United States     Office of Research and  EPA/600/R-94/038d
      Environmental Protection  Development     March 1995
      Agency       Washington DC 20460
svEPA Quality Assurance
      Handbook for
      Air Pollution
      Measurement
      Systems

      Volume IV: Meteorological
      Measurements
      (As Revised March, 1995)

-------

-------
                                      EPA/600/R-94/038d
     QUALITY ASSURANCE HANDBOOK
                       FOR
AIR POLLUTION MEASUREMENT SYSTEMS

    Volume IV: METEOROLOGICAL MEASUREMENTS
                   as revised March 1995     ;
          U. S. ENVIRONMENTAL PROTECTION AGENCY
               Office of Research and Development
        Atmospheric Research and Exposure Assessment Laboratory
            Research Triangle Park, North Carolina 27711
                                      Printed on Recycled Paper

-------

-------
                               ACKNOWLEDGEMENTS

       This document represents the second revision of the Quality Assurance Handbook on Air
 Pollution Measurement Systems, Volume IV:  Meteorological Measurements.  This document is
 essentially the same as its predecessor (EPA-600/4-90-003) but with a new section added on the
 quality assurance of ground-based remote profilers which include Doppler sodars, Doppler radars,
 and radio acoustic sounding systems (RASS).  In addition, an appendix has also been added on
 meteorological monitoring guidance for the Photochemical Assessment Monitoring Station (P AMS)
 network.  This edition was critically reviewed by Desmond T. Bailey, Gennaro H.  Crescenti, Peter
 L. Finkelstein, and John E. Gaynor. Special thanks to Robert A. Baxter, Jean-Michel Page, John S.
 Irwin, Louis M. Militana, Charles E. Riese, and Kenneth H. Underwood for their comments and
 suggestions.                                                 ,

       The first revision of the this handbook was prepared by Thomas J. Lockhart.  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 minor modifications'
 The contributions by Peter L. Finkelstein, Daniel A. Mazzarella, Thomas I Lockhart, William J. King,
 arid Joseph H. White are gratefully acknowledged. This edition was produced under contract 68-02-
 4553, Jack A. Bowen, Project Officer.  This edition was critically reviewed by C. Bruce Baker, Jack
 A. Bowen, Gerald C. Gill, Harold L. Crutcher, and Alvin L. Morris.
                                    DISCLAIMER

       Mention of trade names or commercial products does not constitute U. S. EPA endorsement
or recommendation for use.
                                          i-2

-------

-------
                                TABLE OF CONTENTS
Section                                                                          Pages

4.ii    Acknowledgements                                                         1
4.ii    Disclaimer                                             ;                    i
4.iii    Table of Contents                                      ;                    2

4.0    General Program Requirements for Meteorological Measurements               25
       4.0.0   Outline, Purpose and Overview                                       5
       4.0.1   Glossary                                                            2
       4.0.2   State of the Art                                                      6
       4.0.3   Data Requirements                              ;                    3
       4.0.4   Measurement Requirements                                          7
       4.0.5   References                                                          2
                                                             i
4.1    General Aspects of Quality Assurance for Meteorological Measurements         33
       4.1.0   Outline and Summary                                                3
       4.1.1   Planning for a Quality Assurance Program                              3
       4.1.2   Organization of Authority and Responsibility                           1
       4.1.3   Quality Control Guidelines                       '                    2
       4.1.4   Traceability Protocol                                                4
       4.1.5   Estimating Precision and Accuracy                                    6
       4.1.6   System Audits                                                       2
       4.1.7   Performance Audits                                                  1
       4.1.8   Data Validation Protocol                                             7
       4.1.9   QA Reports and Corrective Action                                    2
       4.1.10  References                                                          2

4.2    QA for Wind Speed, Wind Direction and Turbulence   •    j'                    83
       4.2.0   Outline and Summary                                                4
       4.2.1   Types of Instruments                                                4
       4.2.2   Specifications                                                       32
       4.2.3   Acceptance Testing                                                  5
       4.2.4   Installation                                                          8
       4.2.5   Calibration                                      !                    4
       4.2.6   Operations, Maintenance and QC                                     2
       4.2.7   Performance Audit Methods                                          20
       4.2.8   Estimating Accuracy and Precision                                    1
       4.2.9   References                                      i                    3

-------
4.3    QA for Temperature and Temperature Gradients (A T)
       4.3.0   Outline and Summary
       4.3.1   Types of Instruments
       4.3.2   Specifications
       4.3.3   Acceptance Testing
       4.3.4   Installation
       4.3.5   Calibration                   •    '        '
       4.3.6   Operations, Maintenance and QC
       4.3.7   Performance Audit Methods
       4.3.8   Estimating Accuracy and Precision
       4.3.9   References

4.4    QA for Precipitation Measurements

4.5    QA for Relative Humidity or Dew Point Temperature

4.6    QA for Solar Radiation Measurements

4.7    Q A for Atmospheric Pressure

4.8    QA for Remote Sensors

4.A   Appendix A - PAMS Meteorological Monitoring Guidance
36
3
4
4
2
1
9
3
8
1
1

7

10

8

2

29

20

-------
 Section
                                   Section 4.0
                        GENERAL PROGRAM REQUIREMENTS FOR
                           METEOROLOGICAL  MEASUREMENTS  ;
                                    OUTLINE            !
4.0.0  OUTLINE,  PURPOSE AND OVERVIEW
       OF THE QUALITY ASSURANCE HANDBOOK

4.0.1  GLOSSARY

4.0.2  STATE OF  THE ART

       4.0.2.1   AUDITOR SURVEY
       4.0.2.2   INTERVIEW SUMMARY

4.0.3  DATA REQUIREMENTS

       4.0.3.1  REGULATORY PROGRAMS
       4.0.3.2  RESEARCH PROGRAMS
       4.0.3.3  CONTINGENCY PROGRAMS

4.0.4  MEASUREMENT REQUIREMENTS

;       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
                                                              Section No.  4.0.0
                                                              Revision No.     0
                                                              Date:    17 Sep 89
                                                              Page:     1  of  5
Pages Rev.  Date


5     0     9/89

2     0     9/89

6     0     9/89
            9/89
            9/89
                                                                         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
                                   .    I
    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
cowinicate without aabiguity.  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
•easurement  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.
Exanples 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  CQC) 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  /
     /ConfidenceX
     I   To The   )
     \Management7
                        QUALITY MANAGEMENT ASPECTS'
                                   QUALITY
                                   CONTROL
                                   ASPECTS-
                      - INTERNAL
                        QUALITY
                      ASSURANCE
                       ASPECTS
                                                           \
                                                             (Policy)
   \
 / Operational \
 I  Techniques  I
 \And Activities/

   (When Required
    By A Contract)
 EXTERNAL QUALITY
ASSURANCE ASPECTS
E7"
                                                           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 confusioi
 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.            '                         .
  •
  1   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. i  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 systen.  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.
iapose  a known condition such as the rate of rotation for an anemometer, and
aeasure the  system response.  This  type of  audit assumes  that  the
aanufacturer'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 control1
 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
 ccaposed 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
 tbs 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.    o
                                                           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.  CCUrate>   Such

    The letter described situations where a single company can be structured to
provide^both calibration and auditing services, but caStioned that  the
        Jro^ £ t^auditor requires a management structure insulating  the
        f ??     budgetary concerns of the operating organization.  Responses
        J!K  IK ^^ fSW ^ ln n° CaSe WSS the distinction challenged     All
acwled^  ,h6 H°n^Pt.0f indePendence for QA audits.   Regulators  should
acknowledge  the distinction and require true  independence.

    Situations arise where the fundamental  principle of independence between




                             ^ ssss             risgsr •
instruments  for Agency A might  be asked  to audit the  Agency B  instruments  in

-------

-------
                                                              Section No. 4.0.1
                                                              Revision No.    0
                                                              Date:   1^ 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 valu«, 100(X-T)/T  or
 sometimes expressed as a ratio, X/T.

 CALIBRATION - is a MEASURE of conformance to or discrepancy from a
 Z"5S£Sii!E °J E* °f Titer^a f°r an instrument °r 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 rati° is calc«lated from the overshoot ratio
DELAY DISTANCE  (D) - The  distance  the  air  flows past  a wind vane during the
ti*e it takes the vane  to return to  50 percent of  the initial  displacement.  [2]
      of,                  deSigned  to Provide  the Purchaser
     confidence in the quality of what  is being purchased,
                 out.
                          ^    y
                                                      to
                                           °peratin« ande management  policy
INVERSION (+AT) - is the . inverted lapse rate or an increase of air
temperature with height.  There is no general limit for inversion strength.
           th                    decrease °C air temperature with height
        by the auto convection rate of  3.4 C/100 m
OVERSHOOT (fl) - The ratio of the amplitude of two successive deflections of a
off sedition :?1UarfteS ^^ ^ ^Uilibrium Position after relea^f rom an
oiiset position of ten degrees, as expressed by the equation
                          e
                      £2=
                           (n+l)
                          0
where

respectively.
        and  9(n+i)  are  the amplitudes of the ri and n-fl  deflections,

-------
                                                             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, Xr

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,  til

QUALITY CONTROL - The operational techniques and  activities that  are used  to
fulfill requirements for quality, [il

QUALITY MANAGEMENT - That  aspect of  the  overall management  function  that
determines and  implements  the quality policy.  CiJ

QUALITY POLICY  - The overall quality intentions and direction of an
organization as regards quality, as  formally expressed by top management,  til

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 (SQ, 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°.  [a]
               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]

 til  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.

 121  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.

 [31 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

       r   rSP°nSJbility °f an identifiable part of an organization   Tne
 practice of elevating quality to a management staff level is relatively new
 The value of quality control and the umbrella management structure of quaUty
 assurance became clear when products, purchased against a specification  were
 rejected by the purchaser.  When the cost of rewo?k or scrap absorbs the
 profit  an alternative will be found.  The alternative is to do it right the
     ™   ^ Profession grew during World War II and thereafter as the U.S.
     TST  ??Came a 2J?°lf iC3nt purchaser usi"g comprehensivespecifications
      the well known Mil-Specs.  In the '60sand '70s the practice of planned
 oosolescence and using the customer as the final inspector set up our
 s^arri!:3  £    1Ure ^^ f°reign comPetition with higher quality
 standards   The successful foreign producers,  using the quality principles
 developed in the United States, caused a resurgencJ of quality awareness
 ,fa                    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

 JSSEn'S the  Cl6a? e^ ACt*   In 19?6 a Quality Assurance  H^dbooi for A£
 In ^77  v  TaSUr?T6n! ?ystems:  Volume I-  Principles was published (EPA,  1976)
 It^ll'    *™e  "i A-bl"nt Air SPecific Methods  (EPA,  lS77a)  and Volume III
 Stationary  Source  Specific Methods (EPA,  1977b) were  published   This program
 adoressed the Criteria Pollutants which were covered  by federal  law   P   *
 ^7^° 1°glcal raeasurements were recognized  as supportive  to  the Criteria
 Pollutant measurement  program but they were  secondary.
 B--h r,    ulr AGt WaS amended> Permission for growth of source
 s-ength (and thereby growth of  industry) was granted as a con- sequence of
 dizfusion model predictions based on input meteorological data   Now  tL law
   °eZn1edhf ^^T f°r Valid and ^P-sent- ftive mit^orologlca? data
      vn?n   T?r M ftructure to Provide documented assurance of validity.  In
         "      Mete°rolo«ical Measurements (EPA, 1983) was added to the
           miy
                      ^ work Provided by the private sector was geared to air
   t  a varir   i S" meteorolo«y was ^ded to the technical requirements
Ja&tl If    J    TiU i°nS WSre applied by a varietv of individuals with a
variety of technical backgrounds.  The original Volume IV.  was like a
background guidebook for taking meteorological measurements and general
SS   Th?J r^  ?W QA,^d QC mlght be applied to the requirement for valid
                     -
-   and
                                              rot          an

-------
                                                             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
neteorological 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      NUMBER

Monrovia, CA         4
Atlanta, GA          2
Reno, NV             2
Thousand Oaks, CA    1
Fort Collins, CO     1
Concord, MA          4
E. Syracuse, NY      1
Fox  Island, WA       1
Chapel Hill, NC
RTP, NC              1
West Chester, PA     1
Denver,  CO           1
Aurora,  CO           1
Muscle Shoals, AL    1
E. Hartford, CT
          * 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
                                               i    Date:   17 Sep 89
                                                   Page:    3  of  6
           Survey o-f  Meteorological  Measurement QC/QA People
                   tplease print or type your  responses)
NAME	.	,	COMPANY	
ADDRESS	    ADDRESS	    	
PHONE   <     )                             <      )
I_5? *„•—meteorologist,    chemist,  	environmental  scientist,   QA/QC
P 2i*Sfi?Ka1' —instrument technician,    electronics tech.,  ' engineer.
T-?*!'*?6*1'* —•odeler. 	manager,  __da€I analyst,  	field hand.    This informat-ion'wni be summSrized
mtteLS2 use °i,names_P!:,eo«nPan>e* or  agencies so please be candid.
         -ii—— -.-••— — —-— — •••«»•"==» *** v.ut«paiiaB9 wr agtCTic^ies so pAease oe canaio>
        r "Onsioer an audit or challenge as a QA observation of  the response
       n»«T»2TT™!mrumen£•to a *nown, input and consider a calibration as an
       OPERATIONAL testing and adjustment, as necessary, of an instrument.!
       1. If one meteorological audit is defined as a challenge  to one
          variable or one variable of a system, about how many audits have you
          performed in 1980-1984 	_, 1985	, 1986 .	, 1987	?
       2. Did you34."sually JZLsometimes ^.never us» a written procedure?
       3. If a calibration is defined as the testing and adjustment of one
          variable or one variable in a system, how many calibrations have you
  :        performed in 198O-1984 	_, 1985	, 198<»     .. 1937  	?
       4. Did you25."sually ££.sometimes _/_never us» the manufacturer's
          calibration procedure?
       5. When you perform an audit, do you require the operator to remove the
          sensors from their mounted position?  "T^Yes £0Ha  /.^.Sometimes
       6. Do you require the operator to re-connect the sensor to  the system
          when it is presented for audit? 2£LYes  {«^»No
       7.  If 6.  is yes.  is the re-connection made with 32the operational
          cable,  _Q-a substitute cable or ^either?
       8.  Do you 2#usually J_/_sometimes J£[never measure the starting torque
          of each anemometer bearing assembly and transducer?
       9.  Do you {^.usually XQsometimes j^never measure the starting torque
          of each wind vane bearing assembly and transducer?
       10. Do you 4^-usuallY /^.sometimes^l-Jnever us«« the collocated transfer
  :        standard method  for auditing a wind instrument?
       11. Do you3JL«sually /2someti0MBS ^.never find the audited  instrument
       ,. meets the required specification?
       12. Do you challenge anemometers with known rates of rotation? 3l5l'es
         /£_No - If yes,  how many speeds	?  	Synchronous or 	measured?
       13. Do you challenge direction vanes with a dividing wheel? ^^?Yes
         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? ^^Yes  _£_No
Figore 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 #1, 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
aanagers 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
Meteorologist
Engineer
QA/QC
Instrument Tech
Manager
Environ. Sci.
Number
11
8
7
5
4
4
Electronics Tech. 4
Chemist
Data Analyst
No indication
Total
Average number
Percent change
3
1
2
49
per year

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 physically inspect the
 sensor or do so by performing the operator's function of climbing the tower and
 reaoving the sensor.   Volume IV. should reduce that percentage to zero.
 Question 6 suggests that most auditors (67%) 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
 aeasurement, 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
 r/^.,   Question 10 shows that only 19% rely on the collocated transfer standard
 ICI5J  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
 no,t 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 Sumnary
   :
        After the survey results were  in,  a  series  of visits  was  planned  to  talk
 to :?rivate  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 arid 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
 namling 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
auait 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
HNSR) in Concord, Massachusetts, perhaps the largest of the five organizations
in terras 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
aethod requires a spare  set of  sensors be available for each client.  The
aethods 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
nethod.

       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
    with most of these organizations.
       This  series  of  interviews  provided valuable insights  and confirmations
 Ebout  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
    ( ^ estimation of  uncertainty before the numbers can be dignified by the
    .e  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                         j
            4^0.3.1.1   PSD
                                                         •I
 fh. nn <=*<•   ?!  ^gulatory program used in this document,  and to some extent  in
 the On-Site Meteorological Program Guidance for Regulatory Modeling
Applications (EPA,  1987bK  is  the Amhieni-  Mr.ntt»r.4n~ <-s,»-i_is	^-_ T,
                    1?87b)> is the Ambient Monitoring Guidelines for Prevention
                Deterioration (PSD) (EPA, 1987a).   This 1. the most explicit
 guideline  and one requiring a quality of instrument performance available only
 0 '  i-*  ^ . *?°St, air Duality dispersion models described in the Guideline on Air
 Quality Models (EPA,  1986)  used for PSD applications are Gaussian models
 ^K^f i 1I1PUt data Whi°h 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:                         i
  .                                                        i
 **'       ^ ^D  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.

 an™  *K    2)  If1?*1 direction - representing  the average wind  direction at  10 m
 calculatedgbvT (fd additional heiShts  *>r  elevated sources)  during each  hour
 calculated by carefully averaging samples of wind direction or  by  calculating
 A" resultant   T to^dir?ction usin* ™"y as  the wind speed for each sample8
 samoTei wM  H       di^ef1On  does not  represent the distribution of direction
 samples which occurred during the hour.
  I                                                        !

m h -  K;   u 3) TemPerature  -  representing the air temperature at the  standard 2
m height  above ground (and  additional heights for elevated sources).   aMa™.2

-------
                                                             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 (sigraa 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
aeasurements 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
 saall 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
              rograms "ay be exploratory to see how on-slte data compare with
a™       r ^^ Sitf (airp°rts or state or local agency statLns)
Such questions of representativeness cannot be convincingly answered if the
0n        ta d°es not come from suifcably sensitive instruments, properly
               maintained and subject to QA/QC effort designed to document data
       It is possible to select, install, operate and document on-site
measurement systems to meet PSD requirements.  Public data from airports may
£2 5i£°" V?         ^ ^^ f°r thrSe reaSOnS*  ^PresentativeneS deals
wjLth different meso-scale structures in the surface layer flow driven, in part
not^enrSt S^°ptic/low'  Jt is c°™°n to find that airport meas^emenK do
not represent other sites just a few miles away because the flow is different
Itis 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-sUe data
for the sa*e hour includes samples from the entire hour.   Finally, -the airport
instruments are selected to serve aviation where low wind speeds are of no
inportance.   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                       I
                                                                •
       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.O.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
oeasurement 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,
nodifications 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.O.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
 oifaer 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
 (WD, 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
 acd, 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
 (WJO, 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) tiaes the height of that obstruction.  An obstruction may  be
man-made (such as a building) or natural (such as a tree) (Figure 4.0.3.1).
                                JOT
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
usisaily 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
4.3.4.1)  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 Slope
from tower (between)
(m) (%)
0- 15
15- 30
30-100
100-300

Source: TVA,
±
±
±
±

2
3
7
11

Max. obstruction, or
vegitation height
(m)
0.3
0.5-1.0
3.0
10 x ht.

1977

(most veg.<0.3)

must be less
than distance
to obstruction
•                     Towers should be of the open grid type of construction,
suca 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-Ghristensen, 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
;t:wer or on booms projecting horizontally out from the tower.  If a boom is
iusad, it should support the sensor at a distance equal to twice  the maximum
:di=aeter 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
wisa. 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
axsy 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
secsors is at a point with a minimum number of diagonal cross members, and away
frcm major horizontal cross members.  Even with these precautions, data
ob^ained 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. 4.0.5
                                                             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-6OO/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-45O/2-78-O27R  (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.
                                                       i
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-6OO/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
                                                     I
4.1.0  OUTLINE AND SUMMARY                                   3     0  9/89

4.1.1  PLANNING FOR A QUALITY ASSURANCE PROGRAM              3     0  9/89
                                                     i
       4.1.1.1  PROJECT DESCRIPTION                  '
       4.1.1.2  PROJECT ORGANIZATION
       4.1.1.3  QA OBJECTIVE                         i
       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               i
       4.1.1.9  QA PROCEDURES                        i
       4.1.1.10 CORRECTIVE ACTION AND REPORTS

4.1.2  ORGANIZATION OF AUTHORITY AND RESPONSIBILITY  j        1     0  9/89

4.1.3  QALITY CONTROL GUIDELINES                     i        2     0  9/89

4.1.4  TRACEABILITY PROTOCOL                         j        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                        i
       4.1.4.8  ATMOSPHERIC PRESSURE                 j

4.1.5  ESTIMATING PRECISION AND ACCURACY                     6     0  9/89

       4.1.5.1  DEFINITIONS    ' .                     j
       4.1.5.2  COLLOCATED TRANSFER STANDARDS         i
       4.1.5.3  OTHER CONSIDERATIONS     .            |
                                                     I
4.1.6  SYSTEM AUDITS                                 !        2     0  9/89
                                                     j
4.L.7  PERFORMANCE AUDITS                                    1      o  9/89
                                                     i
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

       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
9/89
9/89

-------
          .                                     .  "            Section No.  4.l.O
                                                              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 facade 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 uses 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
sane.
       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 facade 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 seeded data.
                     1

       4.1.1.3 QA Objective                              j

  ;     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
doc'jnenting and defending claims of data validity.

       4.1.1.4 Calibration Method and Frequency          j

       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 pf 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. 4.2.1
                                                             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
aethodology 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.
                                  ^                     I
       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 measureinont 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 aind 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), Feigeiibaum (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
co»iittees 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 (ANS5I/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
tjie 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
comunicate in writing with the operators to report questionable data and
rleceive 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 staJces 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.
Ibis 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.    o
                                                            Date:   17 Sep 89
                                                            Page:    1  of  4
 4.1.4  TRACEABILITY PROTOCOL
                       practice in QA/QC to use a hierarchy of standards
        i   ,                l discuss how this hierarchy relates to
 meteorological measurements.

     Most meteorological measurements contain some 'sensing element which
             ^ °f ^'* ** ^ USUal transducer outputs of voltage,
                            rmS °f accur^> the r^pcam. of the sensfng


                    r— ^^^
 Protocols for  "traceability to NBS"  for voltage will be ; discussed first.
       4.1.4.1  Vpltage

 NBS forbraon        1f Sfmaintain transfer standards which are sent to
 woa zor caiioration.  These in turn are used to calibrate  the lab's
 sources which in turn are used to calibrate a subject voUmJter   Tnis
                                     s            ua


DVOM S i?» il .     y  ^6 manufacturer w"h a transfer standard,  and another


SrtSlS   burjrbecof6 Ph°C?Sf yleldS  a  relaUVe a"Uira^ if -iSer DvSn is
DV£MS             becomes absolute as comparisons are made with certified

                                                                        f
              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
•K0.00004 •«• 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..9S 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 0.10O ±  (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  l'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 -
1.000
1.000
1.000
Met.
0.0
360
-5.0
Units
- 50.0
- 360
- 15.0
Accuracy (0.1% FS)
Volts
m/s
deg.
C
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."

-------
                                                          I   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
nanometer.  Anyone can build a wind tunnel and measure itsi 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).                                            j

      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  resis'tance 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
nethod and test facilities are not usually certified by NEtS and so the
calibration may not deserve the inferred NBS .authority.

      Differential temperature is nothing more than two or more temperature
neasurements 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 orgamizations 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 tempetrature 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, j

                    G = D cos 9 + 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
aeasurements 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
sonetimes 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

                    n            •        .                             ,

                      :X. - T )  = X - T                      (1)
    where

      E  is the average error  (accuracy)

      Xj^ 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, 1,2,3...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."1  This is more difficult to fit to meteorological
neasurements 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:    4  of  6
                                 n
                                                             .  (7)
                                1=1
             where

               Xai is  the  ith measurement of the subject output
               X. , is  the  ith simultaneous sample from the CTS


The systematic difference, d, is calculated from (4) substituting
X , and X.for Xi and  T,,  respectively.  The estimated standard deviation of
the difference,  s, is  calculated from  (8).
            - d
                                                               (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 aeasurements must be greater than four times the response time of the
instruments to assure sample independence.
                     n
                                           (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
ti»e 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
           3 x 5}
- 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
                                                         i     Date:    17 Sep 89
                                                         i     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 perioii is 24 hours from the
standpoint of  dynamic range variation.                   \

  \     Assume  the CIS will provide the true value in a test in the free
atmosphere within the limits of the calibration of the CIS 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:                                      I
                                                         I
                 Wind speed	•	s at 0.2 m/s

                 Wind direction	•	s a 2 deg.
                                                         i '
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.    T

       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  C1,000  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
11,1 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 have 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
aeasurement.  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
aeasurement data.

       Random errors identified for each component of. a system can be combined
to estimate the total system error by the RSS (root sum 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, validation, 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.               j
                                                       I
    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 concetpts 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 z\. 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 Nor 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.  1 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.       i

     A performance audit on variables such as relative humidity (or dew point
te«perature),  solar radiation and atmospheric pressure vfill usually include a
collocated transfer standard comparison.   This is may b« 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
treasuring 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 plaiusibility 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
qf 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                                 '
                                                        j
       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
 exaaples 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 taJcen and when assumptions
 bf 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.
                                                        i           •
                 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 spsice, 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) or that there is some other problem with
the system.

                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
co»«only 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
       Jind Speed:
      Wind Direction:
      Delta  Temperature:
      Stability:
      Temperature:
       Dew Point:
       Pressure:
>25 ra/s (NEC)
>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
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)
                                        1 hour or 20°F in 3 hours (NCC)

-------
                                                                Section No. 4.1.8
                                                                Revision No.     0

                                                                Date:   17 Sep 89

                                                                Page:    5  of  7
  assuring that  th
                                                      ***
 must allow a far wider range of limits
                                         on
                                                   «           c°ncerns, the NCC
 hour or  three hours versus continuous
                                                                        once per
        4.1.8.3 The AREAL  System
to
                                                                       f°r
multistation network design  nor  s it

of rapid identification of field problems  with
individual data points  thus
                                                              baSiC g°al  is that
                                                     value assigned to
 randomly  selected for audit   A second   n
 well as the hour just before and Stir' t5nd^pendent f"e of these values,  as
 original  hard copy   ThS Sl^in  I      hour'.w"l b. created from the

 discrepancies no?ed.   If there are ontv^T1"^ ^ ^ master file and
 points will  be eliminated from the system   If ^T discrepancie^  these
 seems to  be  a systematic pattern of  *rr^'   It        &re several»  or there

 responsible  for^athering^nd redScin^ thl' ^l^iTl  ^r  ^  °ffiCe
 can correct  and re-enter the  data ano^correcf iL £1    noi-ified  so that  they
 are next passed through  a screeninTprSram wM.h  ? /"-^  SyStem'  ^  data
 questionable  values.  Flagged dat?UT?     I  ?u    iS desi«^ed  to  note and flag
 office for review.  Thereby will either  be a   lab°rat°ry  meteorologiCal   g
 to the project  officer if  there is a lar^I S accePted, discarded, or returned
 officer may accept, discard?  or correct Ke d^f ^h^estionable  data.  That
given in Table  4.1.8.1.  They offer I oomh?  f,         screening values are

                                      c

-------
                                                          Section No.  4.1.8
                                                          Revision No.     0
                                                          Date:    17 Sep 89
                                                          Page:     6  of  7


 anticipated  that some good data will be flagged  but that most data handling
       instrument failure problems will be caught.
("
         New Data
                     C   Start   J

Edit/Merge
Program
I

' 1


Correct Data
Acquisition/
Processing
Procedures
                     I Unvalidated
                     \    File
                                          a


Unresolved
Data

Data Corrections,
Approvals, or
Deletions.
                    Validated
                    Data File
             Schematic flow of decisions in the
              AREAL data validation scheme.
     Figure 4.1.8.1
       Data that pass the screening program will go


selection process, one day  and one hour will  be =no^en     20 will be randomly
stations in the network will  be audited.   One day  in every 20 will D         J




the use of the validators.

       The data  generated by the audit programs will first be compared with
NationafSeather 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
.ay, have to be made, depend!,* upon  the needs  and

-------

-------
                                                               Section No.  4.1.9
                                                               Revision No.     o
                                                               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
        4.1.9.2  Calibration Reports


















       4.1.9.3  Audit Reports









suitable for use in a court of law    report'  The documentation must be

-------
                                                           Section No. 4.1.9
                                                           Revision No.    0
                                                           Date:   17 Sep 89
                                                           Page:    2  of  2
     4 1.9.4  Reports to Management









                         in operations and quality assurance.
         SLe Snimu^ costs do not include either office or storage space.

                                                            $100,000
                                                              25,000
  One  tower,  installation and equipment
  Annual  maintenance cost
  Annual  cost of surveillance and quality assurance
      (including personnel and supplies,
     magnetic tape,  paper etc.)
  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
                                                              50,000
                                                              60,000
the same . "
    Of ^articular interest in this reasonably accurate estimate (ten years  ago)



       4.1.9.5  Discrepancy Reports
       Some systems report discrepancies as a section of another report and
                                                                      -
ofcontr'ol isTecessar^trkeep track of open reports and monitor
toward completion (called follow-up or needling).
                                                                  progress

-------
                                                             Section No. 4.1,10
                                                             Revision No.     0
                                                             Date:    17 Sep 89
                                                             Page:     1  of  2
  4.1.10.     REFERENCES
  ANSI£A?2c;.1987a:   Q90>  Quality Management and Quality Assurance Standards -
                wstion -d use-                        cards
 ANSI/ASQC,  1987b:   Q91,  Quality Systems - Model for Quality Assurance in
                                          '-"--  - ~   £^ -  -
 ANSI/ASQC,  1987c:  Q92, Quality Systems - Model  for Quality Assurance in
     Production and Installation. Amer. Soc.  For  Qu^V^tiSTSSJL..  WI
                   lilt                                               n   «
                and Test.  Amer. Soc.  For Quality Control, Milwaukee,  WI 53203.
                 A              fna«ement  and Quality System Elements
                 Amer. Soc. For Quality  Control, Milwaukee,  WI 53203.
                       Test Method f°r MEASURING HUMIDITY- WITH A
    PSYCHROMETER (THE MEASUREMENT OF WET- AND DRY-BULB TEMPERATuks
    c. 337-84.  Amer. Soc. for Testing and Materials, Philadelphia,  PA  19103


                                                       Co. ,
                                        r
F.ir=b.nt, J. E., J979,  (personal, communication) Nuclear Regulatory Commission.

-------
                                                             Section No.  4.1.10
                                                             Revision No.      0
                                                             Date:    17 Sep 89
                                                             Page:     2  of  2


Filimxw  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  EL   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  Techno1.,  6,  525-526.

 Lockhart, T. J. , 1989b:  Accuracy of the  collocated transfer  standard method
     for wind  instrument  auditing.  J. Atmos. Oceanic Techno1.,  6,  715-/^J.

 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.

 Revnolds  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 Techno 1'. , 4,  435-453.

-------
                                                             Section No. 4.2.0
                                                             Revision No.    0
                                                             Date:  Sep  17 89
                                                             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                                   40     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
                                                       I

       4.2.2.1  WIND SPEED
            4.2.2.1.1  Threshold                       i
                4.2.2.1.1.1  Definition
                4.2.2.1.1.2  ThreshoId 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                                     ,

i       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
       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

       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

       4.2.4.1  GENERAL CONSIDERATIONS
       4.2.4.2  WIND SPEED
       4.2.4.3  WIND DIRECTION
            4.2.4.3.1  Exposure
                       Orientation
                             True Solar Noon Method
                             Solar Azimuth Method
                                               Pages  Rev.
                                                  Date
                                                             9/89
                                                             9/89
4.2.4.3.2
    4.2.4.3.2.1
    4.2.4.3.2.2
4.2.5  CALIBRATION

       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

       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

       4.2.7.1  GENERAL CONSIDERATIONS
                                                 4
                                                  9/89
                                                             9/89
                                                20
                                                  9/89
            4.2.7
            4.2.7
            4.2.7
            4.2.7
            4.2.7.1.5
      1.1
      1.2
      1.3
      1.4
Who.
What
Where
When
How

-------
Section
       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, d>
       4.2.7.4  TURBULENCE OF 

4.2.8  ESTIMATING ACCURACY AND PRECISION

       4.2.8.1  MEASUREMENTS
       4.2.8.2  SUMMARIZED DATA

4.2.9  REFERENCES
  Section No. 4.2.0
  Revision No.    0
  Date:  Sep  17 89
  Page:    3  of  4

Pages  Rev.    Date
              9/89
                                                                         9/89

-------
                                                             Section  No.  4.2.:
                                                             Revision No.     -
                                                             Date:  Sep  17 89
                                                             Page:     4  of  4


               QUALITY ASSURANCE FOR HIND 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
sill dXcusslon of secondary effect sensors is provided but the handbook is nc.
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  Censors
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 use.
to verify  or  to  judge the work  of others who verify conformance  to

SPeCOnceathensPecifications  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 ^
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 »ethods 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 »eth°fs
 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. :
                                                              Revision No.     C
                                                              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:
       o write procurement specifications,
       o adopt and apply acceptance testing methods,
       ° site  the sensors in the representative flow of interest,
       ° perform calibration and maintenance  services,  j
       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.
  1
    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.    C
                                                             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
 the range 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.!.
                                                              Revision No.     I
                                                              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.
         L|        '      "       •                •         \

                 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-thnsshold-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% error in the X propeller because of threshold
 nonlinearity would cause an insignificant 0.014  m/s error in the wind speed aoi
 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 gjood 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 hofc wir« 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






                                                         "1
of the SODAR.

-------
                                                               Section  No.  4.2.2
                                                               Revision No.     j
                                                               Date:    17  Sep  S~
                                                               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.   Thi«=
 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 m/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
        2.0
                                 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
                                   Wind  Speed (m/s)  .

 Figure 4.2.2.1  A hypothetical  cup anemometer  threshold analysis.

-------
                                                              Section N.o.  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 Tg is torque (g cm2/sec2)                      ,
i                 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).
                                                        •I
;                 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 MRI Model 1022 (48.3 g) K= 2.7
#3 R.M. Young Model 21180 (9.7 g)
#4 MRI Model 1074 (186.8 g)
K= 4.9
K=14.8
                                                        I
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 O.O03 oz-in
(0.216 g-cm) to 0.03  oz-in (2.16 g-cm).   To convert oz-in to g-cm, multiply by


                      The method of  using  a starting torc[ue  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

-------
 propeller.  The range of the torque watch may hot  be as
                                                              Section No.  4.2.2
                                                              Revision No.    0
                                                              Date:    17 Sep 89
                                                              Page:     5   of 32

                                                        sensitive as  one would
 like.  If the model 366-3 is used on anemometer #1 in lable 4.2.2.1,  the
 threshold (0.003 oz-in or 0.216 g-crn) 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
iplane as  the shaft.   Calibration results from
 weighing  the weights and measuring the distances.
                                                     Fi;gure 4.2.2.3 Vaters
                                                        \            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 is 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-CK 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)
                                                   •& f*f    ..

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
                                                        i     Page:    7  of 32
                                                        i

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 (31 operations.
Assume the test shows  that the A(3) value is uniformly 10/4 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 All) 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 f low 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 UVW 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 HStull, 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 (Lcckhart, 1987).  A  vane
oriented propeller  will have small v-errors knd w-errors from misalignment.
These will be small because the propellers respond nearly as the cosine of the
misalignment angle, 2% for a iO degree misalignment. The u-error is too small
to measure for light  weight helicoid propellers.  These are all A(la) 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 inclUde 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 functi-on 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 HCAR 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
t Institute for Arctic and Alpine
a
b
E
(m/s) (m/rev. )
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»
than zero
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

Research, Boulder,
1.03
1.03
0.93
1.00
0.98
0.97

1.01


1.03
0.97


Colo.
* Analysis of NBS data provided by L. Petralli
tt Data from Lockhart (1977)
E is the ratio of the test result

to the mfg. '


s calibration.
Baynton lists the value of "a" for the 12 anemometers h« 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 ssome 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 MR I Model 1074

Test
No.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Test date
— — — NB
Output
Freq.
(hz)
14
28
65
88
112
240
380
500
625
755
880
1005
1255
1500
1755
•1995
2265
2530
S 	 Y
Tunnel Tunnel
Speed Speed
dph) (i/s)
0.9 0.4
1.3 0.6
2.8 1.3
3.7 1.7
4.7 2.1
9.6 4.3
15.. 1 6.8
19.9 8.9
24.9 11.1
29.8 13.3
34.7 15.5
39.6 17.7
49.5 22.1
59.3 26.5
69.6 31.1
79.6 35.6
90.2 40.3
100.7 45.0
X 	 Linear Regression 	
Output
/132 X'
(rps) (c/s)
0.11 0.28
0.21 0.53
0.49 1.19
0.67 1.59
0.85 2.02
1.82 4.29
2.88 6.77
3.79 8.90
4.73 11.12
5.72 13.43
6.67 15.64
7.61 17.86
9.51 22.30
11.36 26.64
13.30 31.17
15.11 35.42
17.16 40.21
19.17 44.91
- 11/18/75 HRI, Altadena, Calif. •

X'-Y
(*/5)
-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
Hodel 1074
r
X" :'-Y Regression Output: No. 1-1E
(s/s) (e/s) Constant 0.032035
0.25 -0.15 Std Err' of Y Est '. 0.106914
0.50 -0.08 R Squared 0.99994?
1.15 -0.10 No. of Observations IS
1.56 -0.09 Degrees of Freedos 16
1.99 -0.11
4.26 -0.03 X Coefficient(s) 2.34165s
6.75 -0.00 Std Err of Coef. 0.004174
8.88 -0.02
11.10 -i.03 lm
13.41 0.09 Regression Output: No. 1-1E
15.63 0.12 Constant C
17.85 C.15 Std Err of Y Est 0.106039
22.29 0.16 R Squared 0,99994e
26.64 0.13 No. of Observations IB
31.17 0.05 Degrees of Freedot 17
35.43 -0.15
40.23 -0.10 X Coefficient (s} 2.344297
44.93 -0.08 Std Err of Coef. 0.002764
8 132/rev.
Linear Regression of NBS Data

















5
£
i5







.5 -
0.4-
0.3 -
0.2-
0.1 -

-
-0.2-

-0.3-

-0.4-
—0.5 -


t-




/"""•^







Residual


.^-E 	
rf^T








Error Analysis






"\
\










i - +
KEY .
	 v>

+- X"












_ __ i ft. • cn
                            10         20         30
                                     Wind Speed (m/s)
Figure 4.2.2.7  Residual  Errors from HRI Model 1074
40
                                                                       50

-------
                                                             Section No. 4.2.2
                                                             Revision No.    0
                                                             Date:   17 Sep 89
                                                             Page:   II  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 csilculated.  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.
                                                        I
                     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 anemoaeter can
be calibrated to an accuracy sufficient for most applications.

                     4.2.2.1.2.3  Application of accura<:y 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
ispeeds,  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)             *(0.2 o/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 MBS,
              showing rate of  rotation vs. wind speed at 1O speeds."
By  implication, the  latter specification  refers  to accuracy type A(2),  although
the  expectation is that A(l) 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 (m/s)  = 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
	 AES 	
Test
No.

1
2
3
4
5
6
7
a
9
10
11
12
13
14
15
16
17
18
19
20
Output
Count
U)
3660
3675
10889
10911
18108
18207
25118
25433
33006
33003
40163
40139
47636
47577
55102
55105
62613
62785
10930
10931
Y
Tunnel Tunnel
Speed
(•/s)
1.14
1.14
3.27
3.27
5.42
5.42
7.59
7.61
9.82
9.83
11.96
11.95
14.13
14.14
16.35
16.30
18.53
18.56
3.29
3.28
Speed
(l/S)
1.14
1.14
3.27
3.27
5.42
5.42
7.55
7.61
9.82
9.83
11.96
11.95
14.13
14.14
16.35
16.30
18.53
18.56
3.29
3.28
X
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
	 Linear Regression 	

X'
(t/s)
1.15
1.15
3.28
3.29
5.41
5.44
7.57
7.58
9.81
9.81
11.92
11.92
14.13
14.11
16.33
16.34
18.55
18.60
3.29
3.29

X'-Y
(fi/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

X"
U/s)
1.09
1.09
3.23
3.24
5.37
5.40
7.54
7.55
9.80
9.80
11.92
11.91
14.14
14.12
16.35
16.36
18.58
18.63
3.24
3.24

X'-Y
(t/s)
-0.05
-0.05
-0.04
-0.03
-0.05
-0.02
-0.05
-0.06
-0.02
-0.03
-0.04
-0.04
0.01
-0.02
0.00
0.06
0.05
0.07
-0.05
-0.04
UJ 10 pulses oer revolution counted for 100 seconds.
fttaospheric
Environment
Service of Canada
April 12,
1983 with
I. Earle

Regression Output:
Constant
Std Err of Y Est
R Squared
No. -of Observations
Degrees of Freedoi

X Coefficient(s)
Std Err of Coef.



Regression Output:
Constant
Std Err of Y Est
" R Squared
No. of Observations
Degrees of Freedoe

X Coefficient (si
Std Err of Coef.



No
0.
0.
0.




X'
. 1-20
067574
023965
9999S4
20.00
18.00

0.295224
0



000279


X"
No. 1-20
0
0
0



0
0


.000000
.043528
.999944
20.00
19.00

.296803
.000268


Chapaan, Young 27106 6 10/rev.
  Linear Regression of AES-RR Data
               Residual Error Analysis
O.5 -
0.4-
0.3-
0.2-
0.1 -
-0.1 -
-0.2-

-0.3-
-0.4 -











I 	 1 	 I 	 1 	





V K~




I 	 1 	 r — —i 	 r- Ml




•.:.._! 	 ^ —
' —~ — .t—




i i ii
f\ 1
"•."•. i

'. ' - '

- 	 _f'


KEY
v y"
* . /\ . '
^^— X'
1 :l ,f \ '
C . 1
                                     10         . .  . 15
                              Wind Speed (m/s)

Figure 4.2.2.8  Residual Errors from Helicoid Propeller
                                                .19

-------
                                                              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 oyerestimation 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
Al lowed
(ra/'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.ll 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.    D
                                                             Date:    17 Sep 83
                                                             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 ra tower in the atmosphere can be expressed as some function of speed  [f(u)],
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 qn 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 ra 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  sizies
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 haves 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 jtimes.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 Jixis 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 propeUL&r_jnay 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.
 -60  -50  -40, -30 .  -20  -10   0   10   20   30   40   50   60

              6 ELEVATION ANCLE (DEGREES)
                                                       	(1) eo« 6 CURVE
                                                       	(2) co«3 0  CURVE

                                                       —I——— (3) STANDARD SMALL CUPS

                                                       —«-' — —— (4) SENSITIVE SMALL CUPS

                                                       	— {5} TOTAL SPEED SENSOR

                                                       • •••	(6) VELOCITYVANE (J«P)

                                                       .......	(7) VECTORVANE (pnrt
                                                         I AXIS CONSTRAINED HORIZONTAL)
Figure 4.2,2.9  Anemometer response to off-axis flow

-------
                                                             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 ra/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 '
                                                        i
       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 e^ in the definition is the equilibrium direction of the vane
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  6  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)."  oDoes 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)              s  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  109  to 5°  was to
relate the starting, threshold to accuracy.  With wind sp
-------
                                                             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/s.  One standard methodology is
defined in ASTM (1985b) and described in Finkelstein (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 measurementodevice 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
Offset Angle
5°
K
Wind Sentry (RMY 03301)
Wind Monitor (RMY 05103)
Wind Monitor AQ (RMY 05305)
Propvane (RMY 08003)
Microvane (RMY 12302)
Blvane-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.
1
10
16
15
25
14
17
19
3
2
13
19
16
of
.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
e
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
0.
0.
0.
0.
0.
0.
0.
0.
0.
0,
0.

0.

E
017
114
260
304
760
367
457
378
049
029
394

497

r
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.

1.


1
2
2
4
3
6
7
4
2
2
1

9
K(10°)/K(5°)
                       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

-------
                                                  '    .  i     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  - Twoi Vanes
                       1975 Data - (see Lockhart 1978)
                KEY:
             MRI Model  1074
             TG Model 53.2
                                                                       "500
                                                                       -100
                                                                           S
                                                                           o
                                                                           I
                                                                       -50
                                                                           CD
                                                                           Z5
                                                                           CT
                1.0
                      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-cra, 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:    2O  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 0 (deg. )
»\
T T
(g-cm) (g-cm)
72 69
216 217
»V
T T*.
(g-cm) (g-cm)
72 76
166 177
562 557
«'*
T I3
(g-cm) (g-cm)
108 97
324 278
1152 1163
2°\
T T4
(g-cm) (g-cm)
144 150
432 430
T is the measured torque holding the vane at offset angle 0
K is the linear regression coefficient when a - 0
T is the predicted torque using K from the following:
A 1 ,. A 1 ,
TX= 19.55 U T2= 50.22 IT
A 2 A 2.
T = 57.47 U T * 88.81 IT
J ^t
                     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 U  expression quite well.  The 5 degree and 10 degree data for the Model
1074  define a different slope than U2 on the log-log plot of Figure 4.2.2.11.
An expression of U1'6 fits  the data vjell  enough  to use to extrapolate  the
experimental data for this vane des^n 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. '978)
    £
    o
    CD
    Z3
    CT
    o
       7 -
       6 -
5 -
           1.3 m/s     v  20°     2.2 m/s
            KEY:X 15
                A 10°
                +  5°
       T =  88.81  U2
                                                        4.5  m/s
               T = 57.^7 U
                                                                        -1.000
                   T = 50.22  U1'6
        0.2
Figure 4.2.2.11
          0.4
—i—
 0.6
                                       T = 19.55 U
                                                             1.4
                            0.8 .      1
                    In Wind Speed  (m/s)
          Torque measurements as a function of wind speed.
                                                                             £
                                                                             o
                                                                             I
                                                                     c
                                                                     13
                                                                     CT
                                                                        - 100
                                                                        - 50
                                             1.6
                     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 gra» scale
at 10 cm from the axis of rotation will yield g-c» after dividing by 1O   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.
8£&S£S8BS8gKSS£S?^W':¥*«'^^             .•^•m.L-WWivw*-.V, .V,, .«.«,	, •»,».,„.*.	_.		   . .             *"^  '             "
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 measurenent
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  Cworst 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
arid 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  (errorMl)      ^3 degrees relative to  the sensor
                               mount or index  (^5 degrees
                               absolute error for installed  system)

      (1) as determined by wind  tunnel  tests conducted on production
     samples  in  accordance  with  ASTM D-22.11 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 so»e
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 Wo. 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 systeas,
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 jthan -3 to +3 degrees
relative, or 6 degrees if  the bias has  not been removed.
                                                        i-
                      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 oer
volt = 540 / 5 =  108)                                   i
                                                        i
 :                     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 for<>the "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 ccw
120 ccw
060 ccw
360 ccw
300 ccw
240 ccw
180 ccw
120 ccw
180 cw
240 cw
300 cw
360 cw
060 cw
120 cw
180 cw
240 cw
System
Output
B
(deg. )
177
114
054
001
298
238
176
115
177
239
292
001
056
115
177
239
Nominal
Voltage
B/108
(volts)
1.639
1.056
0.500
0.009
2.759
2.204
1.630
1.065
1.639
2.213
2.704
3.343
3.852
4.398
4.972
2.222
average error C
cw
is clockwise, ccw
Error

E=B-A
(deg. )
-3.0
-6,0
-6.0
1.0*
-2.0
-2.0
-4.0
-5.0
-3.0
-1.0
-2.0
1.0*
-4.0
-5.0
-3.0
-1.0
= -3.4 (*
Normalized
Error
E-C
(deg. )
0.4
-2.6
-2.6
4.4*
1.4
1.4
-0.6
-1.6
0.4
2.4
1.4
4.4*
-0.6
-1.6
0.4
2.4
values
Output
B-C
(deg. )
180.4
117.4
057.4
005.4
301.4
241.4
179.4
118.4
180.4
242.4
301.4
005.4
059.4
118.4
180.4
242.4
excluded)
is counterclockwise

-------
                                                             Section No. 4.2.2
                                                             Revision No.    2
                                                             Date: .   17 Sep 8~
                                                        >     Page:   25  of 32

                     An analysis of this type helps to optimize the accuracy cf
the orientation. QIf 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 N3S
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 averageodirections 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 centerline.

-------
                                                             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)                 £5 m at 1.2 kg/m  (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  (ft) as the ratio of the
amplitudes of two successive deflections of a  wind vane as it oscillates about
8  after release from the offset position, as  expressed by the equation
 B                        .                .              ,
                                    8-  :•..;,..-.        .  •     • •   -
where 8  and 8,      are  the  amplitudes of  the  n  and  (n+1)  deflections,
       n       (n+l)            r    ,-      ,   '     .
respectively.

The Damping Ratio  (17)  may be calculated approximately  from the overshoot ratio
by the formula

                                 .  V1"
                          TI st • -—	

-------
                                                             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)        aO.4 at 1.2 kg/nf3  or

               Overshoot  (1)            <25% at 1.2 kg/in3

               (1) as determined by wind tunnel tests conducted on
               production samples  in accordance with ASTM D-22.11 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       i
                                                         i
                      The Dynamic Vane Bias (8 .  deg.)  is the displacement of
 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 wheince it coraeth
                and whither  it goeth:...  John 3:8
                 From  the standpoint  of wind  measurement1, 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.

-------
" Delay Distance  (50% recovery)


10"
J_

X

/

x^

Thies (experimental)
697 grams

)*
L_



-i

^
HRI
656

1074
grams
                   ,1hies
                   600 grams
            Climatronics F460
            251 grams
           MSI  (experimental)
           191 grams
           Climet
           170 grams
          Vaisala
           92 grams
          MSI  (experimental)
           72 grams
 012        5

Distance (m) at 5 m/s

     4.2.2.15  /
                                                           Section No. 4.2.2
                                                           Revision No.     c
                                                           Date:    17 Sep 89
                                                           Page:    28  of 32
                                            Meteorological  Standards
                   of

-------
                                                              Section No.  4.2.2
                                                              Revision  No.     C
                                                              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 O.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
Over
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

•shoot
Percent
100
90
80
70
60
50
40
35
30
25
20
15
10
5
! denotes
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
PSD 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

                          Coordinates (001-360 degrees)  to 'cartesian coordinates
  he samDlenerlKin meterS)>  ^ Components are added or subtracted^-
 final coornf  .    "* ^ resultant vector direction is found by converting the
 final coordinate  sums to polar coordinates through an  arc-tangent calculation
 ™rai*hf f^ de!iation of each «»Ple about the resultant  ""Tor direct Jon is "a
 Of course   tMs   """^H^ "" difference s of the «•?!•» from the "average "
 vector sables       ""     ^    USed WUh the trUe  ^^^^ighted wind
                                                          I '

 direction       Some  automatic systems currently available pick an assumed
          n-  ?a   y thS mean °f the  1SSt Period'  take the digital difference in
          H  <  !?  t0 180)  °f eSCh Sample fr°m the assumed mean'  and find ?ne
           about'tH ^ thS ^^ difference fo- ^e period.   The standard
           about the mean  is the same  as the standard deviation  about the mean

                                        plus the
           (see        .
 quality assurance to  know that the method used is being satisfied 'Vlth the
 samples taken from the measurement system.                 isnea witn me
          *K   *     fample size is specified in EPA (1987W as 360 samples to
 apparent biL 8tanfird de^iation to within 5 or 10%.  Lockhart (1988) found an
 f5S So S   I      a rand°m err0r' When the sta^d^d deviation was estimated
 from 120 samples over a 20 minute period.
 Pasquill-Gifford stability class.   The method is
                                                                     a
                 (T
                  A(l-hr)
                                15
-    +CT
*30    A45
                                                     60
                                                         1
                                                         i
   i           where a-2  is calculated between 00 and 15  minutes
                      15

              where 
-------
                                                             Section No. 4.2.2
                                                             Revision No.    0
                                                             Date:   17 Sep  89
                                                             Page:   32  of  32






formulation for o-A(1_hr) is shown below-
 A(l-hr)
   where nltVii3.n4 are the number of samples in periods 1,2.3 and

     and d2,d2,d2,d2 are calculated for each period from
                                VA
                                V  A
               where \ is the average direction for period ^ etc.

                 and A  is the average direction for all four periods.

  See Box et al. (1978) for further discussion of  the calculation of  total
  variance from discrete subset variances  and means.

-------
 4.2.3  ACCEPTANCE TESTING
                                                              Section No.  4.2.3

                                                              Revision No.     0

                                                              Date:    17 Sep 89

                                                              Page:     1  of  5
                 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 ?he Salt will
          h-           °f Perforraance  specifications,  those- which can be
          by simple inspection testing and  those which require unusual test
     e  S.-                                        on    ,
 If  the  manufacturer does not provide such documentation, the choice is be Jwfen
 accepting  the manufacturer's unsubstantiated claim or 'hiving a specif ic test

                                      5

                                   TSASS

                 Wind Speed and Wind direction sensors which predictably
bearn     It   Se"OUS^ ^fluence the performance of the seLors  are the
bearinf condit on^vT t ^ ^.^ exPerienced inspector can "measure-
bearing condition by feeling or spinning the shaft.   Th« receiving  inspection
is a protection against putting defective equipment into the fieW    I? is not
a necessary link in the documentation trail  for data validity purposes  True
      caiahUrMentS T°r daU VaUdity WU1 be most  valu-b^ atPthe  initial
      calibration.   True laboratory conditions may -be chosen however   because

                                         icuit to
       4.2.3.1  Wind Speed



-------
                                                        Section No. 4.2.3
                                                        Revision No.    0
                                                        Date:   17  Sep 89
                                                        Page:    2  of  5
           Table 4.2.3.1 - Anemometer Performance Specification
                                     0.5 to 50 m/s
                                    £ 0.5 m\s

          Accuracy  (error)1'2(t) (*)   *  (0.2 m/s + 5%  of  observed)
          Range
          Threshold1>2(t)
          Distance Constant (t)
                                    2= 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).
             ft)  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
with the anemometer not turning (bel™ sensor threshold) and
anemometer turning at an equivalent 0.
Vinyl Cup Set  (100083) or Heavy Duty  '
for mph and 30 pulses per revolution       ,   .no _t
revolution. At 0.5 m/s the cups should be turning at
                           0.5 * 1.41  =0.35 rps,
                                                                   °UtpUt
                                                                 ==-
                               ^
                           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

                  -era z ss/sr^ ss=
 output  is within the tolerance specification.

-------
                4.2.3.1.3 Distance constant
                                                             Section No. 4.2.3
                                                             Revision No.    0
                                                             Date:   17 Sep 89
                                                             Page:    3  of  5
                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 2s 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 ch;inge 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.                                                 i

              Table 4.2.3.1 - Wind Vane Performance Specification
              Range
              Threshold1>2(t)
                              1 2
              Accuracy  (error) '  (t)
              Delay Distance  (t)
              Damping Ratio (t)

              Overshoot2(t)
   001  to  360  degrees or
   001  to  540  degrees
 £0.5  m\s      !
 :£  3  degrees relative to the
     sensor mount  or index -
"2s  5  degrees relative to
     TRUE  NORTH
                  3
•as  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, I9i35a).  '
                                                       " I
                  1  subject to receiving inspection    I
                  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 £ 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.
                                                  Revision No.
                                                  Date:   17 Sep g
                                                  Page:    5 of
      4.2.3.3 Measurement System
                   ^

                                                 -A
the sensors were providina the innnt vai,,   ""^s-  in 4.2.3.1 and 4.2.3.2
Show perfect results and ?et theTf  f   ?U I    P»«">le for these tests to
show the syste, to be out of oaltbSl™ "S "" c'llbr"I°" "witch., on couid
adjustment. for the substUute^aUb'aUon Suts"?"
the
             rr ^e-    -             -~ - -- 2.
sr is-s^as S^^SI^ r

-------

-------
                                                         •    Section No. .4.2.4.
                                                             Revision No.    D
                                                             Date:   17 Sep S9
                                                         ;    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 sliort 10 a
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
wihd 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.
        •       '                                          I
       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  S


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 starting 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
saae 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.    ;
                                                               Date:   17 Sep 89
                                                               Page:    3  of  S
  CONNECT
                                                           i
          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 (QED)
       H for Historical Epicenter File Searches (EIS)
       M for Geomagnetic Field Values
  Enter program option (Q,  H or M):  m
                                    GEOMAG
                                                                   in
             u                       e™
             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 a?fect«=
 field values for dates between 1945.0 and 1965.0, but not those for ?ater

                           Please contact Norman Peddle, U.S. Geological  Survev
                                                                      *
          ^  Suf8estions?

     7364")* Center>
Press RETURN to continue:
Do you want  information  about  this program  (Y/N)?
Options:   1) Field Values  (D,  I, H, X,' Y, Z, F)
           2) Magnetic Pole Positions
           3) Dipole Axis and Magnitude
           4) Magnetic Center
Display values twice (Y/N)?
   Name of field model:
The following field models are available:
                                     C° 80225' Telephone:
                                                        t ]  N
                                                                        236-.
                                                        [1]
                                                        [Y]
      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
                              i
                              [1/25/89]
                              [  ]  N

                              [  ]  W
                              1010]
                             - 1990.0
                             - 1990.0
                             - 1990.0
                             - 1990.0
                                              J47{
                                     Region
                                   World
                                   48-States
                                   Alaska
                                 ,  Hawaii
                               USCON85
                                 1
                              .25
                                                    3122.6292
                                           {250 feet
Model: USCON85
Date  : 1/25/89

       D
   deg  min
   19  47.9
   19  47.9
Annual change
           Latitude  : 47.25    N
           Longitude: 122.6292 W
I
deg min
69 29.0
69 29.0
H
nT
19526
19526
X
nT
18372
18372
    0
    0
-5.5
-5.5
              0
              0
-0.7
-0.7
-1.9
-1.9
8.7
      Elevation:  250.000 ft

    Y       2       F
   nT      nT      nT
 6613   52181   55715
 6613   52181   55715

-29.8   -39,6   -37.7

-------
                                                              Section No.  4.2.4
                                                              Rtvision No.    Z
                                                              Date:    17 Sep 89
                                                              Page:     4  of  S
                The GEOMAG  program was accessed by calling 1-800-358-2663
0 7  less tnan toe map w^~.   This is consistent with the 0 deg. -5.5  .In.
annual change for the  roughly eight years since the map was revised.

                The  other way to find the direction to TRUE NORTH employs some







 ultMn 1  dee  (without correction)  on  any clear  night.   The  true soiar noon
 metnoS wilfprovide the north-south direction to within  0.1  degree  on any clear
 Say,  given the station longitude,  date and an accurate clock.   A
 Drosram will provide  the azimuth  angle to the sun at  any time of
 station longi?ude,  latitude  and date.   Examples  of the two solar
 given below.                                                •

                       4:2.4.3.2.1  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 ^J«^rth-BOUth.
 t""             ^     ^ ^_ M	^^ Ortlo  4-^0 south Pole and the longitude
 plane passing
 selected.   If
             '£ tL^of  th^^/A^arrnrNoon-^rfrom  the  longitude.   The

 examples  sh™^h£JJ£e  ^lLong.  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  .

 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
?
9
10
T   (WA)

T^CLA)
 LAH
                                          i      ,    ...
            12:00:00 * 4(122:6292  -  [15  x 8])  = 12: 10.52 =( 12: 10;31 PST;-;.

           - 12:00:0(3' ^4(090. 1100^-; [15 x 6]); -12::00.44 - 12xOQ:26;'CST,
                      •'         • •          '        "
  Secondly,  correct for/the Ephemeris of  the sun.
              ....     »  »   1- 1_ _ —**. ~*~*~. f*^ 9 f*v\ ^"/^IIT^*
   TSN
            DI-lBUt,  J.VJI  «,ll-i ^^/..s,...-	-—             •-  •
       T    - A  where A is the correction found in Table 4>;2.4.
-------
                                            Section No. 4.2.4
                                            Revision No.    o
                                            Date:   17 Sep 89
                                            Page:    5  of  8
     Table 4.2.4.2  Ephemeris of the  Sun
	 From the Nautical Almanar
Equation
Date of time
	 	 m. s.
wan. i - 3 24
4 -448
7 - 6 08
10 -724
13 -835
16 -940
19 -1039
22 -1131
25 -12 16
28 -12 54
31 -13 25
Feb. 3 -1349
6 -14 05
9 -1414
12 -1416
15 -1411
18 -1400
21 -13 42
24 -1318
27 -1249
Mar. 2 -12 15
5 -1137
8 -1055
11 -1010
.14 -921
17 -831
20 - 7 38
23 -644
26 - 5 49
29 - 4 54
'A) = 12:10:31 - (-
-A) = 12:00:26 - (•
Equation
Date of time
m. s.
Apr. 1 -400
4 -307
7 -216
10 -126
13 - 39
16 + 6 '
19 + 48
22 +126
25 + 1 59
28 + 2 29
May 1 +253
4 +313
7 +328
10 +338
13 +342
16 +342
19 +337
22 + 3 26
25 +311
28 + 2 52
31 + 2 28
Jun. 3 +200
6 + 1 29
9 + 55
12 +19
15 - 19
18: - 58
21 -136
24 -215
27 -253
30 -330
-4:16) = 12:14:47
-0:08) = 12:00:34

-1989 Yachtsman's Edition
Equation
Date of time
	 	 m.s.
pul. 3 - 4 05
6 -437
9 -506
12 -532
15 -553
18 -609
21 -620
24 - 6 27
27 - 6 28
30 -625
Aug. 2 -615
5 -601
8 -540
11 -515
14 -444
17 -408
20 - 3 28
23 - 2 43
26 - 1 55
29 - 1 03
Sep. 1 - a
4 + 50
7 +150
10 +252
13 +355
1.6 + 5 00
19 +604 '
22 + 7 08
25 ,+811
28 +912
— ^— — — 	 — .
Equation
Date of time
m. s.
Oct. 1 +1011
4 +11 08
7 +12 02
10 +1252
13 +1338
16 +1420
19 +1456
22 +15 27
25 +15 52
28 +1610
31 +1621
Nov. 3 +1625
6 +1622
9 +1612
12 +15 54
15 +1528
18 +14 55
21 +14 14
24 +1326
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
PST = 13:14:47 PDT
CST . :
£

-------
                                                             Section No.  4.2.4
                                                             Revision No.     0
                                                             Date:    17 Sep 89
                                                             Page:     6  of  8
instant the sun is in the cross hair of the
that
                                                                        r
                                                         the base of the
to the mark.
                     The  two  drawbacks  to  the  TSN method are weather and
available for  the  sighting.

                      4 2.4.3.2.2 Solar Azimuth Method

 which makes simple extrapolation Impossible.
                       A Brunton compass,  mounted on a tripod  can be  used for
 solar sighting.   The mirror
 rSSlS*^".^ "addUionrprotror a   polnr 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? 4
  MONTH? 7
  YEAR? 1990
                       SOLAR AZIMUTH ANGLE
                       WEDNESDAY 4 JUL 1990
                 Fox Island, h'A
                 Longitude J22.6292    Latitude 47.25
                 Day of Year 185      Julian Day 2448077
                   Transits Meridian  13  14  53 PDT
     DAY? 25
     KQMTH? 12
     YEAR? 1990
                        SOLAR AZIMUTH ANGLE
                        TUESDAY   25 DEC 1990
                 \ fe* Orleans, LA
                  Longitude 90.11     Latitude 30
                  Diy of Year 359   .  Julian Day 2442:1
                    Transits Meridian 12 0 34'CST
Tiie
HR,MIN? 11.30
HR,MIN? 12,00

HR.MIN? 12,30
HR.MIN? 13,00
HR.MIN? 13,30

HR, HIM? 13,00

Elevation
57.81
61.36

64.00
65s43

65.42
65.42
Azisuth
130.15
141.90

155.86
171.73

188.40
171.73
Ties
&.SIK? 1C. 30
HR.KIK? H.OO ;

HR,HIM? 11.30
KR,MIS? 12.00
i
KR.HIN? 12,30 • ' ! •
HR,HIN? 13,00 !
•
Elrvation
32.35
1i '7'
. .C.
36.11
36.61

36.14
34.73
Aiieuti
155.28

163.07
171.33
179.36

18138
196.65
Figure 4.2.4.1   Screen printouts for two azimuth  examples
                                                                  I
                   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
             where   Lo is  the station longitude
                      La is  the station latitude
                      D  is  the date of interest
         LA  (12/25/90)
   TSN      Almanac   dif.

   CST       .CST     (s)
12:00:34  12:00:: 34    0

-------
                     :.

100D^M22.62Q,<7J5flEM lONGITUDE* MlSHSSi^-
110 READ PI.OB.LO.U.AO.A1.EC.E6
120DATA3.141502654..409095.4.88376619..017202791
130 DATA 6.23471229.,017201970,.016728..00218
140 TR.PI/180:FC.2-PI:REM TO RADIANS: FULL CIRCLE
150 SL.15'INT(LO/15+.5):REM STANDARD LONGITUDE
180 TZ.SL/15-*: REM SELECTS TIME ZONE LABEL
100 LO.LOTRdA«LATR5L-SLTR
210D1S.-SONDAY  MONDAY  TUESDAY WEDNESDAY"
220 D2S.THURSDAY FRIDAY  SATURDAY-
      .-JAfBKWUPRMAYJUNJULAl«3SEPCX:ThK3VDEC-
      .-ASTESTCSTMSTPSTYSTASTADTEDTCDTMDTPDTYDTADr
 260TO.LORXSMM UONGrTUDETIME OFFSET* 12 HH
 270 PRINT •DAY-SINPUTD
 280 PRINT •MONTH'jUNPUT M
 280 IF M>12 THEN PRINT 'INVALID DATE'cGOTO 280
 300 PRINT -YEAR'plNPUT YR
 310 X-1:Y-1:GOSUB 2410
 320 J1-T:REM YEAR DAY 1
 MO X-D:Y-M:GOSUB 2410
 340 YO-T-JU1 «EM DAY OF YEAR
 350 X-INT(IVI )/7:Y-INT(X)
  38S DT-.00059+2JZE-08T :T-
  390 PRINTTAB(28)fSOLAR AZIMUTH ANGLE"
  405 PRINTTABpS);
  410 PRINT MDS(PS.S>WO+1.9):
  420 PRINT D;MIDJ(M$;r(M-1H-1.3):YR
  490X.YD-WD:Y^L-15TR
  500 IF X>80 AND X<293 THEN TZ-TZ+7 : SL-Y
620IFDE>PITHENDE.DE-FC                     _
630 O.ML-RA*EM EQUATION OFT1ME (NOTDISPLAYED)
$40 D3.DE: REM SAVE DECU FOR HEAT BUDGET
680 X—.0145439 : GOSUB 2360
690IFABS(Y>c1THEN720
710 GOTO 780
720 SO^"(1+L1/FC): H-SO: GOSUB 22BO
725TC..00274'SO-SIN(OBrCOS(TL)-SIN(LA)
730 Z-SIN{SO)'COS(LAnCOS(DE)*3)
735TC-TC/Z
740 X -ZTVTC+EO: GOSUB 2310
780 PRINTTAB(23)rrran»lu Mtridian -;
780 IF ABS(LA-DE)>PI/2 THEN PRINT XSPC$;T5
800 H-0: GOSUB 2260
810 X-ZT: GOSUB 2310
820 PRINT X;Y2:TS
 830 PRINT
 850 PRINTTAB(12),-Tin»
 1890 GOTO 4000
 2010 C-0*I-0
 2020 IF YoO THEN 2050
 2030 Z-0.-C-1 UF X<0 THEN N-1
 2040 GOTO 20EO
                             B«v«tion  Aiimuttr
                                                                                   Section  No.  4.2.4
                                                                                   Revision  No.       0
                                                                                   Date:     17  Sep  89
                                                                                   Page:       8   of   8
                                                           2050 Z-XY
                                                           2060 Z-ATN(Z)
                                                           2070 F C-1 THEN Z-PI/2-Z
                                                           2080 F N-1 THEN Z-Z
                                                           2090 F YPI THEN AZ^Z-FC
                                                           2195RETUPJ>I
                                                           2210 H-ZT+SL-RA-LO+ML+PI
                                                           2220 F H>PI THEN H-H-FC
                                                            2230 RETURN
                                                            2260 FOR U-1 TO 5
                                                            2265 ZT-H+RA+U5-SL-ML-PI
                                                            2270 X-SN(ZT) : Y-COS{ZT) : GOSUB 201 0
                                                            2275 M L-LO+L1 -(T,TN+(SL»ZVFC) : NEXT U
                                                            2280 ZT.Z : RETURN
                                                            2310 F 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 ZJNT((Z-Y)-60) : RETURN
                                                            2360Y^X-SIN(LA)-SINpE)y(COS(LA)'COS(DEW
                                                            2370 F ABS(Y)>1 THEN 2390
                                                            2380 X-SOR(1-Y*2) :GOSUB 2010
                                                             2390 RETURN
                                                             2410T-36T(YR-1980)
                                                             24SOT-T-WT(3'(Z+1V4)
                                                             2460 T-T+JNT^S-Y/aH
                                                             2470JO-T+2447689*
                                                                             -T: REM SUN'S MEAN ANOMALY*      '
                                                                             TS REM SUITS MEAN CELESTIAL LONG.TUDE
                                                                 2880 X^W(ML):Y-COS(ML)«OSUB 2010

                                                                 2830 ML-Z
                                                                 2920 RVm-EC«2V(1+EC-COSrTA))flEM RADIUS VECTOR
                                                                 2930 X^«OU*SIN(QB)^-SQH(1-X*2)fiOSUB 2O10
                                                                 2940 DEJaF Z>Pl THEN Z-Z-FC
                                                                 2950 )USWfrL)-COS(OB):YrfOSCTL)SOSUB 2010
                                                                 2960 RA-ZflEM SUN'S RIGHT ASCENSION
                                                                 2970 RETURN            •           ,     .
                                                                 40lOZTHJtfl+MIN/60 :
                                                                 4020 T.T-TN-KZT+SLVFC
                                                                 4060GOSUB2860
                                                                 4070 GOSUB 2210
                                                                 4080 GOSUB 2160
                                                                  4110PRINTTAB(24):
                                                                  4113PR1NTUSING-
                                                                  4117GOTO4000  '
                                                                  9000 END

-------
                                                             Section No. 4.2.5
                                                             Revision No.    0
                                                             Date:   17 Sep 89
                                                             Page:    1  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
                                                           i
                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 Ichallenge 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.                                           i
                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-cra 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.Jos. visible- at- 12.6)  The
true Wind  speed will oe higher tnan 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 (i
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-rfound.   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        '     Figure  4.2.5.2
error of the sensor be distributed equally          A direction va"e clamp
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 be set.   The open space;in .
the potentiometer, if one is used,  can be measured and adjusted for.

-------
                                                             Section No. 4.2.5
                                                             Revision No.    I
                                                             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 a,
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 pf 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 01
 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
  a^,,r-a      impor^ant. asPects of operations, from the standpoint of quality
  Section No*"! ^T^*** ^ P1™' SeCti°n N°' ^' ) a'n* <*°cumen?ation
  (Section No. 4.9.1).  The purpose of operations is to acouire valid data
  wind measurements,  this requires frequent (weekly,  if pSSbl.^!^
  IS at  he ^i^  SHnSOrS-  ™S 1S n0t a "hand-°n" examination bu? 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
          «,         ^ °f an Enal0g recording,  if one is used routinely   wm
         fH   S€T0r 1S Performing'   Ro"tine entries in  the station log' will
         the evidence of attention to support validity claims.

 nr«a .  C^ibrati°ns are a Part of operations.   A member of the operating
 anf wnat Unne±% ° b^°me ^ "eXPert"  °n h°W the '-asurement'Wste^works
 and what it needs to continue "in control"  performance.   Regularly scheduled
 a"curacy10nThe1re ^ T^S* "~ documen tati-  showfng meas'emtn
 accuracy   The frequency of calibrations is  a  variable.   For a new
 ££aiiatS°V rlib^ti0n dUring  the  instaHation is necessary.   A careful
 look at the first week of operation will find  early failures.   If  all sterns  to
       *                  ™    ?  after a m°nth is

                           Calibration is the minimum freq in" If problems are
                90    '      ^ "* C0rrected as <**<*» as .possible
                " -

       4.2.6.2  Maintenance                               |

                4.2.6.2.1 Routine and preventive maintenance

 w :  -,., ^        The only routine maintenance required for the wind
?r analog recorrr       S SSr'SS^* digS r^erf
timely servicing of these requirements  is  a  routine maintenance task
m   f  ,.    ,    Preventive maintenance  must  at  minimum follow the
SS.

-------
                                                             Section No.  4.2.6
                                                             Revision  No.     C
                                                             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 fron
    manufacturer   This is true at least at the sub-component and component
levef  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.  Th     y                                     "SJ^T

                 The next level of spare  part  strategy is  the  sub-component
repaired.
        4.2.6.3  Quality Control
 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 whLS  brings the operators into the  data QC loop.  Their inspection
  n/forrectivfaction is  reported back  to the  QC  i-^^JS^^niSJ-
 Because  of  this  QC  loop,  the  measurement system can be operated   in control
 and valid data produced.

-------
                                              j  Section No. 4.2.7
                                                Revision No.   :
                                                Date:   17 Sep 89
                                                Page:    1  of 2C
    ;

 4.2.7  PERFORMANCE AUDIT METHODS
                                              I
       4.2.7.1 General Considerations



 aaS^aSrSSsSisiKsnr.

                                  of
                                              and/°r
 determined.  0 course  the accuracv
                                                is
 determination has a larger uncertaintv  ThTr-Tc  ??'.  6 accuracV
 anemometer transfer function^ Th« h^«;   J      h°d d°es challe"Se the
 where appropriate   fUnCtlOn'  The best Performance audit uses both methods
            4.2.7.1.1  Who
identify the
-«,« report the
            4.2.7.1.2 What
                -

                             asrw
provided and use^by'the 'audUor ISudin
of last calibration; where re le^nt      g
                                         °f a" the
                                        SSrial numbers

-------
                                                            Section No.  4.2.7
                                                            Revision No.    0
                                                            Date:    17 Sep 89
                                                            Page:     2  of 20
                A. *y 7 1 *?



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 time whence system -J^^TlT^^  *^>
"off-line" and put back "-      . S        /thf complrlson with  the syste*
correlated to the time the
                                                       or correct  time.   Battery
                                                   be,  are generally available.
                 4.2.7.1.5  How
 units compatible with the system output.

        4.2.7.2 Wind Speed
  system operated with interchangeable sensors.

                  4.2.7.2.1 Sensor Control

                  The controlled condition is rate of rotation of the
  which can be converted to     ^ =

  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 the shaft over a period of
 time is the data value.  That value divided by the
 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.
 mtftor 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
                                                               Section No.  4.2.7
                                                               Revision No.     0
                                                               Date:    17 Sep 89
                                                               Page:     3  of 20
                                                           Figure 4.2.7.2
                                                                        17,
and
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 values,
l.b,  4, 7, 17 and 35 rps, both cups and propellers could be challenged at  three
meaningful speeds plus zero.   See Figure 4.2.7.3 for a third d.c. motor system.
Figure 4.2.7.3  An experimental 12-voit d.c.  motor and counter

-------
Climatronics  cup
Teledyne-Geotech cup
 R.ti.  Young propeller
   Figure ,4.2.7.4
                                                             Section  No.  4.2.7
                                                             Revision No.    0
                                                             Date:    17 Sep  =9
                                                             Page:     4  of  23
     A simple d.c.  motor might be made to turr.
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 &
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 CAOO3  (version 8/1/64)

        This method provides for a comparison of the transfer  function
        used with the system to the output of the system.   This  is done
        by causing the anemometer shaft to turn at a known  rate  of
        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 "tffe 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  options!
        determination of starting torque and system time constant.  The
        torque measurement may 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  tp  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, take
        12  samples over a two minute period .and use the average  of the
        samples  to compare with the average rate of rotation measured.

        CAO03—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 cup
        assembly removed and the shaft taped  to assure non-rotation.  A
        second observation may be either a motor driven measured rate of
        rotation for  the operating period  of  the system or a natural tun-
        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  likely true with pulse trains than with generator-
        voltages.
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
       MEASUREMENT SYSTEM - Cup anemometer
            System number 	
            Sensor
            Cup assembly
            Loc at i on      	.__
            Signal conditioner
            Data Channel  	
                                                                    CAOO3
       DATE 	/	/	TIME off line	on  line

       TRANSFER FUNCTION:   (rps to mps)
                                  	 test start
       TEST RESULTS

         CAQQ3-A
          "d.c."

            S CM

            F CW
                            (rps to volts)  ,	__„
                           pulses per revolution
    - .challenge speed -    output   difference

 time   revs,  rps    mps     mps   , mps     '/•

          O     O
            Torque:

         CA003-B


            O

          test
        Or.—In. ,cw,   Time constant
time
expected
   mps
observed
 ;  mps
	seconds

difference
mps     7.
Figure 4.2.7.6  Audit form for the cap anemometer method

-------
                                                         Section No.  4.2.7
                                                         Revision No.    o
                                                         Date:    17 Sep 89
                                                         Page:     7   of 20
       Fixed Axis Propell
                                MSI  method FAP001 (version 8/1/8^
  form  i. provide      W
  known rate of. rotation while
  "linus sign  i. used
                         an
                                          °*
                                                          function
                                                       A Se
                                           the
                                          i1
                                                     p
                                                          th°Ugh  the
                          g,
 transfer function)
                                      t
                                                      fr°"
taneous  Sample voltage may be us^d   S^e
ronous operation o-f the motor is required.
                                                     of
                                                            synch~
                                         system be
distance away).
                                                      are at

measured,
signal conditioner  when the

Figure 4.2.7.7  Audit method for a propeller anemometer

-------
                                                      Section No. 4.2.7
                                                      Revision No.    0
                                                      Date:   17 Sep 89
                                                      Page:    8  of 20
            PERFORMANCE AUDIT REPORT by 	
                                                              FAP001W
  MEASUREMENT SYSTEM - Fixed axis propeller
       System number
       Sensor
       Propeller     	:—
       Signal  conditioner
       Data channel  	
                                      on  line
                                                      test start
  DATE  _ / __ / _ TIME o-ff line

  TRANSFER FUNCTION: 1 rps - O.294 mps  (3  pulses per revolution)
                        CWCvolt.s>-2.5:*4  = m/s

  TEST  RESULTS
FAPOO1W-A
              - challenge  speed -    output

            time    revs.   rps    mps    mps

                                    0  	
                                                     dif-ference

                                                      mps     7.
"d.c."

  S  CW

  S  CCW

  F  CW

  F  CCW

 "sync"

  S  CW

  S  CCW

  F  CW

   F CCW

   Torque:

FAP001W-B


   O

 test
                                               mps
                                                      mps
                                                                7.
              time   rps    mps    volt

                    5.000   1.47	.	
              • 1.1 •  -








                  Oz.-In. cw.	Oz.-In. ccw, T. Const.„	s.
              time
                            expected
                              mps
                                   observed
                                      mps
difference
mps    .7m .
Figure 4.2.7.8  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


 involves mounting S^^cI^^'^St^S th^cSty^ 5T

 calSrattoTrNBS ^'S^^.tS^f^^^1^^^^ ^

 ^mTrabil^y t* belng  US6d> thS CTS needs to be within 10 m of the sul
 ane«ometer in the horizontal and the lesser of 1m or H/10, where H is the
 height above ground in meters, in the vertical.  It is important to sUe the
 CTS to be representative  of the flow at  the subject anemometer   Mutual
 interference should be minimized through siting and through editing out data
 accuracy6  tentiafo?T*' *** Wind'passinS through one to reach the other   The
                           , b method .is  based on data taken in 1982 at the
         H T  -^K   .  °bservatory (BAG) and published by Finkelstein et  al.
       and Lockhart  (1988).  The anemometers for this study were spaced about s
 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
            are connected to the auditor's data logger.  The element of  the CTS
             difference in speed calculated by subtracting the CTS speed from
            speed.   The method requires a sufficient  number of simultaneous and
            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
 f£r rTcT    aSSUme  the subJect  anemometer has a distance constant  of 5 m and
 win be 5?3 - i 7taT C°nstan^of ! m-  If the wind speed is  about  3 m/s,  T
 win be 5/3 - 1.7 s for. the subject and 1/3 =  0.3 s for the  CTS.
        tfill  exist when the sampling rate of the data logger is
 the example^ ^ °'°3 S ^ "^ example'   Most data loggers  are  fast enough for

 „   ,     .    .  ^dependence 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.  fhe A^TM method
aetmes minimum sample size  in terms of the resolution of the  measured or

^±Vid^s^ °f ^-"^tsi^^TT^s.'s^'r
           cne series to  provide the minimum number.   If the  two sensors are
          and properly operating, the variance will be small   The BAO data
          =tanH3^ deviation of the difference to be less than 0.2 m/s
                         "'  1J "" "'    ' iimum condition for a good CTS data
                                        sample size is 900 x 0.04 =  36.

                                                                 o    o

-------
                                                    Section No. 4.2.7
                                                    Revision No.   0
                                                    Date:  17 Sep 89
                                                    Page:  10  of 20








assumption will produce a measure of accuracy.

              The other way is to treat each 15~™i£u^    ide
sample of 15-rainute averaged data.  A 9 hour per o  w             ^ ^^ ^
sa»ple size of 36.  This is an awkward period of^J^          le.  There is
install the CTS and many auditors wo^ * h°u^ d*^J-J  eniarges the dynamic .
no such thing as too many ^^^-^J^^^^or something like 24 hours,
                                 this CTS audit,  look at the data  listed
 in Table 4.2.7.1.
              Each value is a 20-rainute scalar average.  The  subject
 ane^te. is a Conies "-.up ^V-W, .  Th, CTS is a Young Propeller
            sra Lirji J^sjj  -
 nsr s^ofss srsss.^%2 s
 sta.K.ard deviation of  the     ---uo ,ct  aieeter    h respect to

points is calculated by
      s: srss-
                                   is
                                                                and
                                           found by subtracting the .X
                                                              r

                                                                  SS
  the CIS was 4 X high or thej "J"sfthaf thL anaSlls was not deluded in the
plotted as a function of ,Peed to look for correlations.
influence of turbulence on the residual error.
                                                   There seems to be no
              The CTS method provides a measure of accuracy which can be



                                                           si,-"
                                                                       •

-------
                                            Section No.  4.2.7
                                            Revision No.     0
                                            Date:    17 Sep 89
                                            Page:    11   of 20
Table 4.2.7.1 Simulated Audit Data  (BAD 1982)
"/S
Sate/
lie

4 1000
4 1520
4 1840
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 1400
41620
4 1&40
4 1700
4 1720
4 1740
4 1200
4 1E20
4 ''540
4 im
4 1<=20
4 !?40
4 2SOO
4 2C20
4 2040
4 2100
4 2120
4 2140
4 2200
4 1220
4 2240
4 2300
4 n20
4 2340
5 0
5 20
5 .10
5 100
Y X
Subject CTS
Speed Speed
(fi/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
8.338
7.236
5.425
3.989
4.399
4.407
3.B47
3.941
4.344
4.300
2.907
2.861
2.049
1.747
3.796
4.065
3.-B85
4.221
4.097
4.037
3.952
3.559
3.449
3.179
3.840
4.393
(a/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
8.601
7.893
6.600
5.976
3.413
8.693
7.486
5.588
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
4.040
4.617
5
CIS
Sigaa
(1/5)
1.047
2.441
2.780
0.844
1.412
0.883
1.951
3.230
1.115
1.319
1.348
1.120
1.557
1.039
5.887
2.564
1.63S
1.688
2.847
3.537
4.531
1.508
0.751
0.233
0.820
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
0.144
0. 109
0.123
0.035
0.199
0.113
5/1
(I/S)
0.23
0.82
0.73
0.32
0.60
0.37
0.53
0.57
0.24
0.29
0.30
0.34
0.64
0.42
1.33
0.30
0.21
0.26
0.48
1.04
0.52
0.20
0.13
0.06
0.18
0.20
0.0?
0.07
0.03
O.OS
0.15
0.05
0.01
0.09
0.06
0.05
0.03
O.Q5
0.07
0.05
0.03
0.03
0.03
0.01
0.05
0.02
Y-X
tfi/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
-0.1 1
-6.20
-0.22
Y'-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
0.18
0.10
0.21
0.02
-0.02
-0.00
-0.01
0.01
0.00
0.03
0.02
0.00
-0.01
0.05
0.05


5
5
5
5
• 5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5




!
120 3.822 3.955 0.8B1
140 2.198 2.219 0.320
200 1.6SO 1.742 0.059
220 1.484 1.510 0.576
240 1.599 1.624 0.377
300 2.569 2.697 0.076
320 1.662 1.696 0.341
340 1.769 1.934 0.190
400 1.054' 1.114 0.459
420 1.471 1.455 1.217
440 2.181 2.286 0.438
500 1.338 1.419 0.680
520 3.315 3.503 0.180
540 3.060 3.212 0.220
600 1.992 2.074 0.379
620 1.765 1.836 0.336
640 0.920 0.896 0.072
700 0.636 0.523 0.280
720 1.B09 1.856 0.422
740 1.571 1.577 0.173
800 2.058 2.115 0.292
820 1.910 2.000 0.519
840 1.181 1.165 0.345
900 2.880 3.013 0.891
920 4.236 4.402 0.537
940 3.839 4.000 1.752
Average
Standard deviation



0.22
0.23
0.03
0.38
0.23
0.03
0.20
O.iO
0.41
0.34
0.19
0.48
0.05
0.07
0.13
0.18
0.08
0.54
0.23
0.11
0.14
0.26
0.30
0.30
0.12
0.44





-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

Regression Output:
Constant















Std Err of Y 1st
R Squared !
No. of Observations
Degrees of Freedoa

X Coeff. 0.9576
• Std Err 0.0048

•


.

, • ' ' ' •!

































-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












-------
   10
                                                        Section No. 4.2.7
                                                        Revision No.    0

                  SIMULATED CTS  AUDIT      ££    lz "o? II
             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
                    CTS (P-V-W) Wind Speed (m/s)
      4.2.7.9  XY plot of simulated wind speed audit  data
•g 0.2-r	
 o
c/5
 c-0.2
                               A   A*
                   + +
                                                KEY: A =  Sigma/speed
                                                    A  A
                                          :~v    i<"E?r"+""=="c-V-"w residual
                                                                   1.6 
-------
                                                             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 aounted 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 UVW anemometer.

            4.2.7.3 Wind Direction

                4.2.7.3.1 Sensor control                j

                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 fro» 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',
iapose 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. o 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 (5;40@5)
(deg.) (volts)
0
164
66
18
9
2
0
(after Fritschen
0.833
0.981
1.889
2.333
2.417
2.483
2.498
and Gay,
Change1
( % )
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
 aounted oh the tower.   This can provide .positions related to external objects
 rather than constant angle changes.  It is estimated that the accuracy of this
 aethod 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
 noved and clamped when the vane is parallel
 -o the radial line.   If care is taken to
 avoid parallax errors  (non-parallel or
 ^on-perpendicular observations)  this method
 can provide relative accuracy on the order
 sf one degree.

                 The  best  method replaces
 -he vane with a fixture with the capability
 :f holding the shaft in fixed positions
 "•Ith  respect to the  sensor housing.
 Fixtures of this type  can provide
 rspeatable position  accuracy of  < 0.1°.
 Figure  4.2.7.11  shows  such a device.   A
 different application  of  this precise
 nethod  uses a  theodolite  base as  the mount
 f=r the sensor.  With  the  vane or vane
 substitute held  in one position,  the base
 :=n be  rotated  in very accurate  steps.
 Theodolite worm gear assemblies divide a
 rircle  in whole degrees with a vernier
 adjustment  with 0.1 degree  index  marks far
 ?r.ough  apart to allow  easy  interpolation to
 I 02 degrees, a resolution wasted on the
 application of wind direction measurement.
                The audit report form
s-ould contain the transfer function used
•:: convert output voltage to azimuth
agrees.   This may include a 540 format
-r.ere azimuth values greater than 360 are
reduced by subtracting 360.  The report
frrm should also contain the challenge
progression used by the selected method.
Figure 4.2.7.11
                                                        Figure 4.2.7.12

-------
                                                               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
 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 fr6m
 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.              ',"           ' .      '
                                            ;  ,  .   "  L , - •- ' U •"- •     .-       , - , •
                                                          i
  ',               4.2.7.3.2 CIS Method
         ' •                       --                         I
                 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
 saae 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-ninute average direction for  both the CTS  (Young Propeller-Vane, P-V-W) and
 the subject  (Climatronics  F460,  C-V-W).   Also, shown is .the o-Q ,(or 
-------
                                                   Section No.  4.2.7
                                                   Revision No.     0
                                                   Date:    17 Sep  89
                                                   Page:    16  of  20
Table  4.2.7.3   Simulated CIS Direction Audit  Data
s/82
Dace/
Tin*

5
5
5
5
*
4
4
3
4
4
4
4
5
4
4
4
4
4
5
5
4
4
5
5
5
4
5
5
5
4
4
4
S
5
5
S
S
4
4
5
5
5
5
5
5
4
4
5
4
4
5
4
5
5
4
4
4
5
4
4
4
V
5
5
4
4
4
5
4
5
J
4




74O
72O
114O
44O
1OOO
1840
1B2O
1040
22OO
212O
214O
19OO
11OO
1620
172O
17OO
174O
1BOO
102O
62O
164O
192O
5OO
2O
O
232O
112O
1OO
4O
23OO
2220
234O
9OO
12O
92O
600
•420
224O
194O
520
22O
54O
82O
30O
32O
210O
102O
94O
114O
2040
200
2020
4OO
34O
130O
200O
132O
14O
134O
160O
1200
B4O
120O
SCO
124O
122O
1S4O
24O
152O
10OO
144O
1500
Y
Subject
Direction

39.34
20.76
87.87
6. SO
27.02
10.17
3O.55
16.28
7.62
4.O4
10.87
9.00
13.26
60.97
6.10
8.98
6.66
17.58
27.04
32.37
9.24
9.10
30.40
4.44
5.99
6.94
22.84
3.45
5.35
6.21
7.9O
5.88
11.83
9.57
9.31
8.OO
20.18
5.86
8.76
6.29
17. O4
9.38
28.91
8. SO
25.28
35.85
42. 3O
63.08
29.30
12.1O
35.26
11.45
20. 16
12.57
23.11
29.16
22. O9
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
3O.58-
2O. 56
Y-X



2.13
4.57
1.63
1.05
0. 18
0.39
1.42
1.63
O.39
1.22
O.85
O.B7
O.82
O.50
1.23
1.08
O.67
1.22
-O.66
O.85
1.59
1.32
3.85
0.52
0.58
0.33
O.79
0.34
0.50
0.13
-O.45
O.63
-0.13
O.16
-O.05
-0. 17
-1.00
O.O8
O.42
-0.15
4.10
-O.66
O.56
-0.70
2.77
.-0. 19
-0.20
-.1 . 98
-1.25
-O.94
-O.3O
-0.22
2.47
0.77
0.19
O.91
-0.51
O.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
                                        Avar-age
                                        Standard deviation
O.91
1.34
                                                 Regression Output>
                                        Constant                   -0.22090
                                        Std Err of  Y Est             1.33045
                                        R Squared                   0,99939
                                        No. of Observations              72
                                        Degrees of  Freedom               70

                                        X Coefficient
-------
    a>
    c
    o
       360-
    o

   5 270
   -a
    c
   "^
   "P 180
   ^ 090
   O
                       SIMULATED  CTS AUDIT
               Propeller Vone (P-V-W) as Collocated  Transfer Standard
                                                            Section No. 4.2.7
                                                            Revision No.    0
                                                            Date:   17 Sep 89
                                                            Page:   17  of 20
Regression
Y— C\ *7O
Data: Boulder
9/4-5/E
72 - 20 miniH


KEY
y (7-
Sigma Theta
Analysis
L 1 .005 X
Experiment
12
.e averages

S


X



S*


X X
x. # x

jx^
!


X X
^C.-X-x X
X "X X ""•








-75_
-25^,
                                     180           270           360
  :              CTS (P-V-W)  Wind Direction [Theta] (deg.)

Figure 4.2.7.13  Simulated CTS Vind Direction Audit Data I

                4.2.7.3.3  Vertical Wind Direction, 
         ™ i-     K S®"Sor contr°l method is used for bivane auditing.   The
         part of the bivane operation is treated in the same way as the
           part' eePfc di^erent fixtures are used.   Special fixtures are
                           sisn' but *» Principle is the same.   A relative
                K          fiXtUre iS attached to the bivane.  That point is
      fane Sha" 1S Perpendicular to the vertical axis of the sensor.  From
is heldaIn1?l<'PT  "he:*th* 0utPut should ^ the equivalent of 0°,  the vane
              fS^PSJmtl1 its Physical limits are reached,  both tail up { + )
                    Thje*hold i= ^ry hard to measure on a bivane because of
and its  rema     h C°ndlti°nf of the vane"   If the vane is perfectly balanced
                      ^^   " physiGally moved-  a  force gage  measurement at
      4.2.7.4 TURBULENCE OR cr  and s J°b to determine how the algorithm
                      that Process with a known input.  This is also I
                                           sisnal conditioning

-------
                                                            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
range.   The challenge must take into consideration the  wave
variable direction imposed on the system in calculating the
with which the output will be compared.   The effective  time
direction system, calculated from the delay distance  of the
nominal wind speed important to air pollution applications,
aaximum frequency used in the sigma challenge.
some realistic
shape of the
true sigraa value
constant of the
sensor and some
should define the

-------
                                                       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 which 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 SN008)  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 fco 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
 weiaht).   A  1/8" 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  ISO  degree hole.  Rotate
 the clamp until the  output indicates ISO,  either by .equivalent
 voltage or  digital printout.  Since this  is a 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 followina  positions  taking  data at  each point.:  12O,  O6O,
 360. 300.  240.  ISO.  120.  ISO, 24O, 30O,  36O,  06O,  120,  ISO, and
 240 dearees.   This moves the "vane" 42O degrees counter  clockwise
  and then 480 degrees clockwise  to test  "54O" strategies  for the
  angle  discontinuity.                  -            ;

 WVOO4-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.

  WVOO4-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 points 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


-------
              PERFORMANCE AUDIT REPORT by	
                                                         Section No. 4.2.7
                                                         Revision No.    0
                                                         Date:   17 Sep 89
                                                         Page:   20  of 20
WVOO4
   MEASUREMENT SYSTEM - Wind vane
        System number	     _
        Sensor        	___"_      ~~
        Vane          	__ZZZZZZZ      Z   Z
        Location      Z__Z	IZZZ   "
        Signal conditioner ______Z_ZZ_Z
        Data channel	___ZZ  'ZZZ—.IT         ""

   DATE	/—/— TIME off line	on line	test  start	

   TRANSFER  FUNCTION  (volts per degree)	  _      	
                      di scontinui ty strategy~ZZ	ZZ-ZZZZ  Z~Z

   TEST RESULTS

     WV004A      (dif.  »  deg.  - set)

       net   volt deg.  dif.   set  volt deg. dif.   set  volt deg. dif.



       120	ISO  	;	   3*0



       36O	ISO	   120

       30O	240  	   ISO

                                                  24O

     WV004B         1                          2

       volt deg. angle dif.       volt  deg. angle  dif.   sbs«1-2)-360)





       Torque:	Oz.-In. cw. 	Or.-In. ccw, T.  Const.	s.

    WV004C           i                            2

      time     expected    obs. before    obs. after       difference

              volt  deg.    volt   deg.    volt   deg.      (2-1) deg.
Figure 4.2.7.15  A form for the wind direction audit method.

-------
                                                             Section No. 4.2.8
                                                       <      Revision No.    0
                                                             Date:   17 Sep 89
                                                             Page:    1  of  1
                                                         i

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 systens 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
suaraarizations such as a scalar wind speed,  a resultant vector wind speed and
so«e 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
susmarization 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
30O  and O6O  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 Techno 1.,  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)  Araer,  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-O13,  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 Techno 1. ,  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.  Mer.  Meteor.  Soc.,  Boston,  MA
     02108

Kaiaal. J. C.  and J. E.  Gaynor,  1983:  The  Boulder Atmospheric Observatory,
    J. Appl.  Meteor., 22,  pp.  863-880.

Kaioal, 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.  Fujinawa, 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. Ball. 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!., 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 Techno1.,  4, pp.  160-169.

 Lockhart, T  .J.,  1989   Accuracy  of the collocated transfer standard method for
     wind'instrument auditing. J. Atmos.  Oceanic Techno 1. , 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.                         I

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
    Technol., 6, pp. 140-146.
                                                         I
Stearns, C. R.,1985:  Wind-Measurement Calibration,  Response. Ball. Amer.
    Meteor, Soc., 66, p.1545.

Sutton, 0. G. ,  1955:  Atmospheric Turbulence.            j

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
4.3.0  OUTLINE AND SUMMARY

4.3.1  TYPES OF INSTRUMENTS
                                                 : Pages  Rev.
                                                 1   3     0
Date
 9/89

 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
                                                                 9/89
  '.     4.3.2.1  DELTA TEMPERATURE (AT)
       4.3.2.2  TEMPERATURE

 4.3.3 ACCEPTANCE TESTING

 4,3.4  INSTALLATION

 4.3.5  CALIBRATION

        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
  4.3.5
OPERATIONS, MAINTENANCE AND QUALITY CONTROL
        4.3.6.1  OPERATIONS
        4.3.6.2  MAINTENANCE
        4.3.6.3  QUALITY CONTROL

  4.3.7  PERFORMANCE AUDIT METHODS

        4.3.7.1  GENERAL CONSIDERATIONS
                  4.3.7.1.1
                  4.3.7.1.2
                  4.3.7.1.3
                  4.3.7.1.4
                  4.3.7.1.5
                     Who
                     What
                     Where
                     When
                     How
         4.3.7.2  DELTA TEMPERATURE (AT)
                  4.3.7.2.1  Sensor Control Method
                  4.3.7.2.2  CTS Method
                                                     2     0     9/89

                                                     1     0     9/89

                                                     9     0     9/89
                                                      8
                                                                          9/89
   9/89

-------
                                                             Section No.  4.3.0
                                                             Revision No.    0
                                                             Date:    -17 Sep 89
                                                             Page:     2   of  3
Section
       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

       4.3.8.1  MEASUREMENTS
       4.3.8.2  SUMMARIZED DATA

4.3.9  REFERENCES
Pages  Rev.  Date
              9/89
                                                                        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
 IS?6r» beat t0 and from the shield  in a triable 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-.-
x ^      *?*' 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.                                            j

    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 nearly 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 (222fl) as compared to the network thermistor (126Q) or the lOOn
RTD (0.4Q) or the 1000Q RTD (3.8£2).   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 waye length of about 40°C.  The
imbact of this erpbr 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
('«
-10
o

10
20

30

40
J\J
50



Thermistors
	 YSI 	 YSI 	 -
44031
(ohms) (AH)
16600
-684.0
9796
-382.5
5971
-222.3
3748
-133.1
2417
-81.9
1598
-51.7
1081
44203 network
(ohms) (AQ)
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) (AJ2)
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 	
looora Pt
(ohms) (AH)
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
(An) is ohms per degree C for the 10 degree range

MINCO 	 7300 Commerce Lane, Minneapolis, MN 55432
HI-CAL 	 9650 Telstar Ave. El Monte, CA 91731-3093
steepest slope it is 0.025°C Perodegree 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 12.5Q
difference in cable resistance to  two sensors of a AT pair to  cause a 0 1C
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
Shms  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  Ibng
 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

-------
                                                          i    Section No.  4.3. 1
  ;      .                                                      Revision No.     0
                                                              Date:    17 Sep 89
  :                                                        |    Page:     3  of  4

40 gauge (0.08 mm diameter) wire  at 20°C  made of  annealed copper has a
resistance of 3. 4Q/m.   If  it were made  of German  silver  the resistance would be
65.90/m.  Platinum and  iron are each 20.0 0/m while nickel is 15 6  0/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.

oh*™- in AM   ®ecause °f the sma11 resistance change for a 1°C temperature
change (0.4*2), 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.
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.                                             *»y*»«.em
                                                          i
                4.3.1.1.3  Mercury-in-glass thermometers
                                                                              It
.„„• ,  ,     ...   „.     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'
ianS!!oohThey™re 37'9 f (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
To™nLAnKerPOlaUOn m°re accura|e.   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,  i
          -" -           -'  -              -                     !
               ^Therm°COUple  pairs are  wel1  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

Hc5v,aii,,T?Hre*are !  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 sine"  the rangfof'
               temperature  is quite  large,  Signal conditioning circuits may be
               iform  to individual transducers or transducer pairs.  They may
        adjusted to a generic or theoretical curve or transfer function    It is
          to understand the function of the signal conditioner and to treat it
     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            ^or.
where the  shield intake is located is the goal.  A shield protects the sensor
fro. 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 Pfrellel Pit Plae Shield
                                               II
          A£S Steventon Screen
                                           Teledyne Aspirated
                                           Radietian Shield
                               AES Aspireted
                             Stevenson Screen
                          Isreeli Thaller Shield
               TT
              GillNelurtOf
             Ventiltted Shield
                                     Climat Motor Aipireted Shield
                                                         Modified Remo* Shield
AES Mirine Shield
                      AES DuflAtpinted Prrchnmeter


            Curved Pine Shield
                                             EG &G Dew Point
                                            Hygrometer Shield
                             Gill Aspireted Shield
   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 terras 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.
 1                                      •                  !
 i   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 Regula.tory 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)
 !                             i
       The only requirement in EPA (1987a) regarding the vertical temperature
difference is "Errors in measured temperature difference should not exceed
Ol003?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.                                            i
                                                         i

 • o    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.
 i       An inversion,produces a positive AT.             j

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
   C7»
      4 -
      3-
   o
   o
   c
   v  2 H-
    V
    5_
        A
    p  0
    QL
    £
          KEY: 	 MRI AT
                +   MRI-BAO
                            J
                               ^
      -2 -i—i—r
         0     6    12   18   0
                 9/7/82
6    12    18    0
   9/8/82
Figure 4.3.2.1  MRI AT Data, and AT Difference Data for Three Days
                                        12   18   0     6    12   18
                  9/7/82             9/8/82             9/9/82
 Figure 4.3.2.2  Air Temperature at 2 m from the MRI AT Pair
6    12    18    0
  9/9/82

-------
                                                              Section No.  4.3.2
                                                              Revision No.     0
                                                              Date:    17 Sep 89
                                                              Page:     3  of  4
 is the difference between  the  two 20-minuteoaverage AT measurements  from the
 two towers.  Note the dynamic  range of -1.3°C to +4.5°C  (per  6.6m) and the
 agreenent between the two  different instrument systems.  For the three days the
 average difference is -0.028C  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
      Relative accuracy (error)
           -5 to +15 degrees C.
           so.l degrees C."
 1      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
2,  Met One 076
3.   Qualimetrics 8150-A
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.

Radiation error -less than CI.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).

Radiation error - 0.05°C during maximum aspiration
and full sun.  Air speed 360 ft./rain.  (1.8 m/s).

-------
4.  Teledyne Geotech 327C
5.  R. M. Young 43408
                                Section No.  4.3.2
                                Revision No.     0
                                Date:    17 Sep 89
                                Page:     4  of  4


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/m2, 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 th©  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 CIS method.

       4.3.2.2 Temperature

       The accuracy specification for temperature is suggested  in
EPA  (1987b) as
        "Range
        Accuracy (error)
             -40 to +60 degrees C.
            sO.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
siailar 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
»ay 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.

  I  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 nay 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
inversion 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->COLI>* AMBIENT-»HOT-»AMBIENT.
                                                         I
  .  Assume the receiving test will use two temperatures, ambient temperature
and an ice slurry.  Assume a water 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
^o 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.
                                                               "S
    Place the sensors and thernometer into the Thermos.  Use a cork or some
cover to  keep ambient air froa 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 ninutes 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-heat ing
will be snail and the large thermal mass will carry t.he 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
teaperature 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
tenperature 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
reouirement 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.
                                                         j
     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 makins
 the selection.                                                               s

 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
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 iin or  to be supplied by
the calibrator using a decade box or equivalent.  Such procedures are assumed
to have been done and will not be discussed here.  The method recommended in
this handbook is independent of the manufacturer.  It requires only  the generic
transfer function of the resistance element used, that is the ohms vs.
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.

i    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
pontainer 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
                                                       i
    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
jtiae constant goal for the thermal mass.

!A test was conducted to document the performance of the thermal mass assemblies
 and to show the tiae 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 ttl 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
jone square centimeter when the temperature difference is 1°C.  Steel is 0.1,
 rubber is O.0005 and air is 0.00005 (Hodgman, 1955).  The aluminum cylinder
^exchanges heat with its environment (except for the transducers being tested)
   O
       40
              Thermal  Mass Conditioning  Test
                         Meteorological Standards institute
Figure 4.3.5.4
                12    15 ,   18    21    24
              Local Daylight Time
Thermal  mass conditioning and response  test.
through 450 cm  of air,  2.4  cm2 of lucite and 0.1 cm2 of steel.  The difference
in temperature between the holes where the transducers are mounted can be
estimated  from the difference  in conductivity within the aluminum and within
the environment outside  the  aluminum  (99.5%oair), or 10,000:1.  If the mean
temperature of the cylinder  changes at 0.01°C per minute, the gradient change
between hole temperatures could be 0.000001°C per minute.

-------
                                                             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 ti»es 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
theraal 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 thejgain of,the
 difference amplifier.  Pick a common temperature  for the  sitelfrid4substitute a
 fixed resistor for one transducer,  arbitrarily choose  the lower  one.  Assume
 the transducer is a 100JJ platinum type (see Table  4.3.1.1)  and your resister is
 108 ±1D.   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  107.96p.  If 107.79H

-------
                                                              Section No. '4.3.5
                                                              Revision No.    0
                                                              Date:   17  Sep 89
                                                              Page:    5  of  9

 represents 20°C, and if 0.3898 represents a 1°C change  (0.0389Q for 0.1°C
 change or O.01Q = 0.0257°C), the simulated temperature  for both transducers is
 (107.96 - 107.79 = 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
 O.OOO 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.
  i                       .     "     .    - "  ,       • ,    ..''•".
        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
 oh« »eter.  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 0.10Q or 0.26°C.
                                                         ! -  . •
  1      4.3.5.3  Temperature

        The temperature calibration may be achieved cpncurrently 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
 sa»e time the AT calibration is being done.
                                              '-.-'•--:•   !
        It nay 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
 siiipler method.   Liquid baths in a pint or quart insulated bottle with the
 transducer and a good ASTMr or equivalent thermometer mixing the bath together
 will suffice.   Be sure stability.has Keen 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
^cooiuct^drto verify the accuracy of three MINCO 604 ohm iRTDs (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
 teaperature 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.                                   ,i

-------
                                                             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
•inutes, 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  at 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    o
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
C°C)
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
(°0
•-='•• 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 =;^-u-Oi..OOpl4322177
                 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 aniJ 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.
    o
    o
    o
    CO
    
-------
                                                            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.

                          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 ttl 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
Figure 4.3.5.6
 Calibration
      0.5
 O
  o
  c.
      0.4-
      0.3 -
      0.2-
      0.1 -
  2
  
-------
                                                              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 #1 and #3
  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
C!. 02534
-Q. 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 Mined RTDs.  The average and  standard
deviation of the differences  between RTDs is shown in Table 4.3.5.4.
 i                           '            •_..'-.
                  Table 4.3.5.4  Relative difference analysis
Average difference
Standard deviation
	 Minco Sensors (RTDs) 	
#3 - #1
0.2634
0.0183
#3 - «2
0.2753
0.0165
#2 - #1
0.0119
                4.3.5.4.7  Conclusion

                UsingoMinco #1 as a standard of comparison, and adjusting ttl
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 ±0,05°C
                              #2 + 0.010 = temperature ±0.05°C
                              #3 - 0.265 = temperature ±0.05°C
Similar tests in 1984 yielded the following corrections:

                                 0.00  = temperature
SI +
#2 + 0.03
#3 - 0.22
                                       = temperature
                                       = temperature
showing reasonable stability over four years and a reasonable capability to
duplicate relative calibrations.               ,  .
                                                        T '

-------

-------
                                                             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
froa 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 usfe 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.
                                                                      i  - • -
       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
aixed 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
  ;                                                   '    j    Date:   17 Sep 89
  ;                                                           Page:    1  of  8
                                                         I
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 simpily 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                          :

  1              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.
                                                         i
                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
                                                         i

                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. Jieight.  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.  Inplicit 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  HOH

                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
aethod (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.            r  "  '

                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 (version 8/1/84)

     This method provides for a measure of two temperature transducers
     with  electric output used in a differential  application,  often
     called delta T.  The audit equipment includes three thermal mass
     assemblies.  Each is a cylinder of aluminum <6.4 cm diameter by
     17 cm long, 1,341 g in 485 cc>  with holes for different kinds of
     sensors.  This 485 cc mass is suspended in the 1.100 cc inside
     volume of a stainless steel Thermos bottle., which is inside the
     S5OO cc inside volume of an 18.5OO cc cylinder (Gott 2 gallon
     water cooler).  The cold system is filled (4,500 cc) outside the
     Thermos bottle 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 60O watt strip
     heater operating through a 100 F thermostat.  Conditioning for
     the hot and cold masses begins at least 12 hours before planned
     use.   Each mass has a Minco 6O4 ohm (at O C) nickel-iron RTD.
     Resistance is measured and 'converted to temperature using a
     quadratic expression of the Minco Table 14-6O6 (see appendix).
     Relative corrections from intercomparisons made with the three
     RTDs in the same mass are applied yeilding relative accuracies of
     better than O.OS C.  Absolute accuracy is better than O.5 C.

     DLTOO6-A  This method challenges the delta T pair at zero
     difference at three different temperatures.   Place the pair of
     sensors in the hot mass.   Record the mass temperature and the
     sensor temperatures (or the difference if that is all the signal
     conditioner provides)  after about 4O minutes.   Take two more
     samples five to ten minutes apart to verify stability.   Move the
     pair  of sensors to the ambient  mass.   Note:  if DLTOO6-B is to be
     used,  leave the hot mass unplugged and open to the air.   Record
     data  after 4O minutes and again in five to ten minute steps.
     Move  the pair of sensors to the cold mass and  record data after
     4O  minutes and again in five to ten minute steps.

     DLTOO6-B  This method challenges the delta T pair at a small
     temperature difference.   The two masses to use are the ambient
     and the hot,  after the hot has  cooled down from use in DLT006-A
     or has been conditioned by a short input of  heat.   The true
     difference in temperature between the two masses should  be
     between one and  three degrees C. •  Install  one  sensor in  the
     ambient mass (T.I), and. one sensor in  the "hot" mass (T 2).   After
     4O minutes record  the temperatures of  the two  masses and the
     sensors or sensor  difference if that  is all  that is available.
    Record  the data  again  after five to ten minutes.   The true
    difference will  change slowly as the  masses  change temperature.
    Reverse  the sensors and repeat  the method  waiting  4O minutes for
    the first  data point.   If the time constant  of thcj sensor is
    observed to be fast enough  to assure  stability in  less than 40
    minutes, a shorter  period may be used.
Figure 4.3.7.1  Audit method for temperature difference

-------
                                                        Section No. 4.3.7
                                                        Revision No.    0
                                                        Date:   17 Sep 89
                                                        Page:    4  Of  8
               PERFORMANCE AUDIT REPORT by
                                                                 DLT006
     MEASUREMENT SYSTEM -  Temperature  dif-ferenca
          Systea number	„	
          Sensor
          Transducers	
          Locations     •	...
          Si gnal^condi ti oner
          Data channels	„__
     DATE 	/.
      TIME  off  line
                                         on  line 	
                                                        test start
     TRANSFER FUNCTION  (volts per  degree C)  _
                        Conversion  formula	
     TEST RESULTS
       DL.TOO6-A
     cold    sensor  #      1             2	
               •ass
     time    ohms   degC
                output
              volt   degC
                output
              volt   degC
                  1-2
                  dif.
              volt   degC
     average
     ambient   sensor #
                mass
     time   ohms   degC
                output
              volt   degC
                output
              volt   degC
                  1-2
                  dif.
              volt   degC
      average

      hot

      time
   sensor #
   mass
ohms   degC
 1	
  output
volt   degC
 2	
  output
volt   degC
    1-2
    dif.
volt   degC
      a vet-age                   '                               	

        DLT006-B                                        -       -      '
                 T 1       T 2   A       T  1       ,T..2.  •  . B.  .  b-A
      sensor # 1	2	 1-2     1	    2_^___  -.1-2    ..
               mass      mass   di-f.    output    output   dif.  dif.
      time  ohms degC ohms degC degC   volt  degC volt degC degC  degC
      average                                                    ———

                 T 2        T 1    A       T 2       T 1  '   -B    B-A
      sensor #  1	2	 1™2    1	_    2	     1-2
                mass       mass   dif.    output    output    dif.  dit.
      time  ohms degC  ohms degC degC  volt degC vqlt degC  degC  degC
      average •
Figure 4.3.7.2  Audit  form for the temperature difference method

-------
                                                              Section No.  4.3.7
                                                              Revision No.     0
                                                              Date:    17 Sep 89
                                                          i    Page:     5  of  8

                 The  transducers are challenged with a knovm 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, O.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
 •ethod.  This method is described in 4.3.5.   It  is  possible to use  thermal
 •asses 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  prpblems associated
 with mounting parallel instruments are large compared to the value  of the
 •ethod.  Considerable,  but much less,  effort is  required for the Sensor Control
 •ethod which provides  numbers with acceptable  confidence in their accuracy.

       4.3.7.3  Temperature

                 4.3.7.3.1 Sensor  control method
                                                          i "
                 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
 coapared 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.
                                                          I

       4.7.3.2  CTS Method                                |  :

       It is-both-practical  and recommended  to use the CTS method for
teaperature 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 located 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
teaperature system   provides no
information about other temperatures,.   It
may be that the operator calibrated the
systen 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
with a CTS  is useful'.
  Figure 4.3.7.3
Assmann psychrometer

-------
                                                          Section No. 4.3.7
                                                          Revision No.    0
                                                          Date:   17 Sep .89
                                                          Page:    7  of  8
             Temperature - MSI method TEM005  (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 temperaturess  The audit
        equipment includes three thermal mass assemblies.  Each is a
        cylinder of aluminum (6.4 cm diameter by 17 cm long, 1,341 g in
        485 cc) with holes for different kinds of sensors.  This 485 cc
        mass is suspended in the l,lOO.cc inside volume of a stainless
        steel Thermos bottle, which is inside the 850O cc inside volume
        of an 1B.50O cc cylinder (Bott 2 gallon water ;cooler).  The cold
        system is filled (4.5OO cc) outside the Thermos bottle 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 6OO watt strip heater operating throuah a
        1OO F thermostat.  Conditioning for the hot and cold masses"
        begins at least 12 hours before planned use.  Each mass has a
        Minco 6O4 ohm (at O 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
        intei-comparisons made with the three RTDs in the same mass are
        applied yeilding relative accuracies of better than O.OS C.
        Absolute accuracy is better than 0.5 C.
                                                      I
        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.  I

        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.
                                                               .
        TEMOO5-B  Use an  Assmann  aspirated psychrometer  mounted in the
        vacinity of  the  shielded  temperature sensor.   Wind the Assroann
        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		     TEMOOS
   MEASUREMENT  SYSTEM - Temperature
        System  number	
        Sensor         __„--..—	
        Transducer     	
        Location
        Signal  conditioner
        Data channel   	
         	/	/	 TIME o-F-f line	 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
that to a controlled environment such as a wet or dry thermal mass.  The
teaperature 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.
 " "	' • "'  >.---••	- -- -   . .-	•— -  ......		 	 „..,....._	,_...,,,„, _j.   ,,	 -		-

    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.
    iObserv.  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.
    i                                                       i

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.
    i 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
    ^onf. 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
                                                           I
McKay,  D. J.  and J.  D. McTaggert-Cowan, 1977:   An intercomparison of radiation
    shields for  auto stations.  World Meteorological Organization Publication
    ;No.  48O,  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.
                                                          I
 4.4.1   TYPES OF INSTRUMENTS                              \

    Precipitation collectors are of two basic types:   nonrecording and
 recording!

        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
 •ade 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).

 i      4.4.1.2 Record ing  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 t,o NWS  Specification No.  450.2201,  is known as the Universal
gage,  indicating u,sage 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).      \

-------
       4.4.1.3  Instrument Characteristics
                                                              Section No. 4.4.0
                                                              Revision No.    0
                                                              Date:   17 Sep 89
                                                              Page     3  of  7
       The  Host 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.
                                                          I

   :    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 at 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
 wirds on  the balance  mechanism,  weighing gages are fitted;with a damper
 imaersed  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 (O'.Ql inches  of rainfall  per
 bucket).   Rain  rates  are calculated from an  event recorder with pens energized
 sequentially to improve resolution.   The tipping bucket (ai 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/rir 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      I

       Accuracy of measurement for  all  types of  gages  is  influenced perhaps
 mora by exposure than by variations  in design.  Windshields  represent  an
 essential accessory to  improve the  catch of  precipitation, especially,  snow  in
 wicdy 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
 ana 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
otrsr  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  aore 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 drua type can be obtained with monthly rather than
 weekly mechanisms.   Unless  the  tipping bucket gage is-f equipped with a heater,
 it is  of  no use for frozen  precipitation.

        4.4.2.3  Acceptance  Testing

        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
 mechanism assembly is balanced and that the switch is
       4.4.2.4 Calibration
    that the bucket
operational.
       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.

 :                         (48.51/7t)1/2 = 3.93  in. radius

 Ivor 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.
 i                       ' •                        p         i     ' .
 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.  Nonrecordihg
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 3 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
 CTS 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.
i                                             ' ..     •
NCAR,  1984:  Instructor* s Handbook on Meteorological Instrumentation, F. V. '
    Brock, Editor. NCAR Technical Note, NCAR/TN-237-HA.

Weiss,L .L., 1961:  Relative catches of snow in shielded and unshielded gages
    at different wind speeds. Monthly Weather Review, Vol. 89.

   . 1971:  Guide to meteorological instrument and observing practices. World
    Meteorological Organization Ho. 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 TO INT TEMPERATURE

4.5.0  INTRODUCTION                                    ;

       Humidity is a general term for  the water-vapor content of air.  Other,
»ore 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
                                  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
fro» 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
theraometers.  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
  _.     . _ _   .           ,         • the moisture-absorbing sensor.  Other
  Figure 4.5.2  Assmann psychro-      .  .      .    ...          ..u 4.       -.4.
     .    ...       .,      .••.           instruments with sensors that respond to
   sster with parallax guides.          .          ,     ......    ,.    .   ,
              ^                       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 »bst 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 2O 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
hunan 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 aanufacturers.  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
ataospherically 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-tor-
dew-point relationship.             ,,

       In the early I960's, the technique of detecting
the dew point on a cooled mirror surface evolved into a
product ion-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
neteorological type, thermoelectric, cooled-mirror dew
point instrument (Mazzarella, 1977).  Three of these
instruments cover the range of -50 to +50 G.  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
                           d
and T (ambient
temperature).
,(dew point  temperature)
 Two of  the  manufacturers
Figure 4.5.3  A
 typical Dewcel
 sensor housing
and transmitier

-------
                                                             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-rairror 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
•oisture 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
conplaint 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 0.59C 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 sl-ing
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 me.as-
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 ^s  estimated: th^t, the
accuracy  of the LiCl saturated salt technique  is 1.5°Ceover the range-ofj^sb
 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
 airror 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  1:o  the  environmental
 conditions  in which the measurements are to be made.  In most  cases, psychro-
 neters 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
I 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.
'                        '            '                    i
       4.5.1.3   Data  Requirements    .

       Electrical hygrometers for  monitoring applications  have time constants
1 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-
    Figure 4.5.4 A  pair of  tower-      dent of air temperature.   Relative
    mounted Gill aspirated radiation   humidity calculations can  therefore be
    shields for housing temperature    made given  the dew  point  temperature  and
    and dew point sensors (Young).      any temperature  measurement point in  the
; 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.                                                  I

-------
                                                             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 (TD>  °C),,  do the following:
                    exp
                                                            (1)
                     where a =  17.27
                           b » 237.3
       To calculate the dew point temperature (TD>  °C) from air temperature (T,
°C) and relative humidity (RH = 100 r, %) use
                               b  In r +
                                           aT
                                          b + T
                             a -  In r
                                           aT
                                          b + T
                                                            (2)
       To calculate relative humidity (RH = 100 r, %) from air temperature
(T, °C),  wet-bulb temperature (TH> °C), and atmospheric pressure  (P, mb)
through the vapor pressure  (e, mb) and the saturation vapor pressure (eg, mb),
do the following:
                r =
                                       +BTH)(T-TH)
                                                           (3)
                              eoexp
      where A = 6.6  x 10
            B = 1.15 x 10
                         -4
-3
           SW
                    exp
                            aT
                           b + T.
       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 + TD
P + T
19
19
+
+
130r -
130r -
28r
28r
2
2
) '
JJ
                                                            (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 his 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.
 i
        Both the dew cell and the cooled-mirror type instruments have applica-
 tions on lO-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:   "...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 O.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.
                                                        i

        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

       This 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
•anufacturer's operating manual for the instrument.

4.S.3  OPERATIONS

       4.5.3.1   Installation

       Dew point measuring equipment  on a tower should be installed with the
saiae considerations given to temperature sensors.  Reference has already been
•ade to the weatherhood as a shield for the dew cell and to an aspirated shield
for the cooled-rairror instrument.  At some instal la-tions „ -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
*ounting  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 Lid 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.O
                                                             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 ah Assmann-type, motor-
operated psychrometer, providing the instrument is allowed to assume near
aabient conditions prior to use.

       The cooled-mirror instruments require no calibration except  for the
•inor te»perature 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.
   i
4.5.4  PERFORMANCE AUDIT METHODS
   i
   i    Instrument audit procedures for hygrometry systems follow calibration
procedures.  A systems audit should be performed near the beginning of a field
•easurement 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.
1984 and 1985 may be of selective value.
ASTM, 1982, 1983,

-------
                                                              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 (Dew 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 Wet- 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.  s "      \ ,'

Huschke, R.,ed., 1959:  The Glossary of Meteorology. American  Meteorological
    Society, Boston, MA.

Mazzarella, D.  A., 1972:  Meteorological instruments: their selection andA[,use
    in air pollution studies. Proceedings of the Meeting on  Education and :.,
    Training in Meteorological Aspects of Atmospheric  Pollution and Reflated
    Environmental Problems. World  Meteorological Organization, No." 493.'.  .

Middleton,  W. E.;:K.,, arid A. F. Spilhaus, 1953:  Meteorological Instruments;,, .
    University of Toronto Press.                                 ,  •    ,fs', '•'. ">
                                                       •  •  •.••-•-•;  '  ;   h  n> -("-H
Quinn, F.  C. , 1963:  Humidity-the neglected parameter. Testing, ^Engineering.        ^
    The Mattingly Publishing Company,  Inc.                            '._'.', .'>(.,       HH|

: Suntola ai*d Antsoh, 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,
 ^ndeed, 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 •mln.  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 (jim)  and is subdivided as follows:
                      Ultraviolet  (UV)
                       Visible
                       Near-infrared
       0.20 urn to 0.38 /Jim
        0.38 jim to 0.75 /im
        0.75 ion to 4.00 jzm
    Longwave radiation has a wavelength as follows:   '  !

                      Infrared (IR)     4    jum to   100 jjm

and comes from  the Earth and its atmpsphere.   The instruments most commonly
used for environmental  monitoring are discussed below.

4.6.1  TYPES OF INSTRUMENTS

       4.6.1.1  Pyranometers          :         -

       Pyranometers are instruments that measure the solair radiation received
froa the hemispherical  part of
the atmosphere  it sees,      !
including the total sun and
sky shortwave radiation oh a
horizontal surface (Figure
4.6.1).  Most pyranometers
incorporate a thermopile as r ;  •   ?
sensor.  Some use a silicon
photovoltaic ceil 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  f   ;
surface in defined
           Sensing
            Element
     .GuardDisc ,
         '
Precis/on Ground &
Polished Glass Dome
(Frequently Double)
                                        Leveling
                                        Screw
wavelengths.  This is achieved
by substituting one of several
Figure 4.6.1t .features,of a. typical
 >'  '    "   I,' "pyranometer (Carter, et al.,l$77)
  •• •''''  -  '           >•-  •    '    -j-..

-------
                                                             Section No.  4.6.0
                                                             Revision No.     0
                                                             Date:   17 Sep 89
                                                             Page:     2  of 8


colored Schott glass filter dones for the clear glass outer dome.  The smaller
done 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
nane 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
   Pyrheliometer
Declination Adjustment

       24 hr. Dial
       4.6.1.5 Instrument Characteristics               i       ;

       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 pynanometer is not to be confused with the
pyrheliometer, "an instrument for measuring the intensity of direct solar
                                         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 character list ic 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-                        f
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 jinu  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 of ten 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 uV per W/mz, 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 recording 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
unUsTTe., W/m2, cal/cm2.min, etc.  The alternative is a standard millivolt-
m^ter potentiometrie 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 used
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
counts 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
bTpurcnasing  sensors  of  Class 2 or Class  3 as identified by the WMO  (see  Table
4.6.1).                                                          '•.-.,•'•'

       Table 4.6.1 Classification  of Pyranometers According to
                        Physical Response  Characteristics
Sens.
(mW/cm2)
1st Class ±0.1
2nd Class ±0.5-
3rd Class ±1.0
Temp.
( '/. )
± 1
± 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 WMO 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 pyranome.ters, 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 pyranoaeter  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.
                                                         'I
           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 eassy 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                            . i
                                      •           ••        i  -,
                                     ' ^  '         ~        r      -   ,••-'••'•'
       The site selected for an upward-looking  pyrahometer  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.).
                                                         .1  "   ' '

       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 • 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
Holl-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
         w  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
 iMBune 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
                                                           1   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.
    I             •                                 •
       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 Bade 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.
                                                           i  •    •  ..-:,.:..
 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.                                       J
                                                         ".' -1!"  '  ,,''./  ,' i  .'.'•*  .  ' j •
 Carter,  E. A.  etal.,  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 Wo. 2,
                  Field Year  for the  Great Lakes,  Canadian National Commission
    for the Hydrological  Decade.

           I  L    1972-  Survey of instruments for micrometeorology.
    inernitio^l 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
    Hartlook,  Environmental Series. Department of the Army,  Material Command.

WHO   1971-   Guide to meteorological  instrument and observing practices.  World
    'Meteorological  Organization No.  8TP3, 4th edition, Geneva,  Switzerland.
                        t>U.S. GOVEMWENT PRINTING OFFICE:1994-550-001/80371

-------
     QUALITY ASSURANCE FOR ATMOSPHERIC PRESSURE MEASUREMENTS

 4.7.0   INTRODUCTION
   i
        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
 interpretation.  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.                                       I

 4.7.1  TYPES OF INSTRUMENTS
                        '•         '              .                  r •          • -

       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                                        i
                                                                 l
   ;                         '                                 '
       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:
                                                                 i

   i                                    1013.25 hPa                !
                                29.9213 in Hg at 273.15 K
                                      14.6959 Ib in'2
                                                                 I
   ;                                                              i
       Any practical application will be well served by a pressure measurement accuracy of about
10 hPa (= 1% or 100 m in elevation). The best accuracy one can expect to achieve in a monitoring
application is about 0.5 hPa.
                                          7-1

-------
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.

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 normal applications. The accuracy for this small instrument,
when properly used, is 0.2 % or about 2 hPa.
                       Figure 4.7.1 Engineer's altimeter
                       (Weathertronics).
 4.7.5   REFERENCES

 ASTM, 1977:  Standard Methods for Measuring Surface Atmospheric Pressure, D3631-84,
        'American Society for Testing and Materials, Philadelphia, PA.
                                           7-2

-------
      QUALITY ASSURANCE FOR GROUND-BASED REMOTE SENSING DEVICES
  4.8.0  OUTLINE AND SUMMARY

        It has  been common practice  in air quality studies  to use towers  as  platforms  for
  meteorological monitoring.  Such towers range in height from 10 to over 150 m and are typically
  outfitted with in situ sensors at several levels. Often, these towers fall short of the effective stack
  height (ESH) or the height of interest of an air quality study, and thus various techniques must be
  used to estimate meteorological conditions above the tower.  Such techniques are not always  the
  most realistic methods for estimating the vertical structure of the boundary layer. Tall towers are also
  expensive to install and maintain, and commonly, logistical constraints preclude proper siting of such
  a structure.  Meteorological remote sensing provides an alternative to tower based measurements.
    ;                                 -                              i
        Over the past few years, developments in remote sensing technology have made it possible
 to obtain three-dimensional wind velocity (u, v, w) and virtual air temperature (Tv) profiles with the
 precision and accuracy suitable for regulatory applications.  There are three types of commercially
 available remote sensors: SODAR (SOund  Detection And Ranging) which uses acoustic pulses to
 measure horizontal and vertical wind profiles as well as the height above ground of the elevated
 inversion layer and mixed layer; radar (Radio Detection And Ranging) which uses electromagnetic
 (EM) pulses to measure horizontal and vertical winds; and radar/RASS (Radio Acoustic Sounding
 System) which uses both acoustic and EM waves to measure virtual air temperature, wind speed and
 wind direction profiles. Each will be described with detail in the following sections.

    ;    It is important that the user understand  the fundamental differences between remote profiler
 measurements  and in  situ measurements.  In situ  sensors, including temperature probes  cup
 anemometers and wind vanes, are the mainstay of meteorological monitoring.  They are found on
 towers, buildings, bridges or other structures and measure a particular meteorological variable by
 direct contact. However, by their very nature, these sensors disturb the environment in which they
 are sampling. These sensors are easily characterized in wind tunnels or environmental chambers and
 provide the user with a point estimate of the variable in question.

    :   Meteorological remote sensing devices, on the other hand, provide measurements without
 disturbing the environment.  In addition, remote sensing measurements are not restricted in height
 to the extent that m situ, tower-based measurements are.  More importantly, data obtained from a
 remote sensor is represented as a spatial, or more specifically,  a volume average  This particular
 characteristic of these data is described in later sections.

    .   When comparing in  situ sensors  with profiling systems, we must acknowledge various
tradeoffs between accuracy  and the  capability of characterizing the atmosphere   Conventional
methods for obtaining upper air measurements have included aircraft, rockets, tetherballoons  and
rawinsondes.  Of these, rawinsondes released twice per day at some National Weather Service (NWS)
stations have been the principal means for routine upper air observations. Rawinsondes provide point
                                           8-1

-------
measurements of wind direction, wind speed, temperature, and dew point at intervals of about 100
meters   Morning and afternoon  mixing heights are estimated from  the twice daily vertical
Smpe aWfifes using the recommended standard method of Holzworth (1964; 1972). In this
2od  the nixing height is calculated as the height above ground of the intersection of dry adiabat P
tension of the surface temperature with the vertical temperature .profile. Various techniques are
S u ed to interpolate hourly mixing heights between the two rawmsonde  observations. This is no
dways a reliable method since the atmosphere can be extremely variable  especially on time scales
ofjuJt: a few hours. Remote sensors provide indications of the mixing height based on indications of
SableTyer aloft, operate continuously and produce spatially averaged observations that are more
capable of accurately characterizing the atmospheric boundary layer.

       Remote sensors provide profile information based on time averaged observations that are
fixed with height. Instantaneous values acquired by remote sensors may have errors associated with
themTe toTandom interference by sources  such as bugs,  birds or low flying aircraft  These
erroneous values are removed by sophisticated algorithms and then averages are generated _ The e
averages are usually computed for time periods of 15 minutes to one hour, depending uponto data
requifements of a particular field study.  These observations are also spatially averaged due to the
lame sample volume involved.  This type  of averaging over time and space characterizes the
 aTmospheTemore precisely than those values that are interpolated from a data set acquired by twice
 a day rawinsonde launches.

        The following sections in this chapter describe the theory of operation of the various types
 of profiling systems that are commercially available, with an emphasis on system specifications^
 Section Mow on installation procedures and acceptance testing techniques to assure that acqu red
 data ierdSe and representative of atmospheric conditions.  The inherent problems of calibration
 ±edures a,id^performance  audits are discussed in detail.  Standard operating procedures,
 maintenance schedules and quality control issues are outlined in the final sections.
 4.8.1  TYPES OF INSTRUMENTS
        Ground-based meteorological remote sensors have been designed to measure vertical profiles
  of wind velocity and virtual air temperature, as well as the height of the elevated inversion layer.  The
  devebpment aid evolution of these devices over the last several decades have flowed ^two simdar
  but distinct paths: One based on acoustics and the other on electromagnetic (EM) radiation. Wind
  ve och es acquired by sodar are based on the atmospheric effects on the propagation  of acous c
  Inergy  while radars are based on the atmospheric effects on the propagation of electromagnetic
  eS  Rofites of virtual air temperature are obtained by RASS which combines both acoustic and
  EMSchnolodes  Table 4.8.1 provides a summary of typical specifications for the three major types
  ofleTeorolSS remote sensing devices. The theories of operation of all three profiling  systems are
  discussed below.
                                             8-2

-------
  4.8.1.1  SODAR
    • i                                                                I
                                                                           t
        In the late 1960s and early 1970s, remote sensing techniques focused on the development of
  an acoustic-based wind profiling system, commonly known today as a sodar (e g  Beran et al  1971 •
  Beran and Clifford, 1972; Beran, 1975; Balser et al., 1976; Kaimal and Haugen, 1975; Brown and
  Hall,  1978).  The principle of operation is actually quite simple. The mono-static sodar consists of
  a transceiver antenna which is used to transmit and receive acoustic signals. The transducer generates
  a pulse of acoustic energy that is released into the atmosphere, either vertically or at some angle from
  the vertical. As the acoustic wave propagates upward, differences in atmospheric temperature and
  density cause some energy to be  scattered back to the surface. This returned energy is received by
  the antenna and the frequency of the signal is  determined. The difference between the transmitted
  and received frequencies, known as a Doppler shift, is directly proportional to the wind velocity alone
 the beam axis.

        The earliest sodars consisted of a single,  vertically oriented, transceiver antenna approximately
  1.5 m in diameter. The received signal intensity  was recorded on facsimile paper.  This system
 provided the user with qualitative information on the structure of the atmospheric boundary layer to
 heights of up to 1 Km.  Since only one antenna was used to measure the vertical'boundary layer
 structure, the system was termed as operating in a mono-static mode. In this configuration, sodar
 signals are scattered primarily by temperature gradients (Neff, 1988).

        The next step of sodar development lead to a bi-static configuration which uses two antennas.
 One Antenna acts as the  transmitter and is tilted from the vertical, typically, about 30°.  The other
 antenna, which acts as a receiver, is situated away from but tilted toward the transmitting antenna
 Bi-static sodars  obtain wind velocity profiles by measuring the scattering of acoustic signals due to
 temperature and wind velocity fluctuations.  Figure 4.8.1 depicts the mono-static and bi-static
 configurations.                                   ..                   '.     _.

        Further refinement led to the development of three-beam and five-beam sodar systems.  In  a
 three-beam configuration, one antenna is pointed vertically and is used  to measure the vertical wind
 velocity (w). Two other antennas,  which are usually oriented off vertical and at right angles to each
 other,  are used to estimate horizontal components (u and v) of the wind velocity. The five-beam
 configuration is similar to that of the three-beam,  except that two additional antennas are used.  These
 extra^antennas are also oriented off vertical and are at right angles to each other. This configuration
 adds some redundancy, and in theory provides the user with a more reliable estimate of the horizontal
 and vertical wind velocity.

       The phased array Doppler sodar is the latest design of an acoustic profiling system (Figure
4.8.2).  An array of vertically pointing transceivers, in some instances horizontally with "bounce
boards" to direct the beams to the vertical, is utilized.  The number of transducers may range from
20 to over 100, depending upon the requirements of the system.  These transducers are sequenced
slightly out  of phase to electronically "steer" the acoustic beam away from the vertical, thereby
obtaining information required to estimate the horizontal wind velocity.
                                           8-3

-------
       The horizontal components of the wind velocity are measured by releasing two acoustic    -^
signals into the atmosphere at an angle, typically 15 ° to 30°, off the vertical axis. The two acoustic   <]j||j}
signals are typically oriented at right angles to one another. One is usually directed toward the East
or West so that the u component of the horizontal wind velocity can be determined while the other
is directed toward  the North or South for  the v component.  Figure 4.8.3  depicts this beam
configuration. The mean horizontal and vertical wind velocity components can be approximated
using the following simplified equations:
u -
                                         cos a  cos
                                                                                     (1)
                                            v
                                       —     rw>
                                       •w
                                           cos<|>
where Vr is the measured radial wind velocity (m s'1) for the u axis, Vm is the radial velocity in the
vertical,  is the elevation angle of the transmitted beam (degrees), and a  is the azimuth angle
(degrees).  Normally in calculating the mean wind, time averaging is used to eliminate the effect of
variations in the vertical velocity. Some systems correct for mean vertical wind if other than Oms"1.
This is useful in situations where the average vertical wind may not be zero (i.e., in complex terrain).

       Sodars typically operate at frequencies from below 1 KHz to just over 4 KHz with typical
power outputs in the range of 2 to 300 W. In operating  or purchasing a sodar for a particular
application, one should note that the lower the frequency and the higher the power output,  the greater
the range of the sodar. Therefore, sodars can be tuned to obtain the most sound information for a
particular application.

       The vertical range of a typical sodar is 0.5 to 1.5 Km and is a function of frequency, power
output, and atmospheric stability.  The most important factor  affecting range is the presence of
atmospheric turbulence, especially eddies on the scale of 0.1 to 0.3  m since these are the major source
of reflectivity for acoustic waves.  Range resolution  is the distance between reported heights and is
typically between 25 and 50 m. In a typical configuration about 20 to 30 levels are reported.

       A mini phased-array sodar (mini-sodar) is a downsized version of its standard counterpart and
has a height coverage of about 200 m and a range resolution between 5 and 20 m.  The mini-sodar
provides measurements near the surface and is useful for studying local flows and shallow inversions.
Figure 4.8.4 shows a photograph of a mini-phased array sodar.

       Sodar signals are shaped somewhat like a cone as shown in figure 4.8.5,  The half-power
beam width typically ranges from 2° to 10° depending on frequency and the size of the antenna
aperture. The sampling volume increases with height and can be approximated by:.
                                            8-4

-------
                              Sample Volume (m3) - it L (h tana)2
(3)
 where L is the length of the transmitted pulse (m), a is the beam half width (degrees), and h is the
 height of the sample volume (m). For example, if the transmitted pulse length is 20 m, the beam half
 width is 2° and the height of interest is 500 m, then the sample volume would be near 19,000 m3.

 4.8.1.2  RADAR

   !     The principles behind the Doppler radar are similar to sodar except radars use electromagnetic
 (EM) waves to sense turbulent fluctuations in the atmosphere. Because EM signals do not attenuate
 (dissipate) as quickly as sound waves, radars have greater vertical range than  sodar.

   :     The original Doppler radars operated at frequencies that required fairly large reflectors (i.e.,
 water vapor, bugs, or chaff) to reflect the EM signal back to the  receiving antenna. In the 1980s^
 radars were developed which used small scale wind and temperature fluctuations as the source of
 reflection of EM signals (Ye et. al., 1993 and Brown and Hall, 1978).  These so called "clear air"
 radars provided a means to acquire winds aloft without the requirement for large diameter scattering
 mechanisms. Doppler radars today operate at a typical frequency of 915 MHz, and are capable of
 measuring winds to around 3  Km with range resolutions ranging from 60 to  400 m.  Typical
 configurations designed for atmospheric boundary layer studies are capable of measuring to 3 Km
 with a vertical range resolution of 60 to 200 m, thus allowing about 20 to 30 levels to be reported.
 The lowest measurement height (minimum range) is typically between 150 and 200 m.  The most
 important factor effecting the range is the presence of gradients  in the refractive index of the
 atmosphere.  Figures 4.8.3 and 4.8.5 provide a graphical representation of typical Doppler radar.

       Doppler radars emit pulses of EM energy  into the atmosphere. As the EM waves propagate
 outward, some energy  is reflected back to the surface due to the  presence of atmospheric density
 gradients and variations in the refractive index.  The  most common  source of variation in the
 refractive index is the presence of humidity gradients: High humidity in the boundary layer provides
 an ideal environment for the radar to reflect its signal.  In general, the more humid the atmosphere,
 the better the data capture efficiency and greater the range.

       To determine the three-dimensional wind velocity,  three independent EM signals must be
 collected and analyzed. A burst of EM energy is released in the vertical to derive the vertical wind
 velpcity,  Two  other separate  EM bursts, released at  angles from the vertical, are required to
 determine the two horizontal wind components. In the past, three separate antennas were used to
 derive the three components of wind. One antenna pointed vertically and two other antennas tilted
 15° to 30° off vertical and normally at an angle of 90° from one another.  New phased array radars
 are now being manufactured which use one antenna for determining the three components of wind.
The phased array system  has the capability to electronically "steer" the energy pulse away from
vertical,  thereby providing the off axis information required to  determine the  horizontal wind
                                           8-5

-------
components.

       The transmitted half-power beam width of radar is larger than that of the sodar, ranging from
7° to 10°. Equation 4.8.2 may also be used to estimate the size of the sample volume for radar. For
example, if the transmitted pulse-length is 60 m, the beam half width is 4° and the height of interest
is 500 m, then the sample volume is nearly 230,000 m3.

4.8.1.3  RASS

       A Radio Acoustic Sounding System (RASS) is a combination of sodar and radar technology
and is used to obtain profiles of virtual air temperature (Tv), along with the radar's wind profiles (see
Figure 4.8.6). Virtual temperature is the temperature dry air must have to equal the density of moist
air at the same pressure (Stull, 1988).  An acoustic source is added to a Doppler radar, or a sodar
added to a bistatic radar, and used as  a reflective source for back scattering the EM signal. The
variations in temperature produced by the compression and expansion phases of the acoustic wave
provides a refractive index structure from which EM waves can scatter (Gaynor et al., 1993; Neff,
1988).

       The RASS transmits an acoustic pulse vertically into the atmosphere followed by an EM
pulse. Note that some RASS systems use continuous acoustic transmissions to provide the radar with
a well defined reflective source.  Since the EM wave travels much faster than the acoustic wave, the
latter signal intercepts the former, and some EM signal is reflected back to the surface. The returned
EM signal is analyzed to determine the speed of the acoustic pulse, derived from the Doppler shifted
EM signal. The acoustic wave travels at the speed of sound, since this is a function of the ratio of
specific heats, pressure and density pf the air mass, it becomes a relatively simple exercise to derive
an estimate for the virtual air temperature.  The minimum recording height and range resolution of
the temperature measurement are the same as radar, however the height coverage is similar to that
of a sodar, typically 1 Km.

                                        Table 4.8.1
                   Typical Specifications for Meteorological Remote Sensors
System
mini-Sodar
Sodar
Radar
RASS
Variables
u,v, w
U, V, W, Zj
u, v, w
T
*v
Frequency
3-4KHz
1-3 KHz
915MHz
2 KHz
Height
< 0.3 Km
<2Km
<4Km

-------
                                                               Turbulent
                                                                 Layer
 Transceiver


 Mono-Static

   !  Mode
Transmitter
Receiver
                       Bi-static

                          Mode
Figure 4.8.1 Depiction of a mono-static and bi-static sodar.
                                    8-7'

-------
Figure 4.8.2 Photograph of a phased array sodar (Courtesy of Remtech, Inc.),
                                           8-8


-------
                     West   15<
 West
            South
15o    South    Measurement
                    Heights
                                          Side View;
                                                     Surface
Figure 4.8.3 Typical beam configuration for a phased array sbdar and radar.
                                  8-9

-------
Figure 4.8.4 Photograph of a phased array mini-sodar (Courtesy of AeroVironment, Inc.).
                                           8-1,0

-------
Figure 4.8.5 Shape and important components of a sodar and radar beam.
                                       8-11

-------
Figure 4.8.6 Photograph of a radar wind profiler with RASS (Courtesy of Radian Corp.).
                                         8-12

-------
 4.8.2  SPECIFICATIONS
   !
   '     The previous section described the basics of remote sensing devices for meteorological
 monitoring.  Meteorological remote sensing devices, by their very nature, must be configured to
 obtain the most reliable data possible for a given field site. Configuration may include modification
 of the profiler output signal frequency, output signal power, averaging intervals or sampling heights.
 The overall accuracy of an acquired data base is dependent, in part, on the surrounding terrain, nearby
 buildings,  atmospheric  stability, noise sources,  insects and birds.   When compiling  a  set  of
 specifications for the purchase of a remote sensing device, it is important to determine site specific
 information -that will aid the manufacturer in configuring the device to fit the user's  needs.  The
 following sections describe.site specific parameters which need to be identified and provide some
 initial estimates of expected accuracy, precision and data capture efficiency.

 4.8.2.1  SODAR

        The specification for vertical  range will normally determine the appropriate operating
 frequency.  For example, if a user only requires low level winds (< 200 m), then a higher frequency
 (3 to 4 KHz) may be used.  High frequency signals emit little energy in their side lobes and have a
 narrower beam width, thus producing a relatively small sample volume, see Figure 4.8.7. Although
 this provides a relatively cleaner signal, it does have a drawback. High frequency signals attenuate
 faster in the atmosphere then low frequencies. Therefore, more power is required to obtain the same
 vertical range. In situations requiring winds above 200 hi, lower signal frequencies (1 to 3 KHz)
 should be used.  Since low frequencies attenuate more slowly in the atmosphere, less energy is
 required to observe high level winds. The drawbacks with low frequency beams are that they emit
 more energy in .their.side lobes and have a wider beam width, thereby producing a larger sample
 volume with, an increased possibility of generating false echoes;   '      \  ' .

        Sodar signals are shaped somewhat like a cone as shown in Figure 4.8.5.  The beam width
' typically ranges from 2° to 15° depending on frequency. The sampling volume increases with height
 and can be approximated by Equation 3.                          .    I

        Sodars are not usually configured for measurement of the structure of the elevated inversion
 layer above 1.5 Km, due to the enormous power requirements needed to probe the atmosphere to
 these heights'.  For regulatory modeling, atmospheric dispersion models used to derive pollutant
 concentrations and the site climatology will usually dictate whether inversion heights above 1.5 Km
 are required. ,If the user does not require wind velocity information above 1.5 Km, then a sodar
 configured with a low frequency and high output power should be  adequate. Note that radar should
 be considered if wind profiles are needed above 1 Km.          .

   :'   Sodar options usually include  software subroutines that  perform a .variety  of  QA/QC
 functions.   It is important to purchase QA/QC software which provides an. extra.'level of data
 validation.  Care should be taken however, so; as  not to filter out any valid meteorological data.
 Inversion height routines are required if estimates of this level are to be reported.  Software is also
                                           8-13

-------
available for estimating the vertical and horizontal turbulence parameters, ow and ae. However, care
must be taken with how these values are generated since they usually have large errors associated
with them and therefore, are not recommended, for use in regulatory applications at this time. If they
are used, great care must be taken to ensure.that these values are accurate and meaningful.  An effort
is currently underway to investigate the suitability of using sodar derived ow and oe estimates in
regulatory modeling applications.

       Some manufacturers provide routines to correct the horizontal winds for vertical velocity.
In near flat terrain this is usually not a problem unless the system is not perfectly level. However, in
complex terrain the average vertical wind velocity may be large and should be used^to correct the
horizontal winds, if the desired output is the total vector wind speed. During the acceptance test,
discussed later, the wind speed from the sodar should be calculated in the same manner as the test
instrument. During actual monitoring, the operator needs to be careful to supply the wind expected
by the model (i.e., vector or sealer). For example, if a uvw anemometer attached to a  tower is being
used, then both the sodar and the uvw anemometer derived winds should be corrected for the mean
vertical velocity if other than 0.0  ms"1. Information concerning this correction may be found in most
model documentation.

        Gaynor et al. (1992) and Finkelstein et al. (1986) have determined the accuracy of wind speed
estimates generated by sodars to be about 0.2 m s'1, for atmospheric conditions experienced during
the field studies. Sodar observations compared with tower-based measurements indicate the accuracy
ranged from -1.04 to 0.44 m s'1 while the precision ranged from 0.6 to 1.7 m s'1.  These studies also
concluded that the accuracy of the wind direction is about -3.0°, ranging from -6.8° to 4.0°. The
reported precision ranged from  18.4° to 37.6°.                                     j

        Unlike the data from in situ instruments, the quality of data from a sodar is a function of
atmospheric  conditions. When turbulence is low, the signal-to-noise ratio is low and it becomes
increasingly difficult to determine the frequency shift of a return echo. When wind speeds are low,
small errors in the horizontal velocity components can lead to large errors in the  estimate of wind
direction.   The  variability in the estimates of wind  speed is also partially  based on  the
inhomogeneities within the sample volume.  For each sample height, the return frequencies are plotted
and analyzed for a peak frequency, which is used to determine the Doppler shift.  The estimated
Doppler shift is then used to determine the average velocity within the sample volume.  When the
winds are inhomogeneous in the sample volume, the peak in the frequency plot becomes broad, and
thus determination of the peak frequency becomes difficult.  The error in determining the peak
becomes an error in accuracy when compared with the true wind.  The problem is compounded when
determining wind direction. Wind directions are calculated using the u and v velocity information
obtained from the off vertical sodar beams. The errors associated with determining the u  and  v
velocities are accumulated and transferred when computing  wind direction.

        Inversion height calculations, in some systems, are based upon the calculations of aw and oe.
Other  systems use  profiles of reflectivity to  estimate the inversion height.  In  these systems,
sophisticated pattern recognition algorithms are used to determine the height of the inversion layer.
                                           8-14

-------
 They are also capable of detecting multiple layers, if they exist
   i                     "                                           .
        Data recovery of sodars is highly variable and is dependant on atmospheric conditions at the
 various sampling heights. With sodars, it is common to have several levels of invalid or missing data.
 This is typically due to a lack of turbulence at those levels. It is up to the data analyst to interpolate
 or extrapolate missing wind data from the sodar output information. Weber and Wuertz (1991)
 describe a computer program that can be used to validate and fill in these missing data. However,
 care should be exercised so as not to smooth over any real data.  Sodars typically have good height
 coverage during daytime hours when there is strong mixing and there is sufficient turbulence to
 provide an adequate reflective  source.  However, turbulence above the inversion'layer may  be
 suppressed sufficiently to inhibit data capture.  In this case, no data would be recorded. This situation
 occurs frequently at night when the inversion height is below the maximum recording height of the
 sodar. Typical data capture ranges from about 50% to near 90% and is highly variable from hour to
 hour. Data capture here is defined as the percent of valid data received from the sodar during One
 averaging period divided by the total number of levels which the sodar is programmed to sample.
Figure 4,8.7 Sodar beam widths at acoustic frequencies of 1, 2 and 3 KHz with no acoustic absorbers
(Neff, 1988).                              ..;,,'•  :,vi,,   ,-.,.,
4.8.2.2  RADAR            J    ',  ' '''"' "'.'".;"'"••"   ':  ';""';'T^ ';'ii: '    •11'll"-v    v
                                   .  . -   , s ' ,   . , _    ..,.-.•  • •     'K   •           „
   ;    Radars are capable of measuring winds  to several kilometers with a vertical resolution
between 60 and 400 m.  However, range resolutions sHbuldbe kept near 100 m and the lowest
recorded height should be kept to a minimum (i.e., 150 m). Like that of sodar,"many radar options
include software subroutines that perform a variety of functions. It is important to purchase optional
QA/QC software, if available,to provide>an extra'level of data validation.'

       The operating frequencies of all EM devices; incluiding radars, are regulated by the Federal
                                           8-15

-------
Communication Commission (FCC).  The allocated frequency for radar wind profilers for general use
in the United States is 915 MHz, however, other permitted operating frequencies do exist.  This
frequency does provide the radar with good height range and minimum sample height, and should be
adequate for most meteorological applications.

       Like  sodars, data recovery of radars is a function of atmospheric conditions and is highly
variable.  With radar, it is common to have several levels of invalid or missing data.  This is typically
due to a lack of humidity and insufficient levels in the refractive index in the atmosphere at those
heights.  Vertical range of radars is also variable and is a function of atmospheric Conditions.

       During precipitation events, radars measure the fall velocity of the precipitation instead of the
air velocity.  During these events, radars may appear to be generating reasonable wind estimates,
however, it is more likely that the reported wind information is contaminated by the rainfall. During
rain events,  hail, or snow, the data should be flagged as suspect unless corrected  in software.
Assuming the vertical velocity averages to zero during the sampling interval, the vertical velocity
measured during precipitation events represents the fall velocity of the precipitation. Knowing this,
the horizontal winds can be corrected to an acceptable level, but the reported vertical velocity will
be meaningless. During short duration precipitation events, the corrupted data may be pulled out of
the data stream and an average produced using the remainder of the data set, thereby removing the
problem. The radar user should be familiar with how the software handles precipitation events and
should examine the data regularly to.determine if the software handled the  data processing correctly.
Typical data  capture efficiencies range from about 50% to near 90% and are variable from .hour to
hour.                                                                r       :.->;/

4.8.2.3 RASS                                   '"  "   v   '   : '  _     '  '. .;.."'.'.'' !;."'' "../"."

       RASS is an optional component of a radar system with the frequency of the acoustic source
matched with the radar frequency to obtain a maximum reflective source. The power Output of the
acoustic source  should be kept as high as possible to obtain the highest vertical level 6f virtual air
temperature as possible. Data capture efficiencies are usually good, ranging from 70°/o to over 90% .

4.8.3  INSTALLATION
                                                               .  -         •••-(•,   • V  . • .
       The following sections provide information on installation issues related tt> QA/QC concerns.
General information concerning installation and siting of remote sensing devices may be found in the
On-Site Meteorological Program Guidance for Regulatory Modeling Applications (V. S. EPA,
1987).                                                                          i -;>;'.:

4.8.3.1 SODAR                       '   ;'     ;  "   '     ' ?/;':"  ' /  ; ;;'.'T;:.-. '"'.", ............

       Siting of sodars can best be accomplished by vendors or users who have experience .with this
type of remote sensing device.  The complexities of. sodars provide a  challeiige to 'thfe' user who must
optimize the conditions favorable for sodar technology while still making use of available sites in a
                                           8-16

-------
given study area. It is suggested, until the time more data become available on proper installation
procedures, that the vendor or an experienced sodar user be called upon to aid in the site selection
and installation process.

       A problem may exist at some potential monitoring sites due to the presence of acoustically
reflective obstructions. The shapes of emitted acoustic pulses are not completely conical, but have
side lobes, that change shape and energy with frequency (see Figure 4.8.7). Reflective "fixed" echoes
occur when acoustic (sound) waves emitted from sodars are reflected back to the receiver by fixed
objects such as towers, buildings, trees, local terrain features, or other obstructions.  These fixed
echoes are often due to the energy contained in the side lobes of the emitted acoustic pulse.  These
fixed echoes have the effect of biasing the computed wind components u, v, and w.
  i                 '                              -           •      I
                                                                 '•>"•'•         t '
       It is extremely important to determine  if the proposed sampling site has any potential for
prpducing fixed echoes. Printing a facsimile chart sometimes reveals the; presence of fixed echoes.
This  should be  performed shortly  after  system setup,  and repeated seasonally  to aid  in the
determination if fixed echoes exists.  Some fixed echoes may be avoided by  constructing an
acoustically absorbing shelter around the sodar antennas.  These shelters  are designed to absorb most
of the energy released in the side lobes, providing a narrower beam, thuis a cleaner acoustic  signal.
In:  general, it  is recommended that the installer follow guidance provided  in the  On-Site
Meteorological Program Guidance for Regulatory Modeling Applications (U.  S. EPA,  1987).
Section 3.0.  Additional guidance includes the absence of obstructions in an 110°  arc centered  on the
vertical axis or 40° centered on each beam (see Figure 4.8,8).  In addition, if the system is to be
installed near a building,  the antennas should  be oriented off the corners of the building., If the
building does intercept the sound wave, the wave will be reflected away from the  sodar due to the
acute angles of the building's wall.  Some manufacturers provide software routines which can detect
fixed echoes and eliminate them from the consensus output.            |

       All attempts should be made to avoid fixed echoes. However, if a limited  number of sites are
available and all have a possibility of producing fixed echoes, then the fixed echo detection software
should be used to eliminate the problem. Special attention should be used during the acceptance test,
described later, to determine if the fixed echo rejection routines are working properly.

       The antenna does not necessarily have to point in one of the cardinal directions (i.e.,  north,
south, east or west). System software allows the sodar to be setup in almost any direction, allowing
the installer to point the beams away from obstacles that might interfere with the signal.  For example,
if the sodar is to be setup near a tower, the antenna should be oriented so the beams point away from
the tower.

       Another type of interference may occur from objects that emit noise such as local automobile
traffic, nearby construction and overhead aircraft.  Any acoustic source that emits its energy near the
transmission frequency of a sodar has potential for interfering and degrading the quality of the sodar
data. This type of interference is more difficult to .detect because it tends to be seasonal,  sporadic or
raiidom in nature. The potential for this problem may be reduced by installing acoustic absorbing
                                           8-17

-------
shelters around the transceiver arrays.  A simple test to determine if a problem exists at a given site
is to set up the sodar and turn off the transmitter. Analysis of received energy will determine if the
presence  of interfering  noise  exists.   If interference  from remote sources is  detected,  it  is
recommended that the sodar be moved to an alternate site.  The vendor or an experienced  sodar
operator should be consulted during the installation process to decrease the chance of contamination
of these data.

4.8.3.2 RADAR

       Siting a radar is somewhat more difficult than siting a sodar because of an increase  in the
potential for "ground clutter" to interfere with the return signal. Trees, power lines and even terrain
features just a few meters above the radar can produce erroneous data due to reflected EM signals.
Ground clutter often degrades  the signal  enough to render data useless, at least in the first few
reported levels. Obstructions also produce false echoes similar to that of sodars. These false echoes
also degrade the information in the first few reported levels.

       Like sodar, radar beams have side lobes which  emit energy to around 70°  from vertical
(Figure 4.8.9).  These side lobes cause a higher degree  of interference than sodars because radar
return signals are typically very weak, so small amounts of energy reflected back to the receiver may
cause large errors in the estimates of wind.                     '•'.-:•  :•'  Ij . ,   -     j

       Therefore, radars should be setup away from tall buildings, power lines and other obstruction
that may be a potential source of interference. The radar should also be situated on top of a small hill
or building to decrease the potential for ground  clutter contamination. ;The.antenna  does not
necessarily have to point in one of the .cardinal directions (i.e., north, south, east or west). System
software should allow the radar to be setup in almost any direction, allowing the installer to point the
beams away from obstacles that might interfere with the signal. For example, if the radar is to be
setup near a tower, the antenna should be oriented so the beams point away from the tower.

       The vendor or an experienced radar operator should,be consulted during the installation
process to decrease the chance of contamination :of these data.       .,    •     ;            M
4.8.3.3 RASS
       The user of a radar/RASS should follow the'guidelines for installing a radar, as specified in
Section 4.8.3.2. Contamination from external acoustic sources is only a minor problem but should
also be avoided as outlined-for ;so.dars in Section"4.;8.3,l. If a sddar/bistatic radar, is being used to
measure the virtual temperature then the installer should follow the guidelines for installing a sodar,
with the addition of meeting the recommendations, for installing a radar profiler.   ;      i:,  •>'
                                           8-18

-------
Figure 4.8.8  Obstruction free zone recommended for phased array and mono-static sodars.
4.8.4  ACCEPTANCE TESTING                          ;
 ;                                                                j
 !                                                                i • '
       Acceptance testing, as defined in Section 4.2.3, should be designed to determine if newly
purchased or installed equipment is performing according'tb the manufacturer's specifications. The
acceptance test is crucial for profilers since data produced by such instruments cannot be easily
verified by simple tests.  The following acceptance test is suitable for the sodar and maybe easily
modified for radar and RASS.    ..              ::::•<  •;"  .;.. •';,<-, ••-. .:•. •; ..;...•..-,;-.   -  •...-•
 ;      For meteorological remote sensors, an acceptance test should include' comparison of data
from the system to be tested with data from an acceptable in-situ sensor1 on a tower, tethersonde, a
mini-sodar, kite, NWS rawinsonde, or similar systems:  Although in-situ sensors do not qualify as
transfer (or reference) standards, they do possess the required sensitivity to determine if the remote
sensing device is operating normally (within  some broad limits).   The test should  include the
comparison of data at a minimum of three levels; all output generated by the remote sensing device
(e.g., wind speed, wind direction, virtual air temperature), should be included in the comparison. One
level should be the level of interest or application-of the remote sensor data -(i.e,, effective stack
height).    Some manufacturers  correct  the horizontal  wind'components  for vertical  flow
contamination. This correction is suitable in complex terrain where the 'average vertical wind velocity
may be other than Oms'1, indicating up-slope or down-slope fioWiiilt is: important to verify if this
correction is being performed properly by the'system,' If the, correctiori:is:being implemented, then
it should be applied during the acceptance test if one is comparing the device with another remote
sensor or anemometer that does not measure the vertical velocity. Figure 4.8.10 is a work sheet that
may be used for performing an acceptance test on a sodar using a tethersonde as the gauge. The
work sheet may be easily modified for use with other types of systems.
                                           8-19

-------
       Determination of atmospheric stability, by an EPA approved method, should be the first step
 in an acceptance test of a profiling system.  Atmospheric stability is important because it is an
 indication of the degree of turbulence present in the atmosphere.  As discussed in Section 4.8.1,
 atmospheric turbulence provides the mechanism to reflect the transmitted signal back to the receiver.
 Pasquill-Giffbrd (P-G) stability categories of B or C (DOE, 1984) are probably the most desirable
 conditions for performing this test. These two stability classes typically provide a reasonable amount
 of turbulence to reflect sodar signals back to the receiver.  In addition, the turbulence is such that it
 will not significantly "bounce" the tetherballoon, thereby avoiding unnecessary accelerations (which
 can introduce measurement errors) on the instrumentation attached to the tetherline.  Ideally, surface
 wind speeds should be steady at 2 to 5 m s"1. Wind speeds less than 2 m s"1 may be too variable for
 a reliable comparison, while wind speeds greater than 5 m s"1 will cause problems for the tethersonde
 as it is dragged out in more of a horizontal fashion rather than in a vertical profile.

       The tetherballoon should be situated downwind and far enough away from the sodar so that
 it will not interfere (i.e., reflect) with the acoustic signal. It is suggested that a facsimile chart, or
 some indication of signal intensity, be printed during the test to determine if the tethersonde is
 interfering with the sodar.  If a tower or other remote sensing device is being used then printing a
 facsimile chart is not required.  If the tethersonde is  interfering, it will show up on the facsimile chart
 as a solid line, (see Figure 4.8.11). The tetherballoon should be tethered at the first sampling height
 and  data collected for at least  15 to 20 minutes.  The time series information obtained from the
 tethersonde should match the time period for corresponding levels of the sodar sample.  Average
 wind speeds and directions from both systems, along with their  corresponding sample height, should
 then be entered into the work sheet. This procedure should be repeated to obtain similar information
 for at least two other heights.

       The next step is to subtract the time averaged wind speed obtained from the tethersonde from
that obtained from the sodar and record  this  information under the column titled "Wind Speed
Discrepancy ." Repeat this procedure for the  wind direction  information.  Determine the average
discrepancy for each section. If the absolute value of the average discrepancy is less than the sum of
the accuracies of the two instruments for wind  speed and less than the sum of the accuracies of the
two instruments for wind direction, then the profiler passes the acceptance test.  If the test fails, it
may be due to unsuitable atmospheric conditions  at the measurement heights, the winds are not being
corrected for contamination by the vertical velocity, or the average vertical wind velocity is other than
0 m s"1. The test should then be repeated during conditions more favorable for sodar operation, mid-
to-late morning, with  clear skies and 10m wind  speeds between 2 and 5ms"1.  If the sodar still fails
the acceptance test, it may be informative to repeat the test using the u, v, and w components instead
of the direction and speed information. This may reveal a mean vertical flow of something other than
0 m s"1, an error in orientation, or some other problem.

       The performance of meteorological  remote sensors is dependant on  meteorological
conditions. Recognizing this, the meteorological conditions  occurring during the test should be
documented.  This documentation should include the standard hourly observations including, current
weather, ceiling, sky-cover, ambient temperature, wind speed and wind direction.  An estimate of the
                                           8-20

-------
P-G class should also be included. If atethersonde is used, it should be located so as not to interfere
either with the tower sensors or the remote sensor.

       It is very important to make sure the comparison data are processed in the same manner as
the sodar or radar profiler being checked. The data from the tethersonde should be broken down into
its u and v components. At the end of this sampling period, the components should be averaged and
the resultant vector wind speed and wind direction calculated.

       At some sites it may be possible to use National Weather Service rawinsonde data to perform
the acceptance test.  This test is somewhat more difficult to perform but will provide the data required
to complete the test. The rawinsonde should be within 20 Km of the remote sensing site, in simple
terrain, and in  the same meteorological regime as that of the remote sensing instrument.   The
comparison should include  a data time  series long enough to have a  large sample for every
meteorological condition experienced at the site, and only data captured during similar meteorological
regimes at both sites should be used in the comparison. Data at higher elevations should be used for
the comparison since it is less likely that surface features will effect the data.

4.8.5  CALIBRATION AND PERFORMANCE AUDIT METHODS

       Calibration of meteorological remote sensing devices  is problematic since  there is no
correspondence with  calibration of in  situ  instruments.  Direct comparisons with rawinsondes,
tetherballoons, or instrumented towers are not always adequate because of the difficulty in comparing
point estimates with large volume estimates,  as well as the problem of separation in time and space
between the two platforms. Recent advances in QA/QC of sodars have led to the development of a
transponder (responder) unit that simulates returned echoes to a sodar.  Tliis device allows the user
to calibrate the instrument, much like using a constant speed motor to calibrate a cup anemometer.
Due to costs required to build a similar system for EM systems,  no  similar device has been developed
for radars.

       Derivations of the first moments (i.e., wind speed in the direction of the energy pulse) are
based on first principles. If the returned energy is strong enough, then reliable estimates of the radial
wind speed may be obtained.  From these derivations, an estimate of wind speed and direction are
produced.  If the  remote sensing device  has been calibrated, the meteorological conditions are
favorable for the system, has no system problems and the signal-to-noise ratio is high, then the data
produced may be considered of acceptable quality, assuming proper siting and calibration.

       Second moments produced by remote sensing devices  such as aw &nd oe are typically based
on statistics that are generated from wind speed and wind direction time series data.  Statistically,
these second moments  are derived from spatially averaged time series with a data  point being
produced every 5 to 15 seconds. When wind speeds are low, errors in the estimate of wind direction
increase.  Some manufactures use more sophisticated techniques to estimate aw and ae. These
techniques are usually statistically based and provide a more refined estimate of these values, during
certain atmospheric conditions.  However, they still do not provide reliable estimates for all
                                          8-21

-------
meteorological conditions. At this time ow and oe from remote sensors are not recommended for use
in regulatory dispersion modeling. Currently, there is an investigation underway to determine the
usefulness of the reported ow and  oe values for use in regulatory modeling studies.  If this
investigation shows that these values are adequate for modeling, the next revision to this document
will provide for their use in regulatory applications.

       For these reasons, calibration and performance audit techniques for remote sensors should
focus on the instrument electronics and other system components.  If practical, the acceptance test
should be repeated during the calibration process. This will ensure the highest quality data is being
obtained. The following sections provide initial guidance on calibration and performance audit
techniques for sodars, radars, and radars with RASS.

4.8.5.1  SODAR

       To derive an estimate of the  radial wind velocity, the sodar analyzes the frequency of the
returned echo.  The difference between the transmitted  and returned frequency, the Doppler shift, is
then used to derive the estimate of wind speed in the direction of the propagating acoustic wave. If
atmospheric conditions are favorable and the signal-to-noise ratio is high, (i.e., a strong return signal
is received) then an acceptable estimate of the wind  speed within the sampling volume is produced.

       Inherent to  most  sodars is a  subsystem  designed  to identify malfunctions in  the
instrumentation. These subsystems differ with each manufacturer, but are of adequate sophistication     _
to detect most instrument failures.  These subsystems use both software and hardware to check    fUn
system components such as signal  amplifiers, analog-to-digital (A/D)  converters, and voltage
supplies.  In multiple transducer units, the transducers can be checked by comparing  their signal
strength  with their neighbor's signal  strength.  These tools should be used to  determine system
operation on  a  component basis.   To  determine overall system performance,  a transponder
(responder)  should be used to induce signals into  the system to determine if the instrument can
correctly process  the information.   At  a minimum level, the calibration should  include feeding
frequency shifted information into the transceiver array.  If the information is analyzed correctly
(within specified limits), then the calibration can be considered acceptable. If not, then  the system
should be serviced by the manufacturer. Manufacturer's instructions for performing  these system tests
should be followed until guidance is generated to standardize the procedures. In'general, calibrations
should be performed on a semi-annual basis, and whenever the system is moved  or updated.   '•> ••'•'

4.8.5.2 RADAR

       Radar systems use software and hardware similar to that of a sodar to determine individual
component operation. These checks are useful for determining if there are any component failures
in the system and should be performed frequently  enough to prevent long down times.  In most
situations, the software will provide the user with enough information to determine  which  component
is malfunctioning. Due to the immense cost of building a transponder for radar systems, a se'ries'bf
component tests is used to monitor system performance.  A series of test procedures defined to


                                           8-22

-------
thoroughly test  the  functionality of the  radar  should be implemented to  determine system
performance.  Key system components such as gain, power levels, and. noise figures should be
included in the test.  Manufacturer's instructions for performing these systems tests should be
followed until guidance is generated to standardize these tests. Calibrations should also be performed
on radars on a semi-annual basis.                                 .!..,•
                                                                   I
       Tests are currently being conducted to determine a set of minimum requirements for
calibrating  sodars and radar. This section will be revised shortly to address any new requirements.
   1                                    •                              -               '
4.8.5.3  RASS

       RASS systems use acoustic waves to provide the radar with a well defined refractive index
structure for scattering the EM energy back to the receiver. The only real difference between a radar
and a radar with RASS (except for some additional  software) is the presence of acoustic sources.
These acoustic sources typically consist of four transducers,  one  placed on each side of the radar
antenna. Testing the radar component of the system should follow the guidelines discussed Section
4.8r5.2.  This test should be performed on a semi-annual basis or when the system is moved or
updated.                                                 "                      .
-  - i         .            . -.     .     . ••    .  -•.-      •..•;•••,.,••:..-i •••..•"•  :   •,-  •
4.8.6 OPERATION, MAINTENANCE AND QC
   !                  . •    .         -                        .        !          •     •
       Sodars, radars and RASS have automated operating systems and generally require minimal
input from  the user. Variables such as vertical range, range gates, averaging times, frequency, and
power output may be adjusted if needed, but most of the system operations are automatic. The .wind
data should be stored in its u, v, w components, as this will insure minimal loss of information and
more thorough data validation.  This will also, be useful in instances when the wind direction may be
in question.  Statistics such as number of valid return intensities and standard deviation of component
values should also be stored as this information may be useful in detecting instrumentation problems.
If a hard-disk drive is used for storing data, it should  be checked as often as; necessary to insure there
is enough room to store data.  This will avoid the potential for data loss due to insufficient disk.space.

       For the first few weeks after installation, the data should be checked on a daily basis -to
determine if the system is working properly:  Time series plots, of all variables should be produced
and analyzed by a meteorologist or other qualified professional. This step is important for detecting
any bias or anomalies in the data set.  It is usually at this point that false echoes are detected. All
inspections and maintenance activities should be documented  in a site log book.     ,       ,    ;

       After a time when the site operator determines the system to be operating adequately, data
should be plotted and checked on a weekly basis to determine system performance. This information
is useful to aid in the evaluation of the system.  For instance- :data.at certain, heights are not recorded
during particular meteorological conditions but arefine at other times. .This information can also be
used as an  aid in determining .system performance when the system appears to be malfunctioning.
                                           8-23

-------
       Maintenance should include weekly checks of the antenna array, cables, and all connections.
The antenna and antenna shelter should be checked and cleared of any debris.  All cables should be
systematically checked for any breaks due to weathering, animal bites or cuts due to human activities.
If damage is detected, the cable should be immediately replaced. All other connections should be
checked to insure proper operation. If manufacturer supplied hardware diagnostic routines are not
automatically initiated, then they should be performed manually on a weekly basis.

       Systematic routines used to inspect these data provide a level of quality control (QC).  These
QC checks should be performed by a meteorologist or other qualified professional who is familiar
with the physical nature of profiler data sets.  Such a person will more than likely spot and correct
any problems.  Without a qualified inspector, the potential exists for data to be corrupted and go
unnoticed.

       When a problem is found by the QC inspector, a discrepancy report should be issued which
brings the users into the data QC loop.  Their inspection and corrective action is reported back to the
QC inspector closing the loop.  With such a QC loop, the measurement system can be operated "in
control" and valid data produced.

4.8.7  ESTIMATING ACCURACY AND PRECISION

       At the present time, there are no accepted procedures for performing  adequate calibrations
to define system accuracy and precision of sodars, radars, or RASS.  The difficulties were discussed
in previous sections and will not be repeated here. New studies are necessary to.provide valuable
information on sodar and radar performance.  These studies should enlighten our understanding of
remote sensor performance and characteristics. At the completion of these studies, EPA will revise
this Section to include any new information.
                                           8-24

-------
Figure 4.8.9  Radar beam pattern for 924 MHz vertical beam with no clutter fence.
                                           8-25

-------
Instrument Type_
Instalment Serial No.
Acceptance Test Report by_
Specified Accuracy:   Wind Speed
                     Wind Direction
Tethersonde Serial No.
Atmospheric Stability Surface Observations_

Number of Minutes in Average	
                                                 Date:

                                                 Time:
                                    _(deg)

                                       Sonde Type_
Height
 (m)
  Average Sodar
Wind Speed (m s'1)
Average Tethersonde
  Wind Speed (m s"1)
Wind Speed Discrepancy
        (ms-1)
If absolute value of average discrepancy is s 1.0 m s"1, then system passes test (initial)_

If absolute value of average discrepancy is > i.O m s"1, then system fails test (initial)	
Height          Average Sodar
 (m)          Wind Direction (deg)
                         Average Tethersonde
                         Wind Direction (deg)
                               Wind Direction
                             Discrepancy (deg)
Figure 4.8.10  Worksheet for computing sodar discrepancy
                                           8-26

-------
Figure 4.8.11  Sodar facsimile chart depicting tetherballoon interference (shown by arrows).
                                           8-27

-------
4.8.8  REFERENCES

Balser, M., C. A. McNary, A. E. Nagy, R. Loveland, and D. Dickson, 1976:  Remote wind sensing
      by acoustic radar, Journal of Applied Meteorology, 15, 50-58.

Beran, D. W., 1975: Remote Sensing Wind and Wind Shear System. Interim Report. FAA-RD-74-3.

Beran, D. W., and S. F. Clifford, 1972:  Acoustic Doppler measurements of the total wind vector.
      Second Symposium  on Meteorological Observations  and Instrumentation,  American
      Meteorological Society, Boston, MA, pp. 100-110.

Beran, D. W., C. G. Little, and B. C. Willmarth, 1971:  Acoustic Doppler measurements of vertical
      velocities in the atmosphere. Nature, 230, 160-162.

Brown, E. H., and F. F. Hall, 1978: Advances in atmospheric acoustics, Review of Geophysics and
    '  Space Physics, 16, 47-110.

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.  Journal of Atmospheric and
      Oceanic Technology, 3, 594-604.

Gaynor, I E, and G. P. Ye, 1993: Simulation of RASS temperature using fast response temperature
      sensors  on a tall  tower. Eighth Symposium  on Meteorological Observations and
      Instrumentation, American Meteorological Society, Anaheim, CA, January 17-22, pp. 298-
      303.

Gaynor, J., C. B. Baker, andB. D. Templeman, 1992: Fine time scale comparisons between Doppler
       SODAR  and sonic  anemometer-derived winds.  Seventh Symposium on Meteorological
       Observations and Instrumentation,  American Meteorological Society, New Orleans, LA,
      January 13-18, pp. 401-404.

Holzworth, G. C., 1964: Estimates of mean maximum mixing depths  in the contiguous United
       States.  Monthly Weather Review, 92, 235-242.

Holzworth, G. C., 1972: Mixing Heights, Wind Speeds, and Potential for Urban Air Pollution
       Throughout the Contiguous United States, Publication No. AP-101, Office of Air Programs,
       U. S. Environmental Protection Agency, Research Triangle Park, North Carolina.

Kaimal,  J. C., and D. A. Haugen, 1975: Evaluation of an acoustic Doppler radar for measuring winds
      'in the lower atmosphere. 16th Radar Meteorology Conference,  American Meteorological
       Society, Houston, TX, p. 312.
                                         8-28

-------
Kaimal, J. C., and D. A. Haugen, 1977:  An acoustic Doppler sounder for measuring wind profiles
       in the lower atmosphere. Journal of Applied Meteorology, 16, 1298-1305.
   I                        ~~N                                                  '
NefZ, W. D., 1988: Remote sensing of atmospheric processes over complex terrain.  Meteorological
       Monographs, 23.
                                                    '.          '            , '
               ,                   ,"'A",,.         i
Stull, R. B., 1988: An Introduction to Boundary Layer Meteorology, Kluwer Academic Publishers,
   ;    Dordrecht, 666 pp.

Weber, B. L., and D. B. Wuertz, 1991:  Quality Control Algorithm for Profiler Measurements of
       Winds  and Temperatures.   NOAA Technical  Memorandum  ERL WPL-212, Wave
       Propagation Laboratory, Boulder, CO, 32 pp.                 !

Wuertz, D. B., andB. L. Weber, 1989:  Editing Wind Profiler Measurements. NOAA Technical
   i    Report ERL/438-WPL-62, Wave Propagation Laboratory, Boulder, CO, 78 pp.

U. !S.  Environmental Protection Agency, 1987: On-Site Meteorological Program Guidance for
       Regulatory'Modeling Applications.  EPA-450/4-87-013, Research Triangle Park, North
   i    Carolina.

Ye, J. P., D. E. Wolfe, J. E. Gaynor, and D. C. Welsh, 1993:  A detailed comparison between wind
       profiler and tower measurements.  Eighth Symposium on Meteorological Observations and
   i    Instrumentation, American Meteorological Society, Anaheim, CA, January 17-22 pp 298-
       303.                                                     i
                                        8-29

-------

-------
                PAMS METEOROLOGICAL MONITORING GUIDANCE
  4.A.O INTRODUCTION
    !                                                              '
        The following section is an example of meteorological monitoring guidance tailored to a
  specific regulatory application. Most of the information given below is from the Quality Assurance
  Handbook for Air Pollution Measurement Systems, Volume IV:  Meteorological Measurements and
  other EPA documents.

  4.A.1  OVERVIEW

        Title 40 Part 58 of the Code of Federal Regulations (U. S. EPA, 1993) requires the States to
  establish  a network of Photochemical Assessment Monitoring Stations  (PAMS) in ozone
  nonattainment areas which are classified as serious, severe, or extreme. Each PAMS program must
  include provisions for  enhanced monitoring of ozone and its precursors such as nitrogen oxides and
  volatile organic compounds.  In addition, surface and upper-air meteorological monitoring is also
  required.  The Environmental Protection Agency's (EPA) authority for enhanced monitoring is
  provided in Title I, Section 182 of the Clean Air Act Amendments of 1990.

        The importance of high quality meteorological data for these nonattainment areas can not be
  overstated. Meteorology is a critical element in the formation, transport, and eventual destruction
  of ozone and its precursors. Consequently, meteorological data are essential to the development and
  evaluation of ozone control strategies (U. S. EPA, 1991). These evaluations include photochemical
  and receptor modeling, emissions tracking, and trend analysis.  This section provides  guidance for
  meteorological monitoring in support of PAMS. It is intended for use by Regional, State, and local
 EPA  personnel involved in enhanced ozone monitoring activities. An overview of the PAMS
 meteorological monitoring requirements is presented in Table 4. A. 1.

 4.A.2  PAMS SITES                                             !
                   •       .          .
    :    40 CFR Part 58 identifies up to four PAMS site types for a typical urban region. It is intended
 that meteorological monitoring activities will coincide with ozone and precursor sampling at each one
. of these sites.  Site #1 is intended as the upwind/background characterization site and is located in
 the predominant morning upwind direction near the fringe of the urbanized area. Data collected at
 this site are needed to establish ozone and precursor concentrations which may be advected into the
 PAMS area from other regions. Site #2  is the maximum ozone precursor impact site and is typically
 located near the downwind boundary of the central business district where maximum precursor
 concentrations are expected. A second Site #2 may be required in larger urban areas.  This additional
 site would be located near the edge of the central business district downwind of the second most
 predominant morning wind direction.  Site #3  is the maximum ozone concentration site and is
 typically located 15 to 45 km downwind of the urban fringe area.  Data collected at this  site are
 needed to monitor maximum ozone concentrations occurring downwind, of the area of maximum
                                           A-l

-------
precursor emissions. Site #4 is intended as the extreme downwind monitoring site and is located
beyond Site #3.  This site, which is downwind of the predominant afternoon wind direction, is needed
to characterize the extreme downwind transport of ozone and precursor concentrations. Sites #1 and
#2 are required for all PAMS networks whereas sites  #3 and #4 are population dependent.  Further
details on PAMS site types may be found in the Photochemical Assessment Monitoring Stations
Implementation Manual (U.  S. EPA, 1994).

4.A.3  SURFACE METEOROLOGY

       Guidance for surface meteorological measurements is provided in several documents.  They
include the On-Site Meteorological Instrumentation Requirements to Characterize Diffusion from
Point Sources (U. S. EPA, 1981); Guide to Meteorological Instruments and Methods of Observation
(WMO, 1983); Instructor's Handbook on Meteorological Instrumentation (NCAR, 1985); Ambient
Monitoring Guidelines for Prevention of Significant Deterioration (U. S. EPA, \987a); On-Site
Meteorological Program Guidance for Regulatory Modeling Applications (U. S. EPA, 1987b); and
Quality Assurance Handbook for Air Pollution Measurement Systems, Volume IV: Meteorological
Measurements (U. S. EPA, 1989).

       The surface meteorological variables to be measured at all PAMS sites include horizontal
wind speed and wind direction, ambient air temperature, and relative humidity.  Solar radiation,
ultraviolet radiation, barometric pressure, and precipitation are to be measured at only one site (either
Site #2 or #3).  Application areas associated with these measurements are indicated in Table 4.A.2.
A summary of instrument specifications for surface measurements are given in Table 4.A.3.

       Meteorological instrumentation should not be mounted on or near solid structures such as
buildings, stacks, water storage tanks, grain elevators, and cooling towers since they may  create
significant wind flow distortions.  Instead, these instruments should be mounted on an open  lattice
 10 m tower since this structure creates the least amount of wind flow distortion.  There are several
types of open lattice towers:  Fixed, tilt-over, and telescopic. A fixed tower is usually assembled as
a one-piece structure from several smaller sections. This type of tower must be sturdy enough so that
it can be climbed safely to install and service the instruments. Tilt-over towers are also one-piece
structures, but are hinged at ground  level.  This type of tower has the advantage  of allowing the
instruments to be serviced at the ground.  Telescopic 10 m towers are usually composed of three
sections, each approximately 4 m in length. The top section is the smallest in diameter and fits inside
the middle section which, in turn, fits inside the base section.  The tower can be extended to a height
 of 10 m by use of a hand crank located at the lowest section. The top of the tower can be lowered
 to a height of about 4 m providing easy access to the wind sensors. Telescopic arid  tilt-over towers
 are not generally recommended for heights above 10 m. Regardless of which type of tower is used,
 the structure should be sufficiently rigid and properly guyed to ensure that the instruments maintain
 a fixed orientation at all times.
                                            A-2

-------
                                    Table 4.A.1
      Question/answer overview of PAMS meteorological monitoring requirements
Question
Answer
Where?
i
When?


How Long?

How Many Sites?

What Interval?


What Variables?
All serious, severe, and extreme ozone nonattainnient areas.

Routine continuous monitoring during the PAMS monitoring
season (3 months per year minimum).       :

Until area is redesignated as attainment for ozone.

2 to 5 surface sites per network plus one upper-air site.

Surface: Hourly.
Upper-Air: 4 profiles per day (minimum).

Surface:  Wind speed, wind direction, air temperature, and relative
humidity at  all sites.   Solar radiation,  ultraviolet  radiation,
barometric pressure, and precipitation at only one site.

Upper-Air:  Horizontal wind speed and direction required.  Air
temperature highly desired. Vertical wind speed, relative humidity,
and barometric pressure optional.
                                   Table 4.A.2
                     Applications for PAMS meteorological data.
Variable Photochemical
Modeling
Wind Speed /
Wind Direction S
Air Temperature S
Relative Humidity S
Solar Radiation V
Ultraviolet Radiation V
Barometric Pressure /"
Precipitation
Diagnostic Receptor
Analysis Modeling
•s . s
s /
/
/
s ..'.••.
/
/
/ /
                                       A-3

-------
                                       Table 4.A.3
             Summary of sensor requirements for surface meteorological variables.
Variable
Wind Speed
Wind Direction
Air Temperature
Relative Humidity
Solar Radiation
UV-A&B Radiation
Barometric Pressure
Precipitation
Height
(m)
10
10
2
2
any
any
2
1
Range
0.5 to 50 m s-'
0 to 360°
-20 to 40 °C
OtolOO%RH
Otol200Wm-2
Otol2WnV2
800 to 11 00 hPa
0 to 30 mm hr1
Accuracy
±0.2 m S'1 + 5%
±5°
±0.5 °C ,
±3 %RH
±5%
±5%
±lhPa
±10%
Resolution
O.lms'1
1°
0.1 °C
0.5 %RH
IWmf2
0.01 Wm:2
0.1 hPa
0.25mm
Time / Distance
Constants
5 m (63% response)
5 m (50% recovery)
60 s (63% response)
60 s (63% response)
60 s (99% response)
60 s (99% response)
60 s (63% response)
60 s (63% response)
       The objective of instrument siting (horizontal and vertical probe placement) and exposure
(spacing from obstructions) is to place the sensor in a location where it can make measurements that
are representative of the general state of the atmosphere in the region of interest. The choice of a site
for a meteorological tower should be made with an understanding of the regional geography.  Ideally,
a meteorological tower should be located in  an open  level  area away  from the influence of
obstructions such as buildings or trees.  The area surrounding the site should have uniform surface
characteristics.  The specific site  characteristics should be well documented.  This is especially
important where terrain with significant topographic features may introduce different meteorological
regimes at the same time.  Secondary considerations such as accessibility and security must be taken
into account, but should not be allowed to compromise data quality.

       Although it may be desirable to collocate the surface meteorological measurements with the
ambient air quality measurements, this may not be possible at all PAMS sites without violating one
or more of these criteria.  Surface meteorological measurements in urban areas, where compliance
with the above guidance may be precluded by the close proximity of buildings and other structures,
present special difficulties. In such cases, the individual involved in the site selection needs to assess
the likelihood that the data which will be collected at a given location will be valid for the intended
application.  In all cases, the specific site characteristics should be well documented. This is especially
important in areas where surface characteristics and/or terrain are not uniform and whenever standard
exposure and siting criteria can not'be met.

       The recommended sampling interval of-the meteorological  sensors by the data acquisition
system is 10 seconds. Data for all variables should be processed to obtain one-hour averages.  The
observation time should correspond to the time at the end of the averaging period and should be
recorded as local standard time. For example, a recorded time of 1500 (3 p.m.) corresponds to the
                                            A-4

-------
  sampling period from 1400 to 1500;  The data acquisition system clock should have an accuracy of
  ±1 minute per week.                                                   .                y

  4.A.3.1 Wind Speed and Direction

         Horizontal wind speed (m s'1) and wind direction (degrees clockwise from geographical north)
  are,essential to the evaluation of transport and dispersion processes. Measurements of wind speed
  and direction are also important in assessing atmospheric stability and turbulence  Wind speed is
  typically measured with a cup or propeller anemometer; wind direction is measured with a vane.

     ;    The standard height for surface layer wind measurements is 10 m above ground level (WMO
  u 8 , ^ uK 1S mportmi that the tower be located in an area of level and open terrain.  The wind sensor
  should be sited such that the horizontal distance to an obstruction is at least ten times the height of
  the obstruction. An obstruction may be man-made (e.g., building) or natural (e.g., trees).
     |                  '                           '                  '  .
     :   The close  proximity of tall buildings in downtown urban areas will often preclude strict
  compliance with the above exposure guidance. In such cases, the wind sensor should be sited such
 ^measurements are reasonably unaffected by local obstructions and represent, as far as possible
 what the wind-at 10 m would be if there were no obstructions in the vicinity. Site characteristics
 should.always be My documented.  This is especially important when standard exposure and siting
 criteria can not be obtained. Evans and Lee (1981) provide a discussion on the representativeness
 ol 10 m wind data acquired in an urban setting where the average obstruction height is of the same
 order as :the wind measurement height.

        Turbulence in. the immediate wake of the tower (even a lattice type) can be significant Thus
 precautions must be taken to ensure that the wind measurements are not unduly influenced by the
 tower. The wind sensor should be mounted on a mast a distance of at least one tower width above
 the top of the tower, or if the tower is higher than 10 m, on a boom projecting horizontally from the
 tower. The  sensor should be located at a horizontal distance of at least twice the diameter/diagonal
 of the tower 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 (i.e., into the prevailing wind).

       A sensor with a high accuracy at  low wind speeds and  a  low starting threshold is
 recommended for PAMS applications.  Wind  speed measurements should be accurate to ±02 m s'1
 + 5/0 of observed speed from 0.5 to 50 m s'1 with a resolution of 0.1 m s'1. Light weight molded
 plastic or polystyrene foam should be employed for cups  and.propeller blades to achieve a starting
 threshold (lowest speed at  which a rotating anemometer starts and continues to turn and produce a
 measurable signal when mounted in its normal position) of < 0.5 m s'1. Wind vanes or tail fins should
 also be composed of light weight molded plastic or polystyrene. The distance constant (the distance
 of air passage, through the cup or propeller required for sensor to  indicate a 1  - 1/e or 63 2% step
change in the wind speed) should be < 5  m at standard sea level density (1.2 kg m'3) Wind direction
measurement should be accurate to ±5° with  a resolution of 1 °.  The  starting threshold (lowest speed
at which a vane will turn to within 5 ° of the true wind direction from an initial displacement of 10°)
                                           A-5

-------
should be * 0 5 m sl. The delay distance (50% recovery from a 10° deflection) should be < 5 m at
standard sea level density. Overshoot must be < 25% and the damping ratio should he between 0.4
and 0.7.

4.A.3.2  Air Temperature

       Air  temperature  (°C) is strongly  correlated  with  extreme  ozone  concentrations.
Consequently  it is an  essential variable for PAMS  applications.   There are several types of
temperature sensors; these include wire bobbins, thermocouples, and thermistors.  Platinum.resistance
temperature detectors (RTD) provide accurate measurements with a stable calibration over a wide
temperature range and are among the more popular sensors used in ambient monitoring.

       The temperature sensor should be mounted on the tower 2 m above the ground and away
from the tower a distance of at least one tower width from the closest point on the tower. This height
is consistent with World Meteorological Organization (WMO, 1983) and EPA standard monitoring
procedures  The measurement should be made over a plot of open, level ground at least 9mm
diameter  The ground surface should be covered with non-irrigated or unwatered short grass or, m
areas which lack a vegetation cover, natural earth. Concrete, asphalt, and oil-soaked surfaces should
be avoided  As such, the sensor should be at least 30 m away from any paved area. Other areas to
avoid include large industrial heat sources, rooftops, steep slopes, hollows, high vegetation, swamps
snow drifts, standing water, and air exhausts (e.g., tunnels and subway entrances). The sensor should
be located a distance from any obstruction of at least four times the obstruction height.
       Temperature measurements should be accurate to ±0.5 °C over a range of-20 to +40 °C with
a resolution of 0 1 °C. The time constant (63.2%) should be < 60 seconds.  Solar heating is usually
the greatest source of error and consequently  adequate shielding is needed to provide a representative
ambient air temperature measurement.  Ideally, the radiation shield should block the sensor from view
of the sun, sky, ground, and surrounding objects. The shield should reflect all incident radiation and
not reradiate any of that energy towards the sensor. The best type of shield is one which provides
forced aspiration at a rate of at least 3 m s'1  over a radiation range of-100 to +1100 W m .  Errors
 in temperature should not  exceed ±0.25  °C when a sensor is placed inside a.forced aspiration
radiation shield.  The sensor  must also be protected from precipitation and condensation, otherwise
 evaporative effects will lead to a depressed  temperature measurement (i.e., wet bulb temperature).

 4.A.3.3  Relative Humidity                                         •.        .   T .•-.:••
        Measurements of atmospheric humidity are essential to understanding chemical reactions
 which occur between ozone precursors and water vapor. The relative humidity (RH) is defined (List,
 1951) as the ratio of the ambient mixing ratio (w) to the saturation mixing ratio, (w,) at a given air
 temperature and barometric pressure, i.e.,           ,

                                      RH  .'m'—                        '"'   '   '     (1).
                                                ws

 The ambient mixing ratio is defined as the ratio of the mass of water vapor to the mass of dry  air.


                                           • ,A-6

-------
                                                                 The
The saturation mixing ratio is defined as the ratio of the mass of water vapc r
saturated with respect to a plane surface of water to the mass of dry air.
be determined if the relative humidity, air temperature, and barometric
computing the saturation vapor pressure (es) using the relation (Buckj 1981)
                                                         17.502T
                          es = [1.0007  + (3.46x10
   in a given volume of air
  e mixing ratio can easily
pressure are know by first
t~\rt i \
                                                                                     (2)
where 7*is the ambient air temperature (°C) and/? is the barometric pressure (hPa).  The saturation
mixing ratio is then computed using
                                                                                      (3)
where e is 0.622. Substitution of RH and ws into Equation yields the mixing ratio w.

       Other measures of atmospheric humidity include vapor pressure (hPai), dew point temperature
(°C), specific humidity (g kg"1), and absolute humidity (g m"3).  All variables except for the relative
humidity provide a complete specification of the amount of water vapor in the atmosphere.  However,
any  of these  variables  can easily be derived  from the relative humidity given the ambient air
temperature and barometric pressure.

       There are various techniques for measuring atmospheric humidity.  However, the emergence
of capacitive thin-film technology is now producing sensors which are reasonably accurate, reliable,
compact, and inexpensive. Crescenti and Payne (1991) compared thin-film relative humidity sensors
from two different manufacturers  and found that they  performed  quite well.  These sensors are
becoming more common as they are easy to install and operate.
                                                                   i • •
       The relative humidity sensor should be installed using the same siting criteria as that for air
temperature. The sensor should be housed in the same aspirated radiation shield as the temperature
sensor at a height of 2 m above the ground. The accuracy should be ±3 %RR over a range of 10 to
95 %RH (±5 %RH from 0 to 10% RH and from 95 to 100 %RH) and -20 to +40 °C. Resolution
should be 0.5 %RH with a time constant (63.2%) of < 60 seconds.
                                                                   I
   '    The thin-film elements of the humidity sensor must be protected from contaminants such as
salt?  hydrocarbons, and other particulates. These pollutants can easily corrupt the sensing element
and lead to failure of the instrument.  The best protection is the use of a porous membrane filter which
allows the passage of ambient air and water vapor while keeping out particulate matter.
                                           A-7

-------
4.A.3.4  Solar Radiation

       Solar (sometimes called shortwave) radiation is a measure of the electromagnetic radiation
of the sun and is represented as an energy flux (W m"2).  Solar radiation measurements are used in
heat flux calculations, for estimating atmospheric stability, and in modeling photochemical reactions
(i.e., ozone generation). The solar spectrum is comprised of ultraviolet radiation (0.10 to 0.40 um),
visible light (0.40 to 0.73 um), and near-infrared (0.73 to 4.0 um) radiation. About 97% of the solar
radiation incident at the top of the earth's atmosphere lies between 0.29 and 3.0 um (WMO, 1983).
A portion of this energy penetrates through the atmosphere and is received at the earth's surface. The
rest is scattered and/or absorbed by gas molecules, aerosols, various particulates, cloud droplets, and
ice crystals.

       A pyranometer is an instrument used for measuring energy fluxes in the solar spectrum. The
sensor  measures global solar (direct and diffuse) radiation when installed facing  upwards in a
horizontal plane tangent to the earth's surface. The sensing element of the pyranometer is usually a
thermocouple which is protected by a clear glass dome to prevent entry of wavelengths outside the
solar spectrum (i.e., long-wave radiation).

        Solar radiation measurements should be taken in a location with an unrestricted view of the
sky in all directions. In general, locations should be avoided where there are obstructions that could
cast a shadow or reflect light on the sensor. In addition, the pyranometer should not be placed near
light colored walls or artificial sources of radiation. In practice, the horizon should not exceed 5°,
especially from the east-northeast through the south to the west-northwest (65° to 295° azimuth).
A 5° horizon will obstruct only  about 1% of the global  radiation and thus  can  be considered
negligible.

        Pyranometers have no specific height requirement. Consequently, a rooftop usually makes
an ideal location for sensor placement. Lacking a suitable rooftop, an acceptable alternative would
be a location directly south of the meteorological tower. Regardless of where the pyranometer is
sited, it is important that the instrument be level to within 1° of horizontal.  Any tilt from  the
horizontal will introduce significant errors. (Katsaros and DeVault, 1986).  To facilitate leveling, most
pyranometers come with an attached circular spirit level.

        Solar radiation measurements should have a total system accuracy of ±5% of the observed
value with a resolution of 1 W m'2 over a range of 0 to 1200 W m'2. The time constant (99%) should
be i 60 seconds.  Manufacturer's specifications should match WMO (1983) requirements for  either
a secondary standard or first class pyranometer if reliable heat flux and stability parameters are to be
calculated (Table 4.A.4).  Photovoltaic pyrahdmeters" (which usually fall under second class
pyranometers) should notbe.use.d forJPAMS applications. While their cost is significantly less than
that of thermocouple-type pyranometers, their 'spectralresponse is limited only to that of the visible
spectrum.  In essence, these sensors are, nothing more than visible light indicators.     ,

-------
                                        Table 4.A.4                i
                         WMO (1983) classification of pyranometers
Characteristic Units
Resolution Wm"2
Stability "/oFSyear1
Cosine Response %
Azimuth Response %
Temperature Response % ;
Nonlinearity %FS ?
Spectral Sensitivity %
Response Time (99%) seconds
Secondary
Standard
±1
^
- <±3' •;'
<±3
; ±0.5- ;
±2 ,•' •.
, < 25
First ;
Class
'••±5 .1 ' --
• i
±2 |.
<±7 '• ;
<±5 !
.. \ . • r 1
. ±2 ;u'.;
• - ; ±2 .
:• ±5 :,'•
<60
Second
Class
±10
". ±5
<±15
<±10
±5*
±10 ..
<240
.:' '-' . :: :.. • ;. " •••; •. ••.':•>•• •-, • ;>• = •.'••• •••< •- • ' '• • •.. , •,- ' ? ( - '.•'•.
       Ultraviolet (UV) radiation may be divided into three sub-ranges (Table 4. A. 5).  Due to
stratospheric absorption by ozone,,UV radiation that reaches the  surface is usually limited to
wavelengths longer than 0.28 um (UV-A and UV-B ranges). The most inaportant photochemically
active chemical species at these wavelengths are ozone, nitrogen dioxide, and formaldehyde. All
three of these chemical species areimportant in the..chemistry of ozone formation.'

-   "    •;->:  -•••.,-••-! •••'••-•• :.--'-^.-, .-1-;;; ;,-.,:• .f.:-.-   Table4,A.5     i;V      • ,. .-L-. ,'     '-,-„:..i.,-.',...  .•
   ;          .,    ....[.  Ultraviolet radiation classifications(WMO, 1983!);
                                Type
                         Range
       .UV-A-
:;?V!;^rijv-B
 ;r= •;•:;"?. ;-.uv-G-:
                                            0.280100.315^
                                            : Mrs s?;.::-.o.,v'.v'-:'.: •   :
                                            'OdOO to 0.280 jan
      "Ultraviblet pyranometers which have a spectral response spanning both the UV-A and UV-B
(0.286 to 0.4001 m) ranges"afe?recomhie^nded for PAMS applicatibris.  The same siting criteria used
forsolar radiation measurements apply.  The UV sensor should hiaive an a[ccuracy^ofi5%'over the
range of 0 to 12 W m"2, a resolution of 0.01 W m"2, and a time constant (99%) of < 60 seconds.

-------
4.A.3.6  Barometric Pressure

       Barometric pressure (hPa) is useful for examining trends in the weather on the order of several
days or more. It is also essential for the calculation of thermodynamic quantities such as air density,
absolute humidity, and potential temperature.

       There are numerous commercially available pressure transducers which range widely both in
price and performance. Most of these sensors are capable of measuring barometric pressure with an
overall accuracy of ±1.0 hPa over a range of 800 to 1100 hPa, a resolution of 0.1 hPa, and a time
constant (63.2%) of < 60 seconds.

       The sensor can be place at the base of the tower or inside a shelter.  Ideally, the sensor should '
be placed at 2 m above the ground. If needed, the pressure at 10 m (pw) can be derived from the 2
m pressure (p2) by using the hypsometric equation
                                                                                      ,A.
                                                                                      (4)
where z2 and z,0 are 2 and 10 m, respectively, g is the acceleration due to gravity (9.81 m s"2), R4 is
the universal gas constant for dry air (287.05 J kg"1 K"1), and Tv is the mean virtual air temperature
(K) in the layer between z2 and z,0 which is computed by using
                                    rv= r(i +  o.6i w)                           .       (5)


where T\s the mean ambient air temperature (TEC) between z2 and z10) and w is the mixing ratio (g g'1).
The decrease in pressure between 2 and 10 m is 0.9 hPa for a typical ambient air temperature of 20
°C and a mixing ratio of 1 1 g kg"1 (75 %RH).  Altitude of the station above mean sea level and the
height of the pressure sensor above ground level should be  carefully documented.

       If the pressure sensor is placed indoors, accommodations should be made to vent the pressure
port to the outside environment.  One end of a tube should be attached to the sensor's pressure port
and the other ended vented to the outside of the trailer or shelter so that pressurization due to the air
conditioning or heating system is avoided. The wind can often cause dynamical changes of pressure
in a room where a sensor is placed. These fluctuations may be on the order of 2 to 3 hPa when strong
or gusty winds prevail.

4.A.3.7  Precipitation

       The total amount of precipitation which reaches the ground is expressed as the depth to which
it would cover a plane horizontal to the earth's surface in a given period of time. There are several
rain gauge variations, including tipping-bucket, weighing-bucket, capacitive-siphon, and optical. The
most common are the tipping and weighing-bucket which are cylindrical in shape with a 20 cm (8
inch) diameter collection orifice.


                                           A-10

-------
    ;   The rain gauge should be mounted on level ground so that its orifice is horizontal with the
earth's surface.  Obstructions (including the tower) should not be closer than two to four times their
height from the instrument. The ground surface around the rain gauge should be natural vegetation.
It should not be paved since this may cause splashing of rain into the gauge. The orifice of the gauge
should be mounted 1 m above the ground
    i                                                                ;
       Measurement accuracy for all types of rain gauges is influenced more by exposure than by
variations in  sensor, design.   High  winds generally cause an underestimation of precipitation.
Therefore, efforts should be taken to minimize the wind speed at the orifice, especially in open areas.
This is best accomplished with the use of a wind shield.  An example is the Alter type wind shield
which consists of a ring with 32 free-swinging separate metal leaves approximately 1 to 2 cm above
the colleption orifice.

    '•   The rain gauge accuracy should be ±10% of the observed value with a resolution of 0.25 mm
and a time constant (63.2%) of < 60 seconds.
                                                                    j
4.A.4  UPPER-AIR METEOROLOGY
    i
                                                                    I
       40 CFR Part 58 requires at least one upper-air meteorological monitoring system for each
PA1VJIS. affected area.  Profiles of wind speed and wind direction are needed for use in transport and
dispersion modeling. Profiles of air temperature are highly desired since this is a principle indicator
of atmospheric stability. Other variables which can be measured, but not required, include vertical
wind speed, relative humidity, and barometric pressure.  EPA currently does not have any specific
guidance on measurement levels and accuracies for any upper-air data. However, Tables 4. A. 6 and
4.A.7.are WMO (1983) guidelines which can be used, but not required, as a model by those agencies
responsible for implementing PAMS upper-air measurements.
                                           1  ....       •         ,'',*j

         "	'•^•'•-'.'•••\ -'-."••'         Table 4.A.6      .     '     /.         '
                   WMO (1983) observation levels for lower tropospheric
 • ;,.,,-  .,.,...      ;  . s   soundings for operational and research purposes, i
                    Variable
Interval (m) -...   Range (m)
,. Wind Speed and , . . , ... .
Wind Direction


Air Temperature and
'"Relative Humidity
,*.:.!.'. ,-..: ;.. ., :'•;...• i. .,'..., .-; >,• -.. . •
50
160'
200
300
20
50 ;v
'TOO J
' 0 to 300 ;
400 to 600
800 to 1200
1500 to 3000
0 to 300
3 50 to 1000
11 00 to 3 000
                                          A-ll

-------
                                       Table 4.A.7
                 WMO (1983) observation accuracies for lower tropospheric
                      soundings for operational and research purposes.
Variable
Wind Speed
Wind Direction
Accuracy
±0.5 m s'1
±10%
±10°
±5°

WS ^ 5 m s'1
WS > 5 m s'1
WS < 5ms'1
WS > 5 m s'1
                       Air Temperature    ±0.2 °C

                       Relative Humidity   ±5%        RH < 95%
                                          ±1%        RH>95%

       The upper-air measurements are intended, for more macro-scale application than the surface
meteorological measurements.  Consequently, the location of the upper-air site does not necessarily
need to be associated with any particular PAMS surface site. However, for convenience and logistics,
the upper-air site can be  collocated with a surface  meteorology station.   Depending on the
meteorological conditions typically associated with high ozone concentrations in a given PAMS area,
both upwind (Site #1) and/or downwind (Sites #3 and #4) sites may be appropriate locations for the
upper-air  monitoring.  Factors that  should be considered  in selecting a site for the upper-air
monitoring include whether the upper-air measurements for the proposed location  are  likely to
provide the necessary data to describe the meteorological conditions associated with high ozone
concentrations.   Additional upper-air  monitoring systems may be needed in areas  where
meteorological and photochemical processes are complex or where an internal thermal boundary layer
has a significant role in ozone formation and transport.

       A minimum of 4 profiles per day is required. These profiles should be acquired just prior to
sunrise when the atmospheric boundary layer is usually the most stable; during mid-morning when
the growth of the boundary layer is most rapid; during  mid-afternoon when the surface air
temperature is maximum; and during late-afternoon when the boundary layer depth is largest. The
implementing agencies should make every attempt to acquire profiles in the first several hundred
meters of the convective mixed layer.  It is highly desired to obtain profiles of at least 1000 m or to
the top of the convective mixed layer (which can easily exceed 2000 m on summer afternoons).
However,  not all measurement systems are capable of an extended height range.  The implementing
agencies are encouraged to acquire profiles with greater vertical range, higher resolution,  and on a
more frequent basis, if at all possible. Wind, temperature, and humidity profile data obtained by
nearby National Weather  Service (NWS)  radiosondes may be  used to partially fulfill  and/or
supplement the PAMS upper-air monitoring requirement.

       In addition to the above variables, estimates are also required for the depth of the atmospheric


                                          A-12

-------
 boundary layer or mixed layer (i.e., mixing height).  Reliable estimates of the mixing height are
-gssential to dispersion modeling because this is the depth through which vertical mixing of pollutants
 normally occurs.  The  degree  of dispersion within  the mixed layer is primarily a function of
 atmospheric turbulence (i.e., wind flow, surface heating). The mixing height can be determined based
 on air temperature, turbulence, and/or aerosol concentration data.

   ;     The EPA recommended method for estimating mixing height requires measurement of the
 vertical temperature profile (Holzworth,  1964; 1972).  In this method, the mixing height is calculated
 as'the level above the ground in which the intersection of the dry adiabat (9.8 °C km'1) from a mid-
 morning surface temperature and the sunrise temperature profile occurs.  This concept of a mixing
 layer in which the lapse rate is roughly dry adiabatic is founded on thermodynamic principles and on
 operational use in regulatory dispersion modeling over the last two decades. Comparisons of mixing
 height  estimates based on the Holzworth method with several other techniques indicate that all
 methods perform similarly in estimating the maximum afternoon mixing depth (Hanna, 1969; Irwin
 and Paumier, 1990).  The Holzworth method is normally preferred because of its simplicity.

   ;     Another simple method for estimating the mixing height is by using an air temperature profil
 to, derive a potential temperature profile. The potential temperature 0 of an air parcel is defined as
 the temperature which the air parcel would have if it were expanded or compressed adiabatically from
 its existing pressure and temperature to a standard pressure p0, which is generally taken as 1000 hPa
 (Wallace and Hobbs, 1977; Fleagle and Businger, 1980).  An expression for the potential temperature
 can be derived by combining the First Law of Thermodynamics and the Ideal Gas Equation in terms
 of pressure/? and air temperature Tas
                                                                                      (6)
 le
as
                                                                  !    •      '""
where Rd is the universal gas constant for dry air (287.05 J kg'1 K'1), and cp is the specific heat at
constant pressure (1004 J kg'1 K'1). Within a well mixed boundary layer, potential temperature is
nearly a conserved property, i.e., it remains a constant value.  The top of the mixed layer is typically
marked by a rapid increase of potential temperature with height.

   i    There are a variety of platforms for measuring upper-air meteorological data. These include
aircraft, tall towers, balloon systems, and ground-based remote sensors. As with any measurement
system, each has its advantages and disadvantages.,,. The variables that can be measured with each
upper-air system are summarized in Table 4.A.8. .Note that with the exception of aircraft and tower,
no one. upper-air measurement system is capable of acquiring all of the variables listed in the table!
Typical vertical ranges and resolutions for these systems are presented in Table 4. A.9.  The choice
of using any one or more upper-air measurement system is left to the discretion of the implementing
agency. The information presented below provides some general background for each type of upper-
air ;system.                                                        ,
                                          A-13

-------
                                       Table 4.A.8
Meteorological variables that can be measured with various upper-air monitoring systems. Variables
include horizontal wind speed and direction (WS/WD), vertical wind speed (W), air temperature (T),
relative humidity (RH), and barometric pressure (BP).
System
WS/WD
                                        W
                                              RH
                                                                      BP
              Aircraft

              Tower

              Radiosonde

              Tethersonde

              Radar

              Sodar

              RASS

                                        Table 4.A.9
            Typical vertical ranges and resolutions for upper-air monitoring systems.
System
Aircraft
Tower
Radiosonde
Tethersonde
Radar
Sodar
RASS
Range (m)
100 to 10,000
10 to 600
10 to 10,000
10 to 1,000
100 to 3,000
50 to 1,000
100 to 1,500
Resolution (m)
1
1
5
5' . .
60 to 100
25 to 50
60 to 100
 4.A.4.1  Aircraft

        Aircraft (both airplanes and helicopters) are the ultimate mobile observation station.  They
 are capable of traversing large horizontal and vertical distances in a relatively short period of time.
 This platform can be equipped with meteorological instrumentation and an assortment of chemical
 sensors Traditionally, aircraft are used for episodic field studies which often require extensive data
 sets for model evaluation. Lenschow (1986) provides an excellent overview of aircraft measurements
                                            A-14

-------
in boundary layer applications. While an aircraft can provide detailed atmospheric observations over
large areas, the total sampling time per flight (typically 6 to 8 hours) is relatively short because of fuel
considerations.  Aircraft may also be subject to Federal Aviation Administration (FAA) restrictions
on flight paths over urban areas. In addition, the operating cost for this type of platform is extremely
expensive.

4.A.4.2  Tall Towers                                              j
 ••;                 •                         •                      '       •
       In some instances it may be possible to use existing towers which may be located in PAMS
areas to  acquire vertical profiles of atmospheric boundary layer  data.   Radio and  television
transmission towers, which may be as tall as 600 m,  can be equipped with in-situ meteorological
sensors at many levels. An advantage to using a tower is the ability to run an unattended data
acquisition system. Also, data can be collected under all weather conditions.  However, the main
disadvantage of using a tower is the inability to determine the mixed layer height during most of the
day. When moderate to strong convective conditions exist, the mixed layer height easily exceeds that
of the tallest towers.  Another disadvantage is the potentially high cost of maintenance, especially
during instances when the instrumentation needs to be accessed for adjustments or repairs.

4.A.4.3  Balloon Systems                                          i

   :    Balloon-based systems offer  a relatively inexpensive  means for  upper-air  meteorology
measurements. There are two types of balloon systems: Radiosonde (sometimes called rawinsonde)
and tethersonde.

   :    The radiosonde is reliable, robust, light weight, and relatively  small.   The radiosonde is
expendable, and can be mass produced at low cost.  The radiosonde is comprised  of sensors, a
tracking device, and a radio transmitter. This sensor package is suspended from a hydrogen or helium
filled balloon and is released at the surface.  Air temperature is measured with a bimetallic strip,
ceramic semi-conductor, or a wire resistor.  The relative humidity is measured with a carbon hygristor
or a thin-film capacitive chip. The barometric  pressure is obtained with the  an  aneroid capsule.
Ground-based radar is used to determine horizontal wind speed and direction. The radiosonde is
capable of easily traversing the depth of the troposphere and reaching well into the stratosphere.

   ;    A tethersonde system is comprised of a tethered balloon with several sonde packages attached
to the line.  Variables measured include horizontal wind  speed and direction, air temperature, relative
humidity, and barometric pressure.   These data are telemetered to this ground by radio  or by
conductors incorporated within the tethering cable.  The tethersonde is capable of reaching altitudes
up to 1000 m. However, this system can only operate in light to moderate wind conditions (5 m s"1
at the surface, 15ms"1 aloft).  A tethered balloon may also pose as an aviation hazard and is subject
to FAA regulations. A permit must be obtained for permission to operate such a system.

   :    Low cost is the main advantage for these  systems, as well as ease -of transport and relatively
low maintenance. The main disadvantage for balloon systems is that they can be very labor, intensive,
                                           A-15

-------
especially if data are needed on an frequent basis. In addition, vertical wind speed can not measured
by either balloon system.

4.A.4.4  Ground-Based Remote Sensors

       Ground-based remote sensors have become effective tools for acquiring upper-air information
and have played an increasingly important role in atmospheric boundary layer studies.  However,
there is a distinct void in available  guidance needed to help potential users in the regulatory
community.  Because of their unique nature and constant evolution, EPA guidance for remote sensors
is more generic than that which already exists for many of the well established in-situ meteorological
sensors.  Efforts are underway to provide more clearly defined guidance and standard operating
procedures  which will appear in the next edition of the Quality Assurance Handbook for Air
Pollution Measurement Systems,  Volume IV: Meteorological Measurements (EPA, 1989).

       There are two basic types of remote sensing systems used to acquire three-component wind
velocity profiles: Radar (radio detection and  ranging) and sodar (sound detection and ranging).
Radars (also called wind  profilers)  transmit an electromagnetic signal (-915 MHz)  into the
atmosphere in a predetermined beam width which is controlled  by the configuration of the
transmitting antenna. Sodars (also called acoustic sounders) transmit an acoustic signal (~ 2 to 5
KHz) into the atmosphere in a predetermined beam width which is also controlled by the transmitting
antenna.  The radar has a range of approximately 100 to 3000 m with a resolution of 60 to 100 m.
The sodar has a range of about 50 to  1000 m with  a resolution of about 25 to 50 m.

       Both systems transmit  their respective signals in pulses.  Each pulse  is both reflected and
absorbed by  the atmosphere as it propagates upwards. The vertical range of each pulse is determined
by how high it can go before the signal becomes  so weak that the energy reflected back to the antenna
can no longer be detected. That is, as long as the reflected pulses can be discerned from background
noise, meaningful wind velocities can be obtained by comparing the Doppler shift of the output signal
to that of the return signal. A  positive or negative Doppler shift indicates whether the radial wind,
velocity is moving towards or away from the transmitting antenna. The attenuation of a transmitted
pulse is a function of signal type, signal power, signal frequency,  and atmospheric conditions.  Radar
signal reflection  depends primarily on the presence of an index of refraction gradient  in the
atmosphere which varies with temperature and humidity. Sodar signal reflection depends primarily
on the presence of small scale atmospheric turbulence. The reflected signals received by either a radar
or sodar are processed in a computer by signal conditioning algorithms.

       In order to obtain a profile of the three-component wind velocity (U, V, W), one vertical
beam and two tilted beams are needed. The two tilted beams are usually between 15 ° and 30° from
the vertical.  These two beams are also at  right angles to each other in azimuth.  Each antenna
transmits a pulse and then listens for the reflected signal in succession.   After all three antenna
perform this fbnction, enough information is available to convert the radial velocities into horizontal
and vertical wind velocities by using simple trigonometric relationships.           ,
                                           A-16

-------
        There are two types of antenna configurations for radars and sodars: Mono static, and phased
 array.  Monostatic systems consist of three individual transmit/receive antennas. Phased array consist
 of a single antenna array which can electronically steer the beam in the required directions.  Vertical
 panels (also known as clutter fences) are usually placed around the antennas.  This effectively acts
 to block out any stray side-lobe echoes from contaminating the return signal of a radar. For sodars,
 these panels cut down on the side-lobe noise which may be a nuisance to nearby residents and also
 prevents any background noise which may contaminate the return signal.

        A radio acoustic sounding system (RASS) utilizes a combination of electromagnetic and
 acoustic pulses to derive a virtual air temperature profile.  A RASS usually consists of several
 acoustic antennas placed around a  radar  system.  The antennas transmit a sweep of acoustic
 frequencies vertically into the atmosphere. As the sound pulses rise, the speed of the acoustic wave
 varies according to the virtual air temperature.  Concurrently, a radar beam, is  emitted vertically into
 the atmosphere. The radar beam will most strongly reflect off the sound wave fronts created by the
 acoustic pulses. The virtual air temperature is computed from the speed of sound which is measured
 by the reflected radar energy. The typical range of a RASS is approximately 100 to  1500 m with a
 resolution of 60 to 100 m.

        Unlike in-situ sensors which measure by direct contact, remote sensors do not disturb the
 atmosphere: Another fundamental difference is that remote sensors measure  a volume of air rather
 than a fixed point in space.  The thickness of the volume is a- function of the pulse length and
 frequency used.  The width of the volume is a; function of beam spread and altitude.

    i    Siting of these profilers is sometimes a difficult task.  Artificial and natural objects located
 near the  sensors can potentially  interfere with the transmission and return signals,  thereby
 contaminating the wind velocity data.      >                                           ;

        Since sodars utilize sound transmission and reception to determine the overlying wind field,
 a clear return  signal with  a sharply, defined atmospheric peak  frequency, is required.   Thus,
 consideration of background noise may put limitations on where a sodar.can be located. External
 noise sources can be classified as active or passive, 'and as broad-band {random frequency) or narrow-
 band (fixed frequency). General background noise is considered active and is broad-band.  If loud
 enough, it can cause the sodar software to reject data because it can not find  a peak or because the
 signal-to-noise ratio is too low. The net effect is to lower the effective sampling rate due to the loss
 of many transmission pulses. A qualitative survey should be conducted to identify any potential noise
 sources. A quantitative noise survey may be necessary to determine if noise levels are within the
 instrument's minimum requirements.            ••..-'..'

    i   Examples of active, broad-band noise sources include highways, industrial facilities, power
plaij.tSj.and heavy machinery. Some of these noise sources have a pronounced diurnal^ weekly, or
even seasonal pattern. A noise survey should at least cover diurnal and weekly patterns. Examination
of land-use patterns and other sources  of.information may be necessary to determine if any seasonal
activities may present problems.
                                           A-17

-------
       Examples of active, fixed-frequency noise sources include rotating fans, a back-up beeper on
apiece of heavy equipment, birds, and insects. If these noise sources have a frequency component
in the sodar operating range, they may be misinterpreted as good data by the sodar.  Some of these
sources can be identified during the site selection process. One approach to reducing the problem
of fixed frequency noise sources is to use a coded pulse, i.e., the transmit pulse has more than one
peak frequency. A return pulse would not be identified as data unless peak frequencies were found
in the return signal the same distance apart as the transmit frequencies.

       Passive noise sources are objects either on or above the ground (e.g., tall towers, power
transmission lines, buildings, trees) that can reflect a transmitted'pulse back to the  sodar antenna.
While most of the acoustic energy is focused in a narrow beam, side-lobes do exist and are a
particular concern when antenna enclosures have degraded substantially. Side-lobes reflecting off
stationary objects and returning at the same frequency as the transmit pulse may be interpreted by the
sodar as a valid atmospheric return with a speed of zero.  It is not possible to predict precisely which
objects may be a problem. Anything in the same general direction in which the antenna is pointing
which is also higher than 5 to 10 m may be a potential reflector.  It is therefore important to construct
an "obstacle vista diagram" prior to sodar installation that identifies the  direction and height of
potential reflectors in relation to the sodar.  This diagram can be used after some 'data have been
collected to assess whether or not reflections are of concern at  some sodar height ranges. Note that
reflections from an object at distance Xfrom an antenna will show tip at height ^fcos(d), where a is
the tilt angle of the antenna from the vertical.

       The radar, sodar, and RASS antennas should be aligned and tilted carefully as small errors in
orientation or tilt angle can produce unwanted biases in  the data. True North should also be
established for antenna alignment.   Installation of the antennas  should not be permanent since
problems are very likely to arise in siting the profilers in relation to the tower1 and other objects that
may be in the area. One final consideration is the effect of the instrument on'its surroundings.  The
sound pulse from a sodar and RASS is quite audible and could become a nuisance to residents who
      "                                .  ,     "..,.,'',•- •".'-'' '- 't   'Tft }£•*•"**••£,' 1 '* ;' f#q .1 4   ,f   ' '•
might happen to live near the installation site.                             .-..,,   ,

4.A.5  REFERENCES                              "   "  '  " "'   -^—v--••

Buck, A. L., 1981: New equations for computing vapor pressure and enhancement factor!  Journal
       of Applied Meteorology, 20,  1527-1532                            ;,    " ,

Crescenti, G. H., and R. E. Payne, 1991:  Evaluation of two types of thin film capacitiveTelative
       humidity  sensors for use on bupj'S and ships,. '-Seventh Symposium tin Meteorological
       Observations and Instrumentation, American Meteorological Society, New Orleans, LA, Jan.
        13-18, pp. 125-128.                             .
              **                -   •.•-,-     =  • i-^ i-jk^rt ,£  .; j,-v '»"*
Evans, R_ A, andB. E. Lee, 1981:  The problems of anemometer exposure in'urBHn'afeas - a wind-
       tunnel study. Meteorological Magazine, 110, 188-199. '          '    ,!:!"*"
                                           A-18

-------
Beagle, R. G, and J. A. Businger, 1980:  An Introduction to Atmospheric Physics.  Second Edition.
       Academic Press, Orlando, 432 pp.                          !           ,

Hanna, S. R., 1969: The thickness of the planetary boundary layer.  Atmospheric Environment, 3
       519-536.
  ]•-''                .                         I

Holzworth, G.  C., 1964:   Estimates of mean maximum mixing depths in the contiguous United
  ;     States. Monthly Weather Review, 92,235-242.                         .      ;

Holzworth, G. C., 1972:  Mixing Heights,  Wind Speeds, and Potential for Urban Air Pollution
       Throughout the Contiguous United States. Publication No. AP-101, Office of Air Programs,
       U. S. Environmental Protection Agency, Research Triangle Park, North Carolina.

Irwin, J. S., and J. O. Paumier, 1990:  Characterizing the dispersive state of convective boundary
       layers for applied dispersion modeling. Boundary-Layer Meteorology, 53, 267-296.

Katsaros, K. B., and DeVault, J. E., 1986:  On irradiance measurement errors at sea due to tilt of
  !     pyranometers.  Journal of Atmospheric and Oceanic Technology, 3, 740-745.

Lenschow, D. H., 1986: Aircraft measurements in the boundary layer. Probing the Atmospheric
       Boundary Layer, American Meteorological Society, Boston, pp 39-55.

List, R. I, 1951: Smithsonian Meteorological Tables. Smithsonian Institution, Washington, D. C.,
       527pp.

National Center for Atmospheric Research,  1985:  Instructor's Handbook on Meteorological
  ;     Instrumentation. NCAR/TN-237+1 A, Boulder, Colorado.

U.  S. Environmental Protection Agency,  1981:   On-Site Meteorological Instrumentation
       Requirements to Characterize Diffusion from Point Sources. EP.A-600/9-81-020, Research
       Triangle Park, North Carolina.                              j  .    -.-•.-..-

U. S. Environmental Protection Agency,  1987a:  Ambient Monitoring Guidelines for Prevention of
       Significant Deterioration (PSD).   EPA-450/4-87-007, Research Triangle  Park, North
       Carolina.

U. S. Environmental Protection Agency,  1987b:   On-Site Program  Guidance for Regulatory
  '     Modeling Applications. EPA-450/4-87-013., Research Triangle Park, North Carolina.

U. S. Environmental Protection Agency, 1989: Quality Assurance Handbook for Air Pollution
       Measurement Systems.  Volume IV:  Meteorological Measurements,  EPA-6QO/4-90-003,
       Research Triangle Park, North Carolina.     ,    ,  .......     '!    .   ~  !    ,
                                         A-19

-------
U S Environmental Protection Agency, 1991:  Technical Assistance Document for Sampling and
       Analysis of Ozone Precursors. EPA-600/8-91-215, Research Triangle Park, North Carolina.

U S Environmental Protection Agency, 1993: Ambient'Air Quality Surveillance; Final Rule. Code
       of Federal Regulations, Title 40, Part 58, Office of the Federal Register, Washington, D. C.

U S  Environmental Protection Agency, 1994: Photochemical Assessment Monitoring Stations
       Implementation Manual.  EPA-454/B-93-051, Research Triangle Park, North Carolina.

Wallace, J. M., and P. V. Hobbs, 1977:  Atmospheric Science. Academic Press, New York, 467 pp.

World Meteorological Organization, 1983: Guide to Meteorological Instruments and Methods of
       Observation (Fifth edition). WMO No. 8, Geneva, Switzerland.
                                          A-20

-------

-------

-------

-------


-------