of the sensor is described by how well the
shaft position is reported by the transducer and signal conditioning circuit.
The accuracy of wind direction must include the accuracy with which the sensor
is sited with respect to TRUE NORTH. Any error in orientation will be a true
bias and can be removed from the data at any time the facts become known. The
"facts" in this case mean a rigorous quality control program which requires a
site log to indicate any servicing of the sensors. The "true bias" can change
if the sensor is removed and reinstalled without "as-found and as-left"
orientation measurements in the log book. Any possible undocumented change can
negate data correction for orientation.
The requirements for accuracy include EPA (1987a) which
states "Wind direction system errors should not exceed 5 degrees, including
sensor orientation errors." In EPA (1987b) it says
"Accuracy (error)(1) s3 degrees relative to the sensor
mount or index (s5 degrees
absolute error for installed system)
(1) as determined by wind tunnel tests conducted on production
samples in accordance with ASTM D-22.ll test methods."
The footnote is in error. There is nothing in the wind tunnel test which
relates to wind direction accuracy.
4.2.2.2.2.2 Measurement of sensor accuracy
The simple procedure for this measurement requires some
fixture which provides for steps in the direction vane shaft position of known
size. There are innumerable devices and methods for this procedure, many of
which will be described in the calibration section (4.2.5.2). One device which
can move the shaft in 60 degree increments is shown in Figures 4.2.2.13. The
important criteria are stability and knowing that the error band for the
fixture is on the order of 0.1 degrees of arc.
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 23 of 32
Figure 4.2.2.13 Wind direction calibration fixtures from Teledyne
Geotech (left) and Met. Standards Institute (right).
Typically, potentiometers used for wind direction will
have a linearity of about 0.5 percent, 1.8 degrees in 360 degrees. A table of
angles and output values will usually fall within a range between -1.8 degrees
and +1.8 degrees when the bias is removed by subtracting the average error from
each error. This statement is true when the open sector of the potentiometer
is ignored (for 360 degree mechanical and about 352 degree electrical systems,
see Section 8.2) or when errors in 540 degree format switching systems are not
considered. Other contributions to sensor error, such as hysteresis,
out-of-round and signal conditioning errors, when added to the linearity error
mentioned above should provide an error band not larger than -3 to +3 degrees
relative, or 6 degrees if the bias has not been removed.
An example of audit data shown in Table 4.2.2.9 describes
the performance of one wind vane when challenged with a 60 degree fixture. The
fixture settings and the displayed digital output ofthe system are listed. The
system had a 540 degree range and a 5 volt full scale output. The output is
converted to nominal voltage to show how the 540 range works. (Degrees per
volt = 540 / 5 = 108)
The average error of -3.4 was calculated without using
the obvious "open section" values marked by "*." When the fixture is installed
the vane substitute is set in the 180 location and then rotated until the
output is about 180. This need not be precise since the average error provides
a means of normalizing the data by removing the initial bias of approximate
setting. The linearity of the potentiometer-signal conditioner can be seen in
Figure 4.2.2.14. Except forothe "open sector" near 360°, the error is within a
± 3 band, including the 540 format switching error of about 1 .
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 24 of 32
Table 4.2.2.9 - Relative Wind Direction vs. Output Direction
Fixture
Setting
A
(deg. )
180
120
060
360
300
240
180
120
180
240
300
360
060
120
180
240
ccw
ccw
ccw
ccw
ccw
ccw
ccw
ccw
cw
cw
cw
cw
cw
cw
cw
cw
cw
System Nominal
Output Voltage
B B/108
(deg.) (volts)
177
114
054
001
298
238
176
115
177
239
292
001
056
115
177
239
average
1.
1.
0.
0.
2.
2.
1.
1.
1.
2.
2.
3.
3.
4.
4.
2.
639
056
500
009
759
204
630
065
639
213
704
343
852
398
972
222
error C
is clockwise
, ccw
Error
E=B-A
(deg. )
-3.
-6.
-6.
1.
-2.
-2.
-4.
-5.
-3.
-I.
-2.
1.
-4.
-5.
-3.
•-1.
= -3.
0
0
0
0*
0
0
0
0
0
0
0
0*
0
0
0
0
4
Normalized
Error Output
E-C B-C
(deg. ) (deg. )
0.
-2.
-2.
4.
1.
1.
-0.
-1.
0.
2.
1.
4.
-0.
-1.
0.
2.
4
6
6
4*
4
4
6
6
4
4
4
4*
6
6
4
4
180.
117.
057.
005.
301.
241.
179.
118.
180.
242.
301.
005.
059.
118.
180.
242.
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
* values excluded)
is counterclockwise
Relative Wind Direction Accuracy
Actual 16-Point Audit Results - 540 Degree Format
4 -
& 3
a>
-o
2 -
o
1 -
LJ
TD
CD
N
— 0
o
E
o -1 -
-2 -
-3
Number indicates point sequence
1 -8 counterclockwise
8-16 clockwise
£4,12
t- 10,16
+6
13
060
300
120 180 240
Fixture Position (deg.)
Figure 4.2.2.14 Results of a wind vane audit using 60° steps.
360
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 25 of 32
An analysis of this type helps to optimize the accuracy of
the orientation. olf an orientation target is at 120° TRUE, when the vane is
pointing from 120° the output should read about 118 . This effectively centers
the error band (see 4.2.4.3.2 on orientation).
4.2.2.2.2.3 Measurement of orientation accuracy
Orientation error is an important part of the measurement
error, but it cannot be considered until the sensor is installed in the field.
The accuracy of the orientation includes the accuracy in finding TRUE NORTH and
the accuracy with which the vane is aligned to TRUE NORTH. Use different
methods for finding TRUE NORTH. Methodology for orientation is given in
4.2.4.3.2.
4.2.2.2.2.4 Expression of Accuracy
An accuracy specification should include enough
information to define the type of accuracy intended and the method by which
accuracy claims may be tested. There is no requirement for traceability to NBS
for wind direction. The measurement of relative direction is a fundamental
division of a circle. The measurement system can be bench tested by basic
methods yielding a clear expression, of the errors associated with a position
angle vs. system output transfer function. The errors are mostly conditional
biases which are small enough (less than one percent of 360 degrees) to ignore.
The orientation error is a pure bias which cannot be bench
tested. The total error, a simple sum of the two parts (root-sum- square, RSS,
combination is only legitimate with random errors, not biases), can only be
found after installation.
4.2.2.2.2.5 Precision
The definition of accuracy describes a bias term and a
variable term akin to precision. A comparability test (see 4.2.2.1.2.4) will
show that two vanes properly sited and perfectly oriented will report the 20
minute scalar average directions with a difference of less than two degrees,
i.e. precision is ± 2°.
An expression of accuracy for a wind vane operating on a
10 m tower in the atmosphere can be expressed as the relative accuracy plus
orientation accuracy and ± 2° for precision. For a collocated test (Lockhart,
1988), the orientation error can be estimated by the average difference between
the subject wind vane and a collocated wind vane perfectly oriented. If the
orientation error is found to be large, and if a quality control system has
provided records of maintenance showing the orientation has not been changed, a
bias correction can be applied^ The accuracy of the data corrected for bias is
then the relative accuracy ± 2 .
4.2.2.2.3 Delay Distance (Distance Constant)
4.2.2.2.3.1 Definition of delay distance
ASTM (1985b) defines delay distance (D) as the distance
the air flows past a wind vane during the time it takes the vane to return to
50 percent of the initial displacement. The value for this sensor
specification is found in wind tunnel tests, as described in Finkelstein
(1981). The initial displacement is 10 degrees and D is the average of a
series of tests at 5 m/s and 10 m/s using displacements on both sides of the
tunnel centerlirte.
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 26 of 32
The specification in EPA (1987a, PSD) says "...the
distance constant should not exceed 5 m." In EPA (1987b, On-Site) the
specification says
"Delay Distance (1) s5 m at 1.2 kg/m3 (standard
sea-level density)
(1) as determined by wind tunnel tests conducted on production
samples in accordance with ASTM D-22.11 test methods."
4.2.2.2.3.2 Measurement of delay distance
Measurement requires a wind tunnel of reasonable size and
quality. The width of the tunnel should be at least three quarters of the
overall length of the wind vane to be tested. With the small displacement
angle of 10 degrees (about 3 percent of full scale), it is hard to conduct this
test in the open atmosphere.
This specification is strictly a sensor dynamic
performance specification. Any time constants in the signal conditioning
circuits will dampen the apparent sensor response and make D larger than it is
for the sensor. One could argue that it is ONLY the combination of D and the
time constant of the signal conditioner that should be considered in meeting
the regulatory requirements for performance. The 5 m maximum for D is roughly
equivalent to a time constant of 0.5 seconds at 10 m/s wind speed and aim
vane. For this and other reasons it is best to keep the time constant of the
signal conditioning circuits to 100 ms or less. For the same reason, it is
necessary to use high speed recording equipment for the wind tunnel tests. At
10 m/s, aim vane reaches the 50 percent D value in 100 ms. If one wants
resolution to find D to 10 percent of the true value (0.1 m), a 5 ms resolution
in the data is desirable.
4.2.2.2.4 Overshoot or Damping Ratio
4.2.2.2.4.1 Definition of overshoot or damping ratio
ASTM (1985b) defines Overshoot (£5) as the ratio of the
amplitudes of two successive deflections of a wind vane as it oscillates about
9 after release from the offset position, as expressed by the equation
D
9
n
where 9 and 9 are the amplitudes of the n and (n+1) deflections,
n (n+1 )
respectively.
The Damping Ratio (TJ) may be calculated approximately from the overshoot ratio
by the formula
in'"1
T, 3
*2 * H-HI
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 27 of 32
The specification in EPA (1987a, PSD) says "The damping
ratio of the wind vane should be between 0.4 and 0.65..." In EPA (1987b,
On-Site) the specification says
"Damping Ratio (1) so.4 at 1.2 kg/m or
Overshoot (1) <25°/. at 1.2 kg/m3
(1) as determined by wind tunnel tests conducted on
production samples in accordance with ASTM D-22.ll test
methods,"
The subject of dynamic wind vane performance is thoroughly
discussed in MacCready and Jex (1964), Gill (1967), Weiringa (1967) and Acheson
(1970).
4.2.2.2.4.2 Measurement of overshoot.
The measurement of overshoot also requires a good wind
tunnel and sensitive, fast response recording systems. A series of tests were
conducted by Lockhart in 1986 in pursuit of a wind vane design with a 0.6
damping ratio. A sketch of the results of this unpublished work is shown in
Figure 4.2.2.15 to provide an example of how various vane designs compare in
overshoot and delay distance. One of the requirements in the ASTM method is an
initial offset of 10 deg.
4.2.2.2.5 Dynamic Vane Bias
The Dynamic Vane Bias (9 , deg.) is the displacement of
B
the vane from the wind tunnel centerline at 5 m/s. This measurement will
identify wind vanes with unbalanced aerodynamic response because of damage
(bent tail) or design. This is a screening specification not needed or used in
any application requirements. The ASTM method measures this difference, if
any, and disqualifies the vane if the difference is greater than one degree.
4.2.2.3 TURBULENCE
4.2.2.3.1 Definition
The Glossary of Meteorology (Huschke, 1970) quotes Sutton
(1955) defining turbulence as a state of fluid flow "in which the instantaneous
velocities exhibit irregular and apparently random fluctuations so that in
practice only statistical properties can be recognized and subjected to
analysis. The situation is, in fact, analogous to that accepted unreservedly
in the field of molecular physics..." The definition is ended with a quote
from the Bible"
The wind bloweth where it listeth and thou hearest
the sound thereof but canst not tell whence it cometh
and whither it goeth:... John 3:8
From the standpoint of wind measurement, then, turbulence is
not measured, it is calculated. From the standpoint of quality assurance,
turbulence is a difficult subject to control. It is possible to define the
measurement samples from which the statistical properties are calculated. It
is possible to define the algorithm by which the samples are summarized. The
relationship between the algorithm and the application or model is also
important, but it is beyond the scope of this handbook.
-------
t
10
10
1
10'
1
10'
Delay Distance (50% recovery)
Thies (experimental)
697 grams
Thies
600 grams
\
£
Climatronics F460
251 grams
MSI (experimental)
191 grains
Climet
170 grams
Vaisala
92 grams
MSI (experimental)
72 grams
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 28 of 32
012 5
Distance (a) at 5 m/s Meteorological Standards
Institute - March 5. 1986
Figure 4.2.2.15 A sample of the dynamic response of some wind vanes
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 29 of 32
4.2.2.3.2 Direction Measurement for Sigma Theta
The most common turbulence property routinely reported is sigma
theta, the standard deviation of a series of horizontal wind direction samples.
Most, but not all, of the following will also apply to sigma phi, the standard
deviation of a series of vertical wind direction samples. Among the
specifications which are important to the direction measurement used to
calculate sigma theta is delay distance which limits at the small end of the
eddy size spectrum the eddy sizes to which the vane can react. If the vane has
a delay distance of 5 m, it will not detect energy from eddys smaller than 5 m
because the vane cannot react to them. If 1 m eddy sizes are important to the
diffusion being estimated, use a wind vane having a delay distance of 1 m or
less.
Another important specification is overshoot or damping ratio.
Vanes will overshoot when correcting for a direction change. If the overshoot
ratio is 0.5 (or 50%), more variability will be reported from the same
turbulent flow than is reported by a vane with an overshoot ratio of 0.25 (or
25%). The relationship between overshoot ratio and damping ratio is given in
Table 4.2.2.10 as calculated by the equation found in 4.2.2.2.4.1.
Table 4.2.2.10 - Overshoot Ratio vs. Damping Ratio
Overshoot
Ratio
1.00
0.90
0.80
0.70
0.60
0.50
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
Percent
100
90
80
70
60
50
40
35
30
25
20
15
10
5
! denotes PSD
Damping
Ratio
0.00
0.03
0.07
0.11
0.16
0.22
0.28
0.32
0.36
0.40
0.46
0.52
0.59
0.69
range
Sigma calculations are biased by any averaging built into the
signal conditioner. They are also subject to error if external noise gets into
the output, a dilemma for circuit designers. A compromise might be to filter
out any noise at frequencies higher than 20 Hz (0.5 m at 10 m/s). In winds
above 20 m/s, this filter would degrade data from a wind vane having a delay
distance of 1 m. Turbulence from mounting structures upwind of the
vane will bias the sigma value. Out-of-balance conditions with a vane
measuring sigma phi will also bias the statistic, particularly at the low wind
speeds. Un-filtered noise from potentiometers will add an error to the natural
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 30 of 32
variability of the wind. These are problems which are best detected by
inspection of strip charts or oscilloscope traces.
4.2.2.3.3 Statistical Summaries
A few basic concepts will help in considering the specifications of the
statistical algorithm used and the representativeness of the value calculated.
Here again, careful definition will help understand what the circuits and
logical networks are doing to the input samples. EPA (1987b) devotes 44 pages
to Meteorological Data Processing Methods.
Representativeness is the important concept to keep in mind
when examining strange or unusual data. Samples of wind direction taken over a
short period of time (seconds to a minute or two) are likely to exhibit nearly
normal or Gaussian distribution. As the time gets longer (a few minutes to an
hour or more), physical dynamics driving the flow in the surface layer may
provide different shapes. The most common of 'these might be the bi-modal
distribution resulting from land-water, mountain-valley, day-night or
meso-scale convective flow systems. Whatever the driving forces, a bi-modal
distribution cannot be usefully represented by a mean and standard deviation.
This is to say that a data sampling and processing system may work perfectly
and produce numbers which have no physical meaning. From a specification
standpoint, tests for "working perfectly" are possible and should be used.
The method used by the wind direction system to describe the
position of the vane in the series to be statistically described must be
thoroughly described and understood. The most common error in the past,
perhaps even made today, is to do nothing. If the output voltage unambiguously
represents an azimuth angle, and if samples of voltage are described with the
statistical parameters of mean and standard deviation, and then expressed in
units of azimuth angle, great errors will result. These errors are a result of
a discontinuous range of output voltage. If 001-360 degrees are represented by
0-1 volt, samples clustered around 360 will contain some near 0 and some near
1. The mean of 0.5 will be 180 degrees away from the mode.
When analog ink recorders were used exclusively with 360 degree
formats, it was common to see the paper painted red by the pen going back and
forth through full scale, effectively obliterating any data. There are several
ways to avoid or minimize the "crossover" problem. System specification should
define how this will be done. The most common method for minimizing this error
is to use a "540 degree" format. Systems were designed with dual
potentiometers or dual wipers 180 degrees out of phase. When the wiper moved
into the gap, circuit switching would change to the center of the other circle.
This switching would be invisible in the output at the 1/3 and 2/3 scale
points, but when the voltage went beyond full scale, it would switch to 1/3
scale and when the voltage went to zero, it would switch to 2/3 scale. This
format completely eliminated the pen painting problem and drastically reduced
the output voltage switching, but some large pulses remained to occasionally
bias sigma calculations.
With the advent of microprocessors and digital computers it
became possible to combine the samples without any large pulses. One method
uses a unit vector sum to find the resultant vector direction (average
direction). With an assumed wind speed of 1 m/s, each sample of wind direction
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 31 of 32
is converted from polar coordinates (001-360 degrees) to cartesian coordinates
(N-S and E-W components in meters). The components are added or subtracted over
the sample period and the resultant vector direction is found by converting the
final coordinate sums to polar coordinates through an arc-tangent calculation.
The standard deviation of each sample about the resultant vector direction is a
straight forward process on the differences of the samples from the "average."
Of course, this same method can be used with the true speed-weighted wind
vector samples.
Some automatic systems currently available pick an assumed
direction, usually the mean of the last period, take the digital difference in
degrees (limited to 180) of each sample from the assumed mean, and find the
standard deviation and the mean difference for the period. The standard
deviation about the mean is the same as the standard deviation about the mean
plus a constant. The mean difference plus the assumed direction is the true
mean (limited to 360 degrees).
There are also other algorithms which estimate the standard
deviation (see Turner, 1986). It is only necessary from the standpoint of
quality assurance to know that the method used is being satisfied with the
samples taken from the measurement system.
The sample size is specified in EPA (1987h) as 360 samples to
estimate the standard deviation to within 5 or 10%. Lockhart (1988) found an
apparent bias, not a random error, when the standard deviation was estimated
from 120 samples over a 20 minute period.
Most models accept data representing one hour. Sigma theta for
60 minutes is influenced by the changing wind direction during the hour. It is
recommended (EPA, 1987b) that four 15-minute sigma theta calculations be
combined to provide a "one hour" value for the purpose of selecting a
Pasquill-Gifford stability class. The method is
-------
Section No. 4.2.2
Revision No. 0
Date: 17 Sep 89
Page: 32 of 32
is excluded from this value but included in the true standard deviation of the
direction about the hourly mean. If the standard deviation about the hourly
mean is required for a concentration distribution analysis, the correct
formulation for tr is shown below.
A(l-hr)
A(1-hr)
n (T + n
-------
Section No. 4.2.3
Revision No. 0
Date: 17 Sep 89
Page: 1 of 5
4.2.3 ACCEPTANCE TESTING
The procurement document, purchase order or contract, should be specific at
least in terms of required performance specifications. "Required" in this
context may only mean that the instrument meets the suggested or specified
regulatory performance. It is another question, beyond the scope of this
handbook, whether "necessary" relates to the application to which the data will
be used.
There are two kinds of performance specifications, those which can be
verified by simple inspection testing and those which require unusual test
equipment and experience. The former should be tested and the latter certified
by the manufacturer. The manufacturer should have either performed these tests
on one or more samples of the model design or arranged for such tests to have
been conducted by some calibration facility. In either case, a test report
should be available to any who require the documentation. The cost of the copy
of the test report should not be much larger than the cost of normal copying.
If the manufacturer does not provide such documentation, the choice is between
accepting the manufacturer's unsubstantiated claim or having a specific test
run.
A good QA Plan will provide for a QA sign-off of the procurement document
in order to assure that equipment capable of the required performance is being
purchased and that the capability can be verified. Purchasing by brand name is
often expedient where the performance of a model has been verified and all that
is required is more units. This practice is also cost effective when
considering spare parts and instrument technician training costs.
The parts of wind speed and wind direction sensors which predictably
deteriorate and seriously influence the performance of the sensors are the
bearings. It is acknowledged that an experienced inspector can "measure"
bearing condition by feeling or spinning the shaft. The receiving inspection
is a protection against putting defective equipment into the field. It is not
a necessary link in the documentation trail for data validity purposes. True
torque measurements for data validity will be most valuable at the initial
field calibration. True laboratory conditions may be chosen, however, because
torque measurements are sufficiently difficult to make.Therefore, a receiving
inspection may be used for this purpose.
4.2.3.1 Wind Speed
An example is found in Table 4.2.3.1. This performance specification
for an anemometer is hypothetical but one which will meet the requirements of
EPA (1987a, PSD). Each attribute of the instrument is identified by a key as
to whether it is a receiving test candidate of not, and the nature of the
testing is briefly discussed below. Each instrument includes a sensor, signal
conditioner, and recorder. When an attribute of the sensor is affected by the
signal conditioner or recorder, a keyed comment will be made.
-------
Section No. 4.2.3
Revision No. 0
Date: 17 Sep 89
Page: 2 of 5
Table 4.2.3.1 - Anemometer Performance Specification
Range
Threshold1>2(t)
Accuracy (error)
0.5 to 50 m/s
0.5 m\s
1.2,
(0.2 m/s
Distance Constant (t)
5% of observed)
3
5 m at 1.2 kg/m (standard
sea level density)
(t) as determined by wind tunnel tests conducted
on production samples in accordance with
ASTM D-22.11 test methods (ASTM, 1985a).
(*) aerodynamic shape (cup or propeller) with
serial number to be accompanied by test report,
traceable to NBS, showing rate of rotation vs.
wind speed at 10 speeds with 0.1 m/s resolution.
1 subject to receiving inspection
2 transducer with signal conditioner
4.2.3.1.1 Threshold
The threshold receiving test should examine the system output
with the anemometer not turning (below sensor threshold) and with the
anemometer turning at an equivalent 0.5 m/s. If the cup is a Climatronics F460
Vinyl Cup Set (100083) or Heavy Duty Cup Set (101287), the published constant
for mph and 30 pulses per revolution is 9.511 which converts to 1.41 meters per
revolution. At 0.5 m/s the cups should be turning at
0.5
1.41 = 0.35 rps,
or one revolution in 2.8 s. The cup assembly can be turned by hand to
approximate that rate of rotation. If the anemometer is a Young propeller
(08234, 18 cm polypropylene), the turning factor is 0.294 meters per
revolution. At 0.5 m/s the propeller should be turning at
0.5 * 0.294 =1.70 rps,
or two revolutions in 1.2 s. This also can be approximated by turning the
propeller by hand.
The key measurement for threshold, however, is starting torque.
This requires knowledge of the K value (see 2.1.1.2) which should be available
from the manufacturer.
4.2.3.1.2 Accuracy
The receiving test for accuracy is the conversion of rate of
rotation to output in units of wind speed. The transfer function, supplied by
the manufacturer, should be in terms of rate of rotation (rps) vs. wind speed
(mps). The receiving inspector simply turns the anemometer shaft at a few
known rates of rotation to see if the systemoutput compared to the predicted
output is within" the tolerance specification.
-------
Section No. 4.2.3
Revision No. 0
Date: 17 Sep 89
Page: 3 of 5
4.2.3.1.3 Distance constant
The distance constant determination requires a special wind
tunnel test and is beyond normal receiving inspection capability. The time
constant of the anemometer circuitry will influence the effective system
performance. Assume the manufacturer's value for distance constant of the
sensor is one meter. At a wind speed of 10 m/s the sensor will have a time
constant of 0.1 s. Assume the time constant of the system electronics is 2 s.
Then, at a wind speed of 10 m/s, the system time constant is 2 s or 20 m. In
this example, the system electronics would need a time constant of 0.1 s or
less if the response capability of the sensor is to be fully available. At a
wind speed of 0.5 m/s the sensor responds in 2 s, the same as the system
electronics. In this example, the system electronics dominates the sensor
response at all speeds. If the sensor response is to be available at all
speeds (up to 25 m/s), the electronics time constant must be 0.04 s or less.
The time constant can be measured at the receiving test by timing how long it
takes for the output to reach 63.2% (1-1/e) of a step change in speed. The
step change can made by turning the anemometer shaft at a known rate of
rotation and then instantaneously stopping its rotation.
4.2.3.2 Wind Direction
An example of a wind direction specification is found in Table 4.2.3.2
(see EPA, 1987b). This performance specification is also hypothetical but it
is one which will meet the requirements of EPA (1987a, PSD). Each instrument
includes a sensor, signal conditioner, and recorder. When an attribute of the
sensor is affected by the signal conditioner or recorder, a keyed comment will
be made.
Table 4.2.3.1 - Wind Vane Performance Specification
Range
Threshold1>2(t)
Accuracy (error) ' (t)
Delay Distance (t)
Damping Ratio (t;
Overshoot2(t)
001 to 360 degrees or
001 to 540 degrees
a 0.5 m\s
£ 3 degrees relative to the
sensor mount or index -
s 5 degrees relative to
TRUE NORTH
3
s 5 m at 1.2 kg/m (standard
sea level density)
a 0.4 at 1.2 kg/m3
* 25% at 1.2 kg/m3
(t) as determined by wind tunnel tests conducted on
production samples in accordance with
ASTM D-22.11 test methods (ASTM, 1985a).
1 subject to receiving inspection
2 transducer with signal conditioner
-------
Section No. 4.2.3
Revision No. 0
Date: 17 Sep 89
Page: 4 of 5
4.2.3.2.1 Threshold
The threshold receiving test is a starting torque measurement
(see 4.2.2.2.1). To relate the torque measured to wind speed and off-set
angle, a K value is required, either from the manufacturer or from an
independent test. The torque measurement may be made with the vane assembly
removed or with the vane assembly in place. If the latter is chosen,
verticality is essential to negate any out-of-balance in the vane assembly from
biasing the test. Also there must be no air motion. Very small air motions
will bias the test. Use a smoke puff to be sure the air is still and refrain
from breathing in the direction of the vane surface.
4.2.3.2.2 Accuracy
The receiving inspection is the best place to. establish the
true non-linearity, if any, of the direction vane transducer. A test using
some circle dividing fixture capable of fine resolution, 1 deg. for example,
will provide a record which can be referenced in future field spot checks.
Without such a test it is hard to prove accuracy of s 3 deg. If several units
show the same pattern of non-linearity, it should be acceptable to sample
future units and accept a generic shape of the error. When a long series of
samples is planned, there is a tendency to devise methods which are quick. The
time constant of the signal conditioning circuit must be known to establish the
minimum time between position change and output reading or recording. If the
step is small, like 1 deg., three time constants will deliver 95% of 1 deg.,
which is good enough. If the step is large, like 180 deg., three time
constants will deliver 95% of 180 or 171 deg. which is not good enough. It
takes seven time constants to deliver 99.9% or 179.8 deg.
The receiving inspection cannot include the orientation error.
The manufacturer does not deliver orientation. There may be orientation
fixtures, however, which assume that an optical centerline is parallel to the
line set by an orientation pin. This assumption can be tested. Field
orientation may be based on the orientation of a crossarm with the assumption
that the output angle when the vane is parallel to the crossarm is known. This
assumption can be tested or the alignment fixture set in laboratory conditions
to the desired output.
4.2.3.2.3 Delay distance and overshoot
These dynamic characteristics require a special wind tunnel
test and their determination is beyond normal receiving inspection capability.
The time constant of the wind direction circuitry can influence the system
performance as it can with wind speed. Assume the manufacturer's value for
delay distance is one meter and the time constant of the electronics is 2 s.
At wind speeds of 10 m/s the time constant of the wind vane is 0.1 s and at 0.5
m/s it is 2 s, the same as the system electronics. For this example,
therefore, at all speeds above 0.5 m/s the performance of the wind vane is
being limited by the time constant of the electronics. The time constant can be
measured at the receiving test by timing how long it takes for the output to
reach 63.2% (1-1/e) of a step change in direction. For example, the step
change can made by quickly turning the vane from 000 to 180 deg. The time
constant is the "time required for the system output to change from 000 deg. to
113.8 deg.
-------
Section No. 4.2.3
Revision No. 0
Date: 17 Sep 89
Page: 5 of 5
4.2.3.3 Measurement System
All the elements of a system of signal conditioners, recorders and
monitors will require checking for correct function. The receiving inspection
should include testing these various sub-systems. There may be a calibration
switch which replaces the sensors with simulated conditions. Assume a system
has a calibration switch which substitutes the equivalent of 25 m/s and 180
deg. to the input of the signal conditioning boards. In 4.2.3.1 and 4.2.3.2
the sensors were providing the input values. It is possible for these tests to
show perfect results and yet the outputs with the calibration switches on could
show the system to be out of calibration. The reason would be that the
adjustments for the substitute calibration inputs were off.
After the calibration inputs have been adjusted and the "output" shows
the system to be in calibration, a parallel analog recorder may show incorrect
values. This event could be caused by an incorrect adjustment in the interface
which drives the analog recorders from the output. So far the "output" is
assumed to mean the voltage which goes into the data logger and becomes the
archived data. There may be monitoring meters or digital displays on the
system panel. These monitoring meters may differ from the "output" because
they have individual adjustments. All the sub-systems should tell the same
story and the receiving inspection should verify that they do. In fact, it is
rare when a system arrives in receiving with the various outputs in
disagreement, but they must be checked.
It goes without saying that the receiving function records the model
numbers and serial numbers of the component parts and checks the parts received
against the purchase document and the shipping document.
-------
Section No. 4.2.4
Revision No. 0
Date: 17 Sep 89
Page: 1 of 8
4.2.4 INSTALLATION
4.2.4.1 General Considerations
From a QA point of view, there are aspects of the installation which
should be considered. Perhaps the most important of these is siting. See
4.0.4.4 for general siting criteria and discussion. From a QA point of view,
however, failure to meet the recommended siting criteria may be necessary. If
the general site is selected for other measurements for good reason, the wind
sensor siting, may be only a best compromise. There are considerations which
set the options for the compromise. Among these are technical and budgetary
considerations. The qualitative judgments which go into siting are briefly
discussed here.
If buildings or trees are likely to interfere with the wind speed or
direction sensing, try to locate the tower or pole such that the wind sensors
will most faithfully record speed and direction for the direction of primary
concern, e.g. for directions that would take an effluent toward a residential
area.
Another important technical consideration is accessibility of the
sensors. There should be no hesitation in-taking a hands-on look at the
sensors whenever a performance question arises. Yet this is the most difficult
task at most sites. Some sites require special "climbers" to retrieve a sensor
and wait to return it to its installed position. These field people may not be
trained to handle delicate instruments. It may be so difficult or expensive to
get a sensor down that suspected bad data will be accepted rather than facing
the problem. This reluctance causes mounting hardware to become corroded to
the point that the sensor cannot be removed without damage. Most of the
breakage of delicate sensor parts results from handling while climbing on a
tower. If the direction sensor does not have an alignment fixture, it may not
be possible to remove the sensor without going through the orientation
procedure at re-installation.
There are several ways to overcome most of these problems. They all require
design forethought in installation. First, the sensors need to be easily
removed from the mounting structure. They need to be as easily connected to
the rest of the measurement system when they are down, either with the same
cable or a suitable substitute cable. One popular cup and vane design uses a
crossarm which contains non-removable cabling. Either the whole crossarm
assembly must be taken down with the sensors or there has to be a substitute
crossarm to plug the sensors into at ground level. Some towers, the short 10 m
types, can be tilted to access the sensors while still connected. In this case
the sensors are about 90 deg. from their operating position. Some tests
require the sensors to be vertical. In these cases the sensor still needs to
be removed and re-connected. Ideally, the crossarm is left exactly as is so
when the sensor is replaced, nothing physical has changed with respect to
verticality or orientation. Some towers telescope for access to the top where
wind sensors usually are mounted. This makes access easier but removal and
reconnection is still necessary. Some towers have elevators which transport
the sensors up and down the tower. How they deal with cables depends on the
elevator design.
v
An ideal installation is one where the operator can get hands on the
sensor, perform a test adjacent to the electronics and recorder, and re-install
-------
Section No. 4.2.4
Revision No. 0
Date: 17 Sep 89
Page: 2 of 8
the sensor, safely, alone and within one hour. There are no technological
reasons why this cannot be done, except for tall towers where an elevator may
take 20 minutes for a one-way trip. There probably is not a budgetary reason
for avoiding something like this ideal installation, once the cost of invalid
data and true operating costs are factored into the formula. It is usually not
done simply because the need for service is overlooked and the method of access
is not pre-planned.
4.2.4.2 Wind Speed
The wind speed sensor is most susceptible to error from shadowing and
interference. Aside from the need to have the anemometer properly exposed, the
only other consideration is verticality (for cup anemometers). If the cup
wheel is well balanced, a small angle (1 deg. or less) in mounting is not
important. If the cup wheel is not well balanced, the starti-ng threshold will
be degraded.
4.2.4.3 Wind Direction
4.2.4.3.1 Exposure
The problem with verticality for the direction vane is just the
same as with wind speed. For a well balanced vane assembly, a small angle from
vertical is not important. If the vane assembly is not well balanced, the
starting threshold is raised and a predominant direction for light winds may
not have any basis in fact.
Wind vanes are often used for dispersion applications by
calculating the standard deviation of the wind direction about the mean
direction, sigma theta. Unless the wind vane is at the tower top, there will
be some direction where the wind goes through and around the tower before it
gets to the vane. The farther from the tower the wind vane is mounted the
smaller the sector with tower interference. The interference sector can be
selected by placing the vane on the appropriate side of the tower.
4.2.4.3.2 Orientation
Of all the sources of error for a wind direction measurement,
the orientation of the vane to TRUE NORTH has the potential and reputation of
being the largest. A bad orientation provides a fixed bias to the data which
can be removed. If the vane is moved and the constancy of the bad orientation
is in question, the data may not be recoverable. The method of wind vane
orientation must be capable of 1 deg. accuracy with 2 or 3 degrees as the upper
limit of the error. Two steps are necessary to achieve an oriented wind vane.
First, the location of TRUE NORTH must be found to an accuracy of less than 1
deg. Secondly, the wind vane "index" must be aimed at that location with an
accuracy of better than 2 deg. (see 4.2.2.2.2.3 for a discussion of the
location of the normalized error "index").
TRUE NORTH as distinguished from magnetic north is usually
found by reading a magnetic compass and applying the correction for magnetic
declination. The declination can be read from a USGS map. The Fox Island
Station declination, according to the 1959 (revised in 1981) map, is 20.5°.
The USGS is now providing a computer service called GEOMAG. See below:
-------
Section No. 4.2.4
Revision No. 0
Date: 17 Sep 89
Page: 3 of 8
CONNECT
Unauthorized use of this U.S. Government computer system
is punishable under PL98-473
Welcome to the USGS Branch of Global Seismology and Geomagnetism
On-line Information System
Type Q for Quick Epicenter Determinations (QETJ)
H for Historical Epicenter File -Searches (EIS)
M for Geomagnetic Field Values
Enter program option (Q, H or M): m
GEOMAG
The International Geomagnetic Reference Field (IGRF) was revised in August,
1987. The models for 1945.0, 1950.0, 1955.0, and 1960.0 have been superseded
by new definitive models (see, for example, EOS Transactions, American
Geophysical Union, vol. 69, no. 17, April 26, 1988, pages 557-558). The new
models were installed on June 21, 1988. Please note that the revision affects
field values for dates between 1945.0 and 1965.0, but not those for later
dates.
Problems or suggestions? Please contact Norman Peddle, U.S. Geological Survey,
MS-968, Federal Center, Box 25046, Denver, CO 80225. Telephone: (303) 236-1364
(FTS 776-1364).
Press RETURN to continue:
Do you want information about this program (Y/N)? [ ] N
Options: 1) Field Values (D, I, H, X; Y, Z, F)
2) Magnetic Pole Positions
3) Dipole Axis and Magnitude
4) Magnetic Center [1] { ]
Display values twice (Y/N)? [Y] [ ]
Name of field model: [ ] ?
The following field models are available:
Name
IGRF85
USCON85
USALA85
USHAW85
Name of field model:
Date:
Latitude:
North or South (N/S):
Longitude:
East or West (E/W):
Elevation:
Type
Spherical Harmonic
Spherical Harmonic
Spherical Harmonic
Spherical Harmonic
Date range
1945.0
1985.0
1985.0
1985.0
[ 1
[1/25/89] [
[
[ 1 N
Region
1990.0 World
1990.0 48-States
1990.0 Alaska
1990.0 Hawaii
USCON85
]47{.25
1122.6292
{ ] W
[0.0]
[250 feet
Model: USCON85
Date : 1/25/89
Latitude : 47.25 N
Longitude: 122.6292 W
D
deg min
19 47.9
19 47.9
I
deg min
69 29.0
69 29.0
H
nT
19526
19526
X
nT
18372
18372
Annual change:
0 -5.5 0 -0.7 -1.9
0 -5.5 Q -0.7 -1.9
Elevation: 250.000 ft
Y Z F
nT nT nT
6613 52181 55715
6613 52181 55715
8.7 -29.8 -39.6 -37.7
-------
Section No. 4.2.4
Revision No. 0
Date: 17 Sep 89
Page: 4 of 8
The GEOMAG program was accessed by calling 1-800-358-2663
through the modem of an "XT clone" using a "ONE TO ONE" communications program.
Using the capture feature of ONE TO ONE, the following communication was
recorded. Note that the Fox Island Station declination for 1/25/89 is 19.8° or
0.7° less than the map indicated. This is consistent with the 0 deg. -5.5 min.
annual change for the roughly eight years since the map was revised.
The other way to find the direction to TRUE NORTH employs some
astronomical observation. While the compass method is clearly easiest, it is
also the most prone to error. Good training and equipment will reduce these
errors to an acceptable level, but not the "less then 1 deg." advised above.
Training will minimize errors from the influence of nearby metal objects and
the mis-application of the declination correction, but local variation in the
isogonic field is unknown. On the other hand, the observation of astronomic
bodies can be unambiguous. Polaris, the north star, will provide TRUE NORTH to
within 1 deg. (without correction) on any clear night. The true solar noon
method will provide the north-south direction to within 0.1 degree on any clear
day, given the station longitude, date and an accurate clock. A simple Basic
program will provide the azimuth angle to the sun at any time of day given the
station longitude, latitude and date. Examples of the two solar methods are
given below.
4J2.4.3.2.1 True Solar Noon Method
The True Solar Noon (TSN) method finds the time at some
particular date at some particular longitude when the sun is in the north-south
plane passing through the North Pole, the South Pole and the longitude
selected. If the sun is not directly overhead (elevation 90 deg.) the azimuth
line to the sun is TRUE SOUTH or TRUE NORTH. Two calculations are required.
First, find the time of the Local Apparent Noon (LAN) from the longitude. The
examples shown here are for:
Fox Island, WA (Long. 122.6292, Lat. 47.2500), 07/04/90 and
New Orleans,LA (Long. 90.1100, Lat. 30.0000) 12/25/90.
T = 12:00:00 + 4(Long. - 15n), where n is the number of time zones from
LAN
Greenwich. Table 4.2.4.1 is a list of n values for United States time zones.
Table 4.2.4.1 Time Zones
Time Zone
Eastern
Central
Mountain
Pacific
Yukon/Alaska
Hawaii
n
5
6
7
8
9
10
T (WA)» 12:00:00 + 4(122.6292 - [15 x 8]) = 12:10.52 = 12:10:31 PST
LAN
T (LA) = 12:00:00 + 4(090.1100 - [15 x 6]) = 12:00.44 = 12:00:26 CST
LAN
Secondly, correct for the Ephemeris of the sun.
TSN
T - A, -where A is the correction found in Table 4.2.4.2.
LAN
-------
Section No. 4.2.4
Revision No. 0
Date: 17 Sep 89
Page: 5 of 8
Table 4.2.4.2 Ephemeris of the Sun
From the Nautical Almanac - 1989 Yachtsman's Edition
Equation
Date of time
m. s.
Jan. 1 - 3 24
4 -448
7 -608
10 -724
13 -835
16 -940
19 -1039
22 -1 1 31
25 -12 16
28 -1254
31 -13 25
Feb. 3 -13 49
6 -14 05
9 -1414
12 -1416
15 -1411
18 -1400
21 -1342
24 -13 18
27 -1249
Mar. 2 -12 15
5 -11 37
8 -1055
11 -1010
14 -921
17 -831
20 - 7 38
23 -644
26 - 5 49
29 - 4 54
Equation
Date of time
m. s.
Apr. 1 - 4 00
4 -307
7 -216
10 -126
13 - 39
16 + 6
19 + 48
22 +1 26
25 + 1 59
28 + 2 29
May 1 +2 53
4 +313
7 +328
10 +338
13 +342
16 +342
19 +337
22 +326
25 +311
28 + 2 52
31 + 2 28
Jun. 3 + 2 00
6 +1 29
9 +55
12 + 19
15 - 19
18 - 58
21 - 1 36
24 -215
27 - 2 53
30 - 3 30
Equation
Date of time
m. s.
Jul. 3 - 4 05
6 -437
9 -506
12 -532
15 - 5 53
18 -609
21 - 6 20
24 - 6 27
27 - 6 28
30 -625
Aug. 2 - 6 15
5 -601
8 -540
11 -515
14 -444
17 -408
20 - 3 28
23 -243
26 - 1 55
29 - 1 03
Sep. 1 - 8
4 +50
7 +1 50
10 +252
13 +355
16 +500
19 +604
22 + 7 08
25 +811
28 +912
Equation
Date of time
m. s.
Oct. 1 +1011
4 +11 08
7 +1202
10 +1252
13 +1338
16 +1420
19 +1456
22 +1527
25 +1552
28 +1610
31 +1621
Nov. 3 +1625
6 +1622
9 +1612
12 +15 54
15 +1528
18 +1455
21 +1414
24 +13 26
27 +1231
30 +11 29
Dec. 3 +1021
6 +908
9 +751
12 +629
15 +504
18 +337
21 + 2 08
24 + 38
27 - 51
30 - 2 20
T (WA) = 12:10:31 - (-4:16) » 12:14:47 PST = 13:14:47 PDT
T5N
TTsN(LA) = 12:00:26 - (-0:08) » 12:00:34 CST
Once the time of TSN is known, all that remains is to observe
the position of the line to the sun at TSN. An easy way is to use a loosely
mounted theodolite set at 180 deg. to track the sun. [CAUTION: EYE DAMAGE MAY
RESULT FROM LOOKING AT THE SUN WITHOUT SUITABLE PROTECTION - Remember that
harmful UV rays can be present when visual light "looks" safe - USE AN EYE
SAFE FILTER] When a watch (one second resolution set to WWV or equivalent
source for the correct time) shows TSN, tighten the theodolite mounting. At
-------
Section No. 4.2.4
Revision No. 0
Date: 17 Sep 89
Page: 6 of 8
that instant the sun is in the cross hair of the theodolite and the theodolite
correctly labels the azimuth angle as 180. Once set, the theodolite can be
used to find the bearing to any distant feature which might be selected as an
orientation target. Another method is to mark the end of the shadow of a
vertical tower at TSN, thus establishing a N-S line from the base of the tower
to the mark.
The two drawbacks to the TSN method are weather and
schedule. If the sun is obscured at TSN the observation cannot be made. Also,
if other activities command higher priorities, the time of TSN might not be
available for the sighting.
4.2.4.3.2.2 Solar Azimuth Method
The azimuth angle to the sun can be found at any time if
the latitude is also known. A Basic program (Blackadar, 1985) which contains
the necessary subroutines has been edited to provide the outputs shown in
Figure 4.2.4.1. These are the same two examples as are used in 4.2.4.3.2.1.
The program listing is given as Figure 4.2.4.2. Notice in Table 4.2.4.3 that
the two methods do not agree. The differences are trivial. Even at the fast
angular motion of July, the sun moves about 0.5 deg. per minute. The roughly
quarter minute difference in methods represents only a little over 0.1 deg.
uncertainty. Notice also the nonlinearity difference between winter and summer
which makes simple extrapolation impossible.
A Brunton compass, mounted on a tripod, can be used for
solar sighting. The mirror can be set to project the sun and the sighting
points and lines on a white piece of paper. The compass needle can be used as
a reading index or an additional protractor and pointer can be added to the
compass mounting hardware.
-------
Section No. 4.2.4
Revision No. 0
Date: 17 Sep 89
Page: 7 of 8
DAY? A
MONTH? 7
YEAR? 1990
DAY? 25
ffiTH? il
YEAR? 1990
SOLAR AZIHUTH ANGLE
iEDHESDAY 4 3UL 1990
Fox Island. MA
Longitude 122.6292 Latitude 47.25
Day'of Year 185 Julian Day 2448077
Transits Meridian 13 14 53 PDT
SOLAR AZIMUTH ANSI
TUESDAY 3 DEC 1990
te* Orleans, LA
Longitude 90.11 Latituss 30
Eiy of iear 359 Julian Dav 244S251
Transits flendian 12 0 34 C3T
Tiie
HR,MIN? 11.30
HR.MIN? 12.00
HR,HIN? 12,30
HR.MIN? 13,00
HR.HIN? 13,30
HR.HIN? 13,00
Elevation
57.81
61.36
64.00
65.43
65.42
65.42
Azuuth
130.15
141.90
155.86
171.73
188.40
171.73
Tise
s.aiN? ic.:o
HR.KIN? 11.%
HR.KIN? 11.30
HK.HIN? i:.oo
ffi,HIH? 11.30
®,HIN? 13,00
Elsvation
3135
34.67
:*.n
36.61
-; 14
.Ml *~
34.73
A:iautt
1SS.2S
163.07
(T TT
i / * . JV
179.36
193.38
196.65
Figure 4.2.4.1 Screen printouts for two azimuth examples
Table 4.2.4.3 Solar Method Comparisons
Lo.La.D WA (7/4/90)
Method TSN Almanac dif.
Units PDT PDT (s)
TSN 13:14:47 13:14:53 -6
LA (12/25/90)
TSN Almanac dif.
CST CST (s]
12:00:34 12:00:34 0
where Lo is the station longitude
La is the station latitude
D is the date of interest
-------
50 STAJ.-MSI, Fox Wind. WA-.REM STATION NAME _^
60 READ LO.LA
1 00 DATA 1 22.629,47.2S:REM LONGITUDE & LATITUDE
110 HEAD PI.OB,LO.L1,AO.A1,EC,EO
1 20 DATA 3.1 41 592654..409095.4.8837661 8..01 7202791
130 DATA 6.23471 229..01 7201 970..016728..0021 8
1 40 TR.PI/180fC-2-PI:REM TO RADIANS: FULL CIRCLE
150 SL-15'INT(LO/15».5):REM STANDARD LONGITUDE
180 TZ-SL/15-4: REM SELECTS TIME ZONE LABEL
190 LO-LOTR:LA.LATR:SL-SLTR
210D1J--SUNDAY MONDAY TUESDAY WEDNESDAY"
220 D2J-THURSOAY FRIDAY SATURDAY"
2300$.01$*D2S:X$.'n>
240 MS-'JANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDEC-
250 Z$--ASTESTCSTMSTPSTYSTASTADTEDTCDTMDTPDTYDTADr
260 TN-LO/FC+.5:HEM LONGITUDE TIME OFFSET *• 12 HR
270 PRINT 'DAY-::INPUT 0
280 PRINT •MONTH';:INPUT M
290 IF M>12 THEN PRINT •INVALID DATE'rGOTO 280
300 PRINT "YEAR-::INPUT YR
310 X.1:Y.1:GOSUB 2410
320 JI-TflEM YEAR DAY 1
.130 X-0:Y-M:GOSUB 2410
340 YD-T-JU1 flEM DAY OF YEAR
350 X-INT(T+iy7:Y-INT(X)
360 WD-INT(7"(X-Y)+.5):REM DAY OF WEEK
380 T-T+3449.5+TN:HEM T IS NOW TIME OF LOCAL MEAN NOON
385 DT-.OOOS9+2JE-08T : T.T+OT : REM EPHEMEfllS TIME
390 PRINT TAB(28);"SOLAR AZIMUTH ANGLE"
405 PRINT TAB(28);
410 PRINT MIQ$(D$,9*WCM,9);
420 PRINT D*HO$(M$,3'(M-1)+1,3);YR
423 PRINT TAB(20);STA$;
430 PRINT TAB(20);"D«y o» Y««r-;YD;TA8(48).-Juli«n D«y:INT(JD+1)
490 X.YD-WD: Y-SL-15TR
500 IF X>90 AND X<298 THEN TZ-TZ+7 : SL>Y
510TS-MID$(ZJ.3TZ+1J)
610 GOSUB 2860: REM FIND SUN AT LOCAL MEAN NOON
620 IF DE>PITHEN DE.OE-FC
630 Q-ML-RAJ1EM EQUATION OF TIME (NOT DISPLAYED)
640 D3-DE : REM SAVE DECL FOR HEAT BUDGET
660 X-.01 45439 : GOSUB 2360
690 IF ABS(Y><1 THEN 720
710 GOTO 780
720 SO-Z-{UL1/FC) : H-SO : GOSUB 2260
725 TC-.00274-SO-SIN(OB)-COS(TD-SIN(LA)
730 Z-SIN(SO)'COS(LAnCOS(DE)«3)
73STC-TC/Z
740 X-ZT+TC*EO : GOSUB 2310
780 PRINT TAB(23)rrran«iu Meridian •;
790 IF ABS(LA-DE)>PI/2 THEN PRINT XJJ($;TJ
800 H-0 : GOSUB 2260
810 X.ZT : GOSUB 2310
820 PRINT X;YZ:TJ
830 PRINT
850PRINTTAB(12)^Tim« Elcvnon Azimum'
1990 GOTO 4000
Section No. 4.2. 4
Revision No. 0
Date: 17 Sep 89
Page: 8 of 8
2050Z.X/Y
2060 Z-ATN(Z)
2070 IF C-1 THEN Z-PV2-Z
2080 IF N-1 THEN Z~Z
2090 f Y<0 THEN Z J»PI
2100 F Z<0 THEN Z^+FC
21 10 RETURN
2160CZ^IN(LA)'SIN(DE)fCOS(LA)'COS(DE)-COS(H)
2165 SZ-SOR(1-CZA2): ZA^TN(SZ/CZ)
2170 f ZA<0 THEN ZA-ZA»PI
2175 X-COSpE)'SIN(HySZ
2180Y.(SIN(LA)'CZ-SIN(DE))/(SZ'COS(LA»
2185 GOSUB 2010
2190 AZ^ : IF AZ>PI THEN AZ-AZ-FC
2195 RETURN
2210 hUZT+SL-RA-LO+ML+PI
2220 IF H>PI THEN H-H-FC
2230 RETURN
2280 FOR U.1 TO 5
2265 ZT-H+RA+LO-SL-ML-PI
2270 X^IN(ZT) : Y-COS(ZT) : GOSUB 2010
2275 ML-LO+Lr(T-TN+(SL»Z)/FC) : NEXT U
2280 ZT^: RETURN
2310 IF X<0 THEN X-X+FC : GOTO 2310
2315 W.X-24/FC : X-INT(W)
2320 Z-(W-X)-60 : Y.INT(Z)
2330 Z-INT((Z-Y)'60) : RETURN
2380Y.(X-SIN(LA)-SIN(DE)X(COS(LA)'COS(DE))
2370 F ABS(Y)>1 THEN 2390
2380 X^OR(1 -Y«2) : GOSUB 201 0
2390 RETURN
2410TJ6r(YR-1980)
2420 T-T-INTfT (YR*INT((Y+9V12))/4)
2430 S^GN(Y-9):A-A8S(Y-9)
2440 Z-INT((YR+S'INT(A^)yi 00)
2450 T-T-JNT(3'(Z+1 )/4)
2460 T-T+JNT(275*Y/9)+X-.5
2470 JO-T+2447689*
2480 RETURN
2860 MAWkO+A1 T: REM SUN'S MEAN ANOMALY
2870 ML-LO+L1T: REM SUN'S MEAN CELESTIAL LONGITUDE
2880 X-S1N(ML):Y-COS(ML):GOSUB 2010
2890 ML-Z
2910 TA-MA+OL:TL^IL*DLflEM TRUE ANOMALY & LONGITUDE
2920 RV^1.£C»2V(1+EC-COS(TA)):REM RADIUS VECTOR
2930 X.S(N(TL)-SIN(OB):Y.SQH(1-X*2):GOSUS 2010
2940 DE-ZJF Z>PI THEN Z^-FC
29SO X^IN(TL)*COS(O8):Y^:OS(TL):GOSUB 2010
2960 RA^REM SUN'S RIGHT ASCENSION
2970 RETURN
4000 INPUT •HR.MIN'lHR.MIN:
2020 IF YoO THEN 2050
2030 Z-0.-C.1 :IF X<0 THEN N.I
2040 GOTO 2060
4020 T.
4060 GOSUB 2860
4070 GOSUB 2210
4080 GOSUB 21 60
4090AL-Pt/2-ZA
41 10 PRINT TAB (24);
4113 PRINT USING •
41 17 GOTO 4000
9000 END
»#f .**-.ALTR, 1 80+{AZ/TH)
Figure 4.2.4.2 basic program listing for finding solar azimuth as a
function of Longitude, Latitude, Date and Time
-------
Section No. 4.2.5
Revision No. 0
Date: 17 Sep 89
Page: i of 4
4.2.5 CALIBRATION
Calibration, as defined on page 3 of the Purpose statement in the beginning
of this handbook, qualifies the process as both a measurement and adjustment,
if necessary, of the performance of the system and its components.
Manufacturers usually include in their manuals the details of all the available
calibration or adjustment points. From a QA standpoint, the important
consideration is how the system is working as a whole. Since only parts of the
system are adjustable, the relationship of these adjustments to the whole
system must be known. This brief section will focus on documentation of
calibrations and methods to verify the system response to subcomponent
adjustments.
4.2.5.1 Wind Speed
4.2.5.1.1 System accuracy
The part of a calibration which challenges the entire system,
except for the coupling or reaction of the sensor to the wind,relates the rate
of rotation of the anemometer shaft to output speed. It does not matter if the
rate of rotation is caused by a synchronous motor or a d.c. motor with a
provision for shaft revolution counting. What does matter is the accuracy of
the determination of AVERAGE rate of rotation and a common averaging PERIOD
used by the system and the challenge. The operators may choose to conduct this
calibration with the sensor installed on the tower. When multiple outputs
exist, the calibration should record values from each of them, but the critical
output is the one used to produce the official archieved data.
The accuracy determination depends on both the method used in
the challenge and the accuracy of the measurement of the input. If a
synchronous motor is used, there must be some reason to believe the motor was
turning in sync with the commercial power. Repeated samples which do not
change is one form of evidence. Commercial power is generated within a
frequency tolerance of 60 ± 0.1 cps. Synchronous motors which are hand held
with a flexible coupling to the anemometer shaft may go in and out of sync
providing a slightly changing output. Shaft rotation counters can also produce
erroneous outputs. Some evidence of their performance, such as counting a
synchronous motor shaft rotation or simply counting revolutions at a slow rate,
is needed in the documentation of the test equipment, preferably before and
after field use.
4.2.5.1.2 Component accuracy
If the system has built in calibration circuits, they should be
calibrated at the same time as the total system. They are handy to use on a
routine service schedule, but there needs to be some evidence of their
calibration. If panel meters or portable DVOMs are used to check the signal
conditioner or transducer sub-system, there needs to be evidence that they are
in calibration. It is possible to adjust a circuit to provide the required
output on a meter which has a 2% error and thereby introduce a 2% error to the
calibrated system output.
.The calibration of the sensor starting threshold can only be a
measurement. Adjustment is usually impossible. The exception might be the
amount of end play in the shaft-bearing assembly, but that level of sensor
-------
Section No. 4.2.5
Revision No. 0
Date: 17 Sep 89
Page: 2 of 4
repair is usually left to the laboratory or shop for good reason. The accuracy
of the torque measurement, or non-measurement, is also important. Assume a
torque watch, or similar device, with a range of 0.003 to 0.030 oz-in. The
threshold of measurement is 0.003 oz-in or 0.22 g-cm. If a cup anemometer has
a K value of 1.4 (see 2.1.1.2), the torque provided by a 0.4 m/s (0.9 mph) wind
is 0.22 g-cm [from T=Ku2]. The torque provided by a 0.5 m/s (1.1 mph) wind is
0.35 g-cm. If the torque watch cannot measure the starting torque of the shaft
because it turns before the indicator moves, the starting torque is < 0.22 g-cm
and the starting speed is < 0.4 m/s. If, instead, the starting torque reads
0.35 g-cm (0.005 oz-in), the starting speed is 0.5 m/s and within
specification. If the starting torque reads 1.0 g-cm (0.014 oz-in or about
half scale on the torque watch), the starting speed is 1.4 m/s and clearly in
need of service. Documentation of this measurement will tell the data QC
inspector that the data from this anemometer is in error in the indicated 0.2
(assuming a 0.2 m/s offset) to about 3 m/s range. (3 m/s wind provides 12.6
g-cm torque or about an order of magnitude more than that provided at 1 m/s.
The difference between 0.35 and 1.4 is not likely to be visible at 12.6) The
true wind speed will be higher than the indicated speed. At indicated 0.3 m/s
it would be 1.5 m/s and at indicated 3 m/s it would really be 3 m/s.
4.2.5.2 Wind Direction
4.2.5.2.1 System accuracy
The system calibration of a wind vane can be checked on the
tower by aiming the vane to and from known directions, such as a distant
mountain peak or similar feature. If checks are made with respect to a mounted
component, such as a crossarm, the orientation of the crossarm also needs to be
checked. A single distant feature should be the orientation target with a
known bearing with respect to TRUE NORTH. Other targets can be secondary
checks which challenge both the orientation and the performance of the
transducer. For systems using the 540 format, the targets should be reached
after a clockwise revolution and then again after a counter clockwise
revolution to challenge both parts of the transducer.
Before the transducer is removed from the tower, a
documentation of the as-found output with the vane held pointing at or from the
orientation target is essential. This single act provides the basis for data
validity for the period beginning with the previous as-left record and ending
with this as-found reading. Since the sensor should not have been removed from
the tower or adjusted in orientation without the as-found and as-left readings,
these values should be the same, within the capability of pointing the vane (1
deg., 2 deg. at the most). If they are not, the data QC inspector will have
some detective work to do. Usually, when the sensor is removed and used in a
calibration at the location of the rest of the system, replacement in a keyed
fixture will cause the as-left value to be the same as the as-found. Figure
4.2.5.1 shows three examples of how manufacturers provide an orientation key
for wind direction sensors. If there is no keyed fixture, the full orientation
procedure will be required.
-------
Section No. 4.2.5
Revision No. 0
Date: 17 Sep 89
Page: 3 of 4
Figure 4.2.5.1 Methods for keying orientation of direction sensors
One simple orientation procedure requires a clamp which will
hold the vane from turning. A hose clamp will work for some designs. Figure
4.2.5.2 shows a hose clamp used for this
purpose. Tape which does not stretch is • , , / "- ' •-.- •'••-••,•,• -- .'••'
marginally useful. Stretchy tape like
duct tape or electricians tape will only
work on a perfectly calm day. Set the
vane so that the output is the correct
value for the orientation target (see
4.2.2.2.2.1). If the angle of the
orientation target is coincident with a +2
deg. error relative to the average error
of 0 deg., the output should be 2 deg.
higher than the bearing of the orientation
target. Only in this way will the relative
error of the sensor be distributed equally
about TRUE directions. Tighten the clamp so the output is both correct and
constant. Mount the clamped sensor on the tower and turn it until the vane
points at the orientation target. Clamp the vane in place. Verify that the
output is still correct before removing the vane clamp.
4.2.5.2.2 Component accuracy
The same comments regarding calibration circuits, parallel
recorders and panel meters apply to wind direction as they do to wind speed,
mentioned in 4.2.5.1.2 above. With the sensor next to the signal conditioner
(attached with either the operating cable or a suitable substitute) and with a
fixture which holds known relative directions, the signal conditioner can be
adjusted if required. The 540 offset voltage, if one is used, can be tested
and adjusted. The output voltage vs. position can b'e set. The open space in
the potentiometer, if one is used, can be measured and adjusted for.
Figure 4.2.5.2
A direction vane clamp
-------
Section No. 4.2.5
Revision No. 0
Date: 17 Sep 89
Page: 4 of 4
A single potentiometer has an electrical range of something
like 355 deg. with a mechanical range of 360 deg. If the transfer function of
relative direction to voltage output is
9 = 360 x V
where 0 is the angle in deg. and V is the output in volts (0-1 V scale), and
the maximum "full scale" output, set by shorting the potentiometer wiper to the
high side of the potentiometer, is 1.000 V, a small error will have been set
into the system. The error will be +1.4 % of reading. At 355 deg. the output
will be 360. At 180 deg. the output will be 182.5 deg. This adjustment error
added to the linearity error of the potentiometer may be more than is
acceptable. If instead, the signal conditioner is set to output 0.986 V when
the vane is set to 355 deg., the output will be 355 deg. (360 x 0.986). At 180
deg. the output will be 180 (assuming no linearity error). All of the error
between 355 deg. and 360 deg. is in that 5 deg. sector.
Is this acceptable for PSD (EPA, 1987a) applications? The
"wind direction system error" which cannot exceed 5 deg. is the error of the
averaged wind direction samples. If the mean direction were 355 deg. with a
range of ±5 deg., and if the distribution were bi-modal with half the values at
350 deg. and half at 360 deg., and if the output voltage remained at 0.986 V
between 355 deg. and 360 deg., the average output would be 352.5 deg., a -2.5
deg. error. If the dead space were at 0 V, the output would cause the bi-modal
distribution to look like half 350 deg. and half 360 deg. producing the correct
average of 355 deg. This is a maximum error estimate. True distributions
would cause smaller average errors. Even a wind averaging 357.5 deg. with a
range of ±2.5 deg (the vane is always in the dead space), the error is 2.5 deg.
The starting threshold of the wind vane is important to
accurate low wind speed directions. The design of the vane along with the
off-set angle (or error tolerance) provides a K value. The K value along with
the starting torque of the vane assembly provides a threshold wind speed.
Assume a 5 deg. error tolerance and a K value of 15. At 0.5 m/s the available
torque is 3.75 g-cm. At 10 cm out from the axis or rotation, a force of 1/3 of
a gram should move the vane assembly. This is another threshold of the torque
gauge situation. At 1 m/s the torque available is 15 g-cm and at 10 cm the
force is a reasonable 1.5 g. At 1 m/s and 10 deg. error tolerance, K becomes
37.5, the torque available becomes 37.5 g-cm, and the force at 10 cm is an
easily measured 3.75 g.
-------
Section No. 4.2.6
Revision No. 0
Date: 17 Sep 89
Page: 1 of 2
4.2.6 OPERATIONS, MAINTENANCE AND QUALITY CONTROL
4.2.6.1 Operations
The important aspects of operations, from the standpoint of quality
assurance, are planning (see QA Plan, Section No. 4.1.) and documentation
(Section No. 4.9.1). The purpose of operations is to acquire valid data. For
wind measurements, this requires frequent (weekly, if possible) visual
examination of the sensors. This is not a "hands-on" examination but simply a
look at the active shapes, cups, propellers and vanes, to be sure there has
been no physical damage. Sensitive wind instruments can be damaged by hail and
by birds. The nature of an analog recording, if one is used routinely, will
tell how the sensor is performing. Routine entries in the station log will
provide the evidence of attention to support validity claims.
Calibrations are a part of operations. A member of the operating
organization needs to become the "expert" on how the measurement system works
and what it needs to continue "in control" performance. Regularly scheduled
calibrations build the expertise and the documentation showing measurement
accuracy. The frequency of calibrations is a variable. For a new
installation, a calibration during the installation is necessary. A careful
look at the first week of operation will find early failures. If all seems to
be going well, a calibration check after a month is prudent. If no problems
surface, a full calibration at the end of the first quarter is advisable. For
some site environments and some applications quarterly calibrations are
recommended. Semi-annual calibration is the minimum frequency. If problems are
found they must be documented and corrected as quickly as possible. The
requirement of 90% joint frequency of valid wind and stability data does not
permit much down time. The frequency of calibrations or calibration checks
should be determined by the performance of the instrument system. If problems
occur, the week-month-quarter frequency should begin again. When it is
demonstrated that the system is once again "in control," routine calibration
frequency (semi-annual or quarterly) can resume.
4.2.6.2 Maintenance
4.2.6.2.1 Routine and preventive maintenance
The only routine maintenance required for the wind system
should be applied during routine calibrations. Sensors exposed to the elements
need cleaning and protective lubricants applied to their mounting hardware.
When a sensor needs to be removed for close inspection or calibration and it
cannot easily be removed because set screws or nuts are locked to their threads
by corrosion, a failure in routine maintenance is the reason.
If the system has supply requirements, such as ink and paper
for analog recorders or tapes and printer paper for digital recorders, the
timely servicing of these requirements is a routine maintenance task.
Preventive maintenance must at minimum follow the
manufacturer's recommendations. Considerable damage can result by ignoring
this guidance. Some people like to oil anything that moves. Sensitive wind
sensors require specific care if the threshold is to be maintained.
-------
Section No. 4.2.6
Revision No. 0
Date: 17 Sep 89
Page: 2 of 2
4.2.6.2.2 Corrective maintenance
Parts for wind systems are not to be found anywhere except from
the manufacturer. This is true at least at the sub-component and component
level. When a part fails or wears out, the new part usually must come from the
manufacturer. This may take a week or two depending on the part and the
manufacturer. It is prudent to have spare parts on hand to cover some
predictable failures. A component plug-in philosophy is the quickest way to
correct failures. If a bearing or a potentiometer fails in a sensor, a new
calibrated sensor is simply plugged in while the failed one is repaired. If a
circuit card fails, a new calibrated card is plugged in while the failed one is
repaired.
The next level of spare part strategy is the sub-component
level. Critical and difficult to buy parts are stocked and used to repair
sensors or circuit cards. Conventional sensors will always need repair at some
point in time, bearings and direction potentiometers usually. Circuit cards
are becoming so reliable that maintenance is hard to anticipate.
4.2.6.3 Quality Control
The quality control (QC) of a data monitoring program is a loop driven
by routine inspection of the data for validity. The data QC person should be a
meteorologist who is familiar with how wind data should look and what kinds of
variety are provided by the atmosphere. Such an inspector will spot problems
before they are obvious to an observer who may be an expert in another field
but is not a meteorologist. If a technically qualified QC inspector is not
available, the best compromise that is available must be made. It is very
dangerous in terms of lost time if no routine data QC function is followed in
the QA Plan, or if there is no QA Plan.
When a problem is found by the data QC inspector, a discrepancy report
is issued which brings the operators into the data QC loop. Their inspection
and corrective action is reported back to the QC inspector closing the loop.
Because of this QC loop, the measurement system can be operated "in control"
and valid data produced.
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 1 of 20
4.2.7 PERFORMANCE AUDIT METHODS
4.2.7.1 General Considerations
A performance audit is the determination of the instrument system
accuracy made with an independently selected method and by a person who is
independent of the operating organization. To make this determination for wind
measurements, knowledge of the input conditions imposed upon the sensors is
required. Given knowledge of these input conditions, the transfer functions
and the system's data handling method, the output can be predicted. The
difference between the predicted output and the system output is the error of
the system or its accuracy.
The methodology starts with the ways of controlling and/or measuring the
input conditions. When controlled inputs are used, as should always be the
case for starting thresholds, anemometer rate of rotation vs. output and
relative vane position vs. output, the accuracy of the output is easily
determined. Of course, the accuracy of the anemometer transfer function is not
a part of this determination. When the input conditions are not controlled,
as with the collocated transfer standard (CTS) method, the accuracy
determination has a larger uncertainty. The CTS method does challenge the
anemometer transfer function. The best performance audit uses both methods
where appropriate.
A performance audit must follow some written procedures. Since the
procedures must be relevant to the design of the instrument or system being
audited, only general principles will be described below with some specific
examples. The data from the audit should essentially fill out an audit form.
It is important, however, for the auditor to be sufficiently experienced to be
able to deviate from the procedure or the form when the pursuit of truth leads
away from the expected.
The starting point of an audit form is the documentation of the who,
what, where, when, and how the audit values were acquired.
4.2.7.1.1 Who
The performance audit report form should contain a space to
identify the auditor. The audit report which summarizes the audit findings
should report the names and affiliations of the operators of the system.
4.2.7.1.2 What
The form should contain a section to identify the instrument
being audited by manufacturer, model number, and serial number.
Sub-assemblies, such as a cup wheel of an anemometer, should be identified by
number. If they are not numbered, the operator should be asked to mark them
for identification.
The audit report should contain a list of all the equipment
provided and used by the auditor, including model and serial numbers and time
of last calibration, where relevant.
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 2 of 20
4.2.7.1.3 Where
The audit form should have a space to show the location of the
sensor on a tower, including height. A sketch is useful to show the relative
positions of the sensing elements with respect to possible biasing influences,
such as the tower, other sensors and buildings.
4.2.7.1.4 When
The date or dates when the audit affected the system operation
should be listed. The time when the system or a particular sensor was taken
"off-line" and put back "on-line" should be listed. The time, or time period,
when each datum value was taken is vital for the comparison with the system
output. Implicit in this is the need for the time the auditor uses to be
correlated to the time the operator or the system uses. The auditor should
rely on the National Bureau of Standards station WWV for correct time. Battery
operated receivers, such as the Radio Shack Time Cube, are generally available.
4.2.7.1.5 How
The audit form should either contain a copy of the method used
or reference the method number. The audit report should contain copies of the
audit method used. The methods should be detailed enough to identify each step
in the acquisition of the audit value and in the conversion of the value to
units compatible with the system output.
4.2.7.2 Wind Speed
There are two general philosophies in use by those who operate
anemometers in meteorological monitoring systems and networks. The most common
treats the system as a unit where the sensor and signal conditioner and
recorder are calibrated together. The other, often employed by operators with
large numbers of anemometers, considers the sensor as a standard
interchangeable part. In this case two audits are necessary. One to challenge
the sensor calibration method and the other to challenge the system calibration
using a standard signal as a substitute for the sensor. A full system audit
method from sensor input to system output can be used as a challenge for the
system operated with interchangeable sensors.
4.2.7.2.1 Sensor Control
The controlled condition is rate of rotation of the anemometer
shaft. The cup assembly or propeller is removed for this challenge. The audit
form should provide space for fully defining the transfer function used by the
operators (usually supplied by the manufacturer). This should include the
relationship of rate of rotation (R, rps) to wind speed (U, m/s), rate of
rotation to output volts (0, V) and rate of rotation to frequency (f, Hz), for
light chopper or a.c. generator types. See 4.2.2.1.2 for a discussion of the
U=a+bR and U=bR types of transfer functions and how the constants "a" and "b"
are determined. Some manufacturers provide the transfer function in the form
f » 26.439 (U - 0.281)
which can be converted to
U = 0.281 + 1.135R
once the number of pulses per revolution (30 in this case) is known.
-------
The rate of rotation can be imposed
on the anemometer shaft in a number of ways. If the
method is to drive the shaft with a d.c. motor, the
number of revolutions of
time is the data value.
the shaft over a period of
That value divided by the
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 3 of 20
number of seconds in the time period gives average
rate of rotation R (rps). The R is converted to U
by the transfer function and the U is compared to
the system output in the same units for exactly the
same period of time. If the system provides 5
minute average speeds, the count is for 5 minutes
with the start and stop times inclusive of the
system period. If the d.c. motor is reasonably
constant (± 10%), a few seconds out of
synchronization over 300 seconds is acceptable. The
period of time, however, must be exactly 300 seconds
which can be hand timed with a sweep second watch to
about ± 0.2 s. If the system only reports hourly
averages, and cannot be changed to a shorter time,
samples of the signal conditioner output voltage may
be used to estimate the system output. Three rates
of rotation in addition to zero are recommended.
Since the important speeds are low speeds and not
full scale speeds, the use of simulated speeds on
the order of 2, 5 and 10 m/s is acceptable. Using
the transfer function above, these speeds are
simulated by R values of 1.51, 4.16 and 8.56 rps
(90.6, 249.6 and 513.6 rpm). Figure 4.2.7.1 is an
experimental d.c. motor drive used for this kind of
audit. Figure 4.2.7.2 is a second generation d.c.
motor drive capable of being powered by a D cell and
a 9 volt transistor battery.
If a propeller anemometer with a transfer
function of
U = 0.294R
is challenged, the speeds of 2, 5 and 10 m/s will be simulated by 6.9,
34 rps (414, 1020,and 2040 rpm). If the auditor could generate five R
1.5, 4, 7, 17 and 35 rps, both cups and propellers could be challenged
meaningful speeds plus zero. See Figure 4.2.7.3 for a third d.c. motor
Figure 4.2.7.2
17, and
values,
at three
system.
Figure 4.2.7.3 An experimental 12-volt d.c. motor and counter
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 4 of 20
Climatronics cup
Teledyne-Geotech cup
R.M. Young propeller
Figure 4.2.7.4
A simple d.c. motor might be made to turn
the shaft, but the key to the audit challenge is
the measurement of the shaft revolutions. A
light chopper and counter is a straight forward
approach to this measurement. Hand switching
the counter for periods as short as 60 s will
produce better than one percent accuracy in
time. If the light chopper produced 10 counts
per revolution, the count rate required for the
four R values mentioned above is 5 to 60 Hz, an
easy range for simple battery operated counters.
A system such as described above has the
advantage of independence from commercial power,
a significant advantage for some remote wind
systems.
The controlled condition audit requires a
hands-on policy regarding the sensor. The
measurements of the starting torque does not
require the sensor to be connected to the
circuit. It is possible, with proper equipment
and care, to install the speed challenge motor
on the sensor and operate that challenge with it
connected in its operating location. Another
way to challenge an anemometer is to connect the
shaft to a synchronous motor. The assumption
here is that the motor is running in
synchronization and the R value is therefore
known from the specifications of the motor or an
independent measurement of its rate of rotation.
Figure 4.2.7.4 shows three anemometers coupled
to a synchronous motor through a universal
coupler.
The last measurement is the time constant.
With the simulated speed on its highest rate,
and with a meter on the output voltage of the
signal conditioner, turn off the d.c.motor and
measure the time it takes to reach the value of
the simulated speed minus 63% of the simulated
speed. Examples of wind speed audit procedures
and forms are found in Figures 4.2.7.5 through
4.2.7.8.
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 5 of 20
Cup Anemometer - MSI method CA003 (version 8/l/84>
This method provides -for a comparison o-f the trans-fer -function
used with the system to the output o-f the system. This is done
by causing the anemometer shaft to turn at a known rate o-f
rotation and observing the output. The means of turning the
shaft and measuring the rate of rotation are provided by the
auditor and are completely independent of the operating system.
The method does not challenge the transfer function. This can be
done best with a wind tunnel test.
The report form for this method includes space for an optional
determination of starting torque and system time constant. The
torque measurement ma/ be used as an indication of bearing
condition and hence starting threshold of the anemometer. The
time constant is of use if turbulence is measured.
CAOO3—A Remove the cup assembly. Mount a coupler to the
anemometer shaft. A 1/8" shaft is required. If the anemometer
does not use that size or it is not accessible, an interface
fitting will be required. Clamp the drive motor to the support
column of the shaft so that the coupler is engaged with the drive
wheel. Determine if the cup assembly turns the shaft in a clock-
wise or counter clockwise direction, when viewed from above.
Clockwise is common and is used on the form. Operate the drive
motor at two speeds (find the desired rps from the transfer
function) which are important to the application of the wind
speed data. Use a time period synchronous with the system
output. An average of one minute or longer is required. If the
system provides only instantaneous samples of output volts, tat e
12 samples over a two minute period and use the average of the
samples to compare with the average rate of rotation measured.
CAOO3-B This method requires that the system be operating with
all cables in place (short jumper cables may be used with CAOO3-A
to allow simultaneous access to the anemometer and the signal
conditioner for those systems where these two parts are at some
distance away). At least a' zero rate of rotation must be
measured (or observed) with the anemometer in place, the cu.p
assembly removed and the shaft taped to assure non-rotation. A
second observation may be either a motor driven measured rate o-f
rotation for the operating period of the system or a natural (tin—
measured^ non-zero operation to assure that signal reaches the
signal conditioner when the syst=m is in operating position. The
assumption with the later choice is that if the signal is
transmitted at all it will be properly simulated in method A.
This is more lively true with pulse trains than with generator
voltaaes.
Figure 4.2.7.5 Audit method for a cup anemometer
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 6 of 20
PERFORMANCE AUDIT REPORT by
CAO03
MEASUREMENT SYSTEM - Cup anemometer
System number _
Sensor _____ — _____ _
Cup assembly
Loc a t i on
Signal conditioner
Data Channel _
DATE / / TIME o-f-f line on line test start
TRANSFER FUNCTION: (rps to mps)
(rps to volts)
pulses per revolution
TEST RESULTS
CA003-A . — .challenge speed — output di-f-ference
time revs. rps mps mps mps '/.
0 0 0
"d.c."
S CW
F CW
Torque:
CA003-B
Oz.-In. cw« Time constant
seconds
time
expected
mps
observed
mps
di-f-ference
mps 7.
0
test
Figure 4.2.7.6 Audit form for the cup anemometer method
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 7 of 20
Fixed Axis Propeller - MSI method FAF001 (version 8/1/841
This method provides -for a compari son of the transfer -function
used with the system to the output of the system. A seperate
form is provided for W (vertical component) since a different
transfer function is often used for this direction than is used
for U and V. The method causes the propeller shaft to turn at a
known rate of rotation while observing the output. The means of
turning the shaft and measuring the rate of rotation are provided
by the auditor and are completely independent of the operating
system. The method does not challenge the transfer function.
This can be done best with a wind tunnel test. The sign
convention used with respect to clockwise and counter clockwise
is that of the system being challanged. Differences are always
calculated by subtracting the audit challenge value from the
system output. Arithmetic convention is followed even though the
minus sign is used as an indicator of direction. For example,
the difference between a ~1.5 mps audit challenge and a -1.3 mps
svstem output is +0.2 mps even though the system underestimated
the speed
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 8 of 20
PERFORMANCE AUDIT REPORT by
FAPOO1W
MEASUREMENT SYSTEM - Fixed axis propeller
System number
Sensor
Propel1er
Location «_„__
Signal conditioner
Data channel
DATE / / TIME o-f-f line on line test start
TRANSFER FUNCTION: 1 rps = 0.294 mps <3 pulses per revolution)
[WC volts}- 2.51*4 = m/s
TEST RESULTS
FAP001W-A - challenge speed - output di-f-ference
time revs. rps mps mps mps 7.
0 ' 0 .
"d.c."
S CW
S CCW
F CCW
"sync" time rps mps volt mps mps '/.
S CW 5.0OO 1.47
S CCW 5.000 -1.47
F CW 30.00 S.B2
F CCW 3O.OO -B.B2
Torque: Or.-In. cw. Oz.-In. ccw, T. Const. s.
FAPO01W-B expected observed dif-ference
time mps mps mps 7.
0
Figure 4.2.7.8 T Audit form for the propeller Anemometer method
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 9 of 20
4.2.7.2.2 CTS Method
The collocated transfer standard (CTS) method for wind speed
involves mounting a carefully calibrated anemometer in the vicinity of the
subject anemometer being audited. The CTS should have certificates tracing its
calibration to NBS or some other standard facility. If the ASTM (1984) method
for comparability is being used, the CTS needs to be within 10 m of the subject
anemometer in the horizontal and the lesser of 1 m or H/10, where H is the
height above ground in meters, in the vertical. It is important to site the
CTS to be representative of the flow at the subject anemometer. Mutual
interference should be minimized through siting and through editing out data
where the direction shows the wind passing through one to reach the other. The
accuracy potential of the CTS method is based on data taken in 1982 at the
Boulder Atmospheric Observatory (BAG) and published by Finkelstein et al.
(1986) and Lockhart (1988). The anemometers for this study were spaced about 5
m apart. The closer together they are in the horizontal the larger the
direction sector of mutual interference.
The best situation for CTS auditing is one in which both
anemometers are connected to the auditor's data logger. The element of the CTS
audit is the difference in speed calculated by subtracting the'CTS speed from
the subject speed. The method requires a sufficient number of simultaneous and
independent differences. A simultaneous difference is one where the time
between sampling each anemometer output is less than O.lr, where T (s) is found
by dividing the distance constant, D (m), by the wind speed U (m/s).
Independence is achieved when the time between sampled pairs is larger than 4r.
For example, assume the subject anemometer has a distance constant of 5 m and
the.CTS has a distance constant of 1 m. If the wind speed is about 3 m/s, T
will be 5/3 = 1.7 s for the subject and 1/3 = 0.3 s for the CTS. Simultaneous
samples will exist when the sampling rate of the data logger is less than the
shortest O.IT or 0.03 s in this example. Most data loggers are fast enough for
the example.
Independence is achieved when the time between successive
sample pairs is long enough. In the example above, 4r is 20/3 = 6.7 s for the
subject anemometer and 4/3 = 1.3 s for the CTS. If the CTS logger is set for
one sample every 10 s, the data will be independent at 3 m/s. The ASTM method
defines minimum sample size in terms of the resolution of the measured or
reported speed (assume 0.1 m/s for the example, a recommended resolution) and
the standard deviation of the series of differences. It takes 900 times the
variance of the series to provide the minimum number. If the two sensors are
well sited and properly operating, the variance will be small. The BAO data
showed the standard deviation of the difference to be less than 0.2 m/s
(variance of 0.04). This should be the minimum condition for a good CTS data
set. If the variance is 0.04, the minimum sample size is 900 x 0.04 = 36.
Assume the subject anemometer produces a scalar average speed
every 15 minutes and it is not possible to wire the output into the auditor's
data logger. The CTS uses 90 samples, one each 10 s, to assemble its
concurrent 15 minute scalar average. At this point, one can take one of two
paths. One is to assume that the subject is operating well, has a short enough
distance constant and is likely to agree well with the CTS on a sample to
sample basis; not perfectly because a large bias can still have a small
variance. There is no way to verify this assumption unless the audit results
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 10 of 20
show good agreement. Based on this assumption, each 15 minute value is a data
point with a sufficient sample size to compare the anemometers. Each datum
point can be used in a linear regression analysis to define the subject
anemometer's accuracy as a function of wind speed, if enough dynamic range
exists in the data period. In just a few hours, then, the CIS method with this
assumption will produce a measure of accuracy.
The other way is to treat each 15-minute average as a single
sample of 15-minute averaged data. A 9 hour period will provide the minimum
sample size of 36. This is an awkward period of time. It takes some time to
install the CTS and many auditors work 8 hour days like other people. There is
no such thing as too many samples. Added time usually enlarges the dynamic
range for the audit. The optimum CTS audit goes for something like 24 hours,
one diurnal cycle. For an example of this CTS audit, look at- the data listed
in Table 4.2.7.1.
Each value is a 20-minute scalar average. The subject
anemometer is a Climatronics F460 cup (C-V-W). The CTS is a Young Propeller
Vane (P-V-W) located 5 m away. Both are at a height of 10 m. The standard
deviation of the CTS sensor is shown for each 20-minute period. The difference
between the subject anemometer and the CTS is shown in the column headed by
"Y-X." Notice that the average difference is a small -0.12 m/s and the
standard deviation of the difference series is 0.10, half of the maximum
criteria. To express the accuracy of the subject anemometer with respect to
the CTS, a linear regression was run. The constant of 0.02 m/s says there is
no bias of significance. The X Coefficient of 0.96 says that there is a 4 '/,
underestimate of speed at all speeds. The best fit straight line through the Y
points is calculated by multiplying the X (true) value by the coefficient and
adding the constant. The residual error is then found by subtracting the X
value from the estimated Y value, Y', (column headed by Y'-X). Notice that the
average difference is 0.000, as it must be, and the standard deviation is
lower, 0.07 m/s. The audit report for this subject anemometer would report the
error of 4 7, in the slope of the subject transfer function. (NOTICE: This is
an example. The CTS is arbitrarily selected. It is possible that in this case
the CTS was 4 Y. high or they were each off 2 %. That is not important for this
example and the speeds are so close that this analysis was not included in the
BAG experiment.) These data are shown graphically in Figure 4.2.7.9 and Figure
4.2.7.10. The anemometer data are shown on an XY plot with the best fit
straight line through the 72 points. Figure 4.2.7.10 shows an XY plot of the
residuals. Also shown is the normalized turbulence,
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 11 of 20
Table 4.2.7.1 Simulated Audit Data (BAO 1982)
9/82
Date/
Tiae
4 1000
4 1020
4 1040
4 1100
4 1120
4 1140
4 1200
4 1220
4 1240
4 1300
4 1320
4 1340
4 1400
4 1420
4 1440
4 1500
4 1520
4 1540
4 1600
4 1620
4 1640
4 1700
4 1720
4 1740
4 1300
4 1820
4 1840
4 1900
4 1920
4 1940
4 2000
4 2020
4 2040
4 2100
4 2120
4 2140
4 2200
4 2220
4 2240
4 2300
4 2320
4 2340
5 0
5 20
5 10
5 100
Y
Subject
Speed
(l/S)
3.701
2.972
3.557
2.660
2.463
2.376
3.210
5.511
4.546
4.261
4.157
3.127
2.364
2.451
4.176
8.281
7.673
6.371
5.569
3.346
8.333
7.236
5.425
3.989
4.399
1.407
3.847
3.941
4.344
4.300
2.907
2.361
2.049
1.747
3.796
4.065
3.835
4.221
4.097
4.037
3.952
T ffcn
•J . MM 7
3.449
3.179
3.940
4.393
X
CIS
Speed
(I/S)
3.729
2.979
3.555
2.605
2.365
2.399
3.344
5.623
4.722
4.555
4.455
3.250
2.433
2.498
4.270
3.601
7.S93
6.600
5.976
3.413
8.693
7.486
5.538
4.111
4.617
4.615
4.012
4.121
4.474
4.474
3.035
3.156
2.221
2.015
3.957
4.198
4.031
4.376
4.261
4.194
4.135
3.717
3.530
3.290
1.040
4.617
5
CIS
Sigaa
(1/5!
1.047
2.441
2.780
0.344
1.412
0.883
1.951
3.230
1.115
1.319
1.348
1.120
1.557
1.039
5.887
2.564
1.633
1.638
2.347
3.537
4.531
1.503
0.751
0.233
0.320
0.930
0.345
0.291
0.151
0.363
0.442
0.154
0.022
0.190
0.246
0.215
0.137
0.206
0.232
0.194
0.141
0.109
0.123
0.035
0.199
0.113
s/X
!«/s)
0.23
0.82
0.73
0.32
0.60
0.37
C.53
0.57
0.24
0.29
0.30
0.34
0.64
0.42
1.38
0.30
0.21
0.26
0.48
1.04
0.52
0.20
0.13
0.06
0.18
0.20
0.09
0.07
0.03
0.08
0.15
0.05
0.01
0.09
0.06
0.05
0.03
0.05
0.07
0.05
o.o:
0.03
0.03
0.01
0.05
0.02
Y-X
(fi/S)
-0.03
-0.01
0.00
0.05
0.10
-0.02
-0.13
-0.11
-0.18
-0.29
-0.30
-0.12
-0.07
-0.05
-0.09
-0.32
-0.21
-0.23
-0.41
-0.07
-0.36
-0.25
-0.16
-0.12
-0.22
-0.21
-0.16
-0.18
-0.13
-0.17
-0.13
-0.29
-0.17
-0.27
-0.16
-0.13
-0.15
-0.16
-0.16
-0.16
-0.18
-0.16
-0.13
-0.11
-0.20
-0.22
T! -X
(I/S)
-0.11
-0.10
-0.13
-0.14
-0.17
-0.06
0.02
-0.10
-0.00
0.12
0.13
0.01
-0.01
-0.04
-0.06
-0.02
-0.10
-0.03
0.18
-0.05
0.01
-0.04
-0.05
-0.03
0.05
0.04
0.02
0.03
-0.04
0.01
0.02
0.13
0.10
0.21
0.02
-0.02
-0.00
-0.01
0.01
0.00
0.03
0.02
0.00
-0.01
O.C5
0.05
5
5
5
5
5
5
5
C
5
5
5
5
5
C
J
5
5
5
c
J
5
5
5
5
5
5
c
c
•J
120
140
200
220
240
300
320
340
400
420
440
500
520
540
600
620
640
700
720
740
800
820
840
900
920
940
3.822 3.955 0.881 0.22
2.198 2.219 0.520 0.23
1.630 1.742 0.059 0.03
1.484 1.510 0.576 0.38
1.599 1.624 0.377 0.23
2.569 2.697 0.076 0.03
1.662 1.696 0.341 0.20
1.769 1.934 0.190 0.10
1.054 1.114 0.459 0.41
1.471 1.455 1.217 0.84
2.181 2.236 0.433 0.19
1.338 1.419 0.680 0.48
3.315 3.503 0.180 0.05
3.060 3.212 0.220 0.07
1.992 2.074 0.379 0.18
1.765 1.836 0.336 0.18
0.920 0.896 0.072 0.08
0.636 0.523 0.230 0.54
1.809 1.856 0.422 0.23
1.571 1.577 0.173 0.11
2.058 2.115 0.292 0.14
1.910 2.000 0.519 0.26
1.181 1.165 0.345 0.30
2.880 3.013 0.891 0.30
4.236 4.402 0.537 0.12
3.339 4.000 1.752 0.44
Average
Standard deviation
Regression Output:
Constant
Std Err of Y Est
R Squared
No. of Observations
Degrees of Freedoa
X Coeff. 0.9576
Std Err 0.0048
-0.13
-0.02
-0.06
-0.03
-0.02
-0.13
-0.03
-0.17
-0.06
0.02
-0.10
-0.08
-0.19
-0.15
-0.08
-0.07
0.02
0.11
-0.05
-0.01
-0.06
-0.09
0.02
-0.13
-0.12
-0.16
-0.125
0.100
-0.01
-0.05
0.01
-0.01
-0.02
0.04
-0.01
0.11
0.04
-0.05
0.03
0.04
0.06
0.04
0.02
0.02
-0.04
-0.11
-0.01
-0.04
-0.01
0.03
-0.04
0.03
-0.05
0.01
0.000
0.069
0.0233
0.0699
0.9982
72
70
-------
Section No. 4.2.7
Revision No. 0
SIMULATED CIS AUDIT ££ II Ss0pf 20
Propeller Vane (P— V—W) as Collocated Transfer Standard
Regression Analysis
Y = 0.023 + 0.96 X
Data: Boulder Experiment
9/4-5/1982
72 — 20 minute averages
CO
2468
CIS (P-V-W) Wind Speed (m/s)
Figure 4.2.7.9 XY plot of simulated wind speed audit data
en
0.2 -•
CO
1-0.2
A A
KEY: A = Sigma/Speed
A A
KEY: + * C-V-W residual
1 1 I t
r i
QL
'co
1.2'
°-6
CO
CO
(—
o
0.4
0-00
5: 0 2 4 6 8
Wind Speed (m/s) of CIS (P-V-W)
Figure 4.2.7.10 flesiduaj analysis of a simulated speed audit
10
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 13 of 20
4.2.7.2.3 "W" Anemometers
Some stations measure the vertical component of the wind, with
an anemometer sensitive only to the vertical component of the wind. A
vertically mounted helicoid propeller, or "W" propeller, is the most common
instrument for this measurement. The same audit methods can be used as are
used on a propeller anemometer. A synchronous or d.c. motor will challenge the
rate of rotation vs. wind speed and a torque device can be used to find the
starting threshold. The common manufacturer's recommended practice is to use a
different transfer function for the W propeller than the one applied when the
same propeller is used for vane-oriented speed or for the N-S and E-W
components of a UW anemometer.
4.2.7.3 Wind Direction
4.2.7.3.1 Sensor control
The first thing to do on a performance audit of a direction
vane is to record the as-found orientation value. Have the operator hold the
vane so that it points to or from (whichever is most accurate for aiming) the
distant orientation target. Verify the alignment by viewing the vane and
target from the ground. Move back away from the tower or mast on a plane which
passes through the sensor and the target and verify that the vane is in the
plane also. Field glasses or a theodolite, can help make this sometimes
difficult observation. The vane must be held steady or clamped until a
constant output exists for a few minutes. Record this value.
The controlled condition for a wind vane is a relative position
of the vane with respect to the sensor housing. There are several ways to
impose a series of known relative positions on the vane-sensor combination.
They vary in effective accuracy. It is critical to know the time constant of
the direction circuit BEFORE starting the performance audit. It can be
measured by setting the vane to a known direction, simulate a wind from 090
holding the vane steady until the 090° (or voltage equivalent) output is
steady. Move the vane quickly (< 1 s) to 270° and measure the time constant of
the system. Assume that a time constant of 3 s is measured. Table 4.2.7.2
shows the change in output angle and voltage (assuming a 540 format and 5V
output) as a function of time.
Table 4.2.7.2 - Time Constant Effects
Time
Constant
(No. )
0
0.2
1
2.3
3
4.6
6.9
Time
Angle
(sec. )
0
0.5
3
6.9
9
13.8
20.7
Vane
Angle
(deg. )
090
270
270
270
270
270
270
Angle
(deg. )
090
106
204
252
261
268
270
Error (540@5)
(deg.) (volts)
0
164
66
18
9
2
0
(after Fritschen
0.333
0.981
1.889
2.333
2.417
2.483
2.498
and Gay,
Change
( % )
0.0
9
63.2
90.0
95.0
99.0
99.9
1979)
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 14 of 20
Notice that in this example a 180 shift requires waiting 20
seconds for the reading to be representative of the new position. If a 90
shift is used, 14 seconds will provide an output within 1 of the final value.
If measurable time constants are found, suggest to the operator that the
manufacturer be called about steps which might be taken to modify the circuit to
a minimum time suitable for 60 Hz noise filtering.
The least accurate method for challenging the relative position
accuracy of a wind vane is to point the vane in various directions while still
mounted on the tower. This can provide positions related to external objects
rather than constant angle changes. It is estimated that the accuracy of this
method is two to five degrees, with the exception of a parallel alignment. The
tail vane can be located parallel to a cross arm to within one degree, and held
parallel on a calm day.
A second method puts the operating sensor in a controlled
situation like a room where the electronics are located. The sensor can be
placed at the center of a template with radial lines every 60 . The sensor can
be oriented to the template and the vane
moved and clamped when the vane is parallel
to the radial line. If care is taken to , *«
avoid parallax errors (non-parallel or * !fcf> ,
^ r ' . '>-&$ :s>f_,., .
non-perpendicular observations) this method
can provide relative accuracy on the order
of one degree.
The best method replaces
the vane with a fixture with the capability
of holding the shaft in fixed positions
with respect to the sensor housing.
Fixtures of this type can provide
repeatable position accuracy of < 0.1 .
Figure 4.2.7.11 shows such a device. A
different application of this precise
method uses a theodolite base as the mount
for the sensor. With the vane or vane
substitute held in one position, the base .
can be rotated in very accurate steps.
Theodolite worm gear assemblies divide a
circle in whole degrees with a vernier
adjustment with 0.1 degree index marks far
enough apart to allow easy interpolation to
0.02 degrees, a resolution wasted on the
application of wind direction measurement.
The audit report form
should contain the transfer function used
to convert output voltage to azimuth
degrees. This may include a 540 format
where azimuth values greater than 360 are
reduced by subtracting 360. The report
form should also contain the challenge
progression used by the selected method.
Figure 4.2.7.11
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 15 of 20
For example, Fig. 4.2.7.14 and Figure 4.2.7.15.show an audit method^and audit
form for wind direction which specifies 16 relative angles, each 60 from the
last one, in a clockwise rotation for 420° followed by a counterclockwise
o o
rotation of 480 . This tests a 540 format, provides four samples at 180
(duplicates from each direction) and a duplicate counterclockwise 240 pair.
The report of this series describes the range of relative error resulting from
the shaft position measurement of the sensor (see 4.2.2.2.2.3 for an example).
The starting threshold of the bearing and transducer assembly
should be measured by some method (see 4.2.2.2.1.2 and 4.2.2.2.1.3). If the k
value is not available to convert the torque to threshold speed at some
accuracy angle, the operator should be requested to ask the manufacturer to
provide it for the next audit.
The bearing to the orientation target should be independently
challenged with a method capable of better than compass accuracy. A theodolite
is ideal for finding the bearing to other distant objects. A solar observation
is recommended (see 4.2.4.3.2).
The last activity of the sensor control audit is to repeat the
orientation test described above for the as-found value. The as-left value
will represent any changes the operator may have made and the new orientation,
if the sensor was not keyed for orientation.
4.2.7.3.2 CTS Method
There is no technical need for a CTS audit of direction. No
new information is added by this method to that gained in the sensor control
method. As a parallel example to the simulated CTS speed audit, data from the
same period of time from BAO is shown here, structured as a simulated CTS
direction audit. Table 4.2.7.3, sorted for ascending CTS direction, shows the
20-minute average direction for both the CTS (Young Propeller Vane, P-V-W) and
the subject (Climatronics F460, C-V-W). Also shown is the
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 16 of 20
Table 4.2.7.3 Simulated CTS Direction Audit Data
9/82
Date/
Ti me
5
5
5
5
4
4
4
5
4
4
4
4
5
4
4
4
4
4
5
5
4
4
5
5
5
4
5
5
5
4
4
4
5
5
S
5
5
4
4
5
5
5
5
5
5
4
4
5
4
4
1 5
4
5
3
4
4
4
5
4
4
4
5
5
5
4
4
4
5
4
S
4
4
74O
720
1140
44O
1000
1840
1820
1040
2200
2120
2140
19OO
11OO
1620
1720
1 700
1 ?4O
18OO
1020
620
164O
1920
500
20
0
2320
1120
100
4O
2300
2220
2340
9OO
120
920
6OO
420
2240
1940
520
220
540
820
3OQ
320
21OO
1020
94O
114O
2040
2OO
2020
4OO
34O
13OO
200O
1320
14O
1340
1600
120O
84O
1200
800
1240
1220
154O
240
1520
10OO
1 44O
150O
Y
Subject
Direct! on
(deg. )
110.72
126. 18
136.01
149.21
170.93
171.89
175.12
176.16
17S.63
179.81
180.49
181.55
183.77
184.24
186.51
186.90
187.74
189.05
189.53
191.87
193.01
194.01
196.86
197.68
199.46
201.43
202. 16
203.33
205.00
205.77
205.97
207.65
207. 16
209.25
212.67
214.09
214. 12
216.04
217.59
223.97
229.63
227.38
230.43
232.28
239.91
237.55
239.85
239.28
142.97
247.28
253.86
257.50
260.27
264.41
264 . 86
267.01
269.87
273.76
274.96
275.77
277.95
289.50
299. 15
300.55
301.51
310.99
312.43
318.79
324.82
349.62
359 . 1 3
360.42
X
CTS
Direction
(deg. )
1O8.58
121.60
134.38
148. IS
170.74
171.50
173.71
174.52
175.24
178.59
179.64
180.68
182.95
183.74
185.29
185.82
187.07
187.82
J9O. 19
191.01
191.43
192.68
193.01
197. 17
198.88
201. 11
201.37
202.99
204 . SO
205.64
206.42
207.02
207.29
209.08
212.72
214.27
215.12
215.95
217.16
224. 12
225.53
228.04
229.87
232.98
237. 14
237.75
240.06
241.26
244.21
248.22
254. 16
257.72
257.80
263.64
264.68
266. 1O
270.39
273.13
274. O4
275.35
277.67
289,42
297.87
298.01
300. 18
308.24
310. OO
315.62
322.4''
345.54
355.74
356. 8O
s
CTS
Sigma
(deg. )
39.34
20.76
87.67
6.50
27.02
10.17
30.55
16.28
7.62
4.04
10.87
9 . 00
13.26
60.97
6. 10
8.98
6.66
17.58
27.04
32.37
9.24
9. 1O
3O.4O
4.44
5.99
6.94
22.84
3.45
5.35
6.21
7.90
5.86
11.83
9.57
9.31
a.oo
20.18
S.86
8.76
6.29
17.04
9.38
28.91
8.80
25.28
35.85
42.30
63.08
29.30
12.10
35.26
11.45
2O. 16
12.57
23.11
29. 16
22.09
43.68
34.09
21.27
27.46
29.47
11.74
65.04
22.86
17.77
10.46
21.03
12.26
67.66
30.58
20.56
Y-X
(deg. )
2. 13
4.57
1.63
1.05
0. 18
0.39
1.42
1.63
0.39
1.22
0.85
0.87
0.82
0.50
1.23
1 . 08
0.67
1.22
-O.66
0.85
1.59
1.32
3.85
0.52
0.58
0.33
0.79
0.34
0.50
0. 13
-0.45
0.63
-O. 13
0. 16
-O. 05
-O. 17
-l.OO
0.08
0.42
-0. 15
4. 10
-O.66
0.56
-O.7O
2.77
-0. 19
-0.20
-1.98
-1.25
-0.94
-O.3O
-0 . 22
2.47
0.77
0. 19
O.91
-0.51
0.63
0.92
0.42
0.28
0.08
1.28
2.55
1.33
2.75
2.43
3. 17
2.33
4.08
3.4O
3.62
Average
Standard deviation
Regression Output:
Constant
Std Err o+ Y Est
I* Squared
No. o-f Observations
Decrees of Freedom
X Coe-f-f icient 1.OO498
Std Err o-f Coe-f. O.OO297
0.91
1.34
-0.22090
1.33045
0.99939
72
70
-------
SIMULATED CTS AUDIT
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 17 of 20
Propeller Vane (P-V-W) as Collocated Transfer Standard
en
&i
^
x x
*.x . x& x x
^c*A!«C>o< x xx
-
X x
•^-"-X-y-X- X
X % X
-50
090 180 270 360
CTS (P-V-W) Wind Direction [Theta] (deg.)
Figure 4.2.7.13 Simulated CTS Wind Direction Audit Data
4.2.7.3.3 Vertical Wind Direction, <£
The sensor control method is used for bivane auditing. The
vertical part of the bivane operation is treated in the same way as the
horizontal part, except different fixtures are used. Special fixtures are
required for each bivane design, but the principle is the same. A relative
zero point is set when the fixture is attached to the bivane. That point is
where the vane shaft is perpendicular to the vertical axis of the sensor. From
this starting point where the output should be the equivalent of 0°, the vane
is held in 15° steps until its physical limits are reached, both tail up (+)
and tail down (-). Threshold is very hard to measure on a bivane because of
the static balance conditions of the vane. If the vane is perfectly balanced
and its remains where ever it is physically moved, a force gage measurement at
some distance from its axis of rotation will yield the starting threshold just
as the vane begins to move.
4.2.7.4 TURBULENCE OR . The sigma values
result from how the samples are combined to estimate the statistical
parameters. It is a part of the auditor's job to determine how the algorithm
works and to challenge that process with a known input. This is also a
functional way to document the impact of the signal conditioning time constant
on the measurement of direction variability.
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 18 of 20
The challenge should be realistic or at least within some realistic
range. The challenge must take into consideration the wave shape of the
variable direction imposed on the system in calculating the true sigma value
with which the output will be compared. The effective time constant of the
direction system, calculated from the delay distance of the sensor and some
nominal wind speed important to air pollution applications, should define the
maximum frequency used in the sigma challenge.
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 19 of 20
Wind Vane - MSI method WV004 (version 8/1/84)
This method describes the relative performance of the wind vane
as a shaft-position transducer and the orientation of the
transducer with respect to true North. The former is.done with a
fixture, part of t>/hich is mounted to the transducer body and part
mounted to the shaft in place of the vane. The latter requires a
determination of true North (see MSI method SNOO8) and a setting
of the transducer relative to that orientation.
The report form for this method includes space for the optional
method to define the "open space" where relevant to the sensor
and application. Also there is space to record starting torque
measurements and system time constant estimates for turbulence
measurement.
WV004-A Remove the wind vane assembly (vane, shaft and counter-
weight). A 1/S" shaft is required. If the sensor does not use
that size or it is not accessible, an interface fitting is
required. Mount the disc on the vertical shaft. Mount the clamp
to the support column for the shaft so that the pin engages the
disc and the disc is free to move when the pin is withdrawn. Set
the fixture parts with the pin in the 180 degree hole. Rotate
the clamp until the output indicates ISO, either by equivalent
voltage or digital printout. Since this is & position
measurement, the challenge is constant and instantaneous values
may be used, being sure to react to the needs of the time
constant for stable readings. Move the disc (vane substitute) to
the following positions taking data at each point: 120, 060,
360, 300, 240, 180, 120, ISO, 24O, 300, 360, 060, 120, ISO, and
240 degrees. This moves the "vane" 420 degrees counter clockwise
and then 4SO degrees clockwise to test "540" strategies for the
angle discontinuity.
WV004-B Define the "open space" for 36O degree potentiometer
transducers. Install the index line fixture to an appropriate
position with respect to the protractor mounted to the disc.
Disengage the pin. Rotate the disc until the output changes from
maximum voltage to five degrees less than maximum. Record the
angle to 1/2 degree resolution. Rotate the disc back toward
maximum voltage and record the angle when maximum is first
reached. Rotate the disc until the output changes from minimum
voltage to five degrees greater than the minimum. Record the
angle. Rotate the disc back toward the minimum voltage and
record the angle when minimum is first reached.
WV004-C After having found a distant target of known .direction
(see SNO08), set the vane so that the direction is the output of
the sensor. Clamp the vane to the shaft support tube so that the
output stays constantly correct, even with light wind forces on the
vane. In heavy wind, a fixture replacing the vane is required.
Place the sensor In its mount and rotate the sensor body until
the vane counterweight poirts to the target. Clamp the sensor in
place, check the output and remove the vane clamp. Record one
system data point with all cables in place and the sensor clamped.
Figure 4.2.7.14 A method for auditing a. wind direction sensor
-------
Section No. 4.2.7
Revision No. 0
Date: 17 Sep 89
Page: 20 of 20
PERFORMANCE AUDIT REPORT by WV004
MEASUREMENT SYSTEM - Wind vane
System number
Sen sor
Vane ~_
Loc at i on
Signal conditioner
Data channel _
DATE / / TIME o-f-f line on line test start
TRANSFER FUNCTION (volts per degree)
discontinuity strategy
TEST RESULTS
WV004A
-------
Section No. 4.2.8
Revision No. 0
Date: 17 Sep 89
Page: 1 of 1
4.2.8 ESTIMATING ACCURACY AND PRECISION
4.2.8.1 Measurements
Section No. 4.1.5 contains a detailed discussion of methods of
estimating accuracy, precision and bias using wind speed and wind direction as
examples. That material will not be repeated here. The measurement process
begins with an instrument which has some element or part which is sensitive to
the variable of interest. If interest is in air flow with respect to the
surface, there are two variables, wind speed and wind direction. Each part of
this section discusses various aspects of how the measurements might be made,
calibrated, operated, maintained and documented to support a claim of
measurement validity. It is recognized that instruments are usually parts of
data systems with sampling, processing and summarizing routines designed to
produce final elements of a data base to be used for some application. The
earlier parts of this section were devoted to methods of tracking the
measurement process all the way through the system to the system output.
Accuracy was addressed in terms of how well what was designed to be done was
actually done. The second part of this sub-section will deal with how well the
system design serves the application.
4.2.8.2 Summarized Data
Summarization schemes are many and preclude a full discussion here. The
auditor should define the methods used and comment on the appropriateness of
the method to the application of the summarized data. There may be concurrent
summarizations such as a scalar wind speed, a resultant vector wind speed and
some kind of summarized wind direction. The accuracy of the data system should
reflect estimated errors because of an inappropriate summarization program.
A software analysis is required to be sure that the declared method of
summarization is in fact being accomplished by the computer program. For
example, is the scalar average wind direction avoiding the error of averaging a
circular range with a discontinuity? If the average of winds ranging between
300° and 060 turns out to be about 180°, this problem still exists.
-------
Section No. 4.2.9
Revision No. 0
Date: 17 Sep 89
Page: 1 of 3
4.2.9 REFERENCES
Acheson, D. T.,1970: Response of cup and propeller ratios and wind direction
vanes to turbulent wind fields. Meteor. Monogr., No.33, Amer. Meteor. Soc.,
pp.252-261.
Acheson, D. T.,1988: Comments on "Anemometer Performance Determined by ASTM
Methods." J. Atmos. Oceanic Technol., 5 ,pp. 381-382.
ASTM,1984: Standard Practice for DETERMINING THE OPERATIONAL COMPARABILITY OF
METEOROLOGICAL MEASUREMENTS, D 4430-84. Amer. Soc. for Testing and
Materials, Philadelphia, PA 19103.
ASTM,1985a: Standard Test Method for DETERMINING THE PERFORMANCE OF A CUP
ANEMOMETER OR PROPELLER ANEMOMETER (Draft 6 of D22.ll). Amer. Soc. for
Testing and Materials, Philadelphia, PA 19103.
ASTM,1985b: Standard Test Method for DETERMINING THE DYNAMIC PERFORMANCE OF A
WIND VANE (Draft 8 of D22.ll) Amer. Soc. for Testing and Materials,
Philadelphia, PA 19103.
Blackadar, A. F.,1985: Almanac for a weather station. Heldref Publications,
Washington, DC 20016.
Baynton, H. W.,1976: Errors in wind run estimates from rotational anemometers.
Bull. Amer. Meteor. Soc., 57, 1127-1130.
Box, G. E. P., W. G. Hunter and J. S. Hunter, 1978: Statistics for
experimenters. John Wiley & Sons, ISBN 0-471-09315-7.
EPA, 1987a: Ambient Monitoring Guidelines for Prevention of Significant
Deterioration (PSD), EPA-450/4-87-007, Office of Air Quality Planning and
Standards, Res. Triangle Park, NC 27711.
EPA, 1987b: On-Site Meteorological Program Guidance for Regulatory Modeling
Applications, EPA-450/4-87-013, Office of Air Quality Planning and
Standards, Res. Triangle Park, NC 27711.
Finkelstein, P. L.,1981: Measuring the dynamic performance of wind vanes.
J. Appl. Meteor., 20, pp. 588-594.
Finkelstein, P. L.,J. C. Kaimal, J. E.Gaynor, M. E.Graves and T. J.Lockhart,
1986: Comparison of Wind Monitoring Systems. Part I: In-Situ Sensors.
J. Atmos. and Oceanic Technol., 3, pp. 583-593.
Finkelstein, P. L., J. C. Kaimal, J. E.Gaynor, M. E.Graves and T. J. Lockhart,
1986: Comparison of Wind Monitoring Systems. Part II: Doppler Sodars.
J. Atmos. and Oceanic Technol. , 3, pp. 594-604.
Fritschen, L. J. and L. W. Gay, 1979: Environmental instrumentation.
Springer-Verlag, N.Y. ISBN 0-07-033175-8.
-------
Section No. 4.2.9
Revision No. 0
Date: 17 Sep 89
Page: 2 of 3
Gill, G. C.,1967: On the dynamic response of meteorological sensors and
recorders. Proceedings of the First Canadian Conference on
Micrometeorology, Part 1. Meteorological Service of Canada, Toronto.
Gill, G. C.,1973: The Helicoid Anemometer. Atmosphere, 11, 4, pp. 145-155.
Hayashi, T., 1987: Dynamic response of a anemometer. J. Atmos. Oceanic
Techno 1., 4, pp. 281-287.
Hoehne, W. E.,1973: Standardizing Functional Tests. IEEE Transactions on
Geoscience Electronics, Vol GE-11, No. 2, April.
Huschke, R. E.,1970: Glossary of Meteorology. Amer. Meteor. Soc., Boston, MA
02108
Kaimal, J. C. and J. E. Gaynor, 1983: The Boulder Atmospheric Observatory,
J. Appl. Meteor., 22, pp. 863-880.
Kaimal, J. C., J. E. Gaynor, P. L. Finkelstein, M. E. Graves, and
T. J. Lockhart, 1984: A field comparison of in situ meteorological
sensors. NOAA/BAO Report No. Six.
Kondo, J., G. Naito, and Y. Fujlnawa, 1971: Response of Cop Anemometer in
Turbulence. J. Meteor. Soc. of Japan, 49, pp.63-74.
Lockhart, T. J., 1977: Evaluation of rotational anemometer errors. Bull. Amer.
Meteor. Soc., 58, pp. 962-964.
Lockhart, T. J., 1978: A field calibration strategy for rotating anemometers
and wind vanes. Proceedings of the 4th Symposium on Meteorological
Observations and Instrumentation, Denver CO, April 10-14. pp. 57-60.
Lockhart, T. J., 1985a: Some cup anemometer testing methods. J. Atmos. Oceanic
Techno 1. , 2, pp. 680-683.
Lockhart, T. J., 1985b: Wind-Measurement Calibration. Bull. Amer. Meteor.
Soc. , 66, p.1545.
Lockhart, T. J., 1987: Performance of an anemometer determined by the ASTM
method. J. Atmos. Oceanic Technol. , 4, pp. 160-169.
Lockhart, T .J., 1989 Accuracy of the collocated transfer standard method for
wind instrument auditing. J. Atmos. Oceanic Technol., 6, pp. 715-723. 6/88)
MacCready, P. B.,Jr. and H. R. Jex,1964: Response characteristics and
meteorological utilization of propeller and vane wind sensors. J. Appl.
Meteor., 3, pp. 182-193.
MacCready, P. B.,Jr.,1965: Dynamic response characteristics of meteorological
sensors. Bull. Amer. Meteor. Soc., 46, 533-538.
MacCready, P. B.»Jr.,1966: Wind speed measurements in turbulence. J. Appl.
Meteor., 5, pp. 219-225.
-------
Section No. 4.2.9
Revision No. 0
Date: 17 Sep 89
Page: 3 of 3
Middleton, W. E. K. and A. F. Spilhaus, 1953: Meteorological Instruments.
University of Toronto Press.
Natrella, M. G.,1966: Experimental Statistics. National Bureau of Standards
Handbook 91.
Snow, J. T., D. E. Lund, M. D.Conner, S. B. Harley and C. B. Pedigo.1989: On
the dynamic response of a wind measuring system. J. Atmos. Oceanic
Techno!., 6, pp. 140-146.
Stearns, C. R.,1985: Wind-Measurement Calibration, Response. Ball. Amer.
Meteor. Soc. , 66, p.1545.
Sutton, 0. G., 1955: Atmospheric Turbulence.
Turner, D. B.,1986: Comparison of three methods for calculating the standard
deviation of the wind direction. J. Climate Appl. Meteor., 25, pp. 703-707.
Wieringa, J., 1967: Evaluation and design of wind vanes. J. Appl. Meteor.
6, pp. 1114-1122.
-------
-------
Section No. 4.3.0
Revision No. 0
Date: 17 Sep 89
Page: 1 of 3
Section 4.3
QA FOR TEMPERATURE AND TEMPERATURE GRADIENTS (AT)
OUTLINE
Section Pages Rev. Date
4.3.0 OUTLINE AND SUMMARY 3 0 9/89
4.3.1 TYPES OF INSTRUMENTS 4 0 9/89
4.3.1.1 TRANSDUCERS
4.3.1.1.1 Thermistors
4.3.1.1.2 Wire Bobbins
4.3.1.1.3 Mercury-in-glass Thermometers
4.3.1.1.4 Thermocouples
4.3.1.2 SIGNAL CONDITIONING
4.3.1.3 ASPIRATED RADIATION SHIELDS
4.3.2 SPECIFICATIONS 4 0 9/89
4.3.2.1 DELTA TEMPERATURE (AT)
4.3.2.2 TEMPERATURE
4.3.3 ACCEPTANCE TESTING . 2 0 9/89
4.3.4 INSTALLATION 1 0 9/89
4.3.5 CALIBRATION 9 0 9/89
4.3.5.1 EQUIPMENT AND METHODS
4.3.5.2 DELTA TEMPERATURE (AT)
4.3.5.3 TEMPERATURE
4.3.5.4 CALIBRATION REPORT EXAMPLE
4.3.6 OPERATIONS, MAINTENANCE AND QUALITY CONTROL 3 0 9/89
4.3.6.1 OPERATIONS
4.3.6.2 MAINTENANCE
4.3.6.3 QUALITY CONTROL
4.3.7 PERFORMANCE AUDIT METHODS 8 0 9/89
4.3.7.1 GENERAL CONSIDERATIONS
4.3.7.1.1 Who
4.3.7.1.2 What
4.3.7.1.3 Where
4.3.7.1.4 When
4.3.7.1.5 How
4.3.7.2 DELTA TEMPERATURE (AT)
4.3.7.2.1 Sensor Control Method
.4.3.7.2.2 CTS Method
-------
Section No. 4.3.0
Revision No. 0
Date: 17 Sep 89
Page: 2 of 3
Section Pages Rev. Date
4.3.7.3 TEMPERATURE
4.3.7.3.1 Sensor Control
4.3.7.3.2 CTS Method
4.3.8 ESTIMATING ACCURACY AND PRECISION 1 0 9/89
4.3.8.1 MEASUREMENTS
4.3.8.2 SUMMARIZED DATA
4.3.9 REFERENCES 1 0 9/89
-------
Section No. 4.3.0
Revision No. 0
Date: 17 Sep 89
Page: 3 of 3
QUALITY ASSURANCE FOR TEMPERATURE
AND TEMPERATURE GRADIENT (AT)
4.3.0 SUMMARY
The measurement of temperature is standardized in great detail by all those
organizations interested in such procedures, ASTM, ISA (Instrument Society of
America), and IMS (Temperature Measurement Society). The problem with
meteorological applications is that the free air temperature is required. This
means the transducer needs to be exposed to the atmosphere which is in turn
exposed to the sun about half the time and to the very cold outer space the
other half. The atmosphere is conditionally transparent to heat sources (the
sun) and sinks (outer space) so shielding must accommodate a wide range of
radiative conditions. Wind also influences the temperature shield. It
transfers heat to and from the shield in a variable way as a'function of wind
speed. Most effective shields use forced aspiration to expose the transducer
to nearly unmodified outside air. Wind speed may also play a variable role in
the performance of the aspiration system.
This section concentrates on the meteorological applications of air
temperature measurement and the differential temperature measurements which are
interpreted as temperature gradients and applied as a measure of vertical
stability.
Since the application of the measurement should define the accuracy needed,
both the relatively course air temperature and the relatively fine temperature
difference measurements will be considered.
-------
Section No. 4.3.1
Revision No. 0
Date: 17 Sep 89
Page: 1 of 4
4.3.1 TYPES OF INSTRUMENTS
There are several materials and structures which change in some way as a
function of temperature. General books on meteorological instruments such as
Mason and Moses (1984), Middleton and Spilhaus (1953) and particularly Brock
and Nicolaidis (1984) will provide details on a variety of these sensors. From
the standpoint of quality assurance, a few basic principles and a few standard
types will represent the vast majority of instruments in use for air quality
applications.
The measurement of temperature for air quality applications is generally
thought of as either air temperature, T, or a difference between two
temperature measurements, AT. The application of these different measurements
require different specifications and auditing methods. The two types of
temperature measurement will be treated in later sections as separate
measurements.
Temperature instruments are made up of three important parts. The
transducer is the device which changes its electric value as a function of the
temperature of the transducer element. The signal conditioner and cables
convert the electric value to a recordable output, usually volts. The
aspirated radiation shield is the mounting structure which holds the transducer
in the atmosphere where the temperature is to be monitored. Each of these
three parts will be discussed separately since there are various combinations
possible.
4.3.1.1 Transducers
Consider the transducer as the part containing the sensing element. In
most cases, the sensing element is the transducer in air quality monitoring
applications. The element is usually a thermistor (or thermistor network) or a
winding of fine wire on an insulated bobbin. It could also be a thermocouple
or a circuit element like an integrated circuit (see Cole, 1978). The elements
are usually encased in a protective capsule and sealed. From an operational
standpoint it only matters how the sensor reacts to the temperature inside the
aspirated radiation shield.
4.3.1.1.1 Thermistors
The thermistor is an electronic semiconductor made from certain
metallic oxides, such as nickel, manganese, iron, cobalt, copper, magnesium,
titanium and other metals. It is a nonlinear element. One common supplier
(Yellow Springs Instrument Co. [YSI]) sells both the standard thermistor and
the "linear" thermistor. Table 4.3.1.1 shows a typical negative thermal
response curves of raw thermistors and the neariy linear response of the
network thermistor. Also shown is the positive response of two platinum RTD
(resistance temperature detector or resistance thermal device) for contrast.
Notice the large average change per °C with the YSI bead thermistor between
10°C and 20°C (222Q) as compared to the network thermistor (126Q) or the 100Q
RTD (0.4J2) or the 1000Q RTD (3.8«). The raw bead thermistors are included
because, in the future, microprocessor-based data systems can handle nonlinear
transducers as easily as linearized ones. The "linearized" YSI has a small
oscillating error of about ± 0.1°C with a wave length of about 40°C. The
impact of this error on temperature difference systems is small. At the
-------
Section No. 4.3.1
Revision No. 0
Date: 17 Sep 89
Page: 2 of 4
Table 4.3.1.1 - Sensor Resistance vs.Temperature
T
(°C)
-10
0
10
20
30
40
50
Thermistors
YS I YS I
44031
(ohms) (An)
16600
-684.0
9796
-382.5
5971
-222.3
3748
-133.1
2417
-81.9
1598
-51.7
1081
44203 network
(ohms) (Afl)
13438
-127.9
12159
-125.3
10906
-126.1
9645
-128.6
8359
-128.8
7072
-123.6
5836
Platinum RTDs
MINCO
5-100 Pt
(ohms) (AH)
96.09
0.391
100.00
0.390
103.90
0.389
107.79
0.388
111.67
0.387
115.54
0.385
119.39
HY-CAL
looon Pt
(ohms) (An)
961.84
3.816
1000.00
3.804
1038.04
3.792
1075.96
3.780
1113.76
3.768
1151.44
3.756
1189.00
(Afl) is ohms per degree C for the 10 degree range
VC T —-._ — ___ V*
sllow Springs, OH 45387
MINCO 7300 Commerce Lane, Minneapolis, MN 55432
HI-CAL 9650 Telstar Ave. El Monte, CA 91731-3093
steepest slope it is 0.025 C per degree difference. A AT would need to be ±
4 C before the error reaches 0.1 C, at which point the error is moot with
respect to application.
The big advantage to thermistors is the relatively large
resistance of the element with respect to the resistance of the signal cable.
Lockhart and Gannon, 1978, pointed out the fact that it would take a 12.5n
difference in cable resistance to two sensors of a AT pair to cause a 0.1°C
error (bias) in the AT measurement. Most signal conditioning circuits have the
capability of adjustment to eliminate such a bias. Another little known
advantage is stability. A several year stability test conducted at NBS showed
thermistors to be extremely stable. This is contrary to early experience which
suggested that thermistors often failed with a shift in the transfer function
(ohms vs. temperature) gaining them a reputation of instability. Better
packaging designs and better handling practices have eliminated many of these
problems.
4.2.1.1.2 Wire bobbins
The resistance of a wire changes with temperature. If a long
piece of fine wire can be handled in some way, it can be used as a temperature
element. Winding the wire on a non-conductive bobbin is the traditional method
of handling. When the bobbin is large and open, a very fast response sensor is
made. When the bobbin is potted in a stainless steel jacket, a more
traditional slower response sensor is made. The transducer in both cases is a
length of wire. Different metals have different temperature coefficients. A
-------
Section No. 4.3.1
Revision No. 0
Date: 17 Sep 89
Page: 3 of 4
O
40 gauge (O.OS mm diameter) wire at 20 C made of annealed copper has a
resistance of 3.4n/m. If it were made of German silver the resistance would be
65.9Q/m. Platinum and iron are each 20.0 Q/m while nickel is 15.6 fl/m.
Manufacturers, choosing for stability, ease of handling and cost for a suitable
resistance, have settled on a few materials. Platinum (100Q) is the most
common for meteorological applications. Nickel-iron is another common wire
providing a higher resistance at a lower cost with good stability.
Because of the small resistance change for a 1°C temperature
change (0.4T2), the transducer resistance must be measured with both high
resolution and attention to cable resistance. Three and four wire bridge
circuits are commonly used, the latter being best for handling long cables. It
will be shown in the sections on calibration and auditing that many of the
details of the transducer and circuits need not be known. Only the system
performance is important.
4.3.1.1.3 Mercury-in-glass thermometers
These thermometers are not transducers, but they are commonly
used for calibrations. Some styles have enough resolution to be read to 0.02°C
with some care. The ASTM series of Precision thermometers are examples of
these. They are 37.9 cm (15 inches) long and breakable (and expensive - $50 to
$80 each). They are also calibrated for total immersion. The ASTM 62C has a
range from -38 to +32°C and scale divisions of 0.1. The ASTM 63C has a range
of -8 to +32 C and also has scale divisions of 0.1, but they are farther apart
making interpolation more accurate. If higher temperatures are needed, the
ASTM 64C has a range of 25 to 55°C.
4.3.1.1.4 Thermocouple systems
The thermocouple operates on the principle that when two
different metals are joined, a small voltage with a temperature- dependent
magnitude is generated. By comparing this voltage to the voltage generated by
a second thermocouple in a thermally stable environment of known temperature,
the temperature of the first thermocouple can be determined. Because of
complex circuitry and problems with conductors, thermocouple systems are no
longer popular transducers for meteorological monitoring.
Thermocouple pairs are well suited for differential temperature
measurement. They provide the same voltage for any size wire which makes them
ideal for miniature fast response applications.
4.3.1.2 Signal Conditioning
There are a multitude of circuits which will measure resistance.
Usually the transducer and the signal conditioner are purchased as a system,
complete with interconnecting cables. This is advisable since the range of
resistance vs. temperature is quite large. Signal conditioning circuits may be
adjusted to conform to individual transducers or transducer pairs. They may
also be adjusted to a generic or theoretical curve or transfer function. It is
important to understand the function of the signal conditioner and to treat it
as a part of a system along with the transducers and the cables.
-------
Section No. 4.3.1
Revision No. 0
Date: 17 Sep 89
Page: 4 of 4
4.3.1.3 Aspirated Radiation Shields
There are many kinds of shields, as Figure 4.3.1.1 depicts. Most of the
error in measuring air temperature comes from the shield. It is also true that
the magnitude of the error is largely unknown. In Section 4.3.2
SPECIFICATIONS, there is a detailed discussion about accuracy of shield
performance.
The measurement of the temperature of the free atmosphere at the point
where the shield intake is located is the goal. A shield protects the sensor
from radiation and provides the mechanical mount for the sensor on a tower or
mast. If the shield is not aspirated, or designed for effective natural
ventilation, it may become little more than a larger sensor case resulting in
the same radiation errors as an unshielded sensor but with a longer time
constant. To avoid this difficulty, it is necessary to draw the air into the
shield in such a way that it is not modified by the shield temperature but will
come to equilibrium with the transducer at some average time.
Forced aspiration is the only way to minimize radiation error for all
conditions. A fan draws air in past the transducer at a speed suitable for
minimum error. Forced aspiration can be designed to provide a flow in the
right direction under all ambient wind conditions. Insufficient pressure drop
in the fan during strong winds may allow reverse flow to occur transporting the
heat from the fan back to the transducer (Lockhart, 1975).
COW PtnIM fit Plttt Shield
Tflgdynt Aspvttfd
Ridiftton Shiald
AES Str»»n*on Scrim $
Itrtuli Thmlltr Shi»ld
Climtt M&tof Asptritid Shtftd
KmM SeH-Anpirmting ShMd
ModUi«t flume* ShiuU
Gill NtiunOr
HMud ShMd
ACS Mtrint Shnkt
EG &GD*w Point
Hygnmuttr Shitld
AES Oufl Atpmttd PtYCfmmtttr
Cum* Plmt SluuU
Gill Aspirtttd Shnld
Figure 4.3.1.1 Examples of various radiation shields
(McKay and McTaggart-Cowan, 1977)
-------
Section No. 4.3.2
Revision No. 0
Date: 17 Sep 89
Page: 1 of 4
4.3.2 SPECIFICATIONS
The purpose of defining specifications is to give unambiguous meaning to
the terms used by all those who are concerned that the instruments and systems
selected and operated will meet the needs of the application or project. This
starts with procurement specifications and ends with supporting claims of data
quality. These specifications provide the basis for receiving inspection and
testing.
Project and application requirements vary. To make this handbook as
specific as possible, the examples used will be consistent with those presented
in the On-Site Meteorological Program Guidance for Regulatory Modeling
Applications (EPA, 1987b). The specifications for temperature are range and
accuracy. The performance of the radiation shield is not defined by
specification. There is an implication that the accuracy requirements include
this error source, but if they do there is no way suggested to verify the
performance of the shield.
4.3.2.1 Delta Temperature (AT)
The only requirement in EPA (1987a) regarding the vertical temperature
difference is "Errors in measured temperature difference should not exceed
0.003°C/m." This rate is based on a 0.15°C accuracy for a 50m separation.
The requirement came from a time when AT was traditionally measured between the
lower 10m level and the upper 60m level on a tower. If a shorter tower is
used, like a 44m tower, the separation between 10m and 44m, namely 33m, would
show smaller lapse rates and inversions. If the same accuracy were to be
preserved in measuring the equivalent or representative AT from the shorter
tower, the measurement accuracy had to be better, 0.1 C in this case. Some
operators went to even shorter towers and to assure an appropriately accurate
measurement system, the requirement was stated as a per meter error.
The above requirement is impossible to meet with the new 10m towers
including AT. If the aspirated radiation shields are mounted at 2m and 9m (to
avoid interference with the 10m wind), the requirement is to not exceed an
error of 7 x 0.003 = 0.021°C. This is an accuracy which is hard to prove, let
alone achieve.
The dynamic range for a AT installation on a 60m tower might be from
-2 C to + 15 C. Convention for positive and negative AT is:
(-) a lapse rate is the normal decrease of air temperature with
height limited by the auto convection rate of 3.4 C/100 m or
0.034°C/ m. A lapse rate produces a negative AT.
( + ) an inversion is the inverted lapse rate or an increase of air
temperature with height. There is no limit for inversion strength.
An inversion produces a positive AT.
The dynamic range between 2m and 9m is not much different than that between 10m
and 60m. During the EPA-BAO experiment in 1982 (Lockhart, 1988), a pair of AT
sensors was mounted on tower 4 at 2m and 8.6m. Each sensor was a lOOfl Rosemont
platinum RTD in a Young aspirated shield. On tower 3, also at 2m and 8.6m, a
pair of MRI-YSI linearized thermistors mounted in MRI shields (patterned after
the Young shields) were operated. Figures 4.3.2.1 and 4.3.2.2 show three days
of 20-minute average AT data and 2m temperature data from tower 3. .Also shown
-------
Section No. 4.3.2
Revision No. 0
Date: 17 Sep 89
Page: 2 of 4
KEY: MRI AT
+ MRI-BAO
-2
0 6 12 18 0 6 12 18 0
9/7/82 9/8/82
6 12 18 0
9/9/82
Figure 4.3.2.1 MRI AT Data and AT Difference Data for Three Days
CJ
o
o
i_
Q.
10 -
0
6 12 18 0
9/7/82 9/8/82 9/9/82
Figure 4.3.2.2 -Air Temperature at 2 m from the MRI AT Pair
-------
Section No. 4.3.2
Revision No. 0
Date: 17 Sep 89
Page: 3 of 4
is the difference between the two 20-minute average AT measurements from the
two towers. Note the dynamic range of -1.3 C to +4.5 C (per 6.6m) and the
agreement between the two different instrument systems. For the three days the
average difference is -0.02°C with a standard deviation of 0.10°C. The daily
temperature range was about 17 C.
The reason the dynamic range is so high for such a small separation
distance is that the surface is a better radiation receiver and transmitter
than is the air immediately above it. The surface is almost always hotter or
colder than the air above it. Convection and mechanical turbulent mixing drive
almost all of the heat flux between the surface and the air. The closer the
sensors are to the surface the larger the temperature difference per meter of
separation. The drawback is that.the closer the AT pair is to the surface the
more sensitive the differential measurement is to local surface conditions or
character. A black top road will affect a 2-10m AT much more than it will
affect a 10-60m AT. The lower sensor really drives the AT and a 2m temperature
will vary more than a 10m temperature.
For these reasons, the suggested procurement specification in the
On-Site guide (EPA, 1987b) in 8.1.3 reads:
"Range -5 to +15 degrees C.
Relative accuracy (error) sO.l degrees C."
While calibrations and audits of both accuracy and relative accuracy are
usually conducted in controlled environments, the measurement is made in the
atmosphere. The greatest source of error is usually solar radiation. Solar
radiation shield specification is therefore an important part of the system
specification. Motor aspirated radiation shields (and possibly naturally
ventilated shields) will satisfy the less critical temperature measurement. It
is critical that the same motor aspirated shield design be used for both
sensors used to measure AT. The expectation is that the errors from radiation
(likely to exceed 0.2 degrees C) will zero out in the differential measurement.
A motor aspirated radiation shield specification might read:
"Radiation range -100 to 1300 W/m2
Flow rate 3 m/s or greater
Radiation error <0.2 degree C."
Data sheets from five manufacturers (listed alphabetically) specify
their aspirated radiation shields as follows:
1. Climatronics TS-10 Under radiation intensities of 1100 W/m2
measurement errors due to radiation will not
exceed 0.1°C. Aspiration rate 3 m/s at sensor
location.
2. Met One 076 Radiation error -less than 0.05°F (0.03°C)
under maximum solar radiation of
1.6 gm-cal/cm2/min (1100 W/m2).
Flow rate 500 ft/min (2.5 m/s).
3. Qualimetrics 8150-A Radiation error - 0.05°C during maximum aspiration
and full sun. Air speed 360 ft./min. (1.8 m/s).
-------
Section No. 4.3.2
Revision No. 0
Date: 17 Sep 89
Page: 4 of 4
Teledyne Geotech 327C
5. R. M. Young 43408
Shielding - Under test radiation flux density of
1100 W/m errors caused by radiation are less
than 0.1 C. Aspiration rate 6 m/s at sensor
location.
Radiation error - under radiation intensity of
1080 W/m , Ambient temperature - 0.2°C RMS,
Delta T - 0.05°C RMS with identical shields
equally exposed. Aspiration rate - 3 m/s.
It is difficult not to notice the similarity among these very different
designs. An auditor would need a comparative field test to find the relative
error from solar radiation. Such a test can be done using a AT sensor pair
with two (or more) shields collocated at the same level. If the transducers
are well calibrated, the relative temperature of the transducers can be known
to 0.02°C. these transducers in the two aspirated radiation shields will
report the relative performance of the shields to the same relative accuracy.
The one which is coolest in the daytime and warmest at night has the least
radiation error. Several diurnal cycles with sunny days and clear nights are
required. Such a test series could identify the most efficient shield which
could become a standard against which a relative error analysis of any shield
could be made by a CTS method.
4.3.2.2 Temperature
EPA
The accuracy specification for temperature is suggested in
:i987b) as
"Range
Accuracy (error)
-40 to +60 degrees C.
30.5 degrees C."
Some applications such as "PSD" permits without fog problems require an
accuracy of only 1 degree C. For locations with winds generally above 1,5 m/s,
a well designed naturally aspirated shield can provide 1 degree C. accuracy.
When the application requires an aspirated radiation shield, the shield
performance requirement should also be included in the specification. It is
customary to use the lower AT shield to aspirate both the temperature and half
the AT pair. Some designs develop AT by subtracting one temperature
measurement from the- other, in which case there is only one sensor in the lower
shield.
-------
Section No. 4.3.3
Revision No. 0
Date: 17 Sep 89
Page: 1 of 2
4.3.3 ACCEPTANCE TESTING
There are two ranges of temperature to consider. One is the measurement
range and the other is the environmental operating range. The two might be
similar for remote installations without commercial power or air conditioned
shelters. The operating temperature range of the signal conditioning circuits
for remote installations is a function of the radiation shielding of the
electronics and the heat generated by the circuit operation. The electronics
may get both colder and hotter than the air temperature under these
circumstances. The operating temperature range for a station with an air
conditioned shelter is much narrower than the measurement range. The receiving
or acceptance test design must consider these factors. A conventional
temperature chamber is required to control the temperature of the electronics
while the sensors are controlled by a separate thermal environment.
A test which demonstrates the operation of AT and T sensors connected to
their signal conditioners is recommended. The test should include at least two
temperatures. Liquid bath or solid thermal mass devices are recommended to
avoid local gradients in the air. It is necessary to remove the sensors from
the aspirated radiation shield and place them in a bath while they are still
connected to the signal conditioner. This may prove difficult to do with some
designs. Since calibrations and audits are likely to require the same
accessibility of the sensors in the field, it is best to solve this potential
difficulty in the less hostile receiving laboratory space. Perhaps a statement
of accessibility needs to be included in the specification.
Experience has demonstrated that thermal stress can cause sensor failure
when baths are used. The sensors may seem to be sealed, but if a sensor is
submerged in a hot (40°C) bath for enough time to reach equilibrium and take a
series of measurements, and then submerged in a cold (0°C) bath, the pressure
change inside the metal sensor cover may draw water into the element chamber.
It is prudent to assume that the sensors are not hermetically sealed and
to protect them as much as possible. Using solid thermal mass devices is one
way to avoid liquid from wicking or being drawn into the sensor, but it is not
a total protection. In the above example, room air may be drawn into the
cooling sensor as the pressure inside equalizes with ambient pressure. The air
drawn in may be saturated at the new cold temperature or may condense some
water vapor. Whether or not this is a problem depends on how the sensor is
made, but it is best to take what precautions are possible. If complete
immersion is necessary, wrap the sensors in plastic so no liquid can get to the
interface where the wires come out of the sensor. Use partial immersion where
possible keeping the interface dry. Keep temperature changes small and in the
order AMBIENT^COLTw AMBIENT-»HOT-»AMBIENT.
Assume the receiving test will use two temperatures, ambient temperature
and an ice slurry. Assume a watsr bath will be used for the ambient test. If
one to three sensors are in the system, wrap them together along with a
thermometer using a rubber band. Use a Thermos bottle which has been filled
with water several hours earlier. The key to accurate temperature measurements
with sensors of different time constants is in having a thermal mass with
minimal gradients. The value of a Thermos bottle is its long time constant.
It will tend to keep the temperature of its contents constant, but all it can
do is cause the heating or cooling to be slow. In time the contents will be at
-------
Section No. 4.3.3
Revision No. 0
Date: 17 Sep 89
Page: 2 of 2
the same temperature as the surroundings, if the temperature of the
surroundings is constant, like an air conditioned room. When this equilibrium
has been reached, the water in the bottle will be at the same temperature
everywhere in the bottle and stirring is unnecessary. Stirring suggests the
need to mix up parts of different temperature. It is better not to have parts
of different temperature and this can be achieved by reaching equilibrium with
a well insulated mass.
Place the sensors and thermometer into the Thermos. Use a cork or some
cover to keep ambient air from circulating over the top surface of the water
mass. After about 30 minutes, assume the sensors and thermometer are in
equilibrium (they should be, they went in from the same equilibrium
temperature) and start a series of measurements. Take five measurements about
five minutes apart. If the measurements are constant rounded to the nearest
tenth degree C, average the five readings and use them to describe the
temperature with respect to the thermometer and AT with respect to zero
difference. If the measurements are slowly increasing with respect to the
slower thermometer, there is a self-heating error. Any resistance element will
get warm when current flows through it. It is expected that the self-heating
will be small and the large thermal mass will carry the heat away without
detection. It is also possible that the elements are sampled and do not have
current flowing continuously. If self heating is detected, or if you wish to
shorten the time to equilibrium, some mixing of the water in the Thermos may be
useful.
After the response to ambient temperature has been recorded, place the
assembly of sensors and thermometer in a Thermos bottle containing an ice
slurry. The ice should be made with distilled water and crushed into pea sized
pieces and mixed with distilled water until an easily penetrable slurry is
reached. As long as ice is present at the bottom of the Thermos, the
temperature of the slurry will be 0.0 ±0.1 C. Within 15 minutes to one hour,
equilibrium should be reached. A series of five measurements taken five
minutes apart should be recorded. If the measurements support the assumption
of equilibrium, then the five readings are averaged and recorded as the
temperature relative to 0.0 C and the thermometer, and AT with respect to zero
difference.
Accurately reading the meniscus of the thermometer requires two things.
One, is the ability to see the meniscus and the scale at the same time.
Magnification is helpful even for those with good eyes. It makes the
interpolation between scale marks possible and accurate. Secondly, the eye
must be perpendicular to the meniscus to avoid parallax errors. If a mirror is
held against the back side of the thermometer and the center of the image of
the eye moved to the meniscus level when the scale is read, the perpendicular
requirement will be met.
-------
Section No. 4.3.4
Revision No. 0
Date: 17 Sep 89
Page: 1 of 1
4.3.4 INSTALLATION
Each design of aspirated radiation shield has its own installation
requirements. The manufacturer's manual must be used in addition to the
general guidance given here.
The installation location is chosen to represent the temperature relevant
to the application. The height above ground is the first consideration. If
the temperature is to be used for climatological purposes, a 2m height above a
grass surface will do. If the temperature is to be used to describe the air
being drawn into a manifold for chemical analysis, the best location is the one
that represents what the manifold inlet "sees." If a temperature gradient is
to be used to describe the stability of the surface layer, a representative
pair of heights is selected. Siting is best done with the concurrence of the
person who will be using the data for analytical purposes, the person who will
judge the data to be valid, and the person who will accept the data and the
analysis on behalf of the regulating agency. Siting by this committee approach
will benefit from any objective knowledge any member might have, but its strong
point is in the mutual understanding of the criteria which were used in making
the selection.
The second consideration is bias from surrounding structures. The temperature
that is measured is that of the air which is drawn into the aspirator. If
there is a prevailing wind, mount the radiation shield into the wind such that
the wind passes the shield before reaching the tower. The farther the
aspirator inlet is from the tower, the smaller the angle segment which can
contain the tower heat and the more mixing with non-heated air by the time the
inlet is reached. The distance out from the tower should be the maximum
allowed by the mounting hardware. Special booms for temperature may be
necessary if the design does not provide for siting the inlet at least one
tower diameter from the edge, and if 0.5 C accuracy is expected.
-------
Section No. 4.3.5
Revision No. 0
Date: 17 Sep 89
Page: 1 of 9
Such procedures are assumed
The method recommended in
It requires only the generic
that is the ohms vs.
4.3.5 CALIBRATION
The manufacturer's manual will give instruction for the adjustment of
signal conditioning circuits in response to some input specification. Usually
what is required is a precision resistor, either built in or to be supplied by
the calibrator using a decade box or equivalent.
to have been done and will not be discussed here.
this handbook is independent of the manufacturer.
transfer function of the resistance element used,
temperature relationship upon which the measurement system depends for accurate
performance.
4.3.5.1 Calibration Equipment and Methods
The handbook method requires three stable thermal mass assemblies with
temperatures known to about 0.1 C. The three masses may be one mass used three
times with sufficient time allowed for conditioning to a new temperature.
Sufficient effort must go into the determination of thermal stability and
accuracy of the temperature measuring system used for that determination in
order to defend the results of the calibration or audit. The following will
describe one solution to this requirement. This solution is not the only one
but the confidence in stability and accuracy produced by this solution is an
example of the documentation necessary to support claims of accuracy.
The thermal mass design uses a solid aluminum cylinder, shown in Figure
4.3.5.1, chosen for high thermal conductivity. The mass is supported in the
space inside a stainless steel insulated bottle (see Figure 4.3.5.2) by a
lucite tripod on the bottom and by three low thermal conductivity stainless
steel screw spacers at the top. The lid is modified to allow transducer cables
to go through. The insulating bottle is positioned inside a 2-gallon insulated
container modified to allow cables to go through with the top in place as shown
in Figure 4.3.5.3.
Figure 4.3.5.1 Thermal Mass
Figure 4.3.5.2 Insulated bottle
-------
Section No. 4.3.5
Revision No. 0
Date: 17 Sep 89
Page: 2 of 9
Figure 4.3.5.3 Three mass containers: Hot, ambient, cold (right)
The volumes and masses involved with this device are listed in Table 4.3.5.1.
The air space between the inside of the 2-gallon container and the outside of
the insulated bottle is filled with either a cylindrical structure with a
heater strip and thermostat "floated" on stainless steel spacers (#3, hot), air
(#2, ambient) or ice (#1, cold).
Table 4.3.5.1 - Details of a Solid Mass Thermal Device
Component
Aluminum cylinder (less holes)
Holes
Air inside quart bottle
Quart bottle
Air outside quart bottle
2-gallon container
Total for #2 - ambient
Quart bottle (see above)
Ice-water mixture
Air above ice-water
2-gallon container
Total for #1 - cold
Volume
(cm )
485
58
1,100
931
5,882
10,206
18,662
2,574
4,706
1,176
10,206
18,662
Mass
(g)
1,341
0.1
1.3
1,154
7.1
1,600
4,104
2,496.4
4,706
1.4
1,600
8,804
Typical Transducers:
Minco S28F36Y nickel-iron 2.4 7
Rosemont RMT 78-39-7 5 36
-------
Section No. 4.3.5
Revision No. 0
Date: 17 Sep 89
Page: 3 of 9
The ratio of 1,341 grams of thermal mass to about 43 grams of sensors meets
the design goal of a small sensor thermal impact. The several layers of
insulation and the minimization of thermal conductivity paths meets the long
time constant goal for the thermal mass.
A test was conducted to document the performance of the thermal mass assemblies
and to show the time required to condition the hot and cold masses and the time
constant of the ambient mass, see Figure 4.3.5.4. From the beginning of the
test at a little after 10:00 a.m., when ice was put into #1 and the heater was
plugged in for #3, it took about 9 hours for the slope to be flat enough to be
confident that the mass had a homogeneous temperature. The maximum rate of
change of the ambient mass was 0.01°C/min. The stability of the measuring
circuit is shown by the line with triangle symbol. The thermal conductivity of
aluminum is 0.5 cal./sec through a plate one centimeter thick across an area of
one square centimeter when the temperature difference is 1°C. Steel is 0.1,
rubber is 0.0005 and air is 0.00005 (Hodgman, 1955). The aluminum cylinder
exchanges heat with its environment (except for the transducers being tested)
O
OJ
-a
-------
Section No. 4.3.5
Revision No. 0
Date: 17 Sep 89
Page: 4 of '9
Having created a stable environment, there needs to be an accurate method
of measuring both the relative and absolute temperatures of the thermal masses.
A three transducer data system was designed for this purpose. The report of
its calibration is shown at the end of this section in 4.3.5.4.
4.3.5.2 Delta Temperature (AT)
Calibration of a AT system involves two parts. One is the matching of
the transducers at zero difference and the other is the gain of the signal
conditioner for a known difference in temperature of element resistance.
The first part involves placing the AT transducers together in a series
of stable thermal masses. A stable thermal mass is any mass which is at least
25 times the mass of the sensors being conditioned by the thermal mass and
which has thermal gradients of less than 0.01 C/cm throughout the mass. Start
at cold, somewhere in the 0°C to 5°C range, and record the system AT output
after stability has been reached in 30 to 60 minutes. Take readings about five
minutes apart. When the readings stabilize, average the last five. Assume the
output reports a difference of -0.02°C. Then move the two transducers to the
thermal mass at ambient temperature, somewhere in the 15 C to 25 C range. When
stability has nearly been reached, start taking readings about five minutes
apart. When the readings have stabilized, average the last five readings..
Assume the output reports a difference of +0.03°C. Finally, move the two
transducers to the hot thermal mass, somewhere in the 35 C to 40 C range, and
wait 30 to 60 minutes for stability to be reached. Take readings about five
minutes apart. Assuming they are -0.09, -0.07, -0.05, -0.04, and -0.03,
stability has not been reached. After another 30 to 60 minutes take another
series of measurements about five minutes apart. Now they are -0.01, +0.01,
0.00, 0.00 and +0.01. Stability has been reached and the average of the last
five readings, 0.00, is recorded.
This test has confirmed that the two sensors are matched to each other
and to the generic transfer function with which the signal conditioning
circuits have been set. It may be that the matching was done in the circuitry.
It does not matter. It has been shown that the transducers and their parts of
the circuitry agree with each other at three different temperatures.If
agreement is not within ±0.05°C of the true value of 0.00°C, look to the manual
or the manufacturer for guidance in correcting the problem. The AT system
should start off with agreement in controlled conditions of much better than
0.1 C if the atmospheric measurements are to approach that accuracy. The
methods described here for building a stable thermal environment and sampling
the outputs for zero difference are only an example which works. The only
important criteria is that it be documented in terms of stability, whatever
method is used.
The second part of the AT calibration sets to tests the gain of the
difference amplifier. Pick a common temperature for the site and substitute a
fixed resistor for one transducer, arbitrarily choose the lower one. Assume
the transducer is a 100Q platinum type (see Table 4.3.1.1) and your resister is
108 ±1Q. Substitute a precision decade box for the upper transducer. Adjust
the decade box until the AT output is the voltage equivalent of 0.00°C. If the
range is -5 to +15°C for a 0 to 1 volt output, 0.00 C is 0.250 volts. The
output now reads 0.250 volts and the decade box reads 107.96Q. If 107.79Q
-------
Section No. 4.3.5
Revision No. 0
Date: 17 Sep 89
Page: 5 of 9
represents 20°C, and if 0.389Q represents a 1°C change (0.0389C2 for 0.1°C
change or O.Olfl = 0.0257°C), the simulated temperature for both transducers is
(107.96 -o!07.79 = 0.17; 17 x 0.0257 = 0.437) 20.437 ±0.026°C or between 20.41
and 20.46°C.
If the decade box is changed to 108.35 (107.96 + 0.389), the upper
simulated transducer is now 1°C warmer than the lower simulated transducer.
The output should read +1°C or 0.300 volts (1.000 volts + 20 degrees = 0.050
volts/deg.; 0.250 + 0.050 » 0.300). If zero and full scale are to be
challenged, set the decade box to 106.02 (107.96 - [5x0.389] = 106.02) for a
0.000 volt reading and set the decade box to 113.80 (107.96 + [15x0.389] =
113.80) for a 1.000 volt reading. Check the difference in box settings; 113.80
- 106.02 = 7.78Q; 7.78 + 0.389 * 20.00°C.
Beware of rounding errors if enough resolution is not carried or
available on the decade box and the output. Check the decade box with a good
ohm meter. Errors in the tens wheel (when switching from 106 to 113) may be
larger than the smallest wheel. If the tens wheel is only good to 1%, the
uncertainty is O.lOn or 0.26°C.
4.3.5.3 Temperature
The temperature calibration may be achieved concurrently with the AT
calibration. Each thermal mass should have some accurate means of determining
temperature. While the actual temperature is not important for the AT
calibration, it should be recorded on the calibration form. If the temperature
transducer is not one of the AT pair, it can be placed in a thermal mass at the
same time the AT calibration is being done.
It may be that the system does not have a AT measurement but it does
have a temperature measurement. Remembering that the temperature accuracy
requirement is ±0.5°C, the temperature transducer can be challenged with a much
simpler method. Liquid baths in a pint or quart insulated bottle with the
transducer and a good ASTM or equivalent thermometer mixing the bath together
will suffice. Be sure stability has been reached before taking the readings.
Use care or parallax-avoiding devices when reading the thermometer.
4.3.5.4 Calibration Report Example
The following information is reproduced from a report prepared for
Meteorological Standards Institute to document the accuracy of temperature
instrumentation used on audits.
4.3.5.4.1 Introduction
During April 14 to May 8, 1988, a calibration program was
conducted to verify the accuracy of three MINCO 604 ohm RTDs (resistance
thermal devices). Three RTDs, Minco model S28F36Y labeled #1, #2 and #3, were
originally calibrated in 1984 and have been in use for temperature and delta
temperature auditing during the past four years. The earlier calibration was a
relative calibration since the only accuracy of consequence to the application
was the inter- relationship of the three RTDs. The current calibration is both
relative and absolute.
-------
Section No. 4.3.5
Revision No. 0
Date: 17 Sep 89
Page: 6 of 9
4.3.5.4.2 The measurement circuitry
The three RTDs are connected in series to a battery powered
constant current source of 0.500 mA. A fixed resistor (668 ohms) of low
thermal sensitivity was also in the series loop as a reference source (REF).
The voltages across each of the four resistors in the current loop were
connected to input channels of an ADC-1 data logger. A NEC PC-8201A computer
controlled the ADC-1 and collected the data. A program called ADCT was used to
sample all the channels every 5 seconds and to record the average and standard
deviation for periods of time selected through the program. Times of 10
minutes, 30 minutes and an hour were used at different times during the
calibration. The ADC-1 provides an output in tens of millivolts. If the
voltage across a 668 ohm resistor in a 0.0005 ampere current loop is 0.334
volts or 334 mV, the ADC-1 will output 3340.
4.3.5.4.3 The conversion of Minco ohms to temperature
In the Minco Application Aid No. 7, Table 14-604, the nominal
values of resistance for temperatures are given. Nine sets of temperatures and
resistances in five degree steps from 0°C to 40°C were used to find a
mathematical expression for converting resistance to C. A linear regression
was not satisfactory. The quadratic solution to the regression analysis
predicted the temperature at the nine points with an error of less than 0.01 C.
Table 4.3.5.2 shows the input pairs, the predicted temperature and error, and
the constants found and used.
Table 4.3.5.2 - Relationship of Minco Resistance to Temperature
y x
Minco Table 14-604
temperature
(°0
0.000
5.000
10.000
15.000
20.000
25.000
30.000
35.000
40 . 000
resistance
( ohms )
604.00
617.98
632.13
646.46
660.97
675.66
690.52
705.57
720 . 79
A A
y y - y
predicted
temperature
(°C)
0.006
4.998
9.994
14.996
20.001
25.005
30.005
35.004
39.994
error
(°C>
0.006
-0.002
-0.006
-0.004
0.001
0.005
0.005
0.004
-0.006
for y = a + bx + ex2 a - -269. 1531523
R * 1.00000009 b = 0.53213288
n = 9 c « -0.00014322177
A second step in the conversion requires the constant current
to be exactly 0.5 mA so that the recorded voltage can be converted to
temperature. The REF resistor is recorded as temperature using the generic
conversion formula. The difference between the recorded value of REF and the
correct value for 0.5 mA gives the correction. To find the correct value a
series of measurements were made with the Fluke 8060A in the current loop. The
-------
Section No, 4.3.5
Revision No. 0
Date: 17 Sep 89
Page: 7 of 9
least squares straight line was calculated for the data and both the data and
the best-fit line are shown in Figure 4.3.5.5. The REF value for 0.5 mA was
calculated and found to be 22.258 deg. C.
Circuit Current vs. REF (#4)
o
o
o
o"
X
tu
Q.
E
a
5010 -T
5009 -
5008 -
5007 -
5006 -
5005 -
5004 -
5003 -
5002 -
5001 -
5000 -
4999 -
4998 -
4997 -
4996 -
4995 -
4994 -
4993 -
4992 -
4991 -
4990 -
21
Tests of May 3-5, 1988
Regression Analysis
Y = 4497.33 + 22.584 X
22.1
I 22.3
22.258
22.5
REF "Temperature" (deg. C)
Figure 4.3.5.5 "Constant" current versus REF resistor "temperature"
The third step in converting the ADC voltage, V, to temper-
ature, T, comes from the statement in the program ADCT which applies the
quadratic equation to the resistance, assuming 0.5 mA current. That statement
is as follows:T = -269.15 + 0.10642V - 5.728E-06V2.
4.3.5.4.4 Measurement of true temperature
Two methods were used to find the true temperature. The first
was the use of an ice slurry for 0 C. The second was the use of an ASTM 63F
mercury-in-glass thermometer. The Princo Instruments Factory Certificate of
Accuracy Tolerances for s/n 245453, scale range 18°F to 89°F with divisions
every 0.2 F, states ±0.2 F or one division. The thermometer was read with an
optical magnifier with anti-parallax targets to the nearest 0.05 F or about
0.03 C. In ^he relative sense, the temperature should be accurate to 0. 1 C and
in the absolute sense to 0.2°C.
-------
Section No. 4.3.5
Revision No. 0
Date: 17 Sep 89
Page: 8 of 9
4.3.5.4.5 The test facility
The three Minco RTDs were taped to the mercury
bulb of the thermometer after all four devices were threaded through a rubber
stopper. The assembly was submerged in an ice slurry
or in water in a pint Thermos bottle (see Figure
4.3.5.5). The data logger signaled when an average
was being recorded and the thermometer was read at
the same time. The average from the Minco RTDs
represents the middle of the time period while the
thermometer was read at the end of the time period.
A linear extrapolation of the Minco RTD data to the
time of the reading of the thermometer provided
comparable data.
Figure 4.3.5.6
Calibration
0.5
4.3.5.4.6 Results
There were 73 thermometer readings over a nine
day period covering a temperature range of 31.95°F to
89.05°F (-0.03°C to 31.69 C). The differences in
temperature between Minco sensors #1 and #3,
expressed in C, and the thermometer temperature,
expressed in °F, are shown in Figure 4.3.5.7. Sensor
#2 was so close to #1 that it was not plotted. The
calibration correction curves for all three sensors
are shown on the figure. The best fit lines from the
linear regression analyses are drawn on the figure.
The coefficients are listed in Table 4.3.5.3.
MINCO Thermometer Calibration
Meteorological Standards Institute
O
o
c
0)
CD
a
0)
Q.
E
0.4 -
0.3 -
0.2 -
0.1 -I
-0.1 -
-0.2
KEY
a #3 (Regression Analysis
O #1 (Regression Analysis
B
#3 Correction (-0.265° C)
#2 Correction (-0.010° C)
o
o
o
Correction (-0.002° C) •
30
50 70
ASTM-63F (245453) Temperature (deg. F)
90
Figure 4.3.5.7 Calibration of Minco sensors (RTDs)
-------
Section No. 4.3.5
Revision No. 0
Date: 17 Sep 89
Page: 9 of 9
Table 4.3.5.3
Regression output for Minco 81 and 83
versus mercury-in-glass thermometer
Constant (a)
Standard error of y
Coefficient (b)
Standard error of b
Number (n)
Average difference
Standard deviation
Minco #1
0.02262
0.02838
-0 . 00030705
0.00019593
73
0.002
0.028
Minco #3
0.29951
0 . 02534
-0.00041265
0.00017491
73
0.272
0.025
The relative calibration, without consideration of an outside
measurement of temperature, covered a range of 0 C to 44 C. There were 254
averages recorded for each of the Minco RTDs. The average and standard
deviation of the differences between RTDs is shown in Table 4.3.5.4.
Table 4.3.5.4 Relative difference analysis
Average difference
Standard deviation
Minco Sensors (RTDs)
#3 - ftl
0.2634
0.0183
#3 - #2
0 . 2753
0.0165
#2 - tl
0.0119
4.3.5.4.7 Conclusion
Using Minco #1 as a standard of comparison, and adjusting #1
for the bias of 0.002°C as shown in Table 4.3.5.3, the following accuracies of
relative temperature are estimated.
Minco #1
Minco #2
Minco #3
*1 - 0.002 = temperature ±0.05oC
#2 + 0.010 = temperature ±0.05°C
#3 - 0.265 = temperature ±0.05°C
Similar tests in 1984 yielded the following corrections:
#1 -i- 0.00
#2 + 0.03
#3 - 0.22
temperature
temperature
temperature
showing reasonable stability over four years and a reasonable capability to
duplicate relative calibrations.
-------
Section No. 4.3.6
Revision No. 0
Date: 17 Sep 89
Page: 1 of 3
4.3.6 OPERATIONS, MAINTENANCE AND QC
4.3.6.1 Operations
From the standpoint of quality assurance, routine inspection of the
temperature and AT data will help find problems soon after they occur. Routine
inspection of the instrument system is also required. The temperature
equipment is usually free from wear or change, except for the aspirated
radiation shield, which tends to attract homeless critters of all kinds.
Inspection of the temperature shield and transducer should be a part of the
routine site visit and duly noted in the site log book. If there is no routine
visit, a weekly or at the very least monthly visit is recommended. Systems
usually have' some built in calibration feature which substitutes resisters for
the transducers to check or calibrate the signal conditioner." The site visit
should include a temperature comparison with a simple hand-held thermometer for
reasonableness. Guard against radiation errors with the hand held thermometer
when working in direct sun light by keeping the thermometer in shade and away
from body heat, including hot breath. The long time constant of
mercury-in-glass thermometers (minutes) may make the inevitable warming from
biasing heat difficult to observe.
4.3.6.2 Maintenance
Routine maintenance should include cleaning the aspirated radiation
shield and verifying its function. This should not have to be done any more
frequently than the bearing tests of the wind equipment. Since it is necessary
to get the wind equipment down from the tower for the bearing tests, a cleaning
of the radiation shields at the same time would be prudent and economical.
4.3.6.3 Quality Control
The best way to keep a AT system operating "in control" is to use a data
quality control inspection for this variable. The QA Plan should supply the
details of the AT inspection program.
An important aspect of the inspection is the background of the
inspector. Ideally, an experienced QC meteorologist should be used. Lacking
this resource, a training program should be made available to the person who
will routinely perform the data inspections. The training will point out the
nature of AT data as a function of wind speed, cloud cover, and time of day
(solar angle).
Additionally, training will point out that a AT value, that is the
difference between two well calibrated and shielded transducers, is just that.
It is not a gradient measurement unless there is reason to believe that the air
between the two transducers is reacting normally to thermal flux. Cases have
been observed where a 10m to 60m AT averaged in excess of the auto convection
rate for hours. The easy assumption is that there is an instrument error
because autoconvection rates cannot be exceeded for long periods of time. The
often unspoken assumption is that the AT transducers are in the same boundary
layer and the difference in temperature represents the stability condition of
the air. If the site can produce shallow or transitory surface boundary
layers, as can happen with land-water interface regions, one transducer may be
-------
Section No. 4.3.6
Revision No. 0
Date: 17 Sep 89
Page: 2 of 3
in one layer and the other transducer in the other layer. Then the AT value
represents nothing more than single samples in different layers and lapse rate
conclusions are invalid but the data are valid.
Training will show the normal diurnal cycle from a negative AT (lapse
rate) in the daytime to a positive AT (inversion) at night (see Figure
4.3.2.1). An understanding of the physical process will support the data with
other observations of weather conditions. The sun heats the surface much more
than the air above it. The air at the surface is warmed by the now warmer
surface. The warmer air expands and rises and mixes with the air it passes.
This unstable convective process continues until the driving force, the surface
warmer than the air above it, is neutralized. This can happen by either
changes in the radiational heating of the surface or by the effective cooling
of the surface through the heat removal process described above. Considering
the strength of radiant heating (sun angle and sky cover) and the strength of
the mixing process (wind speed), the size of the lapse rate (-AT) can be
imagined.
Conversely at night, the surface is cooled by long-wave radiant loss to
the cold universe. As the surface gets colder than the air above it that
surface air is cooled by conduction. The cooled air is mixed by mechanical
turbulence, caused by the wind flowing over surfaces elements, and slowly cools
the air from the surface up. This very stable process results in the air
above, not yet cooled, being warmer than the air below or an inversion (+AT).
The size of the inversion results from the amount of surface cooling (sky
cover) and the amount of turbulent mixing (wind speed and surface roughness).
Calm clear nights will have larger inversions than cloudy and windy nights. If
the wind is too high, there may not be an inversion at all. All the air may be
mixed so well that there is no measurable difference between the two AT
transducers.
To complicate the picture, temperature measurements also change as
different air with different conditioning history is blown or advected past the
measurement site. Diurnal rhythms can be seen as colder air flows down terrain
features, such as hills and valleys, at night as a consequence of the kind of
surface cooling described above. These vertically stable flow fields often
become decoupled from each other. It is possible for one transducer to be in
one stream of stable air while the other is in another stream. When this
happens, the AT value may not represent a temperature gradient as much as two
separate flow regimes, similar to the boundary layer example.
The trained inspector learns to see these processes in the AT data and
to recognize physically unusual or unlikely data. If the data QC inspector
looks at the data on a weekly schedule, problems will be uncovered shortly
after they occur, thus avoiding long periods of data loss. One week is also
short enough to allow memory of conditions to be correlated with the data.
Discrepancy reports originated by the QC inspector can initiate the testing and
corrective action, if necessary, by the instrument operator. If At
measurements are taken, the purpose is usually the determination of stability
categories or parameters for use in diffusion models. This important variable
deserves careful attention and well documented evidence for validity claims.
The accuracy of the measurement is achievable only with careful calibration.
The accuracy of,the data requires as much care given to data inspection between
calibrations.
-------
Section No. 4.3.6
Revision No. 0
Date: 17 Sep 89
Page: 3 of 3
If only temperature is measured, the accuracy requirements are less
difficult and the data QC inspection can be less rigorous. If data are
collected, however, routine inspection is recommended and a weekly period is
reasonable. The inspection compares the temperatures to reasonable values for
the week and the nature of the temperature change to realistic patterns.
-------
Section No. 4, 3. 7
Revision No. 0
Date: 17 Sep 89
Page: 1 of 8
4.3.7 PERFORMANCE AUDIT METHODS
4.3.7.1 General Considerations
A performance audit is a measurement made by an independent method and
person of the accuracy and precision of the performance of the measurement
system. To make this determination for temperaturemeasurements, knowledge of
the input conditions to the system sensors is required. It is also necessary
to know what the system will do to these input conditions in producing an
output. The output is simply the system output. Given the inputs and the
transfer functions, the output can be predicted. The difference between the
predicted output and the system output is the error of the system or its
accuracy. The methodology starts with the ways of controlling and measuring
the sensor inputs or knowing the inputs in an uncontrolled environment.
Prediction of output from the controlled input requires knowledge of the
transfer function but not necessarily its validity. Knowing the conditions of
the uncontrolled environment does not require knowledge of any intermediate
steps such as transfer functions. Temperature in is simply compared to
temperature out. The method using the latter approach is called the Collocated
Transfer Standard (CTS) method. The CTS method is seldom practical for AT.
The best performance audit uses both methods where appropriate.
A performance audit must follow some written procedures. Since the
procedures must be relevant to the design of the instrument or system being
audited, only general principles will be described below with some specific
examples. The data from the audit should essentially fill out an audit form.
It is important, however, for the auditor to be sufficiently experienced to be
able to deviate from the procedure or the form when the pursuit of truth leads
away from the expected.
The starting point of an audit form is the documentation of the who,
what, where, when, and how the audit values were acquired.
4.3.7.1.1 Who
The performance audit report form should contain a space to
identify the auditor. The audit report which summarizes the audit findings
should report the names and affiliations of the operators of the system.
4.3.7.1.2 What
The form should contain a section to identify the instrument
being audited by manufacturer, model number, and serial number.
Sub-assemblies, such as transducer or RTD and aspirated radiation shield,
should be identified by number. If they are not numbered, the operator should
be asked to mark them for identification.
The audit report should contain a list of all the equipment
provided and used by the auditor, including model and serial numbers and time
of last calibration, where relevant.
4.3.7.1.3 Where
The audit form should have a space to show the location of the
-------
Section No. 4.3.7
Revision No. .0
Date: 17 Sep 89
Page: 2 of 8
sensor on a tower, including height. A sketch is useful to show the relative
positions of the sensing elements with respect to possible biasing influences,
such as the tower, other sensors and buildings.
4.3.7.1.4 When
The date or dates when the audit affected the system operation
should be listed. The time when the system or a particular sensor was taken
"off-line" and put back "on-line" should be listed. The time, or time period,
when each data value was taken is vital for the comparison with the system
output. Implicit in this is the need for the time the auditor uses to be
correlated to the time the operator or the system uses. The auditor should
rely on the National Bureau of Standards station WWV for correct time. Battery
operated receivers, such as the Radio Shack Time Cube, are generally available.
4.3.7.1.5 How
The audit form should either contain a copy of the method used
or reference the method number. The audit report should contain copies of the
audit method used. The methods should be detailed enough to identify each step
in the acquisition of the audit value and in the conversion of the value to
units compatible with the system output. See Figures 4.3.7.1, 4.3.7.2, 4.3.7.4
and 4.3.7.5 for examples of forms and methods. These figures are intended as
aids for writing specific methods and drawing the companion form. These may be
reproduced if they are relevant.
4.3.7.2 Temperature Difference (AT)
4,3.7.2.1 Sensor control method
The audit method should simulate the most complete calibration
method (see 4.3.5). The first step is to condition the thermal mass assemblies
to be used to challenge the transducers to a zero difference environment. If
stirred baths are to be used, be prepared to give ample time for equilibrium.
The amount of time is not well predicted by the transducer time constant.
Stability at the 0.01°C resolution scale is desirable to back up claims of
accuracy to 0.1°C.
The transducers must be removed from the tower along with their
cables. At some installations it will be difficult to impossible to remove the
cables. Substitute cables may be used if care is taken to make sure the
substitute does not change the output more than 0.01°C. This should be
verified by using a fixed resistance representing ambient temperature as a
substitute transducer. Output readings using both the operational cable and
the substitute cable should be recorded on the audit form.
The AT audit usually requires much more elapsed time than
measurement time. It is practical to have equipment which allows the
transducers to slowly reach their equilibrium state and to record this process
through the entire system. This way the elapsed time can be used for the more
labor intensive audit variables.
-------
Section No. 4.3.7
Revision No. 0
Date: 17 Sep 89
Page: 3 of 8
Temperature Difference - MSI method DLT006
-------
Section No. 4.3.7
Revision No. 0
Date: 17 Sep 89
Page: 4 of 8
PERFORMANCE AUDIT REPORT by
MEASUREMENT SYSTEM - Temperature difference
System number
Sensor
Tr an sd uc er s
Loc ation s
Signal^conditioner
Data channels
DLT006
DATE / / TIME o-f-f line
on line test start
TRANSFER FUNCTION (volts per degree C)
Conversion formula
TEST RESULTS
DLT006-A
cold sensor # 1 2
mass output output
time ohms degC volt degC volt degC
1-2
dif.
volt degC
average
ambient sensor #
mass
time ohms deaC
output
volt deaC
output
volt degC
1-2
di-f.
volt degC
average
hot
time
sensor #
mass
ohms degC
output
volt degC
output
volt deaC
1-2
dif.
volt deaC
average
DLTOO6-B
Tl T2A Tl T2 B B-A
sensor # 1 2 1-2 1 2 1-2
mass mass di-f. output output dif. di-f.
time ohms degC ohms degC degC volt degC volt degC degC degC
average
sensor
time
T 2
* 1 •
mass
ohms degC
T 1
2
mass
ohms degC
A
1-2
di-f.
degC
T 2
1
output
volt degC
T 1
2
output
volt degC
B
1-2
dif.
degC
B-A
dit .
degC
sveraae
Figure 4.3.7.2 Audit form for the temperature difference method
-------
Section No. 4.3.7
Revision No. 0
Date: 17 Sep 89
Page: 5 of 8
The transducers are challenged with a known zero difference at
three temperatures. This shows how well the transducers are matched and how
well they follow the generic transfer function. The acquisition of these three
data points may take four hours. Some judgment is required to identify the
point at which stability is achieved. One clue is the reversal of a
progression of data points. If, for example, the five minute average AT values
are 0.09, 0.07, 0.06, 0.06, 0.05, 0.05, 0.06, 0.05, the last 0.06 can be taken
as a signal that equilibrium has been reached. The reported value could be the
average of five stable points (0.05) or the final stable value.
After the zero difference test has been completed, the gain of
the difference circuit can be challenged by using the substitute resistance
method. This method is described in 4.3.5. It is possible to use thermal
masses at different temperatures, but it is not recommended because of the
uncertainties of stability and four transducers measuring two temperatures to
an accuracy of at least 0.025°C. Once the matching at zero difference has been
established, the gain is adequately verified by normal electronic circuit
procedures.
It is not practical to mount a second pair of sensors and
aspirated radiation shields for collocated testing. The interference problems
with the aspirators are hard to overcome. The physical problems associated
with mounting parallel instruments are large compared to the value of the
method. Considerable, but much less, effort is required for the Sensor Control
method which provides numbers with acceptable confidence in their accuracy.
4.3.7.3 Temperature
4.3.7.3.1 Sensor control method
Usually the temperature transducer, if it is different from one
of the AT pair, can be included in the thermal mass with the AT pair, or in
another thermal mass at the same time as the AT pair is being tested. Timing
for stability can include temperature with AT as though it were the same test.
The big difference, however, is that the temperature transducer output is
compared to the thermal mass transducer output as the audit value. A
calibration of the thermal mass transducer is the key to the claim for accuracy
given the challenged system.
If there is no AT system being audited, a simpler method is
appropriate for the temperature system. A two point check using an ice slurry
and an ambient bath is acceptable. Two insulated bottles (pint or quart size)
with a cork to support a calibrated thermometer or a calibrated electric
thermometer are required (see Fig. 4. 3. 5.6).Stability is easier to find since
the readings are only taken to the nearest 0.1°C.
4.7.3.2 CTS Method
It is both practical and recommended to use the CTS method for
temperature audits. The temperature transducer and its aspirated radiation
shield (or even naturally aspirated shield) is usually located at an easily
reached elevation. A CTS such as the Assmann Psychrometer shown in Fig.
4.3.7.3 can be iocated near the temperature sensor. It should be exposed so
the wind reaches the CTS without bias error from other structures. If the wind
-------
Section No. 4.3.7
Revision No. 0
Date: 17 Sep 89
Page: 6 of 8
during the audit is passing through a tower to reach the temperature sensor,
the CIS should not be exposed to sample the same biased air temperature. The
presumption is that the temperature sensor represents the air temperature. Any
error from siting is a part of the measurement error. The CTS should be
mounted to avoid all bias, if possible.
The CTS method should be
used as an additional challenge to the
temperature system. The two point Sensor
Control method is a challenge to the
transducer and signal conditioning circuit.
The CTS method is a challenge to the
radiation shield at one point on the range
scale. A one point challenge of a
temperature system provides no
information about other temperatures. It
may be that the operator calibrated the
system at one point, perhaps the same
ambient temperature as exists during the
audit. There could be a slope error which
causes large errors at near freezing.
Having an accurate temperature measurement
near freezing, accurate to 0.5°C, can be
valuable as it relates to other sensors,
such as wind. If the wind vane does not
show any direction variation and the
temperature system reports 0.0 C, there is
a good chance that ice is on the direction
vane bearing assembly. If the temperature
system reports 3°C, with a 3°C error
because only one point was used in
calibration and audit, a different estimate
of the direction vane problem is necessary.
If audit or calibration records exist
showing the full range performance of the
temperature circuit, a one point spot check Figure 4.3.7.3
with a CTS is useful. Assmann psychrometer
-------
Section No. 4.3.7
Revision No. 0
Date: 17 Sep 89
Page: 7 of 8
Temperature - MSI method TEMOQ5 (version 8/1/84)
This method provides -for a comparison of a temperature transducer
with electric output to a calibrated transducer in a slowly
changing thermal mass at three different temperatures; The audit
equipment includes three thermal mass assemblies. Each is a
cylinder o-f aluminum (6.4 cm diameter by 17 cm long. 1.341 g in
48S cc) with holes for different kinds of sensors. This 485 cc
mass is suspended in the 1.1OO cc inside volume of a stainless
steel Thermos bottle, which is inside the 8500 cc inside volume
of an 18.500 cc cylinder (Gott 2 gallon water cooler). The cold
system is filled (4.500 cc) outside the Thermos b'ottle with a
mixture of ice and water. The ambient system is full of ambient
air. The hot system has a cylindrical frame spaced outside the
Thermos bottle with a 600 watt strip heater operating through a
10O F thermostat. Conditioning for the hot and cold masses
begins at least 12 hours before planned use. Each mass has a
Minco 604 ohm (at 0 C) nickel-iron RTD. Resistance is measured
and converted to temperature using a quadratic expression of the
Minco Table 14-606 (see appendix). Relative corrections from
intercomparisons made with the three RTDs in the same mass are
applied yeilding relative accuracies of better than 0.05 C.
Absolute accuracy is better than 0.5 C.
TEMOO5-A Insert the RTD being challenged in the cold mass. Wait
about 30 minutes or until stable temperature is reached. Record
samples of the RTD temperature from the system output. Record
the resistance measurements of the mass RTD.
Move the RTD being challenged to the ambient mass and repeat the
above procedure.
Move the RTD being challenged to the hot mass and repeat the
above procedure.
TEMO05—B Use an Assmann aspirated psychrometer mounted in the
vacinity of the shielded temperature sensor. Mind the Assmann
and let it run five minutes. Wind again and after an additional
two minutes, begin reading the mercury-in—glass thermometers.
Use the anti-parallax magnifiers. Record the temperatures from
the Assmann and from the system taken at the same time. Be sure
the two sensor systems are sampling from air which has not been
biased by local mounting structures.
Figure 4.3.7.4 Audit Method for Temperature
-------
Section No. 4.3.7
Revision No. 0
Date: 17 Sep 89
Page: 8 of 8
PERFORMANCE AUDIT REPORT by TEM003
MEASUREMENT SYSTEM - Temperature
System number _
Sen sor ___« __
Tr ansducer
Location ~ ~ _
Signal conditioner
Data channel
DATE / / TIME of-f Una on line test start
TRANSFER FUNCTION
-------
Section No. 4.3.8
Revision No. 0
Date: 17 Sep 89
Page: 1 of 1
4.3.8 ESTIMATING ACCURACY AND PRECISION
Section 4.1.5 contains a detailed discussion of methods of estimating
accuracy, precision and bias using wind speed and wind direction as examples.
Temperature transducers (as differentiated from temperature sensors which
may include the radiation shielding) may be calibrated or audited by exposing
them to a controlled environment such as a wet or dry thermal mass. The
temperature of the mass is known either by an installed calibrated transducer
or by collocating a calibrated transducer in the mass. The collocation
alternative requires assurance that the mass is at a homogeneous temperature.
The installed option requires assurance that the installed location is
representative of the homogeneous mass temperature. When such a method is
used, traditional statistical or metrological methods may be used to estimate
accuracy. The precision of the method is the standard deviation about the
series mean value of repeated measurements in the constant and controlled
environment. Such methods are capable of achieving accuracies of 0.1 degree C
and precision of 0.05 degree C or less.
When a collocated transfer standard (CTS) method is used, considerable care
is required in stating the accuracy of the calibration or audit. The different
exposures in the atmosphere of different transducers with different time
constants in different radiation shields puts a larger uncertainty on the
comparison than is found with a controlled environment. If, for example, a 2
meter temperature instrument is compared to an Assmann thermometer mounted
nearby at the same height, the accuracy claim might be no better than the sum
of the two different instrument accuracies. The accuracy of the method might
be 1 degree C if each instrument is capable of a measurement accuracy of 0.5
degree C. There is reason to consider such a method as a comparative
measurement rather than an audit or calibration.
It is possible that a CTS method can have greater accuracy. What is needed
is a body of data which sets the functional precision of the CTS method by
finding the best one can expect from collocated temperature instruments. A
companion requirement is a body of data which compares different radiation
shields as a function of radiation intensity, wind speed and wind direction
relative to the orientation of the aspirator motor.
-------
Section No. 4.3.9
Revision No. 0
Date: 17 Sep 89
Page: 1 of 1
4.3.9 REFERENCES
Brock, F. V. and C. E. Nicolaidis, 1984: Instructor's Handbook on
Meteorological Instrumentation. NCAR/TN-237+1A, National Center for
Atmospheric Research, Boulder, CO.
Cole, H. L., 1978: Air temperature and differential temperature measurement
using 1C temperature sensors. Preprints of Fourth Symposium on Met.
Observ. and Inst., Amer. Meteor. Soc., Boston, MA., pp. 25-30.
EPA, 1987a: Ambient Monitoring Guidelines for Prevention of Significant
Deterioration (PSD), EPA-450/4-87-007. Office of Air Quality Planning and
Standards, Research Triangle Park, NC.
EPA, 1987b: On-Site Meteorological Program Guidance for Regulatory Modeling
Applications, EPA-450/4-87-013. Office of Air Quality Planning and
Standards, Research Triangle Park, NC.
Hodgman, C. D., editor in chief, 1955: Handbook of chemistry and physics.
Chemical Rubber Publishing Company, Cleveland, OH.
Lockhart, T. J., 1975: Variable errors in operational data networks.
Proceedings of the 3rd Symposium on Met. Observ. and Inst., Amer. Meteor.
Soc., Boston, MA., pp. 91-96.
Lockhart, T. J. and M. T. Gannon, 1978: Accuracy and precision of field
calibration of temperature difference systems. Proceedings of the National
Conf. on Qual. Assur. of Environ. Meas., Denver, CO., Nov. 27-29.
Mason, C. J. and H. Moses, 1984: Atmospheric Science and Power Production,
Darryl Randerson, Ed., DOE/TIC-27601, pp. 103-109
McKay, D. J. and J. D. McTaggert-Cowan, 1977: An intercomparison of radiation
shields for auto stations. World Meteorological Organization Publication
No. 480, pp. 208-213.
-------
Section No. 4.4.0
Revision No. 0
Date: 17 Sep 89
Page 1 of 7
QUALITY ASSURANCE FOR PRECIPITATION MEASUREMENTS
4.4.0 INTRODUCTION
By definition, "The total amount of precipitation which reaches the ground
in a stated period is expressed as the depth to which it would cover a
horizontal projection of the earth's surface if there were no loss by
evaporation or run-off and if any part of the precipitation falling as snow or
ice were melted" (WHO, 1971). In any method of precipitation measurement, the
aim should be to obtain a sample that is representative of the fall in the
area. At the outset, it should be recognized that the extrapolation of
precipitation amounts from a single location to represent an entire region is
an assumption that is statistically questionable. A network of stations with a
density suitable to the investigation is preferable.
4.4.1 TYPES OF INSTRUMENTS
Precipitation collectors are of two basic types: nonrecording and
recording'. J'!
4.4.1.1 Nonrecording Gages
In its simplest form, a precipitation gage consists of a cylinder, such
as a can with straight sides, closed at one end and open at the other. The
depth of the liquid in the can can be measured with a measuring stick
calibrated in subdivisions of centimeters or inches (Figure 4.4.1).
To obtain greater resolution, as in the case of the standard 8-inch gage
made to NWS Specification No. 450.2301, the gage is constructed with a ratio of
10:1 between the area of the outside collector cylinder and the inside measuring
tube. The funnel attached to the collector both directs the precipitation into
the tube and minimizes evaporation loss. Amounts in excess of two inches of
rainfall overflow into the outer can, and all measurements of liquid and melted
precipitation are made in the measuring tube with a measuring stick.
The automatic wet/dry precipitation collector, available in several
designs, represents a specialized nonrecording instrument designed for programs
involving the chemical and/or radioactive analysis of precipitation. The
collector is built with a sensor that detects the onset and cessation of
precipitation and automatically releases a lid to open and cover the collector.
In one design, the lid can be made to remain open during either wet or dry
periods. Another model is made with two collectors; the lid is made to cover
one bucket during periods of rain and snow (Figure 4.4.2). In equipment of
this kind involving precipitation chemistry, the volume of water in proportion
to the constituents collected with the water is important, so evaporation must
be kept to a minimum (see EPA, 1985).
4.4.1.2 Recording Gages
Recording gages are of two basic designs based on their operating principles:
the weighing-type gage and the tipping bucket-type gage (Figure 4.4.3). The
former, when made to NWS Specification No. 450.2201, is known as the Universal
gage, indicating usage for both liquid and frozen precipitation. There are
options for the remote transmission of signals from this type of gage. The
-------
Section No. 4.4.0
Revision No. 0
Date: 17 Sep 89
Page 2 of 7
standard National Weather Service Tipping Bucket Rain Gage is designed with a
12-inch collector funnel that directs the precipitation to a small outlet
directly over two equal compartments, or buckets, which tilt in sequence with
each 0.01 inches of rainfall. The motion of the buckets causes a mercury
switch closure. Normally operated on 6 V d.c., the contact closure can be
monitored on a visual counter and/or one of several recorders. The
digital-type impulse can also be used with computer-compatible equipment.
Figure 4.4.1
A typical non-recording
Rain Gage (Belfort
Instrument Co.)
Figure 4.4.2 Automatic
wet/dry precipitation
collector.
Bucket
Platform
Tipping
Bucket
Figure 4.4.3
A Typical Weighing Rain Gage (left) and Typical Tipping
Bucket Rain Gage (Belfort Instrument Company).
-------
Section No. 4.4.0
Revision No. 0
Date: 17 Sep 89
Page 3 of 7
4.4.1.3 Instrument Characteristics
The most accurate precipitation gage is the indicating-type gage.
However, the recording-type gage measures the time of beginning and ending of
rainfall and rate of fall. The Universal weighing gage incorporates a chart
drum that is made to rotate by either an 8-day spring-wound clock or a
battery-powered clock. Recent developments include a unit with a quartz
crystal mechanism with gear shafts for a wide range of rotation periods from
half a day to one month.
The weighing gage is sometimes identified by the name of its designer
(Fergusson) and comes with one of two recording mechanisms. In the single
traverse unit, the pen moves from the base of the drum to the top, typically a
water equivalent of 6 inches. In a dual traverse unit, the pen moves up and
then down for a total of 12 inches of precipitation. A variation of the
weighing gage, a "high capacity" design with dual traverse, will collect as
much as 760 mm or 30 inches. To minimize the oscillations incurred by strong
winds on the balance mechanism, weighing gages are fitted with a damper
immersed in silicone fluid. By incorporating a potentiometer in the mechanism,
the gage is capable of providing a resistance or, as another refinement, a
voltage proportional to the amount of precipitation collected. Linearity of
response is usually a factory adjustment involving the use of calibrated
weights to simulate precipitation amounts. In spite of manufacturer's
specifications, it is doubtful that the gage can resolve 0.01 inches,
especially when the bucket is nearly empty.
In the tipping bucket gage, the balance of the buckets and the leveling
of the bucket frame are critical. Low voltage at the gage is imperative for
reasons of safety. Power is typically 6 V d.c.. The signal is provided by a
switch closure each time the bucket assembly tips (0.01 inches of rainfall per
bucket). Rain rates are calculated from an event recorder with pens energized
sequentially to improve resolution. The tipping bucket (a mechanical device)
takes time to tilt from one position to the next. When the rate of fall is
high, there is spillage and the unmeasured precipitation falls into the
reservoir. Where there is a need for greater accuracy, the collected water is
measured manually, and excess amounts are allocated proportionately in the
record. The accuracy of the gage is given as 1 percent for rainfall rates of 1
in/hr or less; 4 percent for rates of 3 in/hr; and 6 percent for rates up to 6
in/hr.
4.4.1.4 Accessories - Windshields and Heaters
Accuracy of measurement for all types of gages is influenced perhaps
more by exposure than by variations in design. Windshields represent an
essential accessory to improve the catch of precipitation, especially snow in
windy conditions. The improved Alter design, made of 32 free-swinging but
separated leaves supported 1/2 inch above the level of the gage collecting
orifice, is an effective way to improve the catch. In a comparison of shielded
and unshielded 8-inch gages, it has been shown that at a wind speed of 5 mph,
the efficiency of the unshielded gage decreases by 25 percent, and at 10 mph,
the efficiency of the gage decreases by 40 percent (Weiss, 1961).
In below 'freezing conditions when the catch in a gage is snow or some
other form of solid precipitation, it is necessary to remove the
-------
Section No. 4.4.0
Revision No. 0
Date: 17 Sep 89
Page 4 of 7
collector/funnel of nonrecording gages and the funnel in recording gages Some
instruments are available with built-in heater elements that are
thermostatically controlled. An effective heater for conditions that are not
too severe is an incandescent lamp installed in the housing of the gage,
Caution should be exercised, however, as too great a heat will result in
evaporative loss.
4.4.2 SPECIFICATIONS
4.4.2.1 Precipitation Data Requirements
In research studies, especially those related to acid rain, the
instrument used most frequently is the Automatic Precipitation Collector with
one or two collecting buckets and a cover to prevent evaporation. In
operational activities, the choice is between the weighing gage and the tipping
bucket gage. For climatological surveys, the choice might include one of the
above gages as well as a nonrecording type gage. The use of a windshield is
recommended to minimize the errors that result from windy conditions if the
application requires maximum accuracy.
The precipitation measurement made in air quality monitoring stations is
frequently used for descriptive purposes or for episodal analysis. If the
effort required to achieve the level of accuracy specified by most
manufacturers of electrical recording gages is more than the application of the
data can justify, a tolerance of 10 percent may be adequate.
4.4.2.2 Procurement
In purchasing a suitable precipitation measuring system, specify the
type that fits the data application and include a requirement for accuracy
consistent with that application. A variety of gages are available
commercially. In general, the standards established by NWS specifications
result in the fewest problems. For example, there are numerous 8-inch gages
available, but those following NWS specifications are made only of brass and
copper, are more durable, and are reported to rupture less frequently under
extended freezing conditions than those made of galvanized steel.
The procurement of a weighing type gage should include a tripod mounting
base as well as a set of calibration weights. For locations that may not be
readily accessible, or for locations with heavy precipitation, the bucket of
the weighing gage should have an overflow tube. Refer to Section 4.4.3.2 for
antifreeze specifications. If the resolution of time is not too important,
recording rain gages of the drum type can be obtained with monthly rather than
weekly mechanisms. Unless the tipping bucket gage is equipped with a heater,
it is of no use for frozen precipitation.
4.4.2.3 Acceptance Test ing
Except for visual inspection, nonrecording gages do not require
acceptance testing. The weighing gages should be assembled and given a quick
"bench-top" calibration check using standard weights or a measured volume of
water. In addition, the clock mechanism supplied with the gage should be
checked for at least a couple of days, preferably a week. The tipping bucket
-------
Section No. 4.4.0
Revision No. '0
Date: 17 Sep 89
Page 5 of 7
gage should also be bench tested, primarily to be certain that the bucket
mechanism assembly is balanced and that the switch is operational.
4.4.2.4 Calibration
Bench calibrations should follow the recommendation of the manufacturer.
The electrical output gage or the drum recording gage measures weight, whether
total weight in the case of the "weighing gage" or increments of weight in the
case of the tipping bucket gage. Density of water is assumed so the weight can
be expressed in units of volume or depth assuming the area of the collector
opening. Calibrations of the measurement apparatus can be based upon the
introduction of known volumes of water. The area of the collection surface
must be known for the volume collected to be expressed as a depth. For
example, an "eight inch" collector may feed a tipping bucket which tips when
7.95 cc of water has arrived. If this volume of water is to represent 0.01" of
rainfall, the effective collection area must be 48.51 square inches, from the
following calculations:
7.95 cc = 0.485 in.3 - 0.01 in. * 48.51 in.2
If the area is a circle, the diameter should be 7.86 inches.
(48.51/7i)1/2 = 3.93 in. radius
For rate-sensitive systems such as the tipping bucket, the rate of simulated
precipitation should be kept less than one inch per hour. Calibrations require
properly leveled weighing systems (gages) whether on the bench or in the field.
4.4.3 OPERATIONS
4.4.3.1 Installation
Refer to Section 4.0.4.4.2.4 which provides some siting guidance for
precipitation measurement. The support, or base, of any gage must be firmly
anchored, preferably on a level surface so that the sides of the gage are
vertical and the collector is horizontal. The collector can be checked with a
carpenter's level placed at two intersecting positions. The level of the
bucket assembly on the tipping bucket gage is also critical and should also be
checked along its length and width.
Once the weighing gage is installed, the silicone fluid should be poured
into the damping cylinder as required. The pen of the drum recording type is
inked to less than capacity because the ink used is hygroscopic and expands
with increasing humidity, easily spilling over the chart. The final
calibration check with standard weights or suitable substitute should be made
at this time. To check the operation of the tipping bucket, the best approach
is to put a known quantity of water in a can with a small hole so that the slow
flow can be timed. It may be necessary to adjust the set screws, which act as
limits to the travel of the tilting buckets. The average of a minimum of ten
tips should be used. Adjustment is required if a 10 percent or greater error
is found or if greater accuracy is needed.
4.4.3.2 Field Operation of a Precipitation Measurement System
Calibration checks for weighing and tipping bucket gages using the
techniques described above are recommended at 6-month intervals. Nonrecording
gages, whether alone or in a network, should be read daily at a standard time.
-------
Section No. 4.4.0
Revision No. 0
Date: 17 Sep 89-
Page 6 of 7
Although the weighing gage is used for liquid and frozen precipitation,
it requires some special attention for winter operations. First, the funnel
must be removed when snow is expected. Second, the bucket must be charged with
an antifreeze, 24 oz of ethylene glycol mixed with 8 oz of oil. The weight of
this mixture represents the baseline from which precipitation amounts are to be
noted. The bucket should be emptied and recharged when necessary, at about 5
inches in the Universal gage, and at about 10 inches in the punched tape gage.
If the antifreeze mixture is classified as hazardous or environmentally
sensitive, care must be taken in the disposal process. All operational
activities should be recorded in the station log.
4.4.3.3 Preventive Maintenance
Possible leaks in the measuring tube or the overflow container of the
gage are easily checked. The receptacles are partially filled with water
colored with red ink and placed over a piece of newspaper. This procedure is
especially applicable to the clear plastic 4-inch gage which is more easily
damaged. Repairs are performed by soldering the 8-inch gage and by applying a
solvent to the plastic.
A number of pens, some with greater capacity than others, can be used
with the Universal gage. All require occasional cleaning, including a good
soaking and wiping in a mixture of water and detergent. After inking problems,
the next source of trouble is the chart drive; but these problems can sometimes
be avoided by having the clock drive lubricated for the environmental
conditions expected. It is a good practice to have spare clocks in stock.
Routine visual checks of the performance of weighing type gages should
be made every time there is a chart change. The time and date of change, and
site location should be documented. Routine maintenance should include inking
the pen and winding the clock. Battery-powered chart drives will require
periodic replacement of batteries based on either experience or manufacturer's
recommendations. All preventive maintenance activities should be noted In the
log book.
4.4.4 PERFORMANCE AUDIT METHODS
Audits on precipitation measuring systems need be no more frequent than
every 6 months. The irregular occurrence of precipitation makes the use of a
CIS impractical. The performance audit should depend upon the challenging of
the gage with amounts of water known to an accuracy of at least 1 percent of
the total to be used. This method will provide an accuracy of the measurement
system but not the collection efficiency of the gage in natural precipitation.
For tipping bucket gages use a rate of less than one inch per hour and an
amount which will cause a minimum of ten tips.
For weighing gages, it is more convenient to use calibration weights to
challenge the weighing mechanism rather than using the gallons of water
necessary for full scale testing.
All types of precipitation gages should be measured to determine the
effective collection area. This measurement is only required once but the
difficulty of measuring the area of a slightly out-of-round collector may
require several samples to accurately find the area.
-------
Section No. 4.4.0
Revision No. 0
Date: 17 Sep 89
Page 7 of 7
4.4.5 REFERENCES
EPA, 1985: Quality Assurance Handbook for Air Pollution Measurement Systems,
Vol. V, Precip. Measurement Systems, EPA-600/4-82/042a. Office of Research
and Development, Res. Triangle Park, NC 27711.
NCAR, 1984: Instructor's Handbook on Meteorological Instrumentation, F. V.
Brock, Editor. NCAR Technical Note, NCAR/TN-237+1A.
Weiss,L .L., 1961: Relative catches of snow in shielded and unshielded gages
at different wind speeds. Monthly Weather Review, Vol. 89.
WHO, 1971: Guide to meteorological instrument and observing practices. World
Meteorological Organization No. 8TP3, 4th edition, Geneva, Switzerland.
-------
-------
Section No. 4.5.0
Revision No. 0
Date: 6 Aug 89
Page: 1 of 10
QUALITY ASSURANCE FOR RELATIVE HUMIDITY OR DEW POINT TEMPERATURE
4.5.0 INTRODUCTION
Humidity is a general term for the water-vapor content of air. Other,
more specific, terms for humidity include: absolute humidity, relative humid-
ity, specific humidity, mixing ratio, and dew point (Huschke, 1959). This
section discusses the measurement of relative humidity and dew point. Relative
humidity (RH) is a dimensionless ratio of the actual vapor pressure of air to
the saturation vapor pressure at a given dry bulb temperature. Dew point is
the temperature to which air must be cooled, at constant pressure and constant
water vapor content, to be saturated with respect to liquid water. Frost point
is the temperature below 0 C at which air is saturated with respect to ice.
4.5.1
TYPES OF INSTRUMENTS
There are many ways to measure the water vapor content of the atmo-
sphere. These can be classified in terms of the six physical principles
(Middleton and Spilhaus, 1953) listed in Table 4.5.1. Examples of instruments
for each technique are provided.
Table 4.5.1 Principles of Humidity Measurement
Principle
Reduction of temperature
by evaporation
Dimensional changes due to
absorption of moisture,
based on hygroscopic
properties of materials
Chemical or electrical
changes due to absorption
or adsorption
Formation of dew or frost
by artificial cooling
Diffusion of moisture
through porous membranes
Absorption spectra of water
vapor
Instrument/Method
_L
psychrometer
hygrometers with sensors
of hair, wood, natural
and synthetic fibers
electric hygrometers
such as Dunmore Cell;
lithium, carbon, and
aluminum oxide strips;
capacitance film
cooled mirror surfaces
diffusion hygrometers
infrared and UV absorp-
tion; Lyman-alpha
radiation hygrometers
Instruments such as diffusion hygrometers that involve the diffusion of
moisture through porous membranes are used primarily in research programs. The
same is true 'of instruments that utilize the absorption spectra of water vapor,
such as infrared and ultraviolet hygrometers, and Lyman-alpha radiation hygro-
-------
Section No. 4.5.0
Revision No. 0
Date: 6 Aug 89
Page: 2 of 10
meters. This class of instrument requires frequent attention and represents a
major investment in procurement and maintenance costs.
Psychrometry identifies a basic technique for
deriving both relative humidity and dew point temperature
from a pair of thermometers—a dry bulb thermometer that
measures the ambient temperature, and a wet bulb thermo-
meter. The reservoir of the wet bulb thermometer is
covered with a muslin wick. When the wick is moistened
and the thermometer ventilated, the indicated temperature
is related to the amount of evaporative cooling that takes
place at the existing ambient temperature, water vapor
partial pressure, and atmospheric pressure.
The temperature sensors in a sling psychrometer
(Figure 4.5.1) are usually mercury- or alcohol-filled
thermometers. The same is true of portable motor-
operated psychrpmeters (Figure 4.3.7.3), but the psychro-
metric principle has been used with sensors made of
thermocouples, wire-wound resistance thermometers, therm-
istors, and bimetal thermometers. Relative humidity and
dew point are easily determined by observ- ing the
difference between the dry bulb and the wet bulb—the wet
bulb depression—and then referring to psychrometric
tables, charts, or calculators. One must be certain to
use computed values for the atmospheric pressure range of
the location where the observation is taken.
Figure 4.5.1
Sling
psychrometer
More measurements of atmospheric water vapor have
probably been made with the sling psychrometer than by any
other manual method. When properly used and read, the
technique can be reasonably accurate, but it is easily misused. The most
important errors are from radiation, changes during reading, and parallax. The
Assmann psychrometer continuously aspirates the thermometers and protects them
from radiation which allows time and
accessibility for a careful reading to
avoid parallax (a parallax avoiding guide
to keep the eye perpendicular to the
meniscus is best, see Figure 4.5.2). For
good accuracy, particularly where a
variety of observers are taking measure-
ments, an Assmann or equivalent type
psychrometer is recommended. One should
use the psychrometric tables with dew
point values for the altitude (pressure)
where measurements are being made.
Hygrographs, which record relative
humidity, or hygrothermographs, which
record both relative humidity and temper-
ature, usually incorporate human hair as
Figure 4.5.2 Assmann psychro- the moisture-absorbing sensor. Other
meter with parallax guides. instruments with sensors that respond to
water vapor by exhibiting dimensional
-------
Section No. 4.5.0
Revision No. 0
Date: 6 Aug 89
Page: 3 of 10
changes are available. They are made with elements such as wood, goldbeater's
skin (an animal membrane), and synthetic materials, especially nylon.
Instruments that utilize the hygroscopic characteristics of human hair
are used most frequently, primarily because of availability. The hygrograph
provides a direct measure of relative humidity in a portable instrument that is
uncomplicated and is relatively inexpensive. There are limitations in accuracy
below 20 percent relative humidity and above 80 percent that may be unaccept-
able, as well as limitations for applications at low temperatures. Atmospheric
Environment Services of Canada has found that Pernix, a specially treated and
flattened hair element, can be used at temperatures below freezing without
serious errors. The hygrothermograph made to an NWS specification incorporates
human hair as the humidity sensor and bourdon tube (a curved capsule filled
with alcohol) as the temperature sensor.
Dew point hygrometers with continuous electrical outputs are in common
use for monitoring. One dew point hygrometer was originally developed for air
conditioning control applications under the trade name Dewcel (Hickes, 1947)
and was adopted to meteorological use (Conover, 1950). From the trade name,
the generic term dew cell has evolved that now identifies an instrument made by
several manufacturers. This device determines moisture based on the principle
that for every water vapor pressure there is an equilibrium temperature at
which the saturated salt solution neither absorbs nor gives up moisture to the
surrounding atmosphere.
The dew cell, also known by the trade name Dew
Probe, consists of bifilar wire electrodes wrapped
around a woven glass cloth sleeve that covers a hollow
tube or bobbin. The sleeve is impregnated with a
lithium chloride solution (Figure 4.5.3). Low-voltage
a.c. is supplied to the electrodes, which are not
interconnected but depend on the conductivity of the
atmospherically moistened lithium chloride for current
flow. The temperature sensor in the tube is usually a
resistance thermometer, but can be a thermistor,
thermocouple, bimetal thermometer, capillary system, or
any sensor calibrated for the proper temperature-to-
dew-point relationship.
In the early 1960's, the technique of detecting
the dew point on a cooled mirror surface evolved into a
production-type unit. This unit was automatically
operated and had an optical dew-sensing system that
incorporated thermoelectric cooling (Francisco and
Beaubien, 1963). Four manufacturers now produce a
meteorological type, thermoelectric, cooled-mirror dew
point instrument (Mazzarella, 1977). Three of these
instruments cover the range of -50 to +50 C. Linear
thermistors are used to measure the mirror temperature
in three of the units; a platinum wire sensor is used
in the other. All are designed with simultaneous
linear output signals for T
and T (ambient 'temperature).
(dew point temperature)
Two of the manufacturers
Figure 4.5.3 A
typical Devcel
sensor housing
and transmitter
-------
Section No. 4.5.0
Revision No. 0
Date: 6 Aug 89
Page: 4 of 10
make claim to NBS-traceability with stated dew point accuracies ranging from
±0.2° to ±0.4°C and ambient temperature accuracies ranging from ±0.1° to
±0.5 C. All incorporate some form of standardization that involves clearing
the mirror by heating, either automatically or manually. Although complex in
design and operation, this type of cooled-mirror hygrometer is considered to be
a functional standard.
In recent years, two other sensors for humidity have been used on tower
installations for atmospheric pollution studies. One involves the use of an
organic seed, cut and coupled to a strain gage. In principle, absorption of
moisture in the seed results in distortion, which is converted to an electrical
signal by the strain gage assembly. Reports on performance are mixed.
Certainly the applications are limited, and the approach does not represent a
technological advance. By contrast, the thin film capacitor, designed primari-
ly for radiosonde applications, incorporates advanced technology (Suntola and
Antson, 1973). Reports of users in the past have been mixed, with a common
complaint of poor performance in polluted atmospheres. Modern capacitor-type
sensors have achieved a better performance through improved design and user
education.
4.5.1.1 Sensor Characteristics
Although the psychrometer is considered the most practical and widely
used instrument for measuring humidity, two major problems are associated with
wet and dry bulb psychrometry involving the accuracy of the thermometers and
the cumulative errors related to operating technique (Quinn, 1968). An accur-
acy of ±1 percent at 23°C and 50 percent RH requires thermometers with relative
accuracy of ±0.1°C. The commonly used 0.5 C division thermometers introduce an
uncertainty of ±5 percent RH at this condition. This assumes that the readings
were taken at the maximum wet bulb depression, a difficult task with a sling
psychrometer.
It has long been recognized that there are some limitations in using the
dew cell instrument (Acheson, 1963). The lowest relative humidity it can meas-
ure at a given temperature is the ratio of the vapor pressure of a saturated
solution of LiCl to that of pure water. This is calculated to be 11.8 percent
RH. A second limitation is that at -65.6°, -20.5°, +19.0°, and +94.0°C, LiCl
in equilibrium with its saturated solution undergoes a phase change. Errors in
dew point measurements occur at -69 , -39 , -12 , and +40 C. This problem is
inherent in the use of LiCl and cannot be eliminated. It is estimated that the
accuracy of the LiCl saturated salt technique is 1.5°C over the range of -30°
to 30°C.
The optical chilled (cooled) mirror technique of measuring dew point is
a fundamental measurement. No calibration is required for the fundamental dew
generating process. The measurement however is the temperature of the surface
at which the dew forms and as with any electrical temperature measurement
system, calibration is required. The process of periodically heating the
mirror to a temperature above the dew (or frost) point is followed by zeroing
the optical system to correct for the dry mirror reflectance changing due to
contamination. In the better instruments, automatic zeroing is programmable in
terms of frequency and length of time. It can also be accomplished manually.
-------
Section No. 4.5.0
Revision No. 0
Date: 6 Aug 89
Page: 5 of 10
4.5.1.2 Sensor Housings and Shields
Psychrometers of all types should be acclimated to the environmental
conditions in which the measurements are to be made. In most cases, psychro-
meters should be stored in a standard instrument shelter so that the mass of
the thermometers, especially the mass of the housing, adjusts to the temper-
ature of the air. Psychrometers with a stored water supply, such as those on a
tower, must be shielded from solar radiation.
For meteorological applications, the dew cell element should be enclosed
in a weatherhood to protect it from precipitation, wind, and radiation effects.
This type of element functions best in still air. Some aspirated radiation
shields are designed, in keeping with these specifications to house both a
temperature sensor, which requires ventilation, and a dew cell, which requires
only the smallest amount of air flow (Figure 4.5.4). The miniaturization of
the dew cell has created some problems related to excessive air flow and solar
radiation that remain only partially solved.
All'manufacturers of optical cooled-mirror dew point and temperature
monitoring equipment provide housings for the sensors, which include forced
ventilation and shielding from solar radiation.
4.5.1.3 Data Requirements
Electrical hygrometers for monitoring applications have time constants
generally longer than air temperature systems. The usual data of interest are
hourly average values. Data should be reported in terms of the condition
measured, dew point temperature,
relative humidity or wet-bulb and dry-
bulb temperature. Programs may be used
which convert among these if all the
relevant variables are known. The
station elevation may be used to
estimate a nominal pressure if a
measurement is not available. The temp-
erature needed to convert a relative
humidity measurement to dew point temp-
erature is that temperature at the
relative humidity sensor surface. This
may not be the same temperature as that
measured at some other location. On the
other hand, the dew point temperature is
a fundamental measure of the amount of
water vapor in the air and is indepen-
dent of air temperature. Relative
humidity calculations can therefore be
made given the dew point temperature and
any temperature measurement point in the
Figure 4.5.4 A pair of tower-
mounted Gill aspirated radiation
shields for housing temperature
and dew point sensors (Young).
same general air mass. Empirical formulae for the estimation of relative
humidity as a function of dew point temperature and air temperature, relative
humidity as a function of wet-, dry-bulb temperature and pressure, and dew
point temperature as a function of relative humidity and temperature are shown
below. »
-------
Section No. 4.5.0
Revision No. 0
Date: 6 Aug 89
Page: 6 of 10
To calculate relative humidity (RH = 100 r, %) from air temperature (T,
C) and dew point temperature (T , C), do the following:
r = exp
b + T
b + T
(1)
where a = 17.27
b = 237.3
To calculate the dew point temperature (T , C) from air temperature (T,
3C) and relative humidity (RH = 100 r, 7.) use
In r +
aT
b + T
T =
D
a -
In r +
aT
b + T
(2)
To calculate relative humidity (RH = 100 r, %) from air temperature
(T, C), wet-bulb temperature (T , C), and atmospheric pressure (P, mb)
through the vapor pressure (e, mb) and the saturation vapor pressure (e , mb),
do the following:
where A = 6.6 x 10
B = 1.15 x 10
-4
-3
-AP
exp
SW
exp
aT
+BTw)(T-Tw)
(3)
To estimate wet-bulb temperature (T , C) from air temperature (T, C)
dew point temperature (T , °C), relative humidity (r, ratio) and atmospheric
pressure (P, mb), do the following:
P + T
P + T
19 + 130r - 28r
19 + 130r - 28r
(4)
The summarization of these relationships was suggested by A. L. Morris from
material found in Z. Geophysik, 6, 297, 1930, the Smithsonian Meteorological
Tables, Sixth Revised Edition, the Glossary of Meteorology and has been
augmented by hife own derivation of expression (4).
-------
Section No. 4.5.0
Revision No. 0
Date: 6 Aug 89
Page: 7 of 10
Psychrometers are convenient devices for making spot checks of the per-
formance of other devices, especially those that are permanently installed,
providing the checking is done under reasonably steady overcast conditions.
The psychrometric technique built into tower installations presents servicing
problems, especially at temperature extremes. High temperatures cause rapid
evaporation, and low temperatures cause freezing.
Both the dew cell and the cooled-mirror type instruments have applica-
tions on 10-meter or taller tower installations for pollution studies,
providing the sensors are housed in the recommended shields with little, if
any, aspiration for the dew cell and the recommended rate of aspiration for the
cooled-mirror design is selected.
4.5.2 SPECIFICATIONS
4.5.2.1 Procurement
The selection of a humidity instrument is guided by the application to
which the data will be put. The PSD (Prevention of Significant Deterioration)
guideline (EPA, 1987) provides the following: H...If the permit granting auth-
ority determines that a significant potential exists for fog formation, icing,
etc., due to effluents from the proposed facility, error in the selected
measurement technique should not exceed an equivalent dewpoint temperature
error of 0.5 C. Otherwise, errors in equivalent dew-point temperature should
not exceed 1.5°C over a dewpoint range of -30°C to +30°C." This latter toler-
ance allows for the use of lithium chloride dew cells.
Sling psychrometers and aspirated psychrometers with thermometers
shorter than 10 inches do not have sufficient resolution for the accuracies
required for checking other instruments. Equally important, the thermometers
should have etched stems; i.e., the scale markings should be etched on the
glass. Reliable thermometers are factory calibrated at a minimum of two temp-
eratures, and usually at three. Thermometers calibrated with NBS-traceable
standards are preferred.
When patents expired on the original Dewcell, a number of similar units
appeared on the market. In light of problems which have existed in the past,
it is prudent to specify accuracy of the humidity system when it is operating
as a system in the atmosphere. Problems with ventilation rates will be quickly
exposed by this requirement. It is not recommended to purchase components to
patch together in a system. Corrosion in polluted atmospheres can be avoided
by selecting optional 24-carat gold windings, provided cost is not prohibitive.
If dew point alone is to be measured, the standard weatherhood is a proper
choice. If both temperature and dew point are to be measured, it may be
advantageous to purchase a standard shield that provides a housing for the dew
cell and a separate aspirated compartment for the temperature probe.
Optical cooled-mirror dew point systems are now commercially available
from several manufacturers, all of which incorporate either linear thermistors
or platinum resistance temperature devices.
-------
Section No. 4.5.0
Revision No. 0
Date: 6 Aug 89
Page: 8 of 10
4.5.2.2 Acceptance Testing
Test at least the ambient atmosphere at one point in normal wind and
radiation.
4.5.2.3 Calibration
The procedure for calibrating the thermometers in a psychrometer is
essentially the same as any thermometer calibration (See Section 4.3.5).
Both the dew cell and the cooled-mirror hygrometer can be checked for
approximate calibration accuracy with a motor-operated psychrometer. Their
performance should be verified under stable conditions at night or under cloudy
conditions during the day. Several readings taken at the intake of the aspir-
ator or shield are recommended. Bench calibrations of these more sophisticated
units must be made by the manufacturer. The electronics portion of some
instruments may be calibrated by substitution of known resistances in place of
the temperature sensor. This procedure, if appropriate, is described in the
manufacturer's operating manual for the instrument.
4.5.3 OPERATIONS
4.5.3.1 Installation
Dew point measuring equipment on a tower should be installed with the
same considerations given to temperature sensors. Reference has already been
made to the weatherhood as a shield for the dew cell and to an aspirated shield
for the cooled-mirror instrument. At some installations, success has been
reported in mounting these housings so that they are close to the tower frame-
work on the north-facing side. This minimizes the effects of direct solar
radiation and provides a rigid support, especially for the cooled-mirror
sensor, which requires a stable mounting surface. Another consideration in
mounting these devices inboard involves servicing. Inboard mounting makes
recharging the dew cell with lithium chloride and cleaning the reflective
surface of the cooled- mirror hygrometer much easier.
4.5.3.2 Field Operation and Preventive Maintenance
Field calibration checks should be made at least monthly on dew cell
type units. The use of gold wire windings around the LiCl cylinder minimizes
corrosion problems in polluted atmospheres. Periodic removal and washing of
old lithium chloride, followed by recharging with a fresh solution, improves
data reliability.
Once a mercury or alcohol liquid-in-glass thermometer is calibrated,
there is no need for recalibration, unless it is to be used for reference or as
a transfer standard. Errors in wet bulb temperatures are most frequently the
result of an improperly installed or dirty muslin wick, the repeated use of tap
water instead of distilled water, or human error in reading. Wicking material
used on psychrometers must be washed to remove traces of sizing and finger-
prints. Once cleaned, the material is tied at the top of the thermometer bulb
and a loop of thread placed around the bottom so the thermometer bulb is
tightly covered. To prevent solid materials from collecting on the cloth and
preventing proper evaporation, the wick should be wet with distilled water. Of
-------
Section No. 4.5.0
Revision No. 0
Date; 6 Aug 89
Page: 9 of 10
course, slinging or motor aspiration should be done in the shade, away from
reflected or scattered radiation, at a ventilation rate of about 3 to 5 m/s.
Many technique-related errors are minimized by using an Assmann-type, motor-
operated psychrometer, providing the instrument is allowed to assume near
ambient conditions prior to use.
The cooled-mirror instruments require no calibration except for the
minor temperature sensor. Depending on environmental conditions, the mirror is
easily cleaned with a Q-Tip dipped in the recommended cleaning fluid, usually a
liquid with an alcohol base. While the accuracy of a psychrometer is inferior
to that of the optical chilled mirror system, an occasional check at the intake
to the sensor shield is recommended under the provisions specified earlier.
All operational and preventive maintenance activities should be logged.
Data retrieval will be dependent upon program objectives.
4.5.4 PERFORMANCE AUDIT METHODS
Instrument audit procedures for hygrometry systems follow calibration
procedures. A systems audit should be performed near the beginning of a field
measurement program.
The performance audit of a humidity measuring system should be based
upon a comparison with a collocated transfer standard (CTS). Parts of the
system can be tested by conventional electronic tests, but this avoids so much
of the measurement process that it should only be used to augment the total
system test. The CTS may be any qualified instrument. The most accurate type
is the cooled-mirror dew point instrument. The Assmann-type psychrometer with
calibrated thermometers traceable to NBS is acceptable for most data applica-
tions. It is also most convenient since it does not require commercial power
and can be carried to elevated levels on a tower.
Given the qualifier that humidity is a very difficult measurement to
make, a rule of thumb for judging the accuracy of a humidity monitoring system
with an Assmann as the CTS is as follows: when the CTS and the challenged
system agree in dew point temperature to within 1°C, the challenged system is
assumed to be within 0.5 C of the true value. This arbitrarily assigns an
uncertainty in dew point temperature of +0.5°C for the Assmann which is true
for most of the range.
Auditing is best backed by authoritative standards. ASTM, 1982, 1983,
1984 and 1985 may be of selective value.
-------
Section No. 4.5.0
Revision No. 0
Date: 6 Aug 89
Page: 10 of 10
4.5.5 REFERENCES
Acheson, D. T., 1963: Some limitations and errors inherent in the use of
dew-cell for measurement of atmospheric dew points. Monthly Weather Review.
ASTM, 1982: Standard Definitions of Terms Relating to Humidity
Measurements, D4023-82a, American Society for Testing and Materials,
Philadelphia, PA.
ASTM, 1983: Standard Method of Measuring Humidity with CooLed-Surface
Condensation (Dev Point) Hygrometer, D4230-83, American Society for
Testing and Materials, Philadelphia, PA.
ASTM, 1984: Standard Test Method for Measuring Humidity with a Psychrometer
(the Measurement of Vet- and Dry-Bulb Temperatures), E337-84, American
Society for Testing and Materials, Philadelphia, PA.
ASTM, 1985: Standard Practice for Maintaining Constant Relative Humidity by
Means of Aqueous Solutions, E104-85, American Society for Testing and
Materials, Philadelphia, PA.
Berry, F. A., Jr., E. Bollay and N. R. Beers, 1945: Handbook of Meteorology.
McGraw-Hill Book Company, Inc.
Conover, J. H., 1950: Tests and adaptation of the Foxboro dew-point recorder
for weather observatory use. Bulletin of American Meteorological
Society,31(1),13-22.
EPA, 1987: Ambient Monitoring Guidelines for Prevention of Significant
Deterioration (PSD). EPA-450/4-87-007, Office of Air Quality Planning and
Standards, Research Triangle Park, NC.
Francisco and Beaubien, 1965: An automatic dew point hygrometer with
thermoelectric cooling. Humidity and Moisture, edited by A.Wexler,
Reinhold Publishing Company.
Hicks, W. F., 1947: Humidity measurement by a new system. Refrigerating
Engineering. American Society of Refrigerating Engineering.
Huschke, R.,ed., 1959: The Glossary of Meteorology. American Meteorological
Society, Boston, MA.
Mazzarella, D. A., 1972: Meteorological instruments: their selection and use
in air pollution studies. Proceedings of the Meeting on Education and
Training in Meteorological Aspects of Atmospheric Pollution and Related
Environmental Problems. World Meteorological Organization, No. 493.
Middleton, W. E. K., and A. F. Spilhaus, 1953: Meteorological Instruments,
University of Toronto Press.
Quinn, F. C., 1963: Humidity-the neglected parameter. Testing Engineering.
The Mattingly Publishing Company, Inc.
Suntola and Antson, 1973: A thin film humidity sensor. Scientific
Discussions, CIMO VI, World Meteorological Organization.
-------
Section No. 4.6.0
Revision No. 0
Date: 17 Sep 89
Page: 1 of 8
QUALITY ASSURANCE FOR SOLAR RADIATION MEASUREMENTS
4.6.0 INTRODUCTION
Solar energy is the driving force of large-scale atmospheric motion,
indeed, of the general circulation of the atmosphere. Although air pollution
investigators normally consider the measurement of solar radiation secondary to
wind and temperature measurements, solar radiation is directly related to
atmospheric stability. It is measured as total incoming global radiation, as
outgoing reflected and terrestrial radiation and as net total radiation.
Quantitatively, solar radiation is described in units of energy flux,
either W/m or cal/cm «min. When measured in specific, narrow wavelength
bands, solar radiation may be used to evaluate such air pollution indicators as
turbidity, amount of precipitable water, and rates of photochemical reactions.
However, this manual will cover only broadband measurements and sunshine.
The generic term, radiometer, refers to any instrument that measures
radiation, regardless of wavelength. Shortwave radiation has wavelengths less
than 4 micrometers (jjm) and is subdivided as follows:
Ultraviolet (UV)
Visible
Near-infrared
0. 20 urn to 0. 38 fim
0.38 urn to 0.75 /urn
0.75 /jm to 4.00 (urn
Longwave radiation has a wavelength as follows:
Infrared (IR) 4 MID to 100 /im
and comes from the Earth and its atmosphere. The instruments most commonly
used for environmental monitoring are discussed below.
4.6.1 TYPES OF INSTRUMENTS
Sensing
• Precis/on Ground &
Guard Disc
\
element ^~-—^// Polished Glass Dome
(Frequently Double)
4.6.1.1 Pyranometers
Pyranometers are instruments that measure the solar radiation received
from the hemispherical part of
the atmosphere it sees,
including the total sun and
sky shortwave radiation on a
horizontal surface (Figure
4.6.1). Most pyranometers
incorporate a thermopile as
sensor. Some use a silicon
photovoltaic cell as a sensor.
The precision spectral
pyranometer (PSP) is made by
Eppley Laboratories and has
two hemispherical domes
designed to measure sun and
sky radiation on a horizontal
surface in defined
Leveling
Hi |j in Screw
II
wavelengths. This is achieved
by substituting one of several
Figure 4.6.1
Features of a typical
pyranometer (Carter, ei al.,1977)
-------
Section No. 4.6.0
Revision No. 0
Date: 17 Sep 89
Page: 2 of 8
colored Schott glass filter domes for the clear glass outer dome. The smaller
dome suppresses convection, so this type is better sited if tilted from the
horizontal.
4.6.1.2 Bimetallic Recording Pyranometers
Bimetallic recording pyranometers, also known as actin- ometers,
weredesigned by Robitzsch of Germany. These mechanical sensors consist of two
or three bimetallic strips, alternately painted black and white, that
respectively absorb and reflect solar radiation. The resulting differential
heating produces a deformation that is transmitted mechanically through levers
and a pen arm to a clock-wound drum recorder. Although of limited accuracy,
these instruments are useful for locations with no commercial power.
4.6.1.3 Net Radiometers
Net radiometers or net pyrradiometers are designed to measure the
difference between downward and upward total radiation, including the total
incoming shortwave and longwave radiation and the
total outgoing shortwave and longwave radiation.
There are two basic types of net radiometers. The
ventilated plate type, often referred to by the
name of the designers (Gier and Dunkle), is more
popular in research applications than the type
with hemispherical polyethylene domes originally
designed by Funk. Both incorporate thermopiles
with blackened surfaces. Because net radiometers
produce a signal with a positive sign when the
incoming radiation exceeds the outgoing, the
recording equipment must be designed with an
offset zero.
4.6.1.4 Sunshine Recorders
Sunshine recorders are designed to provide
information on the hourly or daily duration of
sunshine. Only one commercially available,
off-the-shelf type of sunshine recorder is now
available. This is the Campbell-Stokes design
(Figure 4.6.2), designated as the interim
reference sunshine recorder "IRSR" by the World
Meteorological Organization. The device consists
of a glass sphere 10 cm in diameter mounted in a
spherical bowl. The sun's rays are focused on a
card that absorbs radiation and changes color in
the presence of sunlight. The recorder is used
infrequently in the United States but extensive-
ly abroad, primarily for the collection of
climatological data. The National Weather Service
routinely uses a Sunshine Switch, which
incorporates one shaded photocell and one exposed
photocell.
Figure 4.6.2 A Campbell-
Stokes Sunshine Recorder
(U.S. Army, 1975)
-------
Section No. 4.6.0
Revision No. 0
Date: 17 Sep 89
Page: 3 of 8
4.6.1.5 Instrument Characteristics
Pyrheliometer
Declination Adjustment
24 hr. Dial
Only the characteristics of pyranometers and net radiometers, the two
types of instruments used most frequently in pollution-related programs, will
be discussed in this section. The pyranometer is not to be confused with the
pyrheliometer, "an instrument for measuring the intensity of direct solar
v , radiation at normal incidence" (WHO,
1971). The pyrheliometer is mounted
in a solar tracker, or equatorial
mount, automatically pointing to the
sun as it traverses from east to west
(Figure 4.6.3). By contrast, the
pyranometer is mounted facing toward
the zenith. Ideally, the response of
the thermopile sensor in the
pyranometer is proportional to the
cosine of the angle of the solar beam
and is constant at all azimuth angles.
This characteristic is known as the
Lambert Cosine Response, an important
characteristic of pyranometers. For
the majority of applications related
to atmospheric pollution, Class 2 and
Class 3 are satisfactory (see Table
Figure 4.6.3 Features of a. 4.6.1).
typical pyrheliometer and track-
ing mount (Carter, et al. , 1977) Most net radiometers now
available commercially are made with a
small disc-shaped thermopile covered by polyethylene hemispheres. In most
units the material used for shielding the element from the wind and weather is
very thin and is transparent to wavelengths of 0.3 to 60 urn. Until recently,
the internal ventilation and positive pressure required to maintain the
hemispheres of net radiometers in their proper shape was considered critical;
however, new designs have eliminated this problem. The plate-type net
radiometer, most often the modified Gier and Dunkle design sold commercially in
the United States, is occasionally used in routine air pollution
investigations. The thermopile heat flow transducer is blackened with a
material that is easily cleaned with water or naphtha. Because the thermopile
is uncovered for total spectrum response, a built-in blower, available for
operation on 115 V 50/60 Hz or 12 V d.c., draws air across the element at a
constant rate eliminating the effects of varying natural winds. The device is
temperature-compensated and typically has a sensitivity of 2.2 fiV per W/m , a
response time of 10 seconds, and a "relative" accuracy of two percent in
calibration. When supplied with a reflective shield on its lower surface, this
plate type net radiometer of the Gier and Dunkle design becomes a total
hemispherical radiometer or unshielded pyranometer.
4.6.1.6 Recorders and Integrators for Pyranometers and
Net Radiometers
The relatively high impedance and low signal of thermopile sensors,
excluding silicon photovoltaic cells, limits their use with both indicating
meters and reco'rding meters. Electronic strip chart millivolt potentiometric
recorders incorporating variable-range rheostats are preferred. The
-------
Section No. 4.6.0
Revision No. 0
Date: 17 Sep 89
Page: 4 of 8
variable-range rheostat permits the exact matching of the recorder scale to
interchangeable sensors so that deflections of the meter represent engineering
units, i.e., W/m2, cal/cm -min, etc. The alternative is a standard millivolt-
meter potentiometric recorder where the data, in millivolts, must be translated
to units of energy, corresponding to full-scale values of 1370 W/m or 1.96
cal/cm «min. It may also be necessary, especially if the signal is to be
as an input to a computer, to combine preamplification with scaling.
4.6.2 SPECIFICATIONS
4.6.2.1
Procurement
In purchasing a solar radiation measurement system, follow the practice
of matching the data requirements to the instrument selection, specifying the
performance required on the purchase order (complete with test method to verify
performance) and testing the performance in receiving. See Section 4.1.4.5 for
comments on traceability protocol. Many types of radiation instruments have
been developed, especially in recent years, because of an increasing interest
in environmental considerations (Gates, 1962), meteorological research
(Monteith, 1972), and solar energy (Carter, et al., 1977). Except for special
studies, the requirements for relating radiation to stability can be satisfied
by purchasing sensors of Class 2 or Class 3 as identified by the WHO (see Table
4.6.1).
Table 4.6.1 Classification of Pyranometers According to
Physical Response Characteristics
Sens.
2
( mW/cm " )
1st Class ±0.1
2nd Class ±0.5
3rd Class ±1.0
Temp.
( % )
± I
± 2
± 5
Lin.
( % )
± 1
± 2
± 3
Max Time
Constant
(sec. )
25
60
240
Cosine
Response
( % )
± 3.0
± 5.7
± 10
Class 2 sensors offer the advantage of providing data comparable to that
collected at National Weather Service stations and at key locations of
Department of Energy (DOE). The sensors to be specified should be commercially
available, field proven by the manufacturer for several years, and have the
technical requirements established by WHO standards. Several American Society
for Testing and Materials (ASTM) standards are available (ASTM, 1984). When
purchasing a recorder or integrator, one must match the calibration factor or
sensitivity of the sensor to the readout equipment. It must be recognized that
the signals from net radiometers, in contrast to pyranometers, require
zero-offset capability to accommodate both negative and positive voltage
outputs.
4.6.2.2 Acceptance Testing
Physical inspection of the relatively fragile pyranometers or net
radiometers immediately after delivery of the instrument is important. One
must be sure that the calibration data have been received and that these data
correspond to the serial number of the instrument. Storage of this information
-------
Section No. 4.6.0
Revision No. 0
Date: 17 Sep 89
Page: 5 of 8
will prove helpful when the time comes to have the calibration of the
instrument checked, or to replace the sensor or readout device. Few
organizations are equipped or staffed to bench-test a radiometer to verify
calibration, but a quick determination can be made indoors as to whether the
sensor and recorder or integrator system is operating by exposing the sensor to
the light of a tungsten lamp. It may be necessary to place the instrument
fairly close to the lamp. Covering the sensor for several hours will ensure
that the system is not "dark counting."
4.6.2.3 Calibration
The user of a pyranometer or net radiometer is normally not equipped to
calibrate the sensor. The best the user can do is to perform field calibration
checks on two cloudless sky days. These checks involve a side-by-side
comparison of the sensor to a sensor of similar design, the calibration of
which can be traced to a transfer standard. Since 1975 all measurements have
been made in accordance with the Absplute Radiation Scale or equivalently the
World Radiometric Reference established at the International Pyrheliometric
Comparison IV at Davos, Switzerland (NCAR, 1984, pp. 4-103). If a side-by-side
comparison is not possible, the device should be returned to the manufacturer
or to a laboratory with the facilities to check the calibration. The frequency
of making comparative readings or having factory calibrations will depend on
environmental conditions. Any indication of discoloration or peeling of a
blackened surface or of scratches on the hemispheres of a pyranometer warrants
recalibration and/or service.
Net radiometers are more delicate and require more frequent attention
than pyranometers. Pyranometers of high quality in a clean atmosphere may
require recalibration annually; net radiometers should be recalibrated at least
yearly. Calibrating the recorder or integrator is an easy task. The standard
method involves the use of a precision potentiometer to impress known voltages
into the circuit. The linearity of the readout instrument may be checked by
introducing a series of voltages covering the full scale, checking first
up-scale and then down-scale. Adjustments should be made as necessary. In the
absence of a precision potentiometer, it may be possible to introduce a
calibrated millivolt source that covers one or two points. Integrators can be
checked the same way, except that the input value must also be timed.
4.6.3 OPERATIONS
4.6.3.1 Installation
The site selected for an upward-looking pyranometer should be free from
any obstruction above the plane of the sensor and should be readily accessible
for cleaning and maintenance. It should be located so that shadows will not be
cast on the device, and away from light-colored walls or other objects likely
to reflect sunlight. A flat roof is usually a good choice; but if such a site
is not possible, a rigid stand with a horizontal surface some distance from
buildings or other obstructions should be used. A site survey of the angular
elevation above the plane of the radiometer surface should be made through 360
degrees (The Eppley Laboratory, Inc.).
The same procedures and precautions should be followed for net
radiometers that are both upward- and downward-looking. However, the
-------
Section No. 4.6.0
Revision No. 0
Date: 17 Sep 89
Page: 6 of 8
instrument must be supported on an arm extending from a vertical support about
1 m above the ground. Except for net radiometers with heavy-duty domes, which
are installed with a desiccant tube in series with the sensor chamber, most
other hemispherical net radiometers require the positive pressure of a
gas—usually nitrogen—to both maintain the shape of the polyethylene domes and
purge the area surrounding the thermopile. In on increasingly popular design,
there is a requirement for internal purging with nitrogen and external
ventilation with compressed dry air through holes on the frame. The compressed
air supply minimizes fogging and condensation.
Precautions must be taken
to avoid subjecting radiometers
to mechanical shock during
installation. They should be
installed securely and leveled
using the circular spirit level
attached to the instrument. Net
radiometers are difficult to
mount and to maintain free of
vibration. Pyranometers of the
Moll-Gorzynski design, used
extensively by Atmospheric
Environmental Sciences (AES) of
Canada, are oriented so that the
emerging leads face north (Figure
4.6.4). This minimizes solar
heat on the electrical
connections of an instrument that
is not temperature compensated,
The thermopiles of these
instruments should be oriented so
that the long side of the
« 10 in
Figure 4.6.4 A Moll-Gorczynskia
Solarimeter (U.S. Army, 1975)
thermopile points east and west (Latimer, 1972). The cable used to connect the
pyranometer to the readout device, recorder, or integrator should be between
16 and 20 gauge and made of shielded, waterproofed 2-conductor copper wire.
The sensor, shield, and readout device should be connected to a common ground.
Potentiometric millivolt recorders are to be used with most high-impedance,
low-signal radiometers. Cable lengths of 300 m or more are practical.
Galvanometric recorders can be used with silicon cell radiometers. Soldered,
copper-to-copper junctions between instrument connectors and/or cables are
essential. Pyranographs or actinographs should be installed on a level surface
immune to shadows. These instruments should be placed in such a way that the
sensitive bimetallic strips lie within 2 degrees of true east and west with the
glass inspection window facing north (in the northern hemisphere).
4.6.3.2 Field Operation of a Solar Radiation System
As part of the quality assurance program, a field calibration check
should be performed at least once every 6 months according to the procedures
outlined in Section 4.6.2.3. Solar radiation instruments require almost daily
attention. The data should be inspected for a reasonable diurnal pattern and
the absence of dark counting. Where strip chart or digital printers are used,
daily time checks are desirable. Data retrieval will depend upon program
-------
Section No. 4.6.0
Revision No. 0
Date: 17 Sep 89
Page: 7 of 8
objectives; but even for climatological programs, data should be collected
monthly. All operational activities during a site visit should be logged.
4.6.3.3 Preventive Maintenance
All types of radiometers require frequent cleaning to remove any
material deposited on the surface that will intercept the radiation. In most
cases, this is a daily operation. The outer hemisphere should be wiped clean
and dry with a lint-free soft cloth, using alcohol.. Any scratching of the
surface will alter the transmission properties of the glass, so cleaning must
be done with care. If frozen snow, glazed ice, hoarfrost or rime ice is
present, an attempt should be made to remove the deposit carefully with warmed
cloths.
Should the internal surface of a pyranometer's outer hemisphere become
coated with moisture, it can be cleaned by carefully removing the outer
hemisphere on a dry day and allowing the air to evaporate the moisture, then
checking the dessicant. If removal of a hemisphere exposes the thermopile
element, extreme care should be taken because it is fragile and easily damaged.
About once each month, the desiccant installed in most pyranometers should be
inspected. Whenever the silica gel drying agent is pink or white instead of
blue, it should be replaced or rejuvenated by drying it out on a pan in 135°C
oven. The level should be checked after each servicing of the radiometer, or
at least monthly. Significant errors can result from misalignment.
Net radiometers require more frequent maintenance attention than
pyranometers. It is necessary to replace the polyethylene domes as often as
twice a year or more before the domes become discolored, distorted, or cracked.
More frequent replacement is necessary in polluted environments due to
accelerated degradation of plastic hemispheres when exposed to pollutants. A
daily maintenance schedule is essential to check on the proper flow of gas in
instruments that are inflated and purged with nitrogen. All PM activities
should be recorded in a log.
4.6.4 PERFORMANCE AUDIT METHODS
A performance audit on a solar radiation system is only practical with a
CTS. The CTS must have the spectral response and exposure equivalent to the
instrument being audited. One diurnal cycle will establish an estimate of
accuracy sufficient for most air quality monitoring applications. The method
of reporting the data from the monitoring instrument (daily integrated value,
hourly integrated value, average intensity per hour, etc.) must be used in
reducing the data from the CTS to provide a meaningful comparison. An audit
frequency of at least six months is recommended.
4.6.5 REFERENCES
ASTM, 1984: Calibration of secondary reference pyrheliometers and
pyrheliometers for field use, E816. American Society for Testing and
Materials, Philadelphia, PA.
Carter, E. A. et al., 1977: Catalog of solar radiation measuring equipment.
ERDA/ORO/5361-1, U.S. Energy and Development Administration.
-------
Section No. 4.6.0
Revision No. 0
Date: 17 Sep 89
Page: 8 of 8
Gates, D. M., 1962: Energy Exchange in the Biosphere. Harper and Row.
Latimer, J. R., 1972: Radiation measurement. Technical Manual Series No. 2,
International Field Year for the Great Lakes, Canadian National Commission
for the Hydrological Decade.
Monteith, J. L., 1972: Survey of instruments for micrometeorology.
International Biological Programs Handbook No. 22. Blackwell Scientific
Publications, Osney Mead, Oxford, England.
NCAR, 1984: Instructor's Handbook on Meteorological Instrumentation,
F. V. Brock, Editor. NCAR Technical Note, NCAR/TN-237+1A.
U.S.Army, 1975: Part 2, natural environmental factors. Engineering Design
Handbook, Environmental Series. Department of the Army, Material Command.
WMO, 1971: Guide to meteorological instrument and observing practices. World
Meteorological Organization No. 8TP3, 4th edition, Geneva, Switzerland.
-------
Section No. 4.7.0
Revision No. 0
Date: 17 Sep 89
Page 1 of 2
QUALITY ASSURANCE FOR ATMOSPHERIC PRESSURE MEASUREMENTS
4.7.0 INTRODUCTION
Surface atmospheric pressure is not generally a required measurement to
make for an air pollution meteorology application. A pressure value may be
required for the calibration or interpreta- tion. Section 4.5.0 lists some
formulas for converting wet- and dry- bulb temperatures to dew point
temperature or relative humidity where a pressure value is required. In many
of these applications, a standard atmosphere pressure for the station elevation
will be good enough. For greater accuracy without measurement, the current
altimeter setting from a nearby airport will provide an adjustment of the
standard atmosphere to present conditions. If measurement is desired, the
following may be helpful.
4.7.1 TYPES OF INSTRUMENTS
The two most common barometers are the aneroid barometer and the mercurial
barometer. These must be read to get a measurement. Most electronic systems
which include pressure as a variable use a sensor which has an aneroid pressure
sensor. The motion of the sensor as a result of pressure changes may be
detected by any number of methods. The latest and most accurate is a capacitor
type.
The Fortin mercurial barometer is used by the National Weather Service as
the official station pressure instrument. Portable precision aneroid
barometers are used to make pressure measurements available at different work
stations. A standard on the measurement of pressure (ASTM, 1977) provides
methods for calibration and height corrections.
4.7.2 SPECIFICATIONS
Meteorologists are familiar with the units of pressure called millibars
(mb). When SI units were adopted internationally, the Pascal (Pa) was chosen
as the pressure unit. The hPa (hecto Pascal) is the common expression of
pressure in the SI units because it is equivalent to millibars. One standard
atmosphere at standard gravity is:
1013.25 mb or hPa
29.9213 in. Hg at 273.15 K
14.6959 lbf/in2
Any practical application will be well served by a pressure measurement
accuracy of about 10 hPa (= \% or 100 m in elevation). The best accuracy one
can expect to achieve in a monitoring application is about 0.5 hPa.
4.7.3 OPERATIONS
If maximum accuracy is the goal, care must be given to the exposure
of the pressure sensor. The sensor is sensitive to both the atmospheric
pressure (weight of the air above the station) and wind pressure. Errors from
wind may be at most about ± 3 hPa under ordinary conditions.
-------
Section No. 4.7.0
Revision No. 0
Date: 17 Sep 89
Page 2 of 2
4.7.4 PERFORMANCE AUDIT METHODS
The audit instrument can be as simple as an aneroid barometer (altimeter)
which has been compared to a calibrated barometer. Figure 4.7.1 shows a pocket
altimeter which will provide all the resolution and accuracy required by
normai applications. The accuracy for this small instrument, when properly
used, is 0.2 % or about 2 hPa.
Figure 4.7.1 Engineer's Altimeter
(VEATHERtronics)
4.7.5 REFERENCES
ASTM, 1977: Standard Methods for Measuring Surface Atmospheric Pressure,
D3631-84, American Society for Testing and Materials, Philadelphia, PA.
NCAR, 1984: Instructor's Handbook on Meteorological Instrumentation, F. V.
Brock, Editor. NCAR Technical Note, NCAR/TN-237+1A.
WMO, 1971: Guide to meteorological instrument and observing practices. World
Meteorological Organization No. 8TP3, 4th edition, Geneva, Switzerland.
-------
-------
------- |